Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Morphometric and geotechnical controls of debris flow frequency and magnitude in Southwestern British… Jakob, Matthias 1996

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1996-147711.pdf [ 28.17MB ]
Metadata
JSON: 831-1.0087740.json
JSON-LD: 831-1.0087740-ld.json
RDF/XML (Pretty): 831-1.0087740-rdf.xml
RDF/JSON: 831-1.0087740-rdf.json
Turtle: 831-1.0087740-turtle.txt
N-Triples: 831-1.0087740-rdf-ntriples.txt
Original Record: 831-1.0087740-source.json
Full Text
831-1.0087740-fulltext.txt
Citation
831-1.0087740.ris

Full Text

MORPHOMETRIC AND GEOTECHNICAL CONTROLS OF DEBRIS FLOW FREQUENCY AND MAGNITUDE IN SOUTHWESTERN BRITISH COLUMBIA by MATTHIAS JAKOB Vordiplom, Universitat Regensburg, 1989 Diplom, Ruprechts-Karl-Universitat Heidelberg, 1991 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Geography) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA October 1996 © Matthias Jakob, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of ^ ° ^ ^ y ^ y The University of British Columbia Vancouver, Canada Date DE-6 (2/88) u A B S T R A C T Morphometric and geotechnical basin characteristics are examined to assess their influence on the frequency and magnitude of debris flows. Frequency is assessed primarily from the dendrochronologic record. Magnitude is determined from a combination of field surveys of deposits and empirical methods, and calculations of probable magnitudes reconstructed from field evidence of peak discharge. Typical magnitudes are determined for basins in which sediment supply and debris transport limit the range of possible debris flow magnitudes. Sediment storage in channels is quantified by using an erodibility index in order to quantify the largest expected magnitude. Although this index requires refinement, it points to the sediment distribution pattern along channels, and in some cases allows extrapolation to unsurveyed channel reaches. Debris flow basins are classified into weathering-limited and transport-limited types according to their debris availability. This classification proves useful when estimates of total event volume are made from peak discharge evidence, and significantly improves the explained variance in the prediction of both frequency and magnitude. Discriminant functions are then developed to formalize the a priori identification of basin type. A sample of 12 basins is used to test the discriminant functions. All test basins were correctly classified which highlights the usefulness and regional applicability of this approach. Morphometric and geotechnical variables were measured on debris flow basins and subjected to regression analysis to identify the most important variables in accounting for the observed debris flow activity. Validation by an independent data set has shown some promising results. However, the overall performance of the regression model indicates the difficulties in reliably estimating debris flow activity. Large deviations between the predicted and observed values probably reflect the insufficient quality of the test data set. The study emphasizes that simple univariate and bivariate models can be used to draw geotechnical envelopes for debris flow activity, but lack higher predictive power. Prediction can be significantly improved by classifying basins as transport- and weathering limited prior to the regression analysis. Both model establishment and model testing require higher quality data than are presently available, which raises the need for more regular monitoring of debris flow channels. Ill Longer records of debris flow frequency and magnitude will benefit geomorphologists in understanding the tempo of geomorphic change, and will aid engineers concerned with the appropriate management of sites and structures threatened by debris flow. iv TABLE OF CONTENTS Abstract ii Table of Contents iv List of Tables vi List of Figures vii Acknowledgement ix Chapter 1 INTRODUCTION 1 1.1 Objectives 1 1.2 Debris flow studies in British Columbia 5 1.3 Thesis outline 8 Chapter 2 T H E O R E T I C A L B A C K G R O U N D 11 2.1 Debris flow frequency 11 2.1.1 The problem of data censoring 11 2.1.2 Climatological controls 13 2.1.3 Non-climatological controls 16 2.1.4 Summary 19 2.2 Debris flow magnitude 19 2.2.1 High resolution magnitude determination 20 2.2.2 Prediction of design magnitude 21 2.2.3 Prediction of mean magnitude 22 2.2.4 Summary 24 2.3 Debris flow frequency-magnitude relationships 24 2.4 Summary 28 Chapter 3 STUDY A R E A 3 0 3.1 Regional overview 30 3.1.1 Geology 33 3.1.2 Geomorphology and Surficial Geology 34 3.1.3 Climate and Hydrology 35 3.2 Study sites at Mount Meager 40 3.3 Study sites at Mount Cayley 41 3.4 Study sites along upper Lillooet River valley 43 3.5 Study sites along the Fraser River valley between Lillooet and Lytton 44 3.6 Study sites in the Hope - Chilliwack area 46 3.7 Study sites at other locations 49 Chapter 4 M E T H O D S OF DETERMINING F R E Q U E N C Y A N D M A G N I T U D E OF DEBRIS FLOWS 5 0 4.1 Introduction 50 4.2 Criteria for basin classification 51 4.3 Determining debris flow frequency 53 4.3.1 Dendrochronology as a measure to determine debris flow frequencies 53 4.3.1.1 Study prerequisites 54 4.3.1.2 Process-response mechanisms 55 4.3.1.3 Dating of tree scars 56 4.3.1.4 Dating via reaction wood and stress-induced growth suppression 59 4.3.1.5 Dating of debris flows using adventitious roots 61 V 4.3.1.6 Presentation of results 63 4.3.2 Historical accounts and air photo analysis 66 4.3.3 Summary 67 4.4 Determining debris flow magnitude 70 4.4.1 Direct estimation or measurement of debris flow volume 71 4.4.1.1 Assessing deposit area 72 4.4.1.2 Assessing deposit depth 72 4.4.2 Empirical approach for estimating debris flow volume 73 4.4.2.1 Flow through channel bends 76 4.4.2.2 Channel geometry for debris flow discharge estimation 78 4.4.3 Typical magnitudes of debris flows 81 4.4.4 Maximum expected magnitude 84 4.4.5 Maximum expected magnitude - typical magnitude relationship 91 4.4.6 Summary 94 Chapter 5 P H Y S I C A L CHARACTERISTICS OF DEBRIS F L O W BASINS 9 6 5.1 Introduction 96 5.2 Use of morphometry in debris flow studies 96 5.3 Choice of morphometric parameters 99 5.3.1 Basin area and percent active area 99 5.3.2 Slope 100 5.3.3 Hypsometric integral 101 5.3.4 Basin relief and relief ratio 103 5.3.5 Roughness and drainage density 105 5.4 Geotechnical classification of source materials 108 5.4.1 Rock mass classifications 109 5.4.2 Point load tests as a surrogate for compressive strength of basin rocks 111 5.4.3 Terrain maps and slope stability mapping 113 5.4.4 Stability mapping in debris flow basins 115 5.4.4.1 Weighted stability number and active area ratio 116 5.4.4.2 Ground checking of stability mapping 119 5.5 Discussion 119 5.6 Summary 124 Chapter 6 STATISTICAL A N A L Y S I S FOR PREDICTING DEBRIS F L O W A C T I V I T Y 126 6.1 Introduction 126 6.2 A priori identification of basin type 127 6.2.1 Discriminant function analysis 127 6.2.2 Summary 132 6.3 Linear-regression model 133 6.3.1 Multicollinearity and variable selection of the regression model 137 6.3.1.1 Forward stepwise regression 138 6.3.2 Model refinement: residual analysis and outlier diagnostics 142 6.3.3 Model validation 145 6.3.4 Final regression model 149 6.3.4.1 Multicollinearity check 152 6.3.4.2 Discussion 155 6.4 Summary 157 Chapter 7 CONCLUSIONS 159 References 165 Appendix A: F R E Q U E N C Y D A T A 180 Appendix B: D E T A I L E D DESCRIPTION OF STUDY SITES 188 vi LIST O F T A B L E S 1.1 A summary of recent debris flow work in British Columbia 10 2.1 Comparison between debris flow frequency analysis and flood frequency analysis 13 4.1 Debris flow erodibility classification for channel beds 86 4.2 Debris flow erodibility classification for channel banks 86 4.3 Estimation of largest expected magnitude in selected debris flow channels 87 5.1 Use of morphometry in debris flow studies 97 5.2 Comparison between interpretive maps and the requirements of this study 114 5.3 Stability criteria of different sediments and bedrock 117 5.4 Calculation of Weighted Stability Number (WSN) for McLeod Creek 117 5.5 Geotechnical and morphometric variables for weathering-limited basins 123 5.6 Geotechnical and morphometric variables for transport-limited basins 123 5.7 Definitions of morphometric and geotechnical variables 124 6.1 Summary of stepwise discriminant analysis for classification of transport-limited 128 and weathering-limited basins 6.2 Squared Mahalanobis distance from group centroids for study basins 129 6.3 Classification scores for test basins 132 6.4 Pearson correlation matrix of dependent and predictor variables 138 6.5 Summary of preliminary regression model 141 6.6 Comparison between univariate and multivariate models using R2-statistics 141 6.7 Observed versus predicted, and data sources for the 12 test basins 150 6.8 Variance Inflation Factors (VIF) for final regression equations 153 6.9 Summary of regression results of the final model 154 Appendix A. Debris flow frequency data 180 vii LIST O F F I G U R E S 3.1 Generalized geologic map showing study sites and control basins in southwestern British Columbia 31 3.2 Mean monthly discharge for selected sites 38 3.3 Precipitation data for selected sites 39 3.4 Lineaments in the Fraser Valley and adjacent ranges 48 4.1 The concept of weathering-limited and transport-limited basins with regard to the occurrence of debris flows 52 4.2 Ponderosa Pine with multiple scars in Fool's Gold Creek debris flow channel 54 4.3 Spruce tree on Ferguson Creek fan, scarred in 1930/31 and 1965 57 4.4 Computer-generated image of tree core sampled on Gunbarrel III debris fan 61 4.5 Dendrochronological summary chart for Collis Creek combining information gathered from reaction wood, tree scars and historical documentation 65 4.6 Record length of all debris flow basins 68 4.7 Frequency distribution of all debris flow basins 68 4.8 Frequency distribution of debris flow magnitudes for all basins 71 4.9 Relationship between debris flow peak discharge and total volume for a worldwide data set 77 4.10 Relationship between peak discharge and total volume of debris flows from southwestern British Columbia 77 4.11 Relationship between peak discharge and channel cross-sectional area 80 4.12 Relationship between peak discharge and bankfull channel width 80 4.13 In-channel sediment storage for selected sites 87 4.14 Relationship between maximum expectable magnitude and typical magnitude 93 4.15 Relationship between maximum expectable magnitude and typical magnitude stratified into weathering-limited and transport-limited basins 93 5.1 Relationship between basin area and debris flow volume 99 5.2 Fresh debris flow deposits at the bottom of Angel Creek cirque, Mount Meager 104 5.3 Construction of the RUG-0 variable 107 5.4 Frequency plot of mean point load strength indices 113 5.5 Air photograph of McLeod Creek showing stability polygons 118 5.6 Influence and control of morphometric and geotechnical parameters on factors determining the frequency and magnitude of debris flows 120 6.1 Development of Final regression model 135 6.2 Log-transformation of the dependent variables FREQ, Qmax, VOL, and DFIA 136 6.3 Cook's distance for the volume variable 144 6.4 Deleted residuals versus residuals for Qmax of the combined data set 151 6.5 Deleted residuals versus residuals for Qmax after omission of Cheekye River basin 151 A-l Debris flow source area in upper Boundary Creek 189 A-2 Sackungen on ridge between Boundary Creek and No Good Creek 190 A-3 Smooth talus slope and erosion caused by debris flow in upper No Good Creek basin 191 A-4 Debris flow failure scarp in talus slope, upper No Good Creek basin 192 A-5 Air photograph of Boundary Creek and No Good Creek basins, on the north flank of Mount Meager 193 A-6 Prominent landslide scar in upper Canyon Creek basin 195 A-7 Deeply weathered and intensively fractured granitic rock, Hotsprings Creek basin 199 A-8 Air photograph showing the lower, active part of Hotsprings Creek basin 200 A-9 Debris flow in Turbid Creek, July 29, 1993 202 A-10 Toppling of heavily jointed bedrock at the head of McLeod Creek basin 208 A-11 Air photograph of Mount Currie northwest face 210 A-12 Overturned toppling bedrock at Mount Currie 211 A-13 Till failures along No Law Creek 212 A-14 Debris flow diamicton and talus slope at Gunbarrel II 216 A-15 Air photograph of Snake Gully debris flow basin 219 viii A-16 Contact between Spences Bridge volcanics and granitic rocks of the Mount Lytton Complex at Fool's Gold Creek 220 A-17 Steeply dipping joints at Kaboose Creek basin 221 A-18 Mount Ludwig debris flow channel along pronounced linear furrow defined by fault zone 224 A-19 Air photograph of Mount Ludwig Creek 225 A-20 Debris flow source area and colluvial channel in upper Ferguson Creek 232 ix ACKNOWLEDGEMENT My first thanks go to my supervisor Michael Bovis, for his lasting support, excellent advice, and constant encouragement through all phases of this study. I also wish to thank my advisory committee; Oldrich Hungr, Lionel Jackson, and Olav Slaymaker, for their interest and many helpful critical discussions. An earlier draft benefited from comments by Greg Henry and John Costa. Funding for the field work was provided by grants to M.J. Bovis from the Natural Sciences and Engineering Research Council of Canada. Living expenses for this period were covered by a scholarship of the Gottlieb Daimler - und Karl Benz - Stiftung. Jerry Dobry, Department of Forestry, had the kindness to let me use their dendrochronological image analysis program. Weldwood Ltd. supplied the record of debris flows for Turbid Creek. Steve Evans provided helicopter access to the debris flow source areas of two basins. Field assistance was provided by Drew Brayshaw, Elizabeth Bronson, Noel Castree, Ernst Jakob, Katherine McLeod, Steve Rice, Deepa Spaeth, and Scott Weston. I particularly want to thank Deepa Spaeth and Scott Weston for working enthusiastically far beyond their call of duty in sometimes dangerous situations. Special acknowledgements go to Martin Evans and Steve Rice, my office mates, for their willingness to discuss vital problems of my thesis, and for putting up with my awkward sense of humor. Brian Waddington, Betsy Fletcher, Markus Kellerhals and David Williams are thanked for taking me along on numerous ski-trips which enabled me to recharge for upcoming academic challenges. I would also like to thank my parents for always supporting my academic path. My four years at U.B.C. would not have been the same without the extraordinary friendship, cheery support, and sustaining motivation of my friend Elizabeth Bronson. CHAPTER 1. INTRODUCTION 1 1.1 OBJECTIVES AND PROBLEM APPROACH Debris flow is the rapid movement of a saturated, poorly sorted mixture of clastic and organic materials in a steep channel. In this thesis, the term debris flow is used to refer to mudflows (containing mostly sand, silt, and clay-sized particles), lahars (volcanic mudflows), and debris torrents (debris flows in steep confined channels with a high amount of organic material, and high water content). Comprehensive general reviews of debris flow have been provided by Costa (1984) and Johnson and Rodine (1984). In the Coast Mountains of British Columbia debris flow is one of the dominant mass movement processes and also constitutes a significant natural hazard. This dissertation concerns the influence of morphometric and geotechnical basin attributes on debris flow activity in southwestern British Columbia. The main objective of this study is to develop a statistical model which can account for documented variations in debris flow activity in 34 debris flow basins in southwestern B.C. and be applied to estimate debris flow frequency and magnitude in other debris flow basins in this region. In the context of this study, debris flow activity is defined as the product of debris flow frequency and magnitude. Frequency is defined as the number of debris flows per year, whereas magnitude is the total volume of material transported by a debris flow event beyond the apex of a debris fan. Although debris flows have been recognized as a frequently occurring and potentially destructive type of mass movement, there has been relatively little work done on frequency-magnitude relationships compared with, for example, hydrology where those relationships form the basis for hydrologic design. A sound knowledge of debris flow frequency and magnitude is important for two main reasons. First, without a knowledge of expected occurrence intervals, debris peak discharge and total debris volumes, neither hazard zonation nor mitigative structures can be designed with any confidence. Second, changes in the frequency or magnitude of debris flows, that might be indicative of changes in land-use or climate, can only be detected if data on past events are available. Apart from this applied consideration, the results obtained from this study can also be useful in landscape evolution models, by providing approximations of the 2 geomorphic work accomplished by debris flow. A comprehensive data base was compiled in this study. This, although incomplete, is a first step toward a regional data base which will only improve as more events are documented. This study tests two hypotheses: (i) frequency and magnitude characteristics of debris flow basins can be predicted from the morphometric and geotechnical characteristics of these basins in this region. (ii) a classification of basins according to their sediment supply conditions significantly improves predictive power. In this context basins are divided into (a) transport-limited basins in which there is an almost unlimited amount of sediment available for debris flows; and (b) weathering-limited basins, in which substantial recharge of sediment has to occur before the next debris flow is possible. This study, incorporates several factors that describe basin geometry and contain information on the stability characteristics of debris flow basins. It differs from previous studies that have attempted to determine frequency and magnitude by correlation with one or two morphometric predictor variables, which can only define rough envelopes at best (Hampel, 1977; Ikeya, 1981a,b; Okubo and Mizuyama, 1981; Takahashi, 1981, Watanabe, 1981; Ikeya and Mizuyama, 1982; Mizuyama, 1982; Kronfellner-Kraus, 1983; Thurber 1983; Hungr et al., 1984; VanDine, 1985, Jackson, 1987). The complex interaction between morphometric and geotechnical parameters within a typical debris flow basin calls for an approach that takes account of the various contingencies occurring in debris flow basins. Reconnaissance throughout southwestern British Columbia, identified 34 basins for detailed study according to the following criteria: - evidence of recent debris flow activity, evident either on air photographs or during reconnaissance field investigations; - potential to obtain dates from scarred and or broken trees along the debris flow path and in the depositional area; - absence of snow avalanche activity and rockfall in the depositional area to ensure that all observed scarring of trees was in fact due to debris flow activity alone; 3 - distance of less than 250 km of a site from Vancouver, B.C. permitting frequent visits should debris flow events occur during the course of the study; - accessibility of site by road or a short hike to minimize logistical difficulties. Implicit in the choice of the first two criteria is an emphasis on recently active debris flow basins that precludes those basins where either no obvious signs of recent debris flows were found, or those basins with debris flow return intervals exceeding 200 years. As a result, it is not possible to quantify the regional probability of debris flow occurrence on a time scale exceeding the period of record. Therefore, the thesis will focus on differences in activity between debris flow basins rather than discriminating between recently active and inactive basins. On a millennial time scale, significant changes in climate or changes wrought by large landslides, may reactivate basins originally classified as inactive. This possibility should be considered when permanent structures are planned on the fans of supposedly inactive debris flow basins. Recently active debris flow basins were identified by geomorphic criteria such as the presence of marginal levees bordering channels and steepfronted terminal lobes consisting of unstratified coarse, poorly to unsorted clasts. These criteria are usually well suited to distinguish between debris flow deposits and those created by water floods and hyperconcentrated flows (Costa, 1988). The selected basins span several lithological groups and include both wet and dry climatic sub-types in southwestern British Columbia. Ideally, it would have been preferable to keep either climate or lithology constant and change the second variable to isolate its influence. Although it is physically possible to stratify basins according to bio-geoclimatic zones, the large number of different lithologies, and the disproportional small number of basins in the dry region, do not allow a statistically meaningful treatment of these groups. Further complication is added by climatic differences between the individual basins that are somewhat difficult to characterize due to the paucity of weather stations in the vicinity of the basins. Precipitation data, which are usually obtained in valleys, have been shown to be very unreliable estimates of actual precipitation intensities in debris flow initiation areas (Thurber, 1983; Schaefer, unpublished reports; Church and Miles, 1987). Given these data limitations, climatic differences were not studied quantitatively, but are evident from both climatic normals and the natural vegetation. As more data 4 become available on debris flow events, particularly in the drier regions, a bio-geoclimatic index might be developable. This would resolve some of the inter-regional variability in debris flow frequency. Debris flow frequencies were determined by a variety of dendrochronological methods, supplemented by historical accounts and the analysis of low-level, high-resolution air photographs (cf. Chapter 4). Debris flow volumes were measured either directly in the field or were calculated from an empirical function, relating peak discharge to total volume (Mizuyama et al. 1992). Determination of peak discharge by the latter method required the surveying of several channel cross-sections within a given debris flow channel. Since no direct measurements of flow velocity were available, evidence of super-elevation of a flow at channel bends was used to back-calculate velocity (e.g. Chow, 1959; Hungr et al., 1984; Jordan, 1994). Maximum expected magnitude, which reflect maximum flow volumes over an extended, but unspecified time period were determined from Hungr et al's (1984) erodibility index for colluvial channels (cf. Chapter 4). Although maximum expected magnitudes were not included in the statistical analysis, they were computed as an alternative method for estimating the volume of debris that could be entrained from a channel during a design event. This approach is appropriate for debris flow systems in which most sediment is derived from the channel itself. In the part of the study dealing with the attributes of debris flow basins, morphometric factors were determined from standard topographic maps at a scale of 1:50,000; geotechnical parameters were determined from low-level air photographs at scales of 1:15,000 to 1:20,000. Field investigation of debris flow source areas served to supplement, confirm or correct the interpretations obtained from air photograph analysis. Areas of similar geotechnical characteristics were digitized from air photographs and classified to obtain indices relevant to the debris supply rate and overall erodibility of source materials. Geotechnical information from parts of rugged basins, inaccessible by foot, was gathered by helicopter landings or by fixed-wing flights. Morphometrical and geotechnical variables were then included in a multivariate regression model to determine predictive equations for frequency, magnitude and activity of debris flows in order to test hypothesis (i). Basins were classified according to their sediment availability, and the regression analysis was repeated to test hypothesis (ii). 5 1.2 DEBRIS FLOW STUDIES IN THE COAST MOUNTAINS OF BRITISH COLUMBIA The number of debris flow investigations in North America has increased significantly over the last 15-20 years. In British Columbia, debris flow research has been motivated primarily by practical considerations. Loss of life and destruction of bridges, roads, and buildings during a series of debris flows in 1981 and 1983 along Howe Sound, generated public awareness of this type of mass movement. This in turn triggered a series of studies that have examined the meteorological conditions responsible for the events; the design of mitigative structures; the development of methods to estimate flow peak discharge and total volume; and various studies of debris flow dynamics and debris flow initiation mechanisms (Thurber Engineering Ltd., 1983; Hungr et al., 1984, 1987; VanDine, 1985; Church and Miles, 1987; Jordan, 1987; Bovis and Dagg, 1988). The steep creeks along Howe Sound had been studied prior to the destructive events of the early 1980s (O'Loughlin, 1972; Russell, 1972), but their debris flow potential was not fully recognized, and in some cases debris flow effects were ascribed to stream floods. This same problem had also been recognized by Costa and Jarrett (1981) in small mountain streams in Colorado. Notable exceptions were the work of Nasmith and Mercer (1979), who recognized the geomorphic and destructive potential of debris flows in their work on debris flows at Port Alice on Vancouver Island, and Miles and Kellerhals (1981), in what must be regarded as a ground-breaking study of debris flow attributes. Another notorious area for debris flow activity is the south flank of the lower Fraser Valley between the towns of Chilliwack and Hope. Debris flow constitutes a significant hazard to the Trans-Canada Highway, the Westcoast Energy trunk gas pipeline, the Trans-Mountain trunk oil pipeline, the B.C. Tel fiber-optics telecommunications line, and several B.C. Hydro power lines that run parallel to the base of the mountain front for over 30 kilometres. A series of debris flows, triggered during a convective storm in 1983, were studied by Evans and Lister (1984), Church and Miles (1987), and Slaymaker et al. (1987). Two major storms affected the area again in November and December of 1995, triggering at least 10 large debris flows, one of which obliterated a restaurant and blocked Highway 3 at the town of Hope. 6 Another important hazard caused by debris flows is channel impoundment of higher order streams that are confluent with debris flow channels, causing rapid flooding upstream followed by extreme flood destruction downstream when the debris flow dam breaks (e.g. Jordan, 1987; Brooks and Hickin, 1991). Debris flow research has also been motivated in the context of logging and logging road construction. Much of the earlier work on this subject was carried out in coastal California, Oregon and Washington (Swanston and Swanson, 1976; Benda and Dunne, 1987; Benda and Cundy, 1990; Buchanan and Savigny, 1990). Wilford and Schwab (1981), Fannin and Rollerson (1993), Millard (1993), and Oden (1994) have investigated debris flows related to logging activities in the Coast Mountains of British Columbia. Most recently, the implementation of the Forest Practices Code of British Columbia (1995) has re-emphasized the importance of debris flows in forested terrain by providing guidelines to adequately manage steep, forested slopes. Much of the above cited work on debris flow activity has been carried out in the plutonic and metamorphic rocks of the Coast Mountains and North Cascades, and the less competent sedimentary and volcanic rocks of Vancouver Island and the Queen Charlotte Islands. Other areas of high debris flow activity in the region include the volcanic complexes of the Garibalidi Volcanic Belt: Mount Garibaldi, Mount Cayley and Mount Meager. Weakly consolidated pyroclastic rocks, tuffs, volcanic breccias and lava rocks frequently fail as landslides and debris flows in these areas (Mathews, 1958; Patton, 1976; Mokievsky-Zubok, 1977; Moore and Mathews, 1978; Clague and Souther, 1982; Jordan, 1987; Brooks and Hickin, 1991; Evans, 1986, 1987, 1992; Evans and Brooks, 1991; Cruden and Lu, 1992; Lu, 1993, Jordan, 1994). Debris flows from volcanic sources (lahars) have also been intensely studied in Japan, where they pose a much larger hazard because of the recent activity of many volcanoes and the extremely high population density in the immediate vicinity of active volcanoes (e.g. SABO, 1986; 1987). In North America, lahars have been studied in the Pacific Northwest by Crandell (1971), Wigmosta (1983), Fairchild (1985), Pierson (1985), Pierson and Scott (1985), Major and Voight (1986), Pierson (1986), Crandell et al. (1979), Scott (1988, 1989) and Scott et al. (1992, 1995). Many of these studies were conducted following the eruption of Mount St. Helens in 1980, the only North American example of a large recent eruption to be followed by a series of lahars. 7 Interest in debris flow studies in southwestern British was recently augmented by development plans on the fan of Cheekye River near the town of Brackendale, approximately 20 km north of Squamish (Thurber Consultants Ltd. and Golder Associates Ltd., 1992; Hungr and Rawlings, 1995; Sobkowicz et al., 1995). It was found that several very large debris flows and debris avalanches had detached from the north face of either Atwell Peak or Dalton Dome on Mount Garibaldi during the last 10,000 years, and there was reason to believe that events of similar magnitude could recur. Following the recommendations of a report by Thurber Consultants Ltd. and Golder Associates Ltd. (1992), further development on the most hazardous parts of the Cheekye fan was halted. Spatial as well as temporal prediction of debris flows requires an understanding of the various initiation mechanisms. Debris flows caused by jokulhlaups (glacial outburst floods) from ice-dammed lakes have been reported by Jackson (1979), Clague et al. (1985) and Jordan (1987). Given the very high number of pro-glacial lakes with unstable moraine crests, this type of debris flow initiation probably occurs quite frequently in the Coast Mountains of British Columbia, but remains largely undetected because of the remoteness of most of this mountain region. Research conducted by Bovis and Dagg (1987, 1988, 1992) advanced understanding on initiation of debris slides, sediment storage and channel sediment recharge. Jordan (1994) listed various initiation mechanisms that can or have occurred in British Columbia. These are listed in Table 1.1. Other work on debris flows from British Columbia has been added for completeness. Finally, a study should be mentioned that focused on large early Holocene debris flow fans because some of these fans are important for the study of contemporary debris flows. In the vicinity of the town of Lillooet, postglacial debris flow deposits constitute a major sediment source for modern debris flows. This study was carried out by Ryder (1971), who concluded that the fans were formed during the geomorphologically very active period immediately following deglaciation in the now semi-arid interior of British Columbia. In two study basins, the majority of sediment transported by contemporary debris flows is derived from reworking of the early Holocene debris flow diamicton. In summary, the southern Coast Mountains and Cascades of British Columbia delineate a region of high debris flow hazard which has led to some of the most intensive and extensive debris 8 flow related research conducted worldwide. Despite these efforts, more work is needed particularly in the fields of debris flow triggering mechanisms, climatic conditions that lead to debris flow initiation, and the terrain attributes which control debris flow frequency and magnitude analysis. The latter topic is the particular focus of this thesis. 1.3 T H E S I S O U T L I N E In Chapter 2,1 review the relevant literature on debris flow frequency and magnitude to summarize present knowledge and highlight major gaps in our knowledge. Chapter 3 provides an overview of the geology, hydrology, geomorphology, and climate of the study basins. The chapter is accompanied by an appendix detailing individual basins. Using this information it is possible to identify several useful predictor variables for debris flow frequency and magnitude. Methods used in this thesis are reviewed in Chapter 4. A preliminary classification of sediment availability is developed to allow the estimation of debris flow volumes via an empirical relation relating peak discharge to total volume. Maximum expected magnitudes are also calculated to provide an alternate means of determining debris flow magnitude. In Chapter 5,1 review the use of morphometric factors in watershed studies, and assess existing geotechnical mapping and classification schemes with regard to their potential use in this study. Several variables that characterize the geomorphic activity, slope stability, and sediment availability within a basin are developed specifically to suit the needs of this study. Interrelations between the variables are examined to understand causal mechanisms and to aid in the identification of data redundancies. The variables extracted and developed in Chapter 5 are then used to develop a multivariate regression model in Chapter 6 that serves to predict the peak discharge, total volume, frequency, and activity of debris flows. Chapter 6 also employs the tentative classification scheme presented in Chapter 4 to provide an objective means of classifying basins according to their sediment availability via discriminant-function analysis. Finally, Chapter 6 examines differences in the predictability of events between weathering-limited and transport-limited basins. Twelve basins with documented debris flow frequencies and magnitudes, provide an independent test sample to test the predictive capabilities of the statistical model. Following model validation, these basins are 9 added to the 34 basins already included in the model, and a new set of regression equations is produced. Chapter 7 summarizes the major findings of the study, and lists recommendations for a future refinement of the statistical model. 10 Table 1.1. A summary of recent debris flow work in coastal British Columbia Source and Initiation mechanism Publication Large landslides causing secondary debris flows Jordan (1987), Clague and Souther (1982), Evans (1992), Cruden and Lu (1992) Volcanic-source debris flows Patton (1976), Moore and Mathews (1978), Clague and Souther (1982), Brooks and Hickin (1991), Evans (1990) Evans and Brooks (1990), Cruden and Lu (1992), Lu (1988, 1992, 1993) Debris flows occurring as a consequence of Ryder(1971) degradation Debris flows originating from jokulhlaups Mokievsky-Zubok (1977), Clague et al. (1985), Jordan (1987), Jackson et al. (1989) Shallow debris slides entering the main debris flow Bovis and Dagg (1987, 1988, 1992), Fannin and ...J""?. Rollerson (1993), Rood (1984,1990) Debris flows triggered by critical stream discharge VanDine(1985) Debris flows and rainfall thresholds Thurber Consulting Ltd. (1983), Church and Miles (1987), Evans and Lister (1984), Miles (1984) Debris flows triggered by rock falls and rock slides Thurber Consulting Ltd. (1983), Bovis and Dagg (1987, 1988, 1992) Other debris flow related work Channel impoundment by debris flows Jordan (1987), Brooks and Hickin (1991), Evans (1986,1992) Debris flows associated with logging activities Wilford and Schwab (1981), Fannin and Rollerson (1993), Millard (1993), Oden (1994) Debris flow hazards and remedial design measures Nasmith and Mercer (1978), Miles and Kellerhals (1981), Martin et al. (1984), Hungr et al. (1984, 1987), VanDine (1985), Hawley (1989), Martin (1989), Morgan et al. (1992), Fannin and Wise (1995), VanDine (1996) Debris flow frequency and magnitude analysis Thurber Consulting Ltd. (1983, 85), VanDine (1985), Church and Miles (1987), Jordan and Slaymaker (1991), Oden (1994), Jakob and Bovis (1996) Documentary work and debris flow classification Evans and Lister (1984), Church (1985), Slaymaker et al. (1987), Slaymaker (1988) CHAPTER 2. THEORETICAL BACKGROUND 11 Magnitude and frequency studies have been widely used in geomorphology to (i) understand and model landform evolution; (ii) manage terrestrial and aquatic resources and (iii) analyze natural hazards. In this Chapter, relevant studies of magnitude and frequency of debris flows are summarized to provide a convenient framework for this study, and to point towards gaps in our knowledge, some of which this investigation attempts to fill. 2.1 DEBRIS FLOW FREQUENCY 2.1.1 The Problem of Data Censoring A sound knowledge of debris flow frequency is an important prerequisite for understanding and quantifying hazards associated with this process. Several approaches have been taken to determine debris flow frequency, and search for controlling factors. For convenience, these factors can be grouped under climatological and non-climatological headings. Common to both groups is the need for high-quality data which is usually severely constrained by a lack of continuous monitoring. Exceptions are several monitored debris flow basins in Japan (Suwa, personal communication, 1995). This limitation requires that debris flows are inferred indirectly from proxy datasets and datable evidence. Botanical-based proxy dating records such as dendrochronology and lichenometry, air photograph analysis, as well as radiocarbon dating, and tephrochronology will only record those (usually high magnitude) events which have left datable evidence. Small debris flows are difficult to identify on air photographs, often remain confined in the channel without causing recordable damage to trees, and may incorporate little dateable organic material. An additional complicating factor frequently encountered in steep mountainous terrain is the discharge of debris flows in higher-order streams leaving none or only a small fraction of the original debris flow material along the channel thus providing inconclusive stratigraphic resolution. Depending on the resolution and underlying assumptions of each dating method, low magnitude/high frequency data are underrepresented (i.e. censored). 12 There are several reasons why data censoring and data limitations increase the difficulties in applying the methods of flood frequency analysis to debris flows. The major differences are summarized in Table 2.1. Both debris flow frequency analysis and hydrologic frequency analysis have the objective of interpreting a past record of events in terms of their probabilities of future occurrence. In hydrology, this involves the selection of an appropriate data series to which a theoretical frequency distribution is fitted. Based on this distribution, inferences are then made about the underlying population. Several statistical criteria should be met to satisfy the requirements for this type of analysis. These criteria are randomness and independence of successive events, and homogeneity and stationarity of the phenomenon. These terms are defined in the context of hydrological frequency analysis by Watt (1989). Table 2.1 indicates that in particular the independence and homogeneity criteria are grossly violated in many debris flow basins. This suggests that a sophisticated statistical treatment of debris flow time series is unwarranted in these cases. Annual maximum series (AMS) are inappropriate for debris flows since events rarely occur every year, and unlike river flows are usually not continuously monitored. Partial duration series (PDS) is preferable because a time interval suitable for debris flow occurrence could be determined by choosing an appropriate magnitude base level. However, PDS requires independence of magnitude exceedance, and assumes an underlying distribution. As indicated in Table 2.1 independence may only be warranted in transport-limited basins. No agreement has been found on underlying theoretical frequency distributions, which depend on the process that is responsible for the frequency of debris flows (Ohmori and Hirano, 1988; Johnson et al. 1991). Costa (pers. comm., 1996) suggests that uncertainties in the magnitude estimates are greater than the assumptions inherent in fitting underlying distributions, and suggests the application of the Weibull formula [P= (n+l)/m]. Finally, debris flows can be initiated by a large variety of processes thus violating the homogeneity criterion. This criterion, however, seems to be of lesser importance given that floods are also created by different processes. This discussion highlights the important relation between frequency and magnitude of debris flows which is further discussed in Chapter 4. It also shows that the difference between flood-frequency analysis and debris flow frequency analysis is primarily one of degree rather than kind, particularly for transport-limited debris flow basins where sufficient data are available. 13 Table 2.1 Comparison between debris flow frequency analysis and flood frequency analysis Statistical prerequisites Debris flow frequency analysis Flood frequency analysis (PDS) Randomness (requirements fulfilled) variable fluctuations caused by natural factors are considered random which is a fair assumption for undisturbed basins non-regulated flows are generally considered random Independence (requirements fulfilled in some basins) cannot automatically be assumed because of supply limitations in weathering-limited basins; requires knowledge of recharge mechanisms independence usually assumed for systems with a lack of large storages; small deviations have little impact Stationarity (requirements fulfilled, but differences in controlling factors) valid only over limited time periods, changes in basin characteristics may be caused by a large landslide, fire, logging, glaciation, changes in long-term precipitation pattern valid for time periods without significant climatic fluctuations or abrupt changes in basin or river systems such as dams Homogeneity (requirements fulfilled in some basins( recurrence interval is controlled by a multitude of processes which are difficult to separate; recognition of these processes important to interpret results of frequency analysis elements from data series may stem from different populations (e.g. snowmelt rainfall, rain-on-snow), unstratified data still valid, but less accurate 2.1.2 Climatological Controls Climatological and meteorological conditions are often investigated as important parameters in the context of debris flow frequency, since it is clear that hydro-climatic events in excess of some threshold are often able to trigger debris flows, to ultimately exert an important control on their frequency. Long-term climatic fluctuations are also important because of their indirect regulation of debris flow frequency. In addition to changes in the type, intensity, and duration of precipitation, climatic fluctuations can cause glacial retreat or advance or destabilize steep talus slopes within the belt of alpine permafrost by melting interstitial ice (e.g. Jackson et al., 1987; Haeberli et al., 1990). These changes, which are caused by climatic amelioration, can ultimately exert control on debris flow frequency. In this section, some of the more recent work on these issues will be discussed to document present understanding of the relation between climatic factors and debris flow frequency. 14 A number of researchers investigated the correlation between various aspects of precipitation and debris flow frequency: Campbell (1975), Nielson et al. (1976), Wolman and Gerson (1978), Suzuki (1979), Caine (1980), Govi and Sorzana (1980), Suzuki and Kobashi (1981), Larson (1982), Hungr et al. (1984), Jackson et al. (1985), Nyberg (1985), Schaefer (1983), Van Dine (1985), Cannon and Ellen (1985), Keefer et al. (1987), Kobashi and Suzuki (1987), Wieczorek (1987), Church and Miles (1987), Cannon and Ellen (1988), Mark and Newman (1988), Wieczorek and Sarmiento (1988), Lips and Wieczorek (1990), Strunk (1991), and Wilson and Wieczorek (1995). In many of these studies, the lack of rain gauges close to debris flow starting areas posed a major limitation. In addition, most debris flows occur at unknown times during rain storms, which increases the difficulty in tying them to precipitation intensity and precipitation summations. However, precipitation intensities measured at distant gauges may not be representative of the study sites due to differences in slope aspect and elevation. To calibrate the rainfall in the study basin, rain gauges should be installed for some time period at representative locations within the basin, then correlated with a station having a much longer record. These correlations have the potential to overcome some of the uncertainties related to localized climatic effects. Using this method, Buchanan et al. (1990) measured precipitation intensities at the research site that were 1.9 times greater than those registered at the nearest recording rain gauge, located 21 km from their study site in northwestern Washington. A somewhat different approach was followed by Caine (1980), who compiled data from different studies and plotted a rainfall intensity-duration envelope, above which debris flows were most likely to occur. Caine's envelope encloses the data analyzed by Bovis (1992) for a major debris flow cycle of November 1990 in the southern Coast Mountains of British Columbia. However, Church and Miles (1987) analyzed debris flows along Howe Sound and the lower Fraser Valley between 1980 and 1984, some of which fell below Caine's envelope; from this they concluded that Caine's precipitation thresholds held little promise for predicting the occurrence of debris flows in their region. Another conclusion drawn by Church and Miles is that debris flows do not always occur during periods of remarkable rainfall, rain-on-snow, or snowmelt, although snowmelt and rain-on-snow as a trigger for debris flows has been suggested by Sharp and Nobles (1953), Gontscharov (1962), Broscoe and Thompson (1969), Sulebak (1969), Azimi and 15 Desvarreux (1974), Morton and Campbell (1974), Owens (1974), Niyazov and Degovets (1975), Kemmerikh (1978), and Rapp and Stromquist (1979). Another factor that is thought to contribute to debris flow initiation is antecedent meteorological conditions, which can cause soil or snow saturation prior to a relatively small triggering storm. Cannon and Ellen (1985), as well as Wieczorek (1987), included antecedent rainfall with existing intensity-duration data, arguing that the degree of saturation of the soil preceding the triggering storm will influence the timing of failures. Wieczorek's threshold plots below Caine's envelope, whereas Cannon and Ellen's threshold for "abundant" debris flows lies above Caine's worldwide curve. Larson (1982) recorded debris flows that occurred after a heavy rainstorm in Spitsbergen. If plotted on Caine's (1980) rainfall intensity - debris flow initiation curve, it is obvious that the values obtained from Spitsbergen are significantly below Caine's threshold. Larson identifies the presence of permafrost as the decisive factor explaining the anomalously lower rainfall intensities responsible for debris flows in this area. Oversaturation of the active layer due to the impermeability of permafrost below was also noted by Rapp (1960). It probably also occurs in non-permafrost environments when extended cold spells during periods of no or little snow cover create a similar quasi-impermeable layer at some time during a rainstorm. The comparison of these data sets illustrates the extreme spatial variability and the contingency of control mechanisms for debris flow initiation, indicating the difficulty of identifying a common precipitation threshold. In summary, many studies have demonstrated a correlation between precipitation factors and the occurrence of debris flows, but the majority of investigations was unable to explain observed frequency or magnitude of debris flows based on precipitation only. The principal problems arising from these studies are (i) the inadequacy of the available monitoring networks; (ii) the lack of data on mobilizable sediment within a basin, which must exert a strong control on the potential for debris flow occurrence; (iii) the heterogeneity of surficial materials, and (iv) the variety of debris flow triggering mechanisms, which may depend on different precipitation thresholds, which themselves are highly variable over time. In light of these problems, it is appropriate to examine the non-climatic factors which have been considered as important debris flow controls. 16 2.1.3 Non-Climatological Controls Prior interest in non-meteorological controls on debris flow frequency can be considered in terms of two closely related issues that focus on (i) the debris supply mechanisms, and (ii) basin disturbance by changes in the vegetation cover. Several studies have examined debris recharge rates for basins in which the amount of erodible sediment is limited (Sharpe, 1974; Hungr et al., 1984; Innes, 1985; Church and Miles, 1987; Bovis and Dagg, 1987). Sharpe (1974) investigated the frequency of debris flows in the San Juan Mountains of southwestern Colorado. His study showed that their occurrence is related to meteorological events, as well as broad scale environmental factors and detailed site controls. Although he did not attempt the quantification of these controls, he emphasized the role of lithology, which in turn affects weathering and the rate of debris input to mudflow source areas. Hungr et al. (1984) made the observation that debris flow occurrence in channels with underlying bedrock is largely controlled by weathering and thus recharge rates, which is in accordance with Innes (1985) findings from southern Norway. In contrast to bedrock-controlled channels, Hungr et al. (1984) suggested that unconsolidated deposits will more likely respond to high precipitation intensity events because of the perennial availability of mobilizable debris. Similarly, Church and Miles (1987) found that bedrock geology, and thus recharge rates due to rock weathering, only provide a significant constraint if rock weathering is the dominant debris supply parameter. Glacial till and poorly consolidated colluvium would therefore mask the effect of bedrock geology in terms of debris flow initiation. Debris accumulation in steep mountain channels was addressed by Bovis and Dagg (1987) who concluded that over time, textural changes occur within the debris flow channels causing a small increase of the angle of shearing resistance, and a large increase in hydraulic conductivity. This information pointed towards the important link between sediment supply mechanisms and frequency and magnitude characteristics of debris flows. According to Bovis and Dagg (1987), changes particularly in hydraulic conductivity could lead to high-magnitude, low frequency debris flows in channels where gradients are too low to produce debris flows directly from hillslope failures. 17 Summarizing these findings, debris supply seems to exert an important control on the frequency of debris flows. It has been included in this study as an important geotechnical component to the statistical model introduced in Chapter 6. However, the direct quantification of debris recharge is very difficult if no long-term measurements are available. For this reason, the percentage and relative stability of easily erodable surficial materials has been used as a surrogate variable (cf. Chapter 5). The other non-climatic control exerted on the occurrence of debris flows is natural and artificial disturbance of the vegetative cover in debris flow basins. Natural disturbances include grazing and fire which can play a significant role in influencing the frequency of debris flows. For example, Jackson (1977), studied debris flow deposits near Big Sur, California, and concluded that mudflows during the previous decade were probably triggered by fires that had denuded a steep drainage basin and had removed protecting vegetation. According to his argument, the frequency of forest fires are a prime determinant of the frequency of mudflows. Similar to Jackson, Innes (1983) observed a substantial increase in the frequency of debris flows in the Scottish Highlands during a period of hill burning and overgrazing which had destroyed the protective moss layers and thus destabilized slopes. Wells (1987) emphasized the link between the occurrence of fire and subsequent channel recharge in a study of fire-debris flow relations in southern California. He concluded that the rate of dry ravel and rill formation accelerated subsequent to fires which enhanced surface erosion and runoff, causing extremely fast material recharge of channels. In addition, runoff is enhanced by the development of a hydrophobic ash layer which further reduces infiltration. According to Wells, these conditions have caused post-fire debris flows to occur even during minor precipitation events. Johnson et al. (1991) applied this knowledge by including a burn factor to their regression model which significantly improved the predictive capabilities in a study of debris flow frequencies in the same physiographic region as Well's study. Changes in vegetation, whether caused by grazing or burning, can dramatically change erosion and recharge rates, which in turn can result in higher debris flow frequencies until revegetation has returned the system to pre-disturbance conditions. Compared with dry regions in the southwestern United States, fire is relatively uncommon in the western Coast Mountains, but 18 was considered as a potential controlling factor within the Clear Range on the eastern side of the Fraser River where abundant fire scars on trees indicate frequent fires. Further investigation revealed that the majority of debris flow source areas are only sparsely vegetated, suggesting that basin disturbance by fire does not warrant the inclusion of this factor in the proposed model (cf. Chapter 6). In the southern Coast Mountains of British Columbia changes in vegetation cover are to a large extent related to logging activities. An increasing body of research is dealing with the impact of timber harvesting on the frequency and magnitude of debris flows and sediment transport (Swanston 1974; Swanston and Swanson, 1976; Wilford and Schwab, 1981; Sidle et al. 1985; Rood 1984, 1990; Fannin and Rollerson; 1993, Millard; 1993; Oden, 1994). Swanston and Swanson (1976) investigated the impact of clearcuts and logging roads on slope stability and the occurrence of mass movements. All of these authors have concluded that timber harvesting dramatically accelerates the frequency of debris flows. Records from the H.J. Andrews Experimental Forest in Oregon revealed that clear-cutting increased the frequency of debris flows by up to 9 times and road related failures were responsible for a frequency increase of up to 133 times (Swanston and Swanson, 1976). A comprehensive study was conducted by Rood (1984, 1990) who inventoried more than 1,300 slope failures on the Queen Charlotte Islands and compared logged and unlogged areas. His results suggest a factor of 30 increase in sediment production from landslides (including debris flows) following logging. This unusually high value may be explained by a combination of easily erodable bedrock and extreme precipitation on the Queen Charlotte Islands, which respond more sensitively to hillslope disturbances. Recently, the Association of US Forest Service Employees for Environmental Ethics (AFSEEE) released a report summarizing the landslide and debris flow response to a large rainstorm in February 1996. A total of 185 landslides and debris flows were documented by aerial survey, of which 114 occurred in clearcuts, 68 were related to logging road failures and 3 were unrelated to human activity, again emphasizing the impact of forest practices on the occurrence of debris flows. Quoted reasons for this increase in mass movement activity include the concentration of unstable debris in the channel caused by harvest, as well as possible post-harvest increases in peak discharge within debris flow gullies. 19 2.1.4 Summary Many researchers have related the frequency of debris flows to meteorological and climatological factors, as outlined in the previous section. Common to these studies is the observation that climatological conditions can only partially explain the occurrence of debris flows. Other factors acknowledged to control debris flow frequency include (i) the rate of recharge of sediment in a channel after a debris flow has occurred, (ii) changes in vegetation cover, caused by grazing, fire, or logging related activities, and (iii) the provision of trigger failures, e.g. by poorly constructed roads or modified drainage. The factor of seismic activity, suggested by Kotarba (1992), is unlikely to be a significant contributor, since saturated or near saturated conditions must coincide with earthquakes for seismic shock alone to be significant. The latter two points are of little significance in this study since vegetation has not been disturbed to a degree that warrants special treatment, and trigger failures related to artificial disturbances have not been observed. Sediment recharge, and sediment availability will therefore be emphasized in this study. 2.2 D E B R I S F L O W M A G N I T U D E Debris flow magnitude is defined in this study as the total volume or as peak discharge of sediment transported down the channel at least as far as the fan apex during a debris flow event. This implies that minor flows that stall within a main channel are not considered events. Debris flow volume and peak discharge can be expressed as either design magnitude or mean magnitude. The first is commonly defined as the largest conceivable event within the lifetime of a structure in the debris flow path (e.g. Hungr et al., 1984), whereas the latter is the arithmetic mean volume or peak discharge of all documented debris flow events. This separation provides a framework for an overview of previous work considering these issues. The reconstruction of debris flow magnitudes is constrained by data censoring as discussed in the first section. Deposits of earlier events are invariably reworked, obliterated or at least partially eroded by streamflow erosion, and the last debris flow, or simply masked by more recent debris flow deposits. While the width and length of a recent deposit can generally be determined, depth estimates on many debris fans are 20 usually approximate because of poor stratigraphic resolution, and a general lack of natural exposures. This general lack of high quality data is one reason why much debris flow work has tended to focus on debris flow frequency. Although this section concerns the prediction of debris flow magnitude, two notable studies are described which provided high resolution volumetric data without attempting to predict future debris flow occurrence. Design magnitude and mean magnitude prediction is the issue of section 2.2.2 and 2.2.3 Studies that investigated magnitude in conjunction with frequency are reviewed in section 2.3. 2.2.1 High Resolution Magnitude Determination Osterkampet al. (1986) used series of layered debris flow deposits exposed by stream incision to determine debris flow volumes for debris flows at Mount Shasta. Additional stratigraphic data was collected by excavating pits with the help of a backhoe. Air photographs and field work supplied areal extents of debris flow deposits which were supplemented by botanical evidence. Magnitudes determined by these methods were then stratified as high, medium and low classes. Podor (1993) employed electronic distance measuring (EDM) techniques to measure the area of debris flow deposits. Photogrammetric analysis of the pre-event fan topography allowed a volumetric assessment of the most recent debris flow. A model which assumes a proportional relationship between area and depth was developed and applied to older debris flows for which areas were determined photogrammetrically. The same debris flows that were assessed with regard to their volume were dated using dendrochronological methods. Both, the study by Osterkamp et al. (1986) as well as the study by Podor (1993) proved successful in quantifying debris flow magnitude, but available data were not used to predict future occurrence, which is the topic of the following two sections. 21 2.2.2 Prediction of Design Magnitude Much of the work presented here has been motivated by debris-flow related disasters that called for engineering solutions, which in turn require the knowledge of the potential size or design magnitude of debris flows. VanDine (1984) compiled a summary of channel parameters and other factors related to debris flows that had been used to determine debris flow magnitude by various authors in Japan, Austria, and Canada. A notable study was conducted by Kronfellner-Kraus (1983) who warned against the application of oversophisticated prediction formulas for the volume of extreme flows. In an analysis carried out on historically documented debris movements in the Eastern Alps, Kronfellner-Kraus found that the volume of extreme flows is related to gradient and catchment area of the debris flow system, but emphasized the importance of the regional geological and physiographic setting as well. Kronfellner-Kraus employed the method of indexing of variables, which introduces biasing multiplicative effects when variables such as gradient and basin area are intercorrelated. Despite his warning against the application of oversophisticated prediction formulas for the volume prediction of extreme flows, several studies have suggested that univariate or bivariate models do not offer a satisfactory degree of predictive power for hazard zonation (Hampel, 1977; Ikeya, 1981a; Ikeya, 1981b; Okubo and Mizuyama, 1981; Takahashi, 1981; Watanabe, 1981; Ikeya and Mizuyama, 1982; and Mizuyama, 1982). Another study that belongs in this group was conducted by Thurber Consultants Ltd. (1983) who compiled a comprehensive dataset on debris flow basins along Howe Sound. The design magnitude of future debris flows was estimated from the amount of debris stored in each creek, assessed from detailed field investigations. Hungr et al. (1984) used this dataset to correlate debris flow magnitude with drainage basin size, for cases in which topographic, geologic, climatic and hydrologic conditions could be considered relatively constant. However, field inspection revealed that these factors are highly variable, even within a relatively small region such as Howe Sound, which led the authors to the conclusion that basin size could only serve a first approximation of debris flow magnitude. Hungr et al. (1984) indicated that where most of the material is entrained from the channel, debris flow magnitude is reasonably well correlated with channel length. The authors suggested a procedure in which design magnitude is determined from 22 an erodibility index for colluvial channels which involves channel surveys with regard to geometry and sediment grain size (cf. Chapter 4). 2.2.3 Prediction of Mean Magnitude In most cases, debris flow magnitude is insufficiently well reconstructed precluding the direct calculation of design magnitudes. However, even short records allow the determination of mean magnitudes by dividing the total volume or peak discharge of all recorded debris flows by their number. Mean magnitudes can then used to map debris flow hazards (e.g. Jordan, 1987) or as dependent variables in regression models which relate them to predictor variables (Mizuyama et al, 1987; Cannon and Savage, 1988; Cannon, 1993). It should be noted in this context that mean magnitude has little significance in cases where development on the debris flow fan requires a calculation of design magnitude, which is defined as the largest magnitude which is conceivable in a basin within the expected lifetime of a structure. This is particularly important in transport-limited basins where debris flow magnitudes can range over orders of magnitude. However, in some weathering limited basins the average magnitude may be similar to the maximum expected magnitude that was recorded in the time frame chosen for this study. This issue is discussed in detail in Chapter 4. Jordan (1987) estimated debris flow magnitude based on field surveys and air photograph analysis in an investigation of the impact of mass movements on river channels in southwestern British Columbia. He grouped debris flow volumes in magnitude classes based on volumetric assessments of the debris flow deposits. Implicit in this treatment is the assumption that the recorded volumes represent an average including older non-recorded events. Prediction of future debris flow activity is based on the assumption that neither frequency nor magnitude will change significantly in the near future. Although some of Jordan's magnitude estimations are associated with large errors due to the lack of data of older events, this study provided an appropriate overview of potential debris flow and other mass movement hazards. Most studies have focused on volumetric assessments of debris flows. A notable exception is a study by Mizuyama et al. (1987) in which the authors attempt to predict peak discharge from 23 total volume data using an empirical relationship between these two measures of magnitude (cf. Chapter 4, section 4.4.2). This approach, of course, requires that at least the volume of the last debris flow is known, from which the peak discharge can then be calculated. Since this relationship is based on a regression model, calculated volumes or peak discharges are means according to least-squares fit along the regression line. Cannon and Savage (1988) and Cannon (1989,1993) suggested an empirical model for the volume-change behavior of debris flows that can be used for mean magnitude determination. This model is based on mass changes caused by channel entrainment or deposition as a debris flow moves along its channel. According to this model, debris flow motion stops when the volume of the flowing debris becomes negligible. Cannon (1989) suggested that the most important factors that control the rate of mass-change are slope, channel morphology, the strength of the moving slurry, and in some cases, vegetation pattern. Using these data (except for strength characteristics), Cannon (1993) developed a regression model that predicted the volume change of debris flows in the Honolulu area. It is important to note that this model only applies to a specific type of debris flow in an area with fairly homogenous rock and soil types. It also assumes that the original landslide scar volume can be used as the initial debris flow volume and that no debris is incorporated by channel erosion. Such a model would be most unrealistic in coastal British Columbia, since channel erosion is often the dominant sediment source. This severe limitation was recognized by Cannon (1993). The volume-change model has found application for debris flow hazard mapping in the Honolulu district (Ellen and Mark, 1993; Ellen et al., 1993). A similar concept was adopted by Fannin and Wise (1995) who presented a model that calculates the cumulative debris flow volume for each channel reach downstream. This model is based on field data from 449 debris flows in the Queen Charlotte Islands, which include the length width, scour depth and deposition depth of each segment along the channel (Fannin and Rollerson 1993). Multiple regression was then used to predict entrainment volumes and deposition volumes given a set of predictor variables describing the geometry of each channel reach. Comparison of results with data from an actual event show high coefficients of determination. It should be noted that these debris flow magnitude calculation are strongly dependent on the time elapsed since the last debris flow, because this determines the amount of material recharged to the channel in this period 24 (Oden, 1994). Recharged material in turn will change the channel geometry as well as the accumulated material volume per channel segment, which have been included as predictor variables in the statistical model. Therefore, it seems to be more appropriate in this case to address the predicted volume as time-dependent magnitude rather than mean magnitude. 2.2.4 Summary Studies that focus on the prediction of debris flow magnitudes are rare due to the poor preservation and stratigraphic resolution of older debris flow deposits, which renders their volumetric assessment very difficult. Two approaches have been followed, both of which are constrained by this problem. Maximum expected magnitudes can be determined statistically by correlating historical data of very large flows to basin parameters, or by determining the amount of potentially erodable material in debris flow channels. Mean or typical magnitude can be computed as the average of all recorded flows and then regressed against basin parameters, an approach that was followed in this study. Implicit in all of these methods is the assumption that magnitude characteristics derived from older events will not change in the near future. 2.3 DEBRIS FLOW FREQUENCY/MAGNITUDE RELATIONSHIPS The last two sections have indicated that frequency and magnitude are closely linked, pointing towards the need for frequency-magnitude relationships for debris flows. The fact that very few studies have focused on this issue can be explained by the difficulties in reliably establishing these relations for a temporally discontinuous process, which in most cases, is not routinely monitored. Volumetric estimation as derived from lake sediments, deltas, and fans can provide information about geomorphic work accomplished in certain time periods, but the individual processes that have acted to create a particular landform can rarely be clearly separated. Only Innes (1985), Ohmori and Hirano (1988), Scott (1989), Johnson et al. (1991), Thurber Consultants Ltd. and Golder Associates Ltd. (1993) and Oden (1994) have explicitly addressed frequency-magnitude relationships for debris flows. 25 Innes (1985) emphasized that theoretically derived frequency-magnitude relations established for hydrological events cannot be transferred to mass movement phenomena. He suggests that debris entrainable by floods is available in almost unrestricted amounts, whereas the amount of sediment for debris flows usually is not. As noted before, this does not necessarily apply in all cases since several basins in southwestern British Columbia contain almost infinite amounts of readily transportable sediment. Innes observed that debris flows were significantly larger in Norway than in Scotland, and exhibited a high spatial variability that would be expected from morphologic, lithologic and climatologic differences between the two study areas. He assumed that, at a given point in time, the frequency of flows will adjust to the rate of debris supply. This has also been suggested for sites in Japan (Okuda et al., 1980a,b). Innes concluded that the frequency of debris flows should decline exponentially with magnitude. This assumption may hold true for transport-limited basins because unlimited amounts of sediment are available, but is a questionable assertion for weathering-limited basins, because the amount of erodible material determines the shape of the frequency distribution. Ohmori and Hirano (1988) investigated a nine-year record of rocky debris flows compiled by the Japanese Ministry of Construction with regard to frequency-magnitude relationships. An examination of the size distribution of 1,739 debris flows showed that they followed an exponential function rather than a log-normal distribution, as is frequently assumed for hydrological events. Due to the lack of volumetric data, the authors used depositional area as a measure of debris flow magnitude. The important conclusion from this study is that low frequency-high magnitude events are most effective in the destruction and formation of mass movement landforms, which conflicts with the traditional assertion that medium frequency-medium magnitude events are most effective in landshaping (e.g. Leopold and Maddock, 1953; Wolman and Miller, 1960). This finding bears important implications for practical applications similar to those carried out by Johnson et al. (1991). The objective of their study was to develop a predictive model to estimate the frequency and magnitude of debris flows in an area where stratigraphic data of earlier events are lacking. They used frequency analysis to estimate the expected debris flow volumes for debris flows in the Los Angeles area. Historical records ranged from 1 to 43 years. A specific 26 debris yield was obtained by dividing the recorded yield by the basin area and by the length of record in years. As opposed to Ohmori and Hirano (1988), who suggested an exponential function, the means and standard deviations of specific debris yields were based on the log-normal distribution, which was assumed to underlie the physical process. Expected annual debris flow volumes for intervals of 2, 5, 10, 25, 50 and 100 years were computed using a power-law regression model based on relief ratio, hypsometric index, fire recurrence interval, and basin area. Sufficient data on frequency and magnitude were available to construct a probability graph. This approach taken by Johnson et al. (1991) is probably the most desirable, but requires high quality data, which in this case, were available because most debris flows deposit in artificial catchment basins that are emptied after each event. This allows a determination of debris flow volume. No firm conclusions can be drawn with regard to the appropriateness of either the log-normal model used by Johnson et al. (1991) or the exponential power model preferred by Ohmori and Hirano (1988). More research is needed to determine whether a single underlying distribution is applicable to all cases, or whether different distributions must be used for different geotechnical and morphometric conditions. Scott (1989) investigated the recurrence intervals and peak discharge of very large lahars in the Toutle-Cowlitz River system, at Mount St. Helens, and associated the occurrence of lahars to periods of volcanic activity over a 40,000 to 50,000 year period. Scott also recognized the problem of data censoring of smaller flows which cannot be recorded, but stressed the importance of hazardous large events for planning purposes. For this reason Scott focused on lahars which traveled at least 20 km. Peak discharge was calculated from superelevations of deposits and lahar frequency was determined from a combination of radiocarbon dates, tree-ring analysis, and historic records. Scott concluded that a major lahar with a travel distance in the order of 50 km and peak discharge of 10,000 m3/s is likely to occur at recurrence intervals around 100 years, suggesting that land-use planning in the Toutle-Cowlitz drainage basins should accommodate these numbers. Large lahars are primarily associated with periods of volcanic activity in this area which implies that frequency and magnitude relations cannot be extrapolated to other areas. Large debris flows and debris avalanches not associated with volcanic activity, but still incorporated volcanic source material, have been observed at Cheekye River which drains the 27 western slopes of Mount Garibaldi in southwestern British Columbia. Magnitude - frequency relationships of debris flows in this basin have been established by Thurber and Golder Associates (1993) in what must be regarded one the most detailed investigations on a single debris flow fan ever conducted. Here, an extensive record of debris flow volumes and frequency was established from extensive stratigraphic investigations on the debris flow fan. Natural and artificial cross-sections at different locations on the fan provided data on the thickness and area of individual debris flows, which were dated by determining the age of fossil wood incorporated in deposits. This method was successful in identifying large (in excess of 1 million m3) debris flows, but failed to determine the frequency of smaller flows which left no or little deposits on the fan. However, historical records of these events indicate that their return interval is less than 50 years. Large events, which occurred over the last 6,000 years, were then ranked according to magnitude and a return interval, R, was calculated from R = (N+l)/m, where N is the number of years of record, and m is the rank. This procedure allowed the construction of a frequency graph in which debris flow volume is plotted against annual probability of exceedance. This report has shown that reasonable frequency - magnitude relationships of large events can be established if datable stratigraphic evidence is available. However, such a study can only be conducted if human life is at risk to justify the cost associated with this investigation. A different approach to establish frequency - magnitude relationships on a much smaller scale was used by Oden (1994), who investigated debris recharge rates in debris flow gullies on the west coast of the Queen Charlotte Islands with respect to changes in land use. Oden measured the accumulated debris in channels after a debris flow had scoured the channel to bedrock and developed a relationship between the time elapsed since the last debris flow and total debris accumulation in the channel. If the date of the last debris flow is known, this relationship provides a means of estimating the magnitude of future debris flows. Several attempts have been made to establish frequency-magnitude relationships for debris flows, and in cases where channels or basins were regularly monitored, or where good stratigraphic records were available, tentative relationships were developed which could have practical application. However, most debris flow channels in the Canadian Cordillera are not regularly monitored, unless they pose an immediate hazard to structures and communications. 28 Thus, in most cases the data record is too fragmentary and thus it is extremely difficult to establish reliable predictions of probable future events. 2.4 S U M M A R Y Neither meteorological nor single physical basin attributes have succeeded in adequately describing frequency and magnitude of debris flows. Due to data censoring by the debris flow process itself, and the lack of regular monitoring, complete records of frequency and magnitude are limited to few cases where frequent debris flows have posed a serious threat to settlements and structures and, as a result, catchment structures have been monitored over some period. Where no direct observations exist, debris flow frequencies and magnitudes must be inferred from depositional evidence. Due to data censoring, very small and very large debris flows remain undetected because they are either too small to be recorded by dendrochronological dating techniques, or occur too infrequently and are largely eroded or covered by younger deposits. Even if continuous records are available, individual debris flow events in many basins no not occur independently from each other. Conventional statistical theory as applied to flood frequency analysis requires both independence and sampling from the same population. Preceding events, however, can substantially alter boundary conditions for subsequent events, thus influencing their time of occurrence and magnitude. Different triggering mechanisms within the same basin are based on different thresholds, creating inhomogeneity within debris flow records because of variations in the causative physical processes. Analytical techniques such as partial duration analysis are also based on underlying distributions which are poorly developed for debris flow. Traditional frequency-magnitude analyses for debris flows as known from flood frequency analysis are in most cases scarcely warranted. Despite these data limitations, existing data are still of importance. For example, small debris flows which may remain unrecorded are irrelevant for most engineering applications, whereas very large events that occur outside the time window of investigation occur over time scales that exceed the life-time of most structures. This study has encountered problems similar to those which have emerged in previous studies dealing with debris flow frequency-magnitude relationships, principally a very fragmentary 29 record of events. However, as will be shown, it is possible to establish a regional data set of debris flow frequencies, and develop surrogate methods for magnitude estimation. Although the issues of frequency-magnitude relationships of debris flows can not be resolved completely, this work provides a substantial contribution to the understanding of frequency-magnitude characteristics in active debris flow basins. 30 CHAPTER 3. STUDY AREA This chapter provides a regional overview of the bedrock and surficial geology, climate, hydrology and geomorphology of the study area. These variables are considered to have a major influence on the occurrence of debris flows. The chapter begins with a summary of the geologic history of the study area, followed by a short account of the surficial geology and geomorphology. Basins are clustered into groups according to their location, and specific features are briefly characterized in the second part of this chapter. Additional physiographic information on each basin is compiled in Appendix B. This detailed description of each basin serves three purposes. First, it describes some basin characteristics that are not easily expressed in terms of numerical parameters. Although some basin attributes were not identified during the original parameter selection process, they can aid in identifying basin idiosyncrasies which are important in the interpretation of statistical results presented in Chapter 6. Secondly, data presented in this chapter and Appendix B help to define criteria appropriate for stratification of the total basin sample into transport-limited and weathering-limited sub-types, for the purpose of improving the predictive power of the statistical model. Thirdly, the information presented here and in Appendix B provides clues on debris flow triggering mechanisms that may, in turn, further our understanding of initiation thresholds and thus ultimately debris flow frequency. 3.1 R E G I O N A L O V E R V I E W The thirty four study basins are located within a mountainous region comprising approximately 25,000 km2. They are located in the drainages of Squamish River, Lillooet River, and Bridge River, as well as the sections of Fraser River between Lillooet and Lytton and between Chilliwack and Hope (Figure 3.1). 31 Figure 3.1 Generalized geologic map of southwestern British Columbia showing the location of study sites and control sites. 32 Legend to Figure 3.1 PRG Lakes, ocean Rivers, creeks Quaternary alluvium, glaciofluvial and lacustrine deposits, till, colluvium Quaternary Garibaldi Group (andesite, basalt and dacite flows and pyroclastic rocks) Eocene Allenby Formation (sandstone, shale, conglomerate, coal) CJB, CJH M g d / O g d Triassic Western volcanic fades, (mafic to felsic volcanic rocks, minor flows, pelite, sandstone, limestone) Carboniferous Bridge River (B) and Hozameen (H) Complexes (chert, argillite, mafic volcanic rocks, carbonate and associated gabbro, and ultramafic rocks, local melange) Carboniferous to Triassic Eastern belt (tectonic melange with chert, carbonates, basalt, gabbro, ultramafic and felsic rocks) Devonian to Permian Chilliwack Group (DPC) and Triassic Cultus Formation (TC) (mafic and felsic volcanic rocks, carbonate, pelite, sandstone, conglomerate, argillite, volcaniclastic sandstone) Miocene granodiorite of Mount Barr (M) and Oligocene Chilliwack (O) Pluton IKN Upper Cretaceous Nanaimo Group (sandstone, shale, conglomerate) Cretaceous Spences Bridge Group (mafic, intermediate, and felsic flows, pyroclastic rocks, sandstone, shale, and conglomerate) Cretaceous Gambier Group (intermediate felsic and mafic rocks, intercalated conglomerate, sandstone and shale) eKJ J KG JH Egd mK/eK/IK Cretaceous Broken Back Hill Formation (interbedded crystal tuff, volcaniclastic sandstone, phyllite, lapilli tuff, rhyolite, andesite and volcanic breccia Early Cretaceous Jackass Mountain Group (arkosic sandstone, siltstone, shale, conglomerate) Jurassic Cayoosh Assemblage (undifferentiated graphitic phyllite, siltstone, Lurbidite, shale, phyllitic quartzite, minor limestone, volcanic tuffs) Jurrasic Harrison Lake Formation (intermediate and felsic flows, pyroclastic rocks, argillite, conglomerate, siltstone, shale, and sandstone) Jurassic Dewdney Creek Formation (sandstone, siltstone, mafic to intermediate volcanic flows and related pyroclastic rocks) Jurassic Bowen Island Group (intermediate and felsic flows.volcaniclastic sandstone and intercalated pelites, minor carbonate and conglomerate ' ( C O L U l M B I A °*2F Study a r e a VICTORIA J mlj PT KTgn Early Tertiary granodiorite and monzogranite of Needle Peak, Mission Ridge, Mount Outram and Texas Creek Plutons Early, middle, and late Cretaceous Coast Belt (quartz diorite, granodiorite, tonalite, diorite, biolite, and tonalite) Early middle Jurassic to Late Jurassic Western Coast Belt (tonalite, quartz diorite, granodiorite, diorite and gneiss of Eagle Plutonic Complex) LateTriassic diorite amphibolite, and granodiorite of Mount Lytton Complex Undivided Permian granodiorite, diorite and amphibolite Late Cretaceous and Tertiary Custer Gneiss (pegmatitic granite, gneiss, with pelitic schist and amphibolite) Early and late Cretaceous Chism Schist (melange of pelitic schist, amphibolite, quartzite, phyllite, minor chert, limestone, and ultramafic rock) Early and late Cretaceous Slollicum Schist (mafic, inter-mediate, and felsic meta-volcanic rocks, pelite, minor volcanic- and carbonate clast conglomerate) Early and late Cretaceous Settler Schist (garnet-biotite, staurolite, kyanite and sillimanite schist, amphibolite meta-pillow basalt, schist) Early and late Cretaceous undifferentiated garnet-biotite, kyanite and sillimanite schist, amphibolite, felsic and mafic metavolcanic rocks Carboniferous, Jurrassic and Triassic undifferentiated rocks — * — - « . Strike-slip fault, arrows indicate relative sense of displacement Fault contact displacement uncertain, contact defined Fault contact displacement uncertain, contact assumed Reverse or thrust fault, teeth indicate upthrust block Source: Monger, J.W.H. and Journeay, J.M. 1994. Geology of the Southern Coast and Intermontane Belt. Geological Survey of Canada. Open file 2490. Scale 1:500,000 33 3.1.1 Bedrock Geology As discussed in the last chapter, various geologic factors have been cited as influencing the frequency and magnitude of debris flows. This section begins with a description of the general geologic history and structure of the study area, followed by a section on surficial geology and geomorphology. Figure 3.1 shows a generalized geological map of the area. The major structures are mid-Cretaceous to early Tertiary in age. They include systems of west and east-vergent contractional faults, right-lateral strike-slip faults, and Eocene extensional faults. The accretion of terranes to the western margin of the North American craton was accompanied by compression and by northwestward displacement of accreted rocks along high-angle reverse faults and strike-slip faults which resulted in a strong northwest-southeast structural and topographic grain (Monger and Journeay, 1994). The majority of sites examined in this study are located within the Coast Plutonic Complex (Wheeler and Gabrielse, 1972). This term, as well as others such as the " Coast Range Batholith" (Le Roy, 1908) or Coast Crystalline Belt (Roddick, 1966) reflect the dominance of plutonic rocks within the orogen. The southern Coast Mountains consist largely of upper Jurassic to lower Tertiary granodiorite and quartz-diorite. Sections of fault-bound stratified rocks, and various metamorphic rocks, can also be found throughout the Coast Mountains. They range in age between Proterozoic to Tertiary in age. Three main Quaternary volcanic centres showing recent volcanic activity are located within the study area: Mount Garibaldi, Mount Cayley and Mount Meager (Figure 3.1). Several sites are located in the Clear Range, to the east of the Fraser River Fault system, a major tectonic boundary forming the eastern limit of the Coast Mountains. The geology along Clear Range between Lillooet and Lytton on the east side of the Fraser River differs notably from that of the Coast Mountains, and comprises Jurassic and Cretaceous sedimentary, and volcanic rocks intruded locally by Mesozoic granitic rocks. Faults, joints, and foliation of Mesozoic and early Cenozoic age form zones of structural weaknesses on steep slopes and are prime determinants of present-day mass movement activity. 34 Bedrock geology of the study sites is very heterogeneous comprises the full range of igneous, sedimentary and metamorphic rock types encountered in the southwestern mainland of British Columbia. i 3.1.2 Geomorphology and Surficial Geology Quaternary glaciation has strongly influenced geomorphic processes in the Coast Mountains of British Columbia. Extensive blankets of basal and ablation till mantle slopes at many research sites. An examination of geology can therefore not be confined to bedrock characteristics only, but must include glacial effects. Mesozoic and late Cenozoic tectonics determined the major alignment of valleys within the study area. Most of the present valley alignments developed from prolonged Cenozoic fluvial erosion exploiting zones of weakness created by the earlier tectonic events. Substantial re-elevation of the Canadian Cordillera occurred in the late Cenozoic time, leading to a cycle of fluvial canyon-cutting that further accentuated earlier erosional alignments. This was followed by multiple glaciations during Pleistocene time, in which most valleys were widened and deepened. At higher elevations, a spectacular suite of alpine glacial landforms developed including cirques, horns, aretes and over-deepened valley heads. Below about 2200 m, glacial erosion developed rounded ridges and roche moutonee features. The erosional legacy of glaciation was an extensive system of steep, undercut slopes, many of which are cut through by discontinuities created by much earlier tectonic events. Many of these joints, foliations and faults are favorably oriented for slope movement, and thus exert an important control on modern-day mass movement (Ryder, 1981b; Muhs et al., 1987). Extensive spreads of morainal materials accumulated in alpine areas during Quaternary time, with basal and ablation till blankets accumulating at lower elevations on most valley side slopes. Deglaciation resulted in the exposure of large amounts of this unconsolidated material, much of which was subsequently eroded by mass movement processes such as rockslides, rockfalls and debris flows as well as continuous agents of denudation such as creep, slope wash, and solution transport (Slaymaker and McPherson, 1972). The rate of delivery of this material seems to have been intensified immediately after deglaciation, causing a peak of sediment delivery 35 to the fluvial system during the early Holocene time referred to as the paraglacial cycle (Church and Ryder, 1972). Neoglacial advances in the past 6,000 years have resulted in renewed morainal accumulations, many of which are highly unstable and are actively feeding mass movement processes today. 3.1.3 Climate and Hydrology As noted in Chapters 1 and 2, climate, and hydrology are important factors in the explanation of debris flow occurrence. However, climatic and hydrologic characteristics vary notably across the study area. The most pronounced climatic difference is found between the perhumid west and the subhumid east flanks of the Coast Mountains. Precipitation on the west flank is dominated by Pacific cyclones, whereas the east flank is in a rain-shadow area thus receiving considerably less precipitation than the coastal regions. Accordingly, mean annual valley-bottom precipitation in the study area ranges from 2,250 mm at Squamish to only 400 mm at Lillooet. There are two dominant hydro-climatic factors that influence the occurrence of debris flows. One factor is intense rainfall during fall storms on the western flank of the Coast Mountains, where Pacific cyclones cause prolonged, orographically enhanced precipitation. The other factor is rain-on-snow, which can substantially add to the total runoff, particularly in late fall and early winter when the snow cover is thin at lower elevations, thus promoting quick melt during abrupt increases in freezing level. Therefore, type of precipitation as well as freezing level elevation are important. Waylen and Woo (1983) note that the position of the regional snow line determines the type of precipitation reaching a drainage basin. Annual floods are related mainly to spring melt, except in those basins that lie below the snowline, where rainfall-induced floods are significant. In a study that included 24 basins in southwestern B.C., Waylen and Woo conclude that 95% of the spring floods are generated by snow-melt and over 95% of the winter floods are caused by rainfall. Debris flows are more responsive to high intensity storms and since snowmelt alone during warm weather is unlikely to produce debris flows, the likelihood of debris flows in the western Coast Mountains is greatest during late fall and early winter. In the following, several 36 specific sites will be discussed with regard to their precipitation and runoff regimes to provide a better understanding of the relations between the occurrence of debris flows and hydro-climatic conditions. Figures 3.2 A, B, C show monthly mean discharges of the Squamish River near Brackendale, the Lillooet River near Pemberton, and Hat Creek between Ashcroft and Lillooet. The shapes of the discharge curves of the Lillooet River and Squamish River are almost identical, indicating similar snowmelt, precipitation and runoff regimes. Both rivers respond primarily to snowmelt indicated by an inverse relationship between mean monthly rainfall and mean monthly discharge. Since both watersheds are heavily glaciated, peak runoff is delayed until midsummer when glacial ice with low albedo is melting. The most severe floods occur during fall, and are frequently associated with rain-on-snow events. A similar climatic pattern is found in the northern Cascade Mountains south of the lower Fraser Valley which is shown by the precipitation records for Hope (Figure 3.3 C). Here, low precipitation in the summer months is followed by a steep increase of precipitation with the arrival of Pacific disturbances, peaking in November, followed by a linear decrease until July. Extreme precipitation events mimic monthly mean rainfall, and snowmelt contributes to runoff between March and July. Snowmelt is also considered a significant factor to runoff production during October, November and December during which warm Pacific air can force the freezing level to rise by 1000 to 2000 metres within several hours (Schaefer, unpublished reports; Church and Miles, 1987). Debris flows in these three areas frequently coincide with fall floods, such as the October 1984 flood of the Lillooet River during which many debris flows occurred in Lillooet River valley (Jordan, 1987). This observation has important implications for decision makers if the most likely time of debris flow occurrence is a required variable for planning purposes. Hat Creek, serves as an example for the hydrologic regime of the dryer climate in the Lillooet-Lytton area. It was selected because of a lack of a suitable gauged creek along the western flanks of the Clear Range. It is located on the western side of the semi-arid Interior Plateau. Discharge data from Hat Creek are considered a reasonable approximation of hydrologic conditions for sites along the Clear Range between Lillooet and Lytton. Hat Creek runoff responds primarily to snowmelt, which is accentuated by an increase in rainfall during the months of May and June. 37 The dominance of snowmelt is shown by the sharp decline of discharge in July and August, which are the rainiest months in this region. In contrast to Hat Creek precipitation, total precipitation in Lillooet is relatively constant over the year with a slight positive trend from March to October (Figure 3.3 A). In Lillooet as well as on the plateau further to the east, extreme precipitation events can be related to either convective or synoptic storms. Lytton, which is located only 60 km south of Lillooet, shows more annual variability in mean monthly precipitation (Figure 3.3 B). Here, precipitation increases in the fall, decreases during late spring and remains low during the summer months which, although not as pronounced, follows the typical pattern for Coast Mountain climate. Extreme precipitation events generally follow the distribution of mean monthly rainfall, with a peak in June and July, which can be explained by thunderstorm activity during those months. Dendrochronologic evidence (see Appendix A) suggests that a larger percentage of debris flows in this region occurs between late spring and mid summer compared to sites in the wetter regions of the Coast Mountains which agrees with the documented precipitation and runoff regime of Lillooet in the vicinity of which most basins of this area are located. 38 Figure 3.2 Mean monthly discharge for selected sites. Data from Historical Stream Flow Summary, British Columbia (1991) Figure 3.3.A. Mean monthly discharge and precipitation, Lillooet River, Pemberton (1914-1988), precipitation, Alta Lake (1951-1980) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec V) •v. £_ D CD k_ ra o T3 c o E E 700 Figure 3.3.B. Mean monthly discharge and precipitation, Squamish River, Brakendale (1922-1988), precipitation, Alta Lake (1951-1980) 3 CD 3 o =3 r-r zr v; TJ a o o Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 0) CO k— ra JZ o TD _>-. -C «a C o E c t) E Figure 3.3.C. Mean monthly discharge and precipitation, Hat Creek (1960-1984 ), precipitation (1960-1981) £ 2 i 4 -• mean monthly discharge H mean monthly rainfall (mm) • mean monthly snowfall (mm) -4- 40 2 4- 20 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 50 30 S 3 CD CO 3 3 o 10 *< "O n o •o' 39 Figure 3.3 Precipitation data for selected sites. Data from Canadian Climate Normals, 1961-90, British Columbia (1993) Figure 3.4.A. Precipitation (Lillooet, Cedar Falls, 1961-1990) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec E £ — • c 'ra i -£= ra V E c o E Figure 3.4.B. Precipitation (Lytton, 1961-1990) 7 0 -60 • monthly mean rainfall B monthly mean snowfall • extreme 24-hr precipitation Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 70 60 50 | h 4 0 30 20 C/l o i. -fx fi) o 3. 40 3.2 STUDY SITES IN THE MOUNT MEAGER VOLCANIC C O M P L E X 1 The Mount Meager volcanic complex consists of several highly dissected volcanic assemblages that were created during eruptions between 1.9 Ma to 2,350 years B.P. This location was chosen because of the high concentration of debris flow basins, high debris flow frequency, and good road access. The Meager Hotsprings recreation site, and recent geothermal exploration by B.C. Hydro have brought attention to the susceptibility of the area to debris flows and flooding (Jordan, 1986, 1987). Jordan (1987, 1994) described the geomorphology and hydrology of the area and lists the chronology of debris flows and other mass movements events in some basins. The physiography of this area is dominated by steep U-shaped valleys with thin glacial deposits and active aggrading floodplains and fans characterizing the valley bottoms (Jordan and Slaymaker, 1991). Sediment is supplied to Meager Creek and Lillooet River primarily by frequent mass movements, as well as glacial meltwater and reworked neoglacial deposits. Apart from debris flows, Mount Meager is notorious for other catastrophic mass movements. Large landslides have been documented by Carter (1931), Patton (1976), Mokievsky-Zubok (1977), Evans (1990, 1992) and Jordan (1986, 1994). The majority of slopes within the volcanic complex show signs of recent landslide activity such as bare landslide scars, toppling failures, tension cracks and steep fresh talus slopes. Jordan (1994) noted a declining gradient in landslide activity from the southern side of the complex to the northern side, which is probably due to the increasing extent of glacial coverage on the northern side. Most of the debris avalanche deposits, thick tephra and pyroclastic flows encountered on the north side of the complex stem from the most recent eruptions, notably the 2,350 years B.P. eruption that is responsible for the distribution of the Bridge River tephra (Read, 1990). On the south side of the complex, deep fills of debris avalanche and debris flow deposits fill the valley of Meager Creek (Jordan, 1987; Jordan and Slaymaker, 1991; Jordan 1994). Even large mass movements sometimes remain unrecognized because they occur in remote valleys or evidence is removed rapidly by erosion. An examination of the air photograph chronosequence of the Mount Meager individual study sites of this and the other areas are described in detail in Appendix B. 41 area, covering the last 25 years, demonstrated that rockslides in the order of 104 to 105 m3 occur frequently. These rockslides can easily be tracked by their fresh deposits on snow and glacier ice. The geology of the complex has been mapped by Read (1978). The southern flanks of Mount Meager, which include the basins of Devastation Creek, Boundary Creek, No Good Creek, Angel Creek, and Canyon Creek consist of hydrothermally altered lower Pleistocene dacitic and rhyolitic lava and pyroclastic deposits overlying Mesozoic plutonic rocks. Mesozoic metasedimentary, high-grade metamorphic rocks as well as Tertiary volcanics and intrusive rocks are occasionally encountered within the basement intrusive rocks (Read, 1978). Hotsprings Creek basin is underlain entirely by plutonic rock and is included here because of its vicinity to the Meager volcanic complex. The central and northern sections of the complex which include the basins of Capricorn Creek, Pothole Creek, Affliction Creek, Job Creek and Mosaic Creek are underlain by mid- and late-Pleistocene dacitic and andesitic rocks. Jordan (1994) noted that the contact between granitic basement rock and overlying volcanic rock occurs at comparatively high elevation (1500 m - 2000 m) in Canyon Creek, Capricorn Creek and Affliction Creek, which accounts for a high proportion of basement rock material in debris flows from discharged by these creeks. 3.3 S T U D Y SITES A T M O U N T C A Y L E Y Mount Cayley is a Quaternary stratovolcano located in the lower Squamish River Valley (Figure 3.1). It is one of 12 volcanic centres that form the Garibaldi Volcanic Belt (Clague and Souther, 1982). Similar to Mount Meager, this site was selected because of its notoriously frequent debris avalanches and debris flows. The Mount Cayley volcanic complex forms a deeply dissected precipitous arrangement of peaks that are glaciated on the north and east. Sheer walls of weak volcanic rock dominate its southwestern slopes descending to Squamish River. Several streams drain from the southwestern flank of Mount Cayley: Terminal Creek, Turbid Creek and Shovelnose Creek. Terminal and Shovelnose basins are partially underlain by Mesozoic and Tertiary basement rock. Accordingly, the debris flows that reach the fan contain a mixture of volcanic and granitic rocks. Turbid Creek, situated between Shovelnose and Terminal Creek, 42 drains one of the most hazardous basins in British Columbia. It is known to have discharged several very large debris flows and debris avalanches during the Holocene and has been the subject of numerous studies (Clague and Souther, 1982; Evans and Brooks, 1991; Brooks and Hickin, 1991; Cruden and Lu, 1992; Lu, 1988, 1992, 1993; Lu and Cruden, 1996). Endurance Creek basin, located outside the volcanic complex on the west side of the Squamish River, consists almost entirely of competent granodiorite basement rock of Jurassic age (Monger and Journeay, 1994). This basin has been included here because of its vicinity to the volcanic complex. The geology of Mount Cayley has been discussed in some detail by Green et al. (1988) and Clague and Souther (1982). The edifice of the Mount Cayley volcanic complex consists of three steep pyramidal peaks: Mount Cayley itself, Wizard Peak and Pyroclastic Peak. Volcanic rock that overlies irregular plutonic and metamorphic basement rock of Mesozoic and early Tertiary age has been dated between 0.31 to 3.8 Ma. (Green et al. 1988). A number of volcanic units are interbedded with till and colluvium (Clague and Souther, 1982). Three eruptive periods have been recognized: The Mount Cayley, Vulcan's Thumb and Shovelnose stage. The earliest, or Mount Cayley stage, produced tuff, breccia and dacite flows that have undergone hydrothermal alteration. The jagged ridge of Mount Cayley is an intrusive spine that is interpreted as the culminating event of the Mount Cayley stage (Green et al., 1988). The Vulcan's thumb stage can be separated into two distinct rock units. The lowermost unit consists of densely jointed porphyritic biotite dacite flows, whereas the upper unit is composed of porphyritic dacite tuff breccia and tuff (Clague and Souther, 1982). The Shovelnose stage is characterized by porphyritic dacite flows, domes and cupolas which occur as discordant masses within the basal members of the volcanic edifice. Clague and Souther identified fractures that were formed during emplacement of the dacite cupola as the principal planes of weakness responsible for the failure of the Dusty Creek debris avalanche. In 1963 a large debris avalanche descended the valley of Dusty Creek, a tributary of Turbid Creek. The source area of this debris avalanche consists of 250 m of pyroclastics rocks dipping steeply towards the southwest, forming a precipitous amphitheater which delineates the Dusty Creek landslide scar. Hydrothermally altered faults and fractures as well as poorly indurated tuff breccia that are characteristic of the failed mass of the Dusty Creek debris avalanche and the Avalanche Creek debris avalanche which occurred in 1984, are widespread in the Mount Cayley 43 area. This observation suggests that the activity of large debris avalanches and associated debris flows has not subsided and future large mass movements are likely to occur. 3.4 STUDY SITES A L O N G UPPER LILLOOET RIVER V A L L E Y The upper Lillooet River valley above Lillooet Lake is a long U-shaped valley, with a wide, flat floodplain and steeply rising valley walls mantled by glacial till and colluvium. Nine debris flow basins were investigated within this area which encompasses the Green River, Rutherford Creek, Ryan River and Birkenhead River drainages (Figure 3.1). The high frequency and spatial density of debris flows along the upper Lillooet River make this area ideal for debris flow related studies. The work of Jordan (1986, 1987, 1994) has greatly contributed to our understanding of debris flow occurrence and debris flow dynamics in this area. Except for McLeod Creek and Mount Currie Creek, all other basins are underlain by granitic rock. The upper part of McLeod Creek basin is underlain by the Triassic Pioneer Formation consisting of green to purple pillow lava and massive greenstone, and minor felsic volcanic flows. The lower third of the basin consists of early and late Cretaceous Gambier Group rock that is characterized by intermediate, felsic and mafic volcanic rocks, intercalated conglomerate, sandstone and shale (Monger and Journeay, 1994). The Mount Currie site is dominated by gneissic rocks of the Pemberton Dioritic Complex (Roddick and Huchinson, 1973). The physiography of upper Lillooet River valley is dominated by glacial landforms which are the legacy of repeated and intensive periods of glaciation during Pleistocene time. From its confluence with Meager Creek to its drainage into Lillooet Lake, the upper Lillooet River flows in an alluvial channel, changing from braiding channel pattern in its upper reaches to meandering near its delta into Lillooet Lake. Unlike many other valleys in southwestern British Columbia, Pleistocene glacial valley fills and fluvial terraces are absent throughout the main valley, suggesting significant Holocene fluvial and colluvial aggradation (Jordan and Slaymaker, 1991). Sediment yield of the basin has been estimated by Slaymaker (1972), Gilbert (1972, 1973, 1975) and Slaymaker and Gilbert (1972), using the advance of the Lillooet River delta. Jordan and Slaymaker (1991) found that frequent debris flows and other landslides from the Mount Meager 44 area contribute significantly to the high rates of sediment production as well as the aggradational environment of the Lillooet River. Glacial recession directly affects the occurrence of debris flows by exposing large amounts of unconsolidated, and unvegetated glacial sediments (Mokievsky-Zubok, 1977; Eisbacher and Clague, 1984; Clague et al., 1985; Jordan, 1987, Haeberli et al. 1990). Several debris flow basins in upper Lillooet River valley have been glaciated during neoglacial advances, and today, one fifth of the basin are still glaciated. Large valley glaciers as well as small cirque glaciers have been receding rapidly during the last 150 years, exposing morainal sediments that are now prone to denudation processes. Debris flow channels are supplied with sediment by small failures along the steep inner slopes of lateral moraines accounting for rapid material recharge in some basins, again emphasizing the importance of neoglacial sediments in the initiation and sediment supply of debris flows in this and other areas in which glacial retreat has uncovered significant amounts of easily erodable sediments. This process has been included quantitatively in this study by measuring the area that is actively shedding material to the debris flow channel. Continuing climatic amelioration is likely to cause further glacier shrinkage and thus make additional material available for debris flow transport. 3.5 STUDY SITES A L O N G T H E FRASER RIVER V A L L E Y B E T W E E N LILLOOET AND LYTTON The six sites in this area differ considerably both in terms of geology and climate from the majority of sites in the Coast Mountains. Bedrock geology in the study basin is dominated by sedimentary rocks of the Jurassic and Cretaceous Relay Mountain Group and the Cretaceous Jackass Mountain Group intruded by lower Cretaceous granitic rock of the Mount Lytton Complex (Monger and McMillan, 1984; Monger and Journeay, 1994). Volcanic rocks of the Spences Bridge Group are locally important. The Fraser-Straight Creek fault is a major Eocene dextral wrench fault system that trends north-south, across the north-northeast regional grain of Later Cretaceous-earliest Tertiary age (Monger and Journeay, 1994). The physiography of this area is strongly influenced by the structural grain of this system. The Fraser River is deeply incised 45 within it. Rights of Holocene alluvial terraces and truncated colluvial and alluvial fans can be found on both valley sides. Occasionally landslide deposits are encountered that have temporarily dammed the Fraser River and may be responsible for glacio-lacustrine sediments in the Lillooet area (Ryder and Church, 1985). Thin veneers of glacial till can be encountered discontinuously on the western slopes of the Clear Range, but glacial till has largely been eroded from the debris flow source areas. This contrasts with the Coast Mountains where till remains in debris flow source areas in many cases. The axes of the debris flow basins are perpendicular to the major fault system and do not appear to be structurally controlled. However, the close vicinity of the basins to a formerly active shear zone may have altered the density and alignment of joints in adjacent bedrock formations. Geotechnical characteristics of bedrock in the area are highly variable, ranging from fairly competent intact clasts of argillite of the Relay Group and the Jackass Mountain formation to very weak volcanic rock of the Spences Bridge Group and brittle granitic and migmatite rock of the Mount Lytton Complex. Bedrock, however, is not always the determining factor in the production of debris flow sediment. In two basins (Gunbarrel II and III), a large part of the source material for recent debris flows is derived from reworked early Holocene debris flow deposits. Debris flow sediment in Gunbarrel I basin is supplied from headward erosion of an extensive talus slope at the foot of the northern flanks of the Fountain Ridge. An analysis of bedrock characteristics in those cases is therefore of limited value since bedrock communition only serves as the first link in the sediment cascade that ultimately leads to sediment entrainment by debris flows. Geotechnical properties of the source material (in this case poorly consolidated postglacial debris flow diamicton) can therefore not be compared to the properties of the source bedrock. The semi-arid climate of this area contrasts with that of the western Coast Mountains. Extreme rainfall events capable of triggering debris flows can occur throughout the year in the basins closer to Lillooet whereas they are more likely to occur more frequently in fall and winter months at Kaboose Creek which is closer to Lytton as indicated by figures 3.4.B and 3.4.C. Temperatures are significantly higher, and precipitation is lower in this region compared to sites in the Coast Mountains, which makes this area susceptible to forest fires. Multiple fire scars are common in trees in this region. However, as noted in Chapter 2, the occurrence of debris flows is 46 not influenced significantly by fire, because the debris flow source areas are largely unvegetated. This contrasts with other more southerly regions, e.g. the Los Angeles area, where densely vegetated source areas are stripped by fire, which enhances dry raveling and ultimately influences the magnitude and frequency of debris flows (Johnson et al., 1991). 3.6 STUDY SITES IN T H E HOPE-CHILLIWACK AREA The physiography of the area is dominated by the floodplain of the Fraser River which becomes increasingly narrow as one approaches the town of Hope. The southern slopes of the Skagit Range east of Chilliwack as well as the southern parts of Mount Hope are notorious for debris flows, many of which have blocked or disrupted the Trans-Canada Highway (Thurber Consultants Ltd., 1985). The Fraser River demarcates the physiographic and geologic boundary between the northern Cascades to the south and the Coast Mountains to the north. The northern slopes of the Cascade Mountains locally reach elevations over 2000 metres. They are generally very steep (30-40°) and are covered by a thin veneer of till and colluvium. Alluvial and colluvial fans have formed at the mouths of the creeks draining these slopes. Some of these extend far into the'Fraser River floodplain or are truncated by the river. The Trans-Mountain oil pipeline, the Westcoast Transmission gas pipeline and associated access roads, the Transcanada Highway and Highway 3 cross these fans and are vulnerable to debris flow impact. Only six historical debris flow events were confirmed from locals and representatives of companies that have worked along this transportation corridor Thurber Consultants Ltd. (1985). Logging, dating back to the 1800s, is the primary land use of the upper portions of some drainage areas in this part of the study area. However, clearcut logging, particularly in upper parts of basins is limited to the last 10-40 years. Dominant geology in the area is granitic rock of the Mount Barr Batholith and Spuzzum Pluton. Cretaceous metamorphic rock (Settler Schist and Slollicum Schist) is locally present as roof pendants within the plutonic rocks. South of the town of Hope, an extension of the north-south striking Hope fault separates gneiss of the Custer Group from sandstones and conglomerates of the Allenby Formation (Monger, 1989; Monger and Journeay, 1994). Study sites in the Hope-47 Chilliwack area are strikingly linked to bedrock structural characteristics, which distinguishes these sites from the majority of basins discussed in this study. Many debris flow gullies in this area are aligned along northwest-southeast trending linears, some of which are known to be faults or geologic contacts. Savigny and Rinne (1991) produced a map showing lineaments in the order of 102 m to 104 m in length (Figure 3.4). Lineaments with known debris flow activity in the Hope-Chilliwack area are marked on this map. Savigny and Rinne (1991) suggested that there is a general correlation with major lineaments and landsliding in the Lower Fraser River valley and that many of these lineaments are associated with regional faults. They also emphasized that the contact zones between plutons and metamorphic pendants are particularly susceptible to slope instability. They argue that the frequency and spacing of discontinuities combined with weathering and joint infillings in these zones significantly decrease slope stability. The same argument can be applied to smaller-scale slope instabilities along these linears such as rock failures that lead to debris flows. Faults acting as geologic contacts are present along Hope Creek, Thacker Creek, and Two-Mile Creek on the northern flanks of Hope Mountain, and at Mount Ludwig Creek east of the Wahleach Power Station. Faults discordant to lithological boundaries are frequently associated with shattered and sheared rocks thus further decreasing slope stability. Shear zones associated to these faults are prone to accelerated weathering, and are likely to weaken the rock along discontinuities. Discontinuities can also have a controlling influence on the movement of groundwater through the rock mass with the potential of further reducing rock strength (e.g. Hencher, 1987). 48 Figure 3.4. lineaments in the Fraser Valley and adjacent ranges (Savigny and Rinne, 1991). Gullies in the Hope I Chilliwack region with known occurrence of debris flows are marked in red. 49 Different aspects of debris flows in this area have been studied by Evans and Lister (1984), Thurber Consultants Ltd. (1985), Slaymaker et al. (1987), Church and Miles (1987) and Bovis and Dagg (1992). Work by Church and Miles (1987) suggested that the majority of debris flows are associated with rainstorms in late fall and early winter when Pacific fronts are funneled into the cone-shaped eastern part of the lower Fraser River. Cycles of debris flows were recorded for late December 1980, mid-July 1983, early January 1984 and, most recently, late November and early December 1995. The July 1983 storm is the only precipitation event that was associated with convection rather than winter frontal precipitation. This storm had a return interval of 10 years for the Agriculture Canada Research Station at Agassiz. Late fall and early winter storms are typically associated with abrupt rises in freezing level and associated snowmelt during fall or winter. Below freezing temperatures prior to the storm may substantially decrease infiltrability, thus adding to rapid runoff (Miles, 1984). The double storm sequence in November/December 1995 which caused several destructive debris flows at Mount Hope, as well as in the Skagit and the Chilliwack valleys re-emphasized the sensibility of this region to debris flow hazard. 3.7 STUDY SITES AT OTHER LOCATIONS Four additional sites were chosen in the northern parts of the study area. Collis Creek, Howe Creek, and Fergusson Creek are located in the vicinity of the communities of Gold Bridge and Bralorne. Roger's Creek is located in the Cayoosh Range and can be accessed via the Duffy Lake Road. Bedrock geology in these basins consists of a variety of sedimentary and volcanic rocks of the Cayoosh assemblage and Bridge River complex with variable geotechnical characteristics. Apart from similar bedrock geology, these basins have few common physiographic characteristics and will be therefore be discussed individually in Appendix B. 50 CHAPTER 4. METHODS OF DETERMINING FREQUENCY AND MAGNITUDE OF DEBRIS FLOWS 4.1 I N T R O D U C T I O N A major precept of this thesis is that the distinction between weathering-limited and transport-limited basins is an important determinant of debris flow behaviour. This chapter begins with a presentation of classification criteria that were used to differentiate between the two basin types to provide a framework on which much of the subsequent analysis and discussion is based. Having provided this background, this chapter will continue with a review of appropriate methods for determining the frequency and magnitude of debris flows. In Chapter 2 it was noted that a magnitude-frequency analysis of debris flows is difficult in many cases using the methods of flood frequency analysis because some statistical requirements cannot readily be applied for debris flow. In cases where these requirements could be satisfied a lack of volumetric data precluded extreme value analysis. Since debris flow channels are rarely monitored, indirect evidence of past debris flows must be sought. Several approaches for determining the dates of debris flow events were used. Dendrochronological methods were employed in the majority of the cases. Historical accounts and air photograph analysis were used to confirm the dates and augment the dendrochronological record. Magnitudes of debris flows were determined either directly from volumetric estimates of the deposit or indirectly by using a relationship between peak discharge and total volume. Maximum expected magnitudes were computed where applicable to determine the volume of the largest conceivable debris flow that could occur. A criterion used in this study for inclusion of a debris flow event is that it traveled beyond the apex of the depositional fan. Usually, this point is associated with a pronounced decrease in slope angle relative to the steeper channel upslope. This sampling criterion means that debris flows that stalled prior to reaching the fan were not considered. From a practical standpoint, a debris flow which neither reaches the fan nor spills out of its confined channel has a limited destruction potential since any development is usually limited to the fan surface and channel margins but not 51 within the active channel. In addition, such small debris flows usually do not leave any datable dendrochronologic evidence, for they do not damage or bury trees along the channel or on the fan. Exceptions to this rule are basins which produce mostly fine grained sediment. Debris flows composed of fine sediment tend to cause tree bark and cambium abrasion rather than the shearing of entire trees. Although rare, some trees may survive debris flow impact in those channels and it becomes possible to date even comparatively small debris flows. Usually, however, such small debris flows are difficult to identify in the field or on air photographs, especially in forested areas. As a result, high frequency and low magnitude events are censored. In southwestern B.C., high magnitude and low frequency events may also be censored since, except from few cases, very large debris flows occur on a time scale exceeding the practical capabilities of dendrochronology, and certainly historical accounts, or air photography (cf. Thurber and Golder Associates 1992). 4.2 CRITERIA FOR BASIN CLASSIFICATION In order to test the second research hypothesis, a classification of basins into weathering-limited and transport-limited is attempted to improve predictions of frequency and magnitude of debris flows. Therefore, criteria must be developed to allow this classification. In weathering-limited basins, the debris source areas and channels contain limited amounts of transportable sediment that are recharged by weathering processes and mass movements after each debris flow. Alternatively, transport-limited basins contain almost unlimited supplies of sediment and are quickly recharged to allow debris flows to be triggered whenever a precipitation, threshold is exceeded (Figure 4.1). In this study, the initial discrimination between weathering-limited and transport-limited basins was based on an examination of the sediment storage in the source area and within the channel. A basin was classified as transport-limited if there seemed to be sufficient sediment available after a given debris flow such that a flow with similar magnitude would be likely to occur whenever a certain precipitation threshold is exceeded. This implies that no significant time period is required for sediment recharge. Alternatively, if there was reason to believe that a significant part of, or the entire stored, removable sediment had been transported by the last debris flow, and that complete recharge would be necessary before another debris flow 52 could occur, the basin was classified as weathering-limited. All basins were discriminated according to these criteria, and total volumes were calculated from the appropriate regression equations (cf. figure 4.10). It is realized that some basins were of intermediate character where not all sediment is discharged in an event, but it is largely depleted. At this stage, intermediate basins were forced into the more appropriate group to avoid oversophistication of the analysis. In Chapter 6 geotechnical and morphometrical variables are used to produce classification functions based on discriminant analysis. These functions allow a more objective differentiation between weathering-limited and transport-limited basins, and point towards basins with intermediate sediment availability character. Figure 4.1. The concept of weathering-limited and transport-limited basins with regard to the occurrence of debris flows. high A weathering-limited system (supply-limited) ris f I 3 'a. low time high A 3 'a. low transport-limited system (supply-unlimited) { debris flow debris 1 1 debr flow r s flow \ { ejtnnsic threshold | time note: thick black columns indicate individual storms 53 4.3 DETERMINING DEBRIS FLOW FREQUENCY Several methods have been applied in past studies to determine debris flow frequencies: i) historical accounts and air photography, which encompass several decades, ii) dendrochronology which can extend the record up to century time scales, iii) lichenometry which, under certain conditions, reaches back into the mid-Holocene and may be used in unforested environments, and finally iv) radiocarbon dating, and v) tephrochronology both of which are able to extend the time window into Pleistocene time. In forested regions, dendrochronology has shown to be a very useful method in establishing debris flow frequencies over a century time-scale since it yields precise and often replicable dates (e.g. Strunk, 1991), and was therefore chosen as the primary dating method in this study. Furthermore, the time span of the dendrochronologic record, from several years to ideally several hundred years is one which is relevant for the majority of engineering and planning purposes. 4.3.1 Dendrochronology as a method to determine debris flow frequencies Alestalo (1971) coined the term "dendrogeomorphology" to describe the use of dendrochronology in geomorphology. Since then, the technique has been widely used, as is evident in reviews of applications presented in Shroder (1980) and Shroder and Butler (1986). In the past decade, several notable studies on debris flow frequencies have been conducted using dendrogeomorphological methods (e.g. Hupp, 1984; Strunk, 1991; Podor, 1993). In particular, a comprehensive study conducted by Strunk (1991) compared and improved dendrogeomorphological methods for determining debris flow frequencies at several locations in the German, Austrian, and Italian Alps. 54 Figure 4.2. Ponderosa pine with multiple scars in Fools' Gold Creek debris flow channel 4.3.1.1 Study Prerequisites In order to successfully apply dendrogeomorphological methods in debris flow basins, several site conditions must prevail. The debris flow fan, and preferably the area adjacent to the channel, must support trees. In most cases, coniferous trees will yield less ambiguous results than deciduous trees since earlywood and latewood boundaries are better defined for most coniferous species, which considerably facilitates dating. Ideally, trees should be growing close to the active 55 channel margins or within the channel because these trees are most likely to contain a good record of debris flows by offering datable scarred wood tissue (Figure 4.2). The age of the trees must considerably exceed the date of several events to provide a means to determine recurrence intervals. The older the trees along the channel and within the zone of deposition, the longer the obtainable record. In order to apply dating using the formation of reaction wood, trees on the debris flow fan must be injured or disturbed to the extent that vertical stress on roots and lateral stress on the lower stem favors the formation of reaction wood. Very thin layers of debris are unlikely to produce a growth reaction particularly in the case of mature coniferous trees with considerable stem diameters. 4.3.1.2 Process-response mechanisms Debris flows can influence tree growth in several ways. In cases of extreme impact, trees can be sheared off or inundated with debris. If a debris flow can be assigned as the sole reason for tree death, the date of the event can be established by cross-dating the dead tree's ring sequence with living specimens at the same site. In many instances, however, trees are scarred or only shallowly buried which generally does not result in death. As a growth reaction to scarring, the tree builds tissue across the injured area, which usually provides an unequivocal date for the event. Lateral stress on the trunk and normal stress on the root system can cause the formation of reaction wood (cf. Alestalo, 1971). Strunk (1991) showed that the critical depth of burial ranges between 1.6 and 1.9 meters for Piceaabies. Stem burial exceeding this amount results in growth termination. Finally, partial burial of the trunk may lead to the growth of adventitious roots. It is usually desirable to use more than one of these tree growth responses for dating and event cross-dating. Error bars on individual events are uncommon in dendrochronologic studies, indicating that this dating method is founded on its replicable precision. By contrast, dating trees that grow on fresh deposits to determine the minimum deposit age is clearly much less precise, and should only be used if no other method can be applied. The time lag between the debris flow event and tree germination (ecesis) can be extremely variable, ranging from 1 year (e.g. Hupp, 1984) to 75 years (Driscoll, 1980, cited in Birks, 1980). 56 4.3.1.3 Dating of tree scars Most debris flows carry boulders that are capable of injuring bark and the underlying cambium by direct collision or progressive abrasion. If the tree survives the disturbance, the scar will start to callous over the damaged part and normal growth will resume. Scar overgrowth will continue until the injured cambium is completely healed. Since the bark is not renewed during the tree's life, disturbance events of previous decades and centuries can be recognized and dated especially if complete discs are available. Usually the method provides a resolution of one year. If the event occurred during the dormant season, the date is usually given as the two possible calendar years (e.g. 1992/93) since it is impossible to determine whether the event occurred in the fall or winter of 1992 or in the spring of 1993. In the case where the event occurred during the growing season, it is usually possible to determine approximately when during the growth season the event occurred if typical growth patterns during the year are known for a given area. Figure 4.3 shows an example of a spruce near the town of Bralorne that has been scarred twice (early growth season of 1965 and dormant season of 1930/31). One problem associated with this method is that severe tree damage might lead to a complete termination of cambial activity, as pointed out by Strunk (1991). If there is evidence for interrupted cambial activity, these the dates of impacts can be determined by cross-dating with other tree sections from undisturbed control populations in the vicinity of the damaged trees. Dating scarred trees is the least ambiguous method, provided that the scar can definitely be attributed to a debris flow event. Snow avalanches, falling trees, gnawing animals, lightning, and fire can also cause scars which appear physiologically to be similar to debris flow scars. A check-list was used in the field that allowed us to attribute a tree's response to a corresponding event. This list included the following criteria: - Position of the tree relative to the debris flow channel: the scars always face either directly upstream or occur on either side of the tree, never on the downstream side. - Location of the scar on the tree stem: usually scars are at or near the basal flare. In some cases, however, sediment around the tree stem may have been eroded which leaves a scar high above the tree base. Similarly, debris flows with large surge waves can injure trees up to several meters 57 Figure 4.3. Spruce tree on Ferguson Creek fan, scarred in 1930131 (point 1) aiui 1965 (point 2). Open arrow points towards pith of tree. Scar above point 1 is of the same date as point 2. above their base. If the debris flow is still mobile at this point, it will continue to move and leave the scar high above the ground. - Location of the scarred tree with regard to other processes acting at the site: if there is evidence that the site is also frequented by snow avalanches, as indicated by the presence of slide alders, by a pronounced avalanche path, or by snow avalanche deposits in the spring, care must be taken 58 in dating debris flow events, since avalanches also carry large clasts or organic debris capable of scarring trees. In this study, several potential study sites that showed signs of frequent debris flow activity were eliminated from the original sample because of obvious snow avalanche activity. - Signs of human activity: as Strunk (1991) pointed out, debris flow gullies are used as timber transport routes in some parts of the Alps. Sliding tree trunks often cause scars with the same characteristics as debris flow scars. In North America, this type of timber transport is not widely used, though yarding down gullies takes place. In areas of active or former logging activities, however, falling trees can injure other trees, but these scars can be discriminated from debris flow scars because they usually produce injuries several metres in length, which are atypical of the types of injuries caused by debris flows. - Fire scars: in areas with frequent forest fires, fire scars can cause considerable problems in dating debris flow events. Forest fires tend to burn in an upslope direction, and on the lee (upslope) side of trees, turbulence created by air circulation supports flames. Therefore, fire scars are often found on the upslope sides of trees but obvious scorching of the wood tissue generally indicates that fire disturbance is the cause. However, at one location along Highway 99 near the town of Lytton (Fools' Gold Creek) it was found that trees injured by debris flows, had subsequently burned. In these cases, fire scars from a nearby fire-damaged control population had to be dated to distinguish between the two scarring mechanisms. In summary, scarred trees provide an excellent means of dating debris flows. One disadvantage is that wedges or entire cross-sections of the tree must be sampled to perform the lab analysis. Depending on the size of the wedge, this operation may severely damage or kill the tree. However, many trees sampled in this manner were moribund at the time of sampling. 4.3.1.4 Dating via reaction wood and stress-induced growth suppression Reaction wood develops in response to external stress on a tree. Two types of reaction wood can be used for dating purposes. Compression wood develops on the side of the stem opposite the stress direction. It can easily be identified by asymmetrical growth and higher wood densities. The onset of compression can then be dated and its cause evaluated. This method was 59 used by Hupp (1984) and Podor (1992) for dating debris flows, but has not received widespread approval. The major problem with this approach is that several processes may lead to growth compression, such as snow and soil creep, logs leaning against the sapling, wind, irregular crown growth, or simply growth on steep slopes. For these reasons, dating of compression wood should only be applied in conjunction with other methods to avoid the possibility of confounding slope movement effects with those of other processes. Debris flows can cause the formation of suppressed growth by partial burial of the tree's stem. The increased normal stress on the root horizon, as well as lateral stress of the compacted debris on the trunk, can lead to pronounced reduction in growth increments. Figure 4.4. shows a good example for growth increment reduction of a tree core sampled on the fan of Gunbarrel III. Here, the debris flow occurred some time between the end of the growth season of the year 1925 and the early growth season of 1926. As in the case of compression wood, growth suppression can be caused by a variety of factors other than disturbance by debris flow, notably climate fluctuations, water and nutrient availability, insect infestations, and earthquakes (e.g. Schweingruber et al. 1991). In addition, most trees display a trend which is the reflection of progressive growth increment reduction with age. Debris flow disturbance does not necessarily lead to growth suppression but can also lead to growth "release" of an individual tree if nearby trees are killed by the flow, which is then due to a competitive advantage for light and nutrients. To avoid confounding these effects with debris flow events, control populations of undisturbed sites with similar conditions were compared with the disturbed samples. The numbers of control trees sampled was dependent on the number of different tree species being used in the study. If different species were analyzed, a control sample of trees was used which reflected the proportion of species of the disturbed trees. Sampling for reaction wood was accomplished by small diameter (typically 4 or 5 mm) increment corers. The cores were mounted and sanded to a high finish. Phipps (1985) and Josza (1988) give detailed accounts of the sampling and core preparation procedures. Precise dating of samples is clearly the most important task in dendrochronology. Simply counting back from the outermost ring can result in error since a variety of factors can cause missing or false rings. Therefore, samples of each sample population were cross-dated. In the 60 southern Coast Mountains of southwest British Columbia, false and missing rings were rarely encountered given the humidity and dependable climatic seasonality in this region, which does not promote the development of missing or false rings. Cross-dating in this environment was accomplished by using Yamaguchi's (1990) method, in which one counts back from the outer ring and lists so called "event years", namely years with pronounced changes in increment width (Schweingruber et al., 1990). Event years were then used to correlate between samples in the same population. If a group of trees showed events in the same year, those dates were termed "pointer years". Dating errors were thus detected easily and corrected. In drier regions, such as the Lytton-Lillooet area, false rings and missing rings become more frequent to the extent that Yamagutchi's method had to be replaced with skeleton plotting. In this method very narrow rings were recorded subjectively on graph paper by assigning a value between 1 and 10 to the narrowness of the ring by comparing each ring with its neighbors. Where a ring was considerably narrower than adjacent rings, a long vertical line was drawn. If only slightly narrower, a shorter line was drawn. Each population consisting of several series was then used to create a master chronology showing only those narrow rings that characterized many individuals at a particular site. The same technique was used to date dead wood. The skeleton plot of the dead sample was moved along the master plot until the best match was achieved by visual correlation. Although somewhat subjective, this method has been used successfully by dendrochronologists for more than 70 years (Douglass, 1919). Detailed accounts of the skeleton plotting method are given in Stokes and Smiley (1968) and Swetnam et al. (1985). Strunk (1991) used suppression wood extensively to date debris flows. He concluded that it is not only the easiest and fastest method but also the most accurate. It should be noted that all of his sample sites were located on low-angle alluvial fans with extensive areal deposition. In the southern Coast Mountains of British Columbia, debris flows frequently deposit directly into higher-order streams that rapidly erode the accumulated sediment. Thus, widespread inundation of fans is not quite as common as in Strunk's field area. Furthermore, it was found that mature Douglas firs and cedars are very resistant to lateral and normal stress on their stem and root systems. For stem diameters exceeding one metre, burial of up to several decimetres did not 61 Figure 4.4. Computer-generated image of a tree core sampled on the Gunbarrel III debris fan, showing ring-width reductions as a typical growth reaction of partial stem obliteration by debris flow sediments. This debris flow occurred during the dormant season of the years 1925/26. appear to have detectable reaction wood. In these cases, dating of debris flow events via suppression wood was used in conjunction with scar dating. 4.3.1.5 Dating of debris flows using adventitious roots Most deciduous trees, and many coniferous species, react to partial burial with the development of adventitious roots. In the case of repeated sedimentation, several root horizons can 62 be encountered and may be used for debris flow dating. Strunk (1991) used spruce (Piceaabies) and found that, even after 540 years, the tree was still able to form adventitious roots. In this study, Douglas fir (Pseudotsuga menziesii ssp. menziesii), western redcedar (Tujaplicata), yellow cedar (Chamaecyparis nootkatensis), western hemlock (Tsuga heterophylla), Engelmann spruce (Piceaengelmanii), subalpine fir (Abies lasiocarpd), ponderosa pine (Pinus ponderosa), and occasionally black cottonwood (Populus balsamifera ssp. trichocarpa), paper birch (Betula papyrifera), and red alders (Alnus rubra) were used for dating purposes. Adventitious roots were encountered in several cases on both of the cedar species, but not in any of the other species. Due to the lack of naturally exposed or excavated tree stems below a former burial level, it cannot be concluded that these other species do not form adventitious roots. However, Heikinheimo (1920), cited in Strunk (1991), reported that a particularly thick tree bark was able to prevent the formation of adventitious roots. Douglas firs, especially older ones, occasionally produce bark sometimes in excess of 20 centimeters in thickness, which might impede renewed root formation. Adventitious roots can be dated by counting the tree rings along root cross-sections or by determining the time of root sprouting. The latter method is more accurate, but requires complete excavation as well as dissection of the tree, which is both time consuming and dangerous if parts of the trunk are deeply buried. For example, during field work in the northern Italian Alps in 1988, the collapse of a timber supported excavation shaft in poorly consolidated debris flow sediments was witnessed. Fortunately, at the time of collapse, nobody was working in the shaft. Both methods of dating adventitious roots are described in detail by Strunk (1991), who made the important observation that, in some cases, several adventitious root horizons may form from one debris flow event. Counting root horizons and dividing this by the tree's age to calculate event recurrence intervals prior to determining their age should therefore be avoided (e.g. Jackson, 1977). If time and money allow, excavation of trees and dating of adventitious root horizons can also yield extremely valuable data on debris flow volumes. If events are dated and each root horizon can be assigned to one specific event, the thickness of each deposit between root horizons can be measured. Repeating this procedure with trees over a large area of a fan allows spatial integration of depth data and thus calculation of flow volumes for several past events. Although very laborious, and only possible under ideal conditions, this method yields accurate data on debris 63 flow frequencies and magnitudes. In this study, several trees were excavated for analysis of adventitious roots. It was discovered that the time needed to excavate trees, the problems associated with data gathering, and the potential dangers of such work did not warrant further excavations. 4.3.1.6 Presentation of results Scar dates can simply be listed, and cross-dated tree ring series can be presented as phase diagrams (e.g. Schweingruber, 1987; Schweingruber et al., 1990), or as curves representing measured increment width changes to infer debris flow disturbances. Since the absolute width of individual growth increments is not essential in this study, it was found that Yamaguchi's (1990) method, in combination with displaying the results as phase diagrams, provided more suitable means than analyzing samples with a high resolution tree-ring scanner, adjusting the ring count, detrending the time series and comparing individual series with a master chronology on the basis of these computer-generated time series. In addition, changes in late-wood densities, which are not recorded by the image analysis tool, were more readily detected by eye, and provided additional information for detecting changes in growth patterns. For these reasons, presentation of data as phase diagrams was preferred. Figure 4.5 is a phase diagram summarizing dendrochronological results from Collis Creek to illustrate the technique. Cores from 14 trees were sampled in summer '94 and examined for growth responses. The right-hand column of the diagram denotes the oldest recorded tree ring. The time series is truncated at 1875 because the number of trees containing event information rapidly decreased beyond that date, rendering the older dates unreliable. Therefore, frequency was calculated for only the past 122 years. Growth increment reductions are indicated along the time axes using the symbols explained in the legend. Depending on the magnitude of a reduction, two degrees of growth reduction are recognized. At the bottom of the sample population a composite plot of the control population is shown. Replications between several individual samples are compared with those in the control population. In the case where the onset of an increment reduction coincides with the onset of a reduction in the control population, debris flows are discarded as the cause for this disturbance. The most likely reason for non-debris flow disturbance 64 seems to be climate variations. Otherwise the correlation of growth reductions among several samples (depending on the sample size, but generally at least 25% of the sample population), suggests that a debris flow had in fact occurred. In a final time - event graph (composite plot of Figure 4.5), scar dates and dates acquired by analyzing reaction wood are combined and compared to detect any dissimilarities which might indicate dating problems. In Figure 4.5, 14 cores from douglas firs growing on the depositonal fan are analyzed with regard to debris flow disturbance during the last 110 years. In this example, there was a pronounced period of growth reduction between 1971 and 1979 which cannot be easily be attributed to debris flows since the composite plot of the control population shows the same pattern. The same is true also for a less pronounced period between 1959 and 1964, as well as the period 1915 to 1920. The latter two periods were visible only in a few trees on the debris flow fan, which indicates that stress caused by factors other than debris flows dominated growth responses. Debris flows were interpreted to have occurred from growth responses alone in the years 1921/22 and 1931/32. Other growth responses were confirmed by scar dates in the years 1901/02, 1950/51, and 1985. The years 1885/86, 1894/95, 1910/11, 1935/36, 1938/39, 1967/68, and 1980/81 were recorded only by scar dates, suggesting that no significant deposition occurred on the fan in the period from 1885-1981, and that the majority of the debris mass remained channelized until they discharged into Hurley River. This conclusion is supported by the position of the scarred trees in the immediate vicinity of the channel. Thirteen events were recorded over the period 1872 to 1994, which gives an average recurrence interval of approximately 9 years, or a frequency of about 0.11 events per year. The same type of data reduction was used to analyze frequencies in the other 34 basins. As mentioned above, the oldest recorded debris flow date is not necessarily suitable for inclusion in frequency calculations if it is substantially older than the majority of sampled trees. Including this date in the frequency calculations would lead to erroneously lower debris flow frequencies, because of the chance of missing debris flows that had occurred in the time period between the last and second-to-last events, during which period no trees were sampled due to age limitations. Figure 4.6 shows the record length available in each basin. Basins BO (Boundary 65 Figure 4.5 Dendrochronolgical summary chart for Collis Creek combining information gathered from reaction wood, tree scars and historical documentation. DF-02 through DF-20 are tree core samples from Douglas firs on the depositional fan. Composite plot summarizes results from all data sources. 00 1- O O CO CO CN CNI <T) r-v sQ i n o 1—« t— ^- i n oo oo ac r>. t>. r>» oo oo r>. ON O CN r H 9 [ [ [ [ • J - -L -c 1 J I-• 0> 8 c 01 TS '> 01 •a o o 5 c o a. a. O 0) CM fO —. Q. E o <-» iS _c c o •a u 3 •a o 01 so -a •a "5 9 Q 9 a <N n u . u . u . u . u-D a. u a. c o XI 3 a. a. 3 (A -a c •a OI -a u O 5 o o -a O ion MO 3 •a XI 01 TS ua c E na incre i other t 1994 "o i other t c c c SO o ssi( c ssi( '5 n pre to c Q a. 01 3 c X o "o ress « •a ress X. c a. in a. O o> 3 c a. CA 66 Creek) and NG (No Good Creek) have very short records, since all debris flow dates were obtained from historical accounts (logging companies) and personal observations. Dendrochronology was not attempted at these locations because the channels are so deeply incised into colluvium that only the very largest events would fill the channel. An additional difficulty is the fact that the fan has been logged and the slash-burned tree stumps do not allow the application of dendrochronologic methods. In these two cases it seemed reasonable to extrapolate frequencies into the past for at least the time of air photograph coverage, since no significant changes seem to have taken place in the basins to suggest significant recent changes in frequency. Figure 4.7 shows the frequency distribution of debris flows in all basins. As expected, there is a notable decline of events with time, which is a function of the loss of record due to a decrease in the probability of survival of trees. Since predictions were not based on this combined frequency plot, but on the individial time series, this data loss did not bias the results. Despite these data limitations, it seems that the periods 1840-45, 1850-55, 1915-1920 and 1980-85 were particularly active, whereas the periods 1855-1860, 1885-90, 1905-10, 1920-25 and 1945-50 showed significantly reduced activity. There are consistent declines in frequency after periods of high activity; this suggests that weathering-limited basins, which account for two thirds of the total sample, require 10-20 years to recharge until an intrinsic threshold is reached and a precipitation event can trigger the next debris flow (cf. Figure 4.7). Therefore, a period of high debris flow activity is likely to be followed by a period of lesser activity even if the magnitude of the triggering climatic events remains constant. However, this observation implies that the majority of weathering-limited debris flows occur in regionally synchronous cycles of events, which is a reasonable assumption given that most of these debris flows occur during synoptic storms in late fall and early winter, and that the recent severe storms of November 1990 and November 1995 exhibited this region-wide tendency. 4.3.2 Historical accounts and air photograph analysis At the commencement of this study, it was anticipated that obliteration or destruction of bridges by debris flows would be recorded by road maintenance companies, forest companies, the 67 Ministry of Transportation and Highways, and the Ministry of Forests or local residents. Unfortunately most events, even including those that caused traffic interruption, were not recorded at all or, at least, little differentiation was made between floods, debris flows, mudslides or rockfalls. This made it impossible in most cases to unequivocally identify debris flows. In several cases, repeated letters and phone calls to government agencies remained unanswered. An exception was the records kept by Weldwood forest company, based in Squamish, who had kept records of debris flows along Turbid Creek in Squamish River Valley. Weldwood provided me with 7 dates for debris flows over the past 30 years that would otherwise not have been available, since the entire debris flow fan had recently been logged and therefore presently contains no usable dendrochronological records. In addition, Turbid Creek channel is deeply incised over most of its length so that only those debris flows with peak discharges exceeding approximately 500-600 m3/s would have been recorded by tree damage. Air photographs were used extensively to identify sites with known debris flow occurrences. Available chronosequences covering the last 40-50 years were then analyzed with regard to the most likely time of debris flow occurrence. Events visible on air photographs were documented as the intervals between years of air photography and a qualitative description of the freshness of the deposit was recorded. These data were then compared and adjusted to results obtained by the other dating methods. Since air photograph analysis of debris flow deposits is limited to large events, small debris flows in forested terrain frequently remain undetected because of dense forest canopies. Therefore, air photographs were only used to supplement data from other sources, and to confirm ambiguous debris flow dates. 4.3.3 Summary Dendrochronology offers several techniques for determining the frequency of debris flows. Considering the large problems involved in sampling and interpretation of adventitious roots and compression wood, dating via suppression wood and tree scars was given preference over the other methods. For the suppression wood method, low angle fans with extensive lateral deposition were found to be ideal. Flat rooting tree species (e.g. Piceaabies, Pseudotsuga menzi-68 Figure 4.6. Record lengths for all debris flow basins c V) m .a Endurance — Terminal _ J » Turbid Deepa Mount Currie Last Day Rainy No Law Petersen McLeod _ Nightmare Boundary No Good Canyon Capricorn Hotsprings Pothole Clearwater Roger's Fool's Gold Snake Kaboose Gunbarrel 1 Gunbarrel2 Gunbarrel 3 19-Mile Mount Ludwig --fjiWjjjjjffj Eureka Patterson 2-Mile Wildcat Howe Fergusson Collis - i 1 1 r - i — i 1 — r - i 1 — i 1 -300 length of record Figure 4.7. Frequency distribution of all debris flow basins t n m m i A i / i m t n m m i n i n m m v> uy i n m i n m u> i n m U g 8 S ? 8 8 2 8 8 S S S 8 ? S ^ 8 8 S S c o c o c o c o c o c b c o c o c o time intervals (5 years) 69 esii) were more sensitive to partial burial than deep rooting species (e.g. Pinusponderosd). Scars caused by impact and abrasion yielded the most accurate and least ambiguous dates and should be used to corroborate other methods, wherever applicable. Historical accounts frequently do not exist or are too ambiguous to be included in the analysis. Air photographs were used to supplement and confirm dates gathered from dendrochronolgic analysis but in any case only provided interval estimates of debris flow dates. Appendix A lists all known debris flow dates in the study area, including the type of data used to establish the event. 70 4.4 DETERMINATION AND PREDICTION OF DEBRIS FLOW MAGNITUDE There are three conceptually different methods for analyzing the total volume of debris flows, the choice of which depends on the physiographic and morphologic setting and the degree of desired detail. These methods can be classified into those approaches that determine the magnitude of debris flows that actually occurred (section 4.4.1 and 4.4.2), and those which aim for the prediction of debris flow magnitude (section 4.4.3 - 4.4.5). Commonly, total volume is determined by measuring the area and average depth of a debris flow deposit. If this is not possible because parts of or all the deposit have subsequently been eroded, an empirical approach can be chosen that relates peak discharge of debris flows to their total volume. Finally, sediment that accumulates in the channel after each flow can be measured in sections and integrated over the whole channel length to provide some estimate of the probable volume of an impending debris flow, or perhaps a typical magnitude for debris flows from a specific basin. The latter method is particularly useful in channels that are rapidly recharged with sediment after each debris flow. In colluvial channels, indices for channel erodibility can be employed to determine maximum expected magnitudes given the present channel conditions. Another approach is the determination of the volume of a failure scar as a surrogate for debris flow volume. This approach is limited to those cases in which either deposition and erosion along the channel are roughly balanced, or where the entire landslide volume is deposited on the fan without significant erosion or incorporation of sediment along its flow path. This rather improbable behaviour has not been observed in any of the basins investigated and will therefore not be considered further. Where possible, both the volumetric analysis and the empirical approach were used jointly and compared to check for consistency. In general, the direct volume measurement method yielded more reliable results if the deposit depth could be approximated with confidence. Irrespective of the choice of the volume-estimation method, magnitudes were often truncated on the very high end as indicated in the introduction for this chapter. This truncation, which can be approximated as 50,000 m 3 , is largely due to erosion of infrequently occurring large events. It is reflected in Figure 4.8 which shows the frequency distribution of magnitudes over the entire sample of debris flow 71 events. As as consequence of erosion and obliteration, debris flow magnitudes are also truncated on the low end with volumes less than 1000 m3. This truncation is not shown in Figure 4.8 because debris flows with less than 4000 m3 are combined into one class. Figure 4.8. Frequency distribution of debris flow magnitudes for all basins. 4.4.1 Direct Measurement or Estimation of Debris Flow Volume A three-dimensional analysis of a debris flow deposit is commonly used to assess magnitude at sites where: 1) little or no information on previous debris flow volumes exists, and 2) the deposit remains well preserved. This method typically involves a survey of the lobe's area and an assessment of average deposit depth. Together these permit an estimation of volume. Where a debris lobe was found to be of simple, symmetrical shape of either uniform thickness or where it 72 tapered distally or laterally in a regular fashion, volume was readily analyzed through multiplication of the typical deposit dimensions. A volume estimate then included the product of lobe length by average width, and either the assumption of uniform depth or some simple function of distal or lateral deposit thinning would be applied. Commonly, however, debris flow deposits were found to be significantly more complex in plan form (e.g. multiple lobes) and deposit depth (e.g. buried channels or extensive levee margins) to be volumetrically assessed with any accuracy through these assumptions (e.g. Podor, 1993). 4.4.1.1 Assessing deposit area Generally, determining a deposit's area was the least complicated portion of three dimensional volume analysis. Small deposits without obscuring vegetation or surface irregularities were effectively surveyed with a compass and measuring tape. Sites that exhibit great deposit relief, or are obscured by vegetation typically require multiple survey base stations throughout the lobe and are more suited to plane table or E.D.M. surveys (Podor, 1993). E.D.M. surveys were found to be inoperable for most deposits because dense stands of trees and underbrush severely limited visibility. For this reason, most volumes were determined by taking width and length measurements using measuring tapes or hip chains and by taking compass readings to reconstruct the two-dimensional geometry of the deposits. One concern is the accurate recognition of the boundaries of debris flow deposits. At sites where event frequency is moderate to high (i.e. recurrence intervals from 1 to 10 years), it was found to be difficult to distinguish the most recent from earlier deposits. Dating the maximum age of trees growing on the deposit and dating of scarred trees where the deposit was unambiguously associated with the tree injury were used to identify the margins of the most recent deposits. 4.4.1.2 Assessing deposit depth Determining the depth of debris flow deposits represents the most difficult aspect of volume analysis. It usually involves the interpretation of limited data, since natural cross-sections are rare and yield ambiguous depths in the case of older deposits. However, channels have frequently incised into the debris flow deposit that allow at least one estimation of the deposit 73 thickness at those sites. The application of one available thickness date can yield errors as large as a factor of two at very complex deposits. The excavation of artificial pits to determine deposit depth is very laborious and was not attempted in this study. Another problem is a lack of information on the fan topography prior to the most recent event which can greatly influence the deposit's estimated depth. In summary, debris flow depth estimations yield errors up to 100% depending on the age, complexity and preservation of the deposit. This implies that the accuracy to which volume is reported can also be associated with a 100% error. In this study, the assumption was made that the process of averaging older deposit volumes will cancel some of the error related to difficulties in depth estimations. 4.4.2 Empirical Approach for Estimating Debris Flow Volume A different approach for estimating debris flow volumes was used by Mizuyama et al. (1992), primarily from studies carried out in China and Japan. Their objective was to predict peak discharge from total debris flow volume. There is a significant lack of data on total flow volumes, but information on peak discharge from geomorphic evidence such as lateral deposits and mud lines is frequently available. Consequently, peak discharge (Qmax) is used here as a predictor of debris flow volume (V), that is the inverse of the approach of Mizuyama et al. Peak discharge is usually computed from the product of cross-sectional area of the wetted channel perimeter and average velocity. Cross-sectional areas are readily measured in the field by means of a measuring tape and stadia rod. Velocity is usually not determined directly due to a general lack of automated measuring systems. However, it can be estimated using the superelevation in channel bends or the runup of debris against obstructions, as discussed later (e.g. Chow, 1959; Hungr et al. 1984). Debris runup against obstructions is not used here since splashing of moving debris can produce mud marks on trees and other obstacles at levels significantly higher than the average flow level. Superelevation and cross-section measurements were preferably taken in bedrock reaches to avoid errors due to scour and fill of the channel. Scour was assumed to be negligible whereas fill was determined by extrapolating the observed channel profile thus determining channel geometry. In cases where there are no distinct channel 74 bends where the superelevation equation can be applied, an empirical relationship relating channel cross-section and peak discharge was employed as a first approximation. The data set by Mizuyama et al. (1992) was extended by adding data from Switzerland (Haeberli et al., 1990), Canada (Hungr et al., 1984; VanDine, 1985, Jordan, 1994, Jakob and Bovis, 1996), and the United States (Pierson, personal communication 1994), (Fig. 4.9). In this study, the Qmax -V relationship was based on a data subset that includes only cases from southwestern British Columbia (Fig. 4.10). These data plot well in the global data set, suggesting that different climatic or geologic conditions do not significantly affect this functional relationship. Mizuyama et al. (1992) noticed that stratifying the total data set into "muddy" and "granular" sub-types yielded better predictions than the combined set alone. This observation still held true after addition of other data points from Canadian debris flows. Mizuyama et al. (1992) suggest that this difference might be explained by greater flow resistance of granular debris flows compared with that of muddy flows. In cases where the sediment composition (as a surrogate for "granularity" and "muddiness") was known, it was possible to estimate debris flow peak discharge from the use of the appropriate regression equation (Figure 4.9). Usually it is not possible to sample sediments during a debris flow event which requires that the physical characteristics of debris flow deposits are used to distinguish between the two flow behaviors. Although these differences in character of deposits are fairly well recognized, there is still some doubt as to the physical mechanisms that justify the differentiation between muddy and bouldery debris flows in the context of the volume-discharge relationship of Mizuyama et al. (1992). Their claim that differences in flow resistance explain the different Qmax/total volume relationships can be justified by the fact that flows with lower flow resistance will produce a smaller surge wave, for a given debris mass, compared with granular flows of the same total volume. However, this argument fails to explain the observed large differences between Qm ax and total volume between the two types of flow. Debris flow volume (Vt) is not exclusively a function of peak discharge Qm ax but also of its duration (t). Since debris flows may discharge in surges with different peak discharge, V t is correctly expressed as: 75 [4.1] VT-fQdt 0 where Q is the discharge at time t.. Debris flow channels are commonly not gauged, which makes it impossible to reconstruct the number of surges if no eye-witness accounts are available. In this case, equation 4.1 can be simplified by replacing Q with Q , the mean discharge for the entire duration of the debris flow event. [4.2] V t - g - f Debris flows with higher clay and water contents are known to travel further and deposit at slope angles significantly lower than granular debris flows with little clay in the matrix. This explains the high mobility of lahars. This behaviour is probably due to the low rate of consolidation of fine-textured debris flows which are easily remobilized by a higher subsequent debris surge. Coarse-textured debris flows, by contrast, resist mobilization because of higher frictional strength and more rapid drainage (Jordan, 1994). This implies that, for a given flow depth, fine-textured debris flows will travel further and convey more sediment to the area of deposition than coarse-grained debris flows, which are more prone to in-channel deposition if flow depth declines below a critical threshold. Despite these rheologic criteria that influence the Qmax versus total volume relationship, total volume is still limited by the amount of sediment that can be mobilized during an event. If the amount of sediment is limited, and only a few surges convey the bulk of the sediment, the duration of the debris flow will be short. If, on the other hand, large amounts of sediment are available and can be entrained into the debris flow during the event, it is likely to be of longer duration. The good agreement between the results presented by the Japanese classification and the classification presented in Figure 4.9 can probably be explained by the fact that muddy debris flows in Japan occur primarily on Holocene volcanoes that can be classified as transport-limited due to the quasi infinite amounts of unconsolidated pyroclastic rocks in the source areas. Similarly, debris flows data from Janjago and Hunsuigo, Yunnan, China, which constitute over 50% of data points of the 76 muddy debris flow type in Figure 4.9, occur in a sparsely vegetated region with unlimited amounts of transportable sediments (Mizuyama, pers. comm. 1995). A debris flow at Turbid Creek that was observed by the author in August 1993, and which lasted over half an hour, serves as an example for this type of basin. Following this argument, Mizuyama et al's (1992) classification of debris flows in muddy and granular was replaced with the transport-limited and weathering-limited classification in Figure 4.9. Debris flow volumes were then calculated by applying the appropriate regression equation from Figure 4.10. 4.4.2.1 Flow through channel bends In absence of direct measurements or reliable eye-witness observations, debris flow velocity is frequently approximated from superelevation in channel bends. The forced-vortex equation for superelevation has been used by Chow (1959), Hungr et al. (1984), Johnson and Rodine (1984), Pierson and Scott (1985), Jordan (1994) and many others. It is given as: [4.3] v = (grcCosGtana)0-5 with tan a = Ah/B where g is the gravitational constant, rc is the radius of curvature of the centre line of a channel bend, B is flow width, Ah is the superelevation between the two sides of the flow, a is the banking angle of the flow, and 0 is the longitudinal channel slope. In the field, the longitudinal slope and superelevation were measured with a clinometer. The radius of curvature was obtained by extending a measuring tape along mid-channel and fixed at certain intervals. Compass readings were made at each section and then plotted on graph paper. A drawing compass was used to construct a best-fit circle to the plotted increments and the radius of the arc was measured. For peak discharge calculations, the velocity calculated from equation 4.3 was multiplied by the cross-sectional area at the superelevation site. This procedure introduced three possible sources of error. First, the application of the superelevation formula which can only approximate velocity, second, channel erosion and fill in colluvial reaches, and third, measurement errors. 77 Figure 4.9 Relationship between debris flow peak discharge and total volume for a worldwide data set. Data from British Columbia are superimposed expressed as diamonds and crosses. 10° 10s I 104 o > 10' 10" y = 1420.4 * xA(0.695) y = 5.6*x*(1.564) o X 4 " cP ° X 0 °o ' ° 0 CP o / <jf 0 0 . » • a o ° • • • ° T 5 cfa't i • 0D — ©— - total vol. transport-limited (global) E>— total vol. weathering-limited (global) ^ total vol. weathering-limited (B.C) X total vol. transpatHimited (B.C) ' 1 1 111 i i i • > i 111 i i i i i i I 10 10" Qmax (m3/s) 10J 10" Figure 4.10. Relationship between debris flow peak discharge and total volume from southwest British Columbia 107 10" E s ~ 10s 4) E O •5 104 103 10" - i 1 — m • y = 22.6 * xA(U45) R = 0.481 - y = 342.7 * xA(0.993) R2= 0.957 ^ * w <£p O O y o / ° : * * o / / ° o 'o / / 0 — e- - total vol. weathering-limited (B.C) \ y < total vol. transrxrHirnited 0.C) — i i i a i 1111 i i i i i 1111 i i i i 11111 i i i i 1 1 1 1 10 10" Qmax (m3/s) 10* 10" 78 Wigmosta (1983) analyzed the assumptions underlying the superelevation formula and concluded that they produce self-compensating errors. The main assumptions are: the square of the cross-sectionally averaged velocity can be substituted for the mean of the squares of the individual filamental flow velocities (Chow, 1959); secondly the cross-channel slope is constant; and thirdly B « rc. Equation 4.3 slightly underestimated the average velocity measured by Pierson (1985), who subsequently calculated velocities from superelevations at several locations on lahars near Mount St. Helens. In their original paper, Hungr et al. (1984) used a correction factor of 2.5 because they assumed that the equation would tend to yield an overestimation of velocity. The correction factor was based on experiments in Japan by Mizuyama et al. (1981) that were conducted with relatively low concentrations of sand and fine gravel in water and therefore might not be typical of many debris flows from granitic source areas. Therefore, it can be concluded that equation 4.3 provides the best available approximation of the velocity of a debris flow at a superelevation location. Better estimates can only be gained if the velocity is measured directly and then compared with the computed value. Equation 4.3 was used in all subsequent velocity and peak discharge calculations in this thesis. 4.4.2.2 Channel geometry for debris flow discharge estimation In the last section it was shown that peak discharge can be used to estimate total volume. In some cases, however, it is not possible to estimate peak discharge because velocity could not be calculated according to the superelevation equation. In these cases, channels either had no distinct bend, or slumping in the channel subsequent to the debris flow made it impossible to reconstruct the wetted perimeter at the superelevation site. In these cases, an empirical relationship between channel geometry and peak discharge was used to approximate peak discharge. Relationships between channel geometry and discharge of alluvial rivers have been investigated in numerous studies which are summarized by Wharton (1995). Relationships between bankfull channel dimensions and bankfull discharges were adopted by the United States Geological Survey following suggestions by Langbein (1960) for ungauged rivers. This concept was used for debris flow peak discharge estimation. Commonly peak discharge is correlated with bankfull channel width or cross-sectional area. For alluvial channels, bankfull channel width can be determined 79 with an acceptable degree of accuracy from air photographs. In debris flow channels, even high resolution, large-scale air photographs are often insufficient to determine channel width since tree canopy obstructs the view. In addition, some debris flow channels are less than five meters wide, which requires sophisticated digital stereoplotters to obtain the required accuracy. For this reason, channel reaches not obstructed by debris dams or logjams were selected in the field. In debris flow channels that are primarily incised into colluvium, the cross-section of the channel can be considerably altered by erosion during the debris flow and by slumping of channel side walls subsequent to the event. To avoid this problem, bedrock reaches with pronounced levees along the channel were sought, which yielded the best results since alteration of the channel cross-section during and after the flow in bedrock channels would be negligible in most cases. Bankfull width and cross-section were then measured with stadia rods and measuring tapes or electronic distance measuring devices between the lateral levees. The highest or most widely spaced levees were used to obtain an indication of the maximum instantaneous peak discharge over the time period since the occurrence of this maximum flow. In some cases, several cross-sections were measured and averaged to obtain more reliable results. Figure 4.11 and 4.12 show the correlations of debris flow peak discharge determined from the superelevation equation with cross-sectional area (A) and channel width (WD) for measurements from 40 debris flow channels in southwestern British Columbia. The QnWA relationship is clearly superior to the Qmax^ Wb relationship (Figure 4.12), which was expected given that the cross-section of the channel is a product term of the equation used to determine Qmax- The good agreement in the Qmax/A relationship can also be explained from the fact that velocity is a function of flow depth, channel slope, and unit weight of the debris. Unit weight in debris flows in the area commonly does not exceed 10% variation (Jordan, 1994). Similarly, channel slope as sin 6 exhibits little variation which suggests that depth of flow, and therfore cross-sectional area is the determining variable. For this reason the cross-sectional area regression equation was applied to obtain an estimate of peak discharge in cases where no other velocity data were available. According to this relationship, Qm ax was calculated as: [4.4] 0^=1 .56*^22 Figure 4.11 Relationship between peak discharge and channel cross-sectional area 81 4.4.3 Typical Magnitudes of Debris Flows The use of an empirical regression method to determine the volume of past debris flows requires one fundamental assumption. In most cases, only the most recent debris flow has been investigated with regard to peak discharge since evidence of older flows has been eroded to varying degrees. Therefore, the volume derived from substituting in the appropriate regression equation will yield a total volume of only this most recent event. In some cases, lateral deposits of older debris flows were found at higher levels than those produced by the last event. In these cases, two or more peak discharge and volume estimates were made. To apply the Qmax versus total volume relationship to all debris flows that occurred within the period of frequency determination, the additional assumption must be made that the volume of the last debris flow is typical of most debris flows that occurred within a give time interval in that basin. This assumption, however, can only be supported when there is no evidence of larger flows along the recent channel upstream from the fan apex, and where there is reason to believe that significantly smaller debris flows have not reached the fan apex. The fan morphology itself is usually too complex to permit this assumption to be made. No doubt, deposits from older debris flows could have been reworked or eroded to an extent where different debris flow events are no longer discernible. Debris flow volumes in weathering-limited basins are governed by a threshold that is determined by the amount of sediment accumulated since time to, which is the time elapsed since the last debris flow. Even very intense rainstorms will not be able to trigger debris flows until an intrinsic threshold controlling material accumulation has been reached. This threshold will determine the minimum volume transportable during a debris flow. Mobilization and deposition of minor amounts of sediment within the channel are common during rainstorms as observed at several locations during repeated visits after heavy rainstorms. However, debris flows in this study were only counted when the majority of sediment was transported at least as far as the fan apex. In order to reach the fan apex, a certain flow depth or volume is required to maintain momentum. It has been observed in several cases that debris flows of small magnitude stalled in 82 the channel without reaching the fan. If the minimum and maximum volume threshold approaches the mean magnitude, the assumption can be made that the last flow can be used as an approximation for average debris flow volume. This average volume in weathering-limited basins will be referred to as the typical volume in the following text. "Typical" in this context is defined as the expected mean magnitude of flows recorded. This implies that very small and very large flows, if they occur, will remain unrecorded because they are either unrecognizable in the field, or occur at time scales outside the window of this study. They can therefore not be included in the frequency-magnitude analysis. In many cases, the magnitude of the last flow was used as the "typical" magnitude, reflecting the observation that the magnitude of recordable minimum and maximum flows are fairly close in value to the last observed flow over the time frame of observation. A possible physical explanation for this behaviour can be extracted from the work by Bovis and Dagg (1988), who suggested that high magnitude-low frequency debris flows should be typical for weathering-limited basins that display streams not steep enough to produce debris flows directly from hillslope failures. According to Bovis and Dagg, small increases in the friction angle of the channel material and large increases in the hydraulic conductivity create more stable conditions for debris in the channel. These circumstances favour the accumulation of large, but marginally stable debris deposits which may ultimately lead to debris flows with high magnitude and low frequency characteristics. Apart from investigating older deposits in the vicinity of the channel, a good indication as to whether the last debris flow was preceded by a considerably larger flow is the growth of mature trees on top of, or in the immediate vicinity of, the last debris flow. If there was a debris flow of significantly higher magnitude preceding the last documented flow, those trees would either be scarred or destroyed. If two or more lateral deposits could be discerned and Qmax could be calculated, the arithmetic mean of these events was used rather than the volume of the last flow alone. For most weathering-limited basins, it seemed reasonable to adopt the concept of "typical" flow magnitudes over the time scales relevant to this study, which then allows the use of the relationship between Q m a x and total debris flow volume. Typical magnitudes, as derived from the arithmetic mean of documented flows, or deduced from field evidence as discussed above, were entered into the statistical analysis for Qmax and total volume. For engineering purposes, 83 computation of return intervals for variable magnitudes is less important when the largest debris flow volume can be determined from field evidence and an approximate time frame given over which no larger debris flow has occurred. For example, the debris flow channel descending from Mount Ross ("Nightmare Creek") northwest of Pemberton is flanked by lateral deposits of the 1984 event, which was only slightly larger than the 1987 event. A n examination of trees and deposits along the channel and fan shows no evidence for flows that significantly exceeded the 1987 debris flow. Mature Douglas firs in the vicinity of the channel are older than 250 years as determined by tree ring counting from cores. Therefore the 1987 event with Qmax of 400 m3/s and a volume of approximately 20,000 m 3 was probably the largest debris flow over a 250 year period even though an event of similar magnitude had occurred only three years previously. Therefore it seems reasonable to adopt the concept of "typical" magnitude for planning purposes as well as geomorphic calculations. A recent debris flow at Hope Creek, which destroyed a restaurant near the town of Hope, showed that even in weathering-limited basins, this concept does not always apply, and available field evidence should be carefully checked before it is adopted. In the case of Hope Creek, an ongoing study revealed that unexpected erosion in a mainly colluvial channel was responsible for about seven times the maximum expected volume (Jakob et al., 1996). An anomalously large debris flow occurred in Pierce Creek in the Chilliwack Valley in December 1995. In this case, the initial failure contributed about 50% of the total debris volume of approximately 60,000 m 3 . This failure occurred along a very steep rock slope covered with a mantle of thick colluvium. These two examples illustrate that the concept of typical magnitude should only be applied to channels which are primarily bedrock controlled, and in basins where there is no evidence for point source failures that could incorporate large volumes of material in the debris flow. In some transport-limited basins, flow volumes can vary over orders of magnitude if debris flow initiation includes large-volume slope failures that are incorporated into the debris flow. Therefore, in these basins, the concept of typical magnitude can not be applied, and more data on the volume of individual flows is needed to determine a meaningful average. Examples of this type of basin are Cheekye River basin, on the southwestern flanks of Mount Garibaldi, Turbid Creek at Mount Cayley, and Devastation Creek at Mount Meager, all located in Quaternary volcanic centres. 84 In the case of Mount Cayley, several estimated volumes were available, including one very large debris flow resulting from the 1984 debris avalanche which allowed a preliminary estimate of average debris flow volume. 4.4.4 Maximum expected magnitude Maximum expected magnitude is defined as the total volume of mobilizable material that can be discharged onto the depositional area during a single debris flow. In contrast to the term design magnitude, no discrete time period or service period of a structure can be attached to this definition since this would assume that the frequency-magnitude relationship of debris flows is known, which is usually not the case. Maximum expected magnitude is included in this study as a means to use erodibility of channel materials and stored sediment to estimate debris flow volumes assuming that no major channel changes occur. Maximum expected magnitude has not been included in the statistical analysis (Chapter 6) because it was impossible to survey the entire channel in most cases, and because of uncertainties involved in the calibration of the erodibility indices. Determination of maximum expected magnitude from sediment stored in the channel, requires that several criteria must be met: i) The debris flow channel must be accessible or at least surveyable by some means over its entire length. ii) The channel bed must consist of non-erodible material (usually bedrock). iii) The channel must be scoured to bedrock after each flow. iv) The sediment recharged into the channel should account for most of the mobilizable debris flow material. If criterion (i) is not met, extrapolation of results will lead to gross errors, unless the channel reaches that were not accessible can be analyzed by using high resolution air photographs. Alternatively, functions can be used to describe the distribution of sediment volume with distance from the fan. Criteria (ii) and (iii) are necessary to establish a time-magnitude relationship. Criterion (iv) is essential, since large side slope failures would falsify the volume calculations. In bedrock channels, cross-sections were surveyed every 50 to 100 metres depending on the 85 variability of the channel geometry and stored sediment. Following the analytical approach developed by Oden (1994), cross-sections were plotted and bedrock outcrops were connected to approximate the bedrock profile beneath the sediment fill. For non-bedrock channels that pass through erodible material, Hungr (pers. comm. 1995) developed a more rigorous empirical approach involving erodibility indices computed for each channel segment, then summed over the entire channel length. This method of analysis assumes that the complete cross-section of the existing channel will be affected during a debris flow event. Equation 4.5 lists the parameters that are to be determined in the field. where M is magnitude in m 3, n is the number of creek reaches, Lj is the length of each reach in metres, Aj is the drainage area adjacent to the reach, ei is a bed erodibility factor, and e2 is an embankment erodibility factor. Index ei is a function of the grain-size of bed material and the channel width, whereas the index e2 is a function of the channel depth, its degree of vegetation, and the gradient of the side slopes. Both indices can be adjusted for channel confinement and channel slope (Tables 4.1 and 4.2). All channels in this study were surveyed according to this method as far as safe access allowed. Hungr (pers. comm., 1995) indicated that this measure of debris flow volume should be termed maximum magnitude since it will most likely yield a maximum rather than an average flow volume. Analyses for maximum magnitude estimates in colluvial channels must meet criteria (i) and (iv), although it is possible to account for slope failures that enter the channel if volumes of scars can be reconstructed. [4.5] 86 Table 4.1. Debris flow erodibility classification for channel beds (Hungr, pers. comm. 1995) Class Channel Material ei 1 bedrock(clean) 0 2 very large boulders D>0.3*B 2 3 large boulders D=(0.1-0.3)«B 4 4 medium boulders D=(0.05-0.1)«B 7 5 small boulders D<0.05«B 8 where B is channel width and D is modal boulder diameter If channel is unconfined, lower class by two grades If slope is less than 15°, lower class by one grade Table 4.2. Debris flow erodibility classification for banks (Hungr, pers. comm. 1995) Class Height1 (m) Stability e2 1 < 1 very stable or low bank 0 2 1-2 stable 1 3 1-2 moderately stable 2 4 1-2 unstable 4 5 2-5 stable 2 6 2-5 moderately stable 4 7 2-5 unstable 10 8 5-10 stable 4 9 5-10 moderately stable 8 10 5-10 unstable 20 1 Height of bank likely to be affected by sliding and slumping and erosion Stability categories: stable: intact, vegetated slope, angle less than 35° moderately stable: vegetated slope with less than 30% of surface affected by shallow sliding unstable: slope with more than 30% of surface affected by shallow sliding, potential sliding, creep or erosion If channel is unconfined, or gradient less than 12°, lower stability category by one grade The volumetric data calculated from Equation 4.5 were plotted against channel length to detect a trend in sediment availability with distance upslope of the debris flow fan. It was anticipated that this trend might be approximated by a negative exponential function since i) the contributing area is increasing as a power function of distance downstream, and therefore more material for accumulation should be available; and (ii) there is a general decrease in slope towards the fan which would facilitate sediment deposition. Figure 4.13 shows examples of volume estimates along several debris flow channels. Gunbarrel I, II, III, Fools' Gold Creek, Pothole Creek and McLeod Creek all show an increase of available sediment with proximity to the fan that can be approximated 87 Figure 4.13 In-channel sediment storage for selected sites 7000 J5 6000 % 5000 •5 4000 i % 3000 I 2000 S IOOO o Fig. 4.13 JV Wildcat Creek ' y j • I 1 I ' I 1 I 1 I 1 I 1 I • I 1 I 1 I 1 Figure 4.13-D Capricorn Creek 100 I 1000 channel length from fan base (m) 25000 I'I'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 'I 1 100 1000 2000 3000 channel length from fan base (m) Figure 4.13.B Turbid Creek 30000 100 1000 2000 channel length from fan base (m) 100 500 1000 channel length from fan base (m) Fig. 4.13.C Clearwater Creek T W T W T W T W T W T W I W T W I W 100 1000 channel length from fan base (m) Figure 4.13.F Snake Gully 14000 100 1000 2000 channel length from fan base (m) 88 Figure 4.13. (cont.) In-channel sediment storage for selected sites Fig. 4.13.H Gunbarrel II I ' I ' I ' I 1 I ' I 1 I 1 I 1 I ' I 1 I 1 I 100 500 1000 1300 channel length from fan base (m) Fig. 4.13.K McLeod Creek 12000 100 1000 2000 channel length from fan base (m) Fig. 4.13.1 Gunbarrel III 30000 I ' | ' I 1 I 1 I 1 I ' I 1 I ' I ' I 1 I ' I ' I ' 100 500 1000 1250 channel length from fan base (m) Fig. 4.13.L Fools' Gold Creek 10000 I i I ' I 1 I 1 I 1 I 1 I 1 I ' I 1 I 1 I 1 I 1 100 500 1000 channel length from fan base (m) 89 by a positive exponential function, which is in contrast with the assumption made above. To investigate this contradiction, a closer examination of the channel morphology is required. Gunbarrel I, II, III, Fools' Gold Creek, and Pothole Creek channels are incised into colluvium, consisting mainly of older debris flow deposits. While sediment is deposited in the lower reaches of these channels, erosion and incision are typical in the upper parts which deepen and widen the channel. McLeod Creek deviates from these characteristics, since much of the measured channel is confined by bedrock. The abrupt increase in sediment availability at around 1500 m upstream from the fan is due to a channel widening and the accumulation of colluvium at this location. There is evidence that much debris is deposited at this location, probably because of water drainage associated with a loss of channel confinement and the decrease in channel slope. The upper parts of the channel were inaccessible because of a waterfall above this accumulation area, but air photographs show that the channel narrows again above the waterfall. Extrapolation based on an exponential function would therefore greatly overestimate the available sediment in the unsurveyed reaches of this particular channel. The implication of this observation is that a function developed for the lower reaches of a debris flow channel are not necessarily applicable over its entire length. Figures 4.13 demonstrate the variability in sediment availability that can be encountered in debris flow channels. The spikes of stored sediment in Turbid Creek (Figure 4.13.B) are due to channel point sources such as rotational and translational slumps of older debris flow deposits that entered the channel at these locations. Spikes that have a similar explanation can be observed at Capricorn Creek and, although less pronounced, in Boundary Creek. At Capricorn Creek an increase in sediment availability up-channel can explained by small debris flow channels that enter the main channel between 1500 and 2500 m upslope from the fan. Large rotational slumps in heavily weathered volcanic rock enter Boundary Creek between 1000 and 1500 m and explain the peak in sediment availability in this reach. Clearwater Creek is also marked by several sharp increases in sediment storage, although the channel is flanked primarily by fairly competent bedrock that is conducive to sidewall stability. Sediment peaks along this channel therefore cannot be associated with prominent sediment sources along the channel as observed at Turbid Creek, Capricorn Creek, and Boundary Creek. Observed sediment peaks at around 150 m, 400 m, 800 m, and 1000 m in Clearwater Creek channel probably represent smaller flows that were deposited 90 intermittently before reaching the fan. These intermittent zones of deposition, which are analogous to sediment wedges in coastal streams, are then remobilized during larger debris flow events. Spikes in sediment accumulation along Snake Gully can be explained primarily by fluctuating channel width in a mainly colluvial channel. Only Wildcat Creek roughly resembles the originally expected negative exponential increase in sediment availability downchannel. Figure 4.13 demonstrates that, although common sediment storage pattern exist among some channels, no single arrangement of sediment can be applied for all channels. Clearly, sediment storage is dependent on the locations of point sources having links to the main channel, and the composition and geometry of channel side walls. Although no common mathematical function exists to explain sediment storage within all debris flow gullies, the positive exponential model shows some promise for channels in deep colluvial material. Simple extrapolation of stored volumes beyond the measured channel reaches is therefore not possible, unless one is confident of the channel type within the inaccessible reaches. If an estimation of maximum expected magnitude is the objective of a study, the proposed procedure is to survey the channel over as far as is possible, then plot cumulative volume against channel distance measured from either the fan apex or fan base. If a function can be fitted satisfactorily to the data, it can be used to extrapolate sediment volumes in the remaining part of the well-defined channel that was not measured directly. Air photographs should be consulted to confirm that no abrupt change in channel geometry gives reason for a significant change in sediment storage characteristics. Figure 4.13.G exemplifies this procedure for the Gunbarrel I debris flow system. Gunbarrel I was surveyed over its entire channel length by approximating sediment storage and erodibility over 19 channel reaches. For the purpose of comparability, only the first 11 data points were included in the analysis. These points alone only indicate a positive exponential function of sediment availability in the channel. Air photographs do not indicate a significant disruption of this trend. An exponential function was fitted to the data and sediment volume (V) was calculated according to: 91 [4.6] V = 458.e0 0 0 3 * L with L denoting the channel length from the fan base. As indicated by Figure 4.13.G, the calculated volumes closely approximate the observed sediment volumes. If no appropriate function can be found to estimate the total sediment stored in the channel, the mean volume per channel reach can be used to estimate volume as a first approximation. Air photographs must be reviewed to confirm the validity of this procedure since it is obvious that this procedure is only applicable in some cases. Most channels in this study Eire underlain by both sediment and bedrock. For this reason the method of determining volumes from indices of erodibility, as well as the sediment volumes stored above the bedrock base were combined. Many channels in this study were accessible over only short reaches because waterfalls, unhikable canyons, and dangerously slippery conditions frequently stopped, or considerably slowed the field work. Table 4.3 shows all basins in which a reasonable estimation of maximum expected magnitude could be made on the basis of the methods outlined above. 4.4.5 Maximum Expected Magnitude - Typical Magnitude Relationship It was noted in section 4.4.4 that maximum expected magnitude is a desired variable in many engineering applications. In section 4.4.2 and 4.4.3 it was pointed out that the estimation of maximum expected magnitudes from field data on peak discharge is problematical. Having determined the maximum expected magnitude (Vm a x) and typical magnitude (V t y p) of several debris flow channels, I investigated the relationship between these two measures of debris flow volume with the view to predicting maximum expected magnitude from typical magnitude data. Figure 4.14 indicates a relatively weak relationship (R2 = 0.39) between the two variables. A stratification into weathering-limited and transport-limited basins yields an improvement of the maximum expected magnitude (Vm a x) for the weathering-limited basins only (Figure 4.14). The larger scatter of data points of transport - limited basins may be explained by the increasing 92 Table 4.3. Estimation of maximum expected magnitude in selected debris flow channels Basin name channel total curve fit function or length channel extrapolation surveyed length (m) (m) Boundary 1700 2600 vd -x*ru 330,000 176,000 510, 000 Capricorn 2900 10000 Vd -x*ru + vm 425,000 890,000 1,300, 000 Fools Gold 1200 1800 Vd - **ru + Vm 52,000 17,000 70, 000 Clearwater 1400 3100 vd - **ru + vm 44,000 53,000 100, 000 Deepa 1200 1500 vd -x*ru + vm 59,000 15,000 74, 000 Fergusson 1400 5400 vd -x*ru + Vm 55,000 160,000 120 000 Gunbarrel I 920 920 entire channe surveyed 36,000 36, 000 Gunbarrel II 1300 1300 entire channe surveyed 100,000 100 000 Gunbarrel HI 1250 1250 entire channel surveyed 150,000 150 000 Hotsprings 3300 5700 Vd - **ru + Vm 240,000 175,000 400 000 Kaboose 1500 1700 Vd m x*ru + vm 15,000 2,000 17 000 McLeod 2100 3900 Vd - x*ru + Vm 73,000 63,000 140 000 Mount Currie 1300 2700 Vd - x*ru + vm 190,000 200,000 390 ,000 No Good 1800 3600 Vd - **ru + vm 380,000 380,000 760 ,000 Pothole 100 2000 y<i • n • £(-2030+l&t) i - l 66,000 260,000 300 ,000 Snake 1850 1850 entire channel surveyed 60,000 60 ,000 Turbid 2430 6900 Vd - x*ru + vm 210,000 390,000 600 ,000 Wildcat 1300 2100 Vd -ru + vm 30,000 2,000 32 ,000 Terminal 1840 11400 Vd + vm 77,000 400,000 480 ,000 where X denotes the mean volume per channel reach and ru the unobserved reaches. measured extrapolated maximum volume V m volume V e expected magnitude (m3) (m3) (m3) complexity of channel morphology, by the more numerous point sources as well as by the increasing uncertainty of mean magnitude estimations. It is interesting to note that an increase in typical magnitude of transport-limited basins (Vdt) is associated with a larger increase in compared to weathering-limited basins, which again emphasizes the point that typical magnitudes for transport-limited basins are of limited value for engineering and planning purposes. Figure 4.14 also shows that the concept of "typical" magnitude for weathering-limited basins cannot be applied if the erodibility indices are used for the calculation of maximum expected magnitudes, 93 Figure 4.14. Relationship between maximum expectable magnitude and typical magnitude c LLl Q P 10 < Q o UJ Cu X < TYPICAL MAGNITUDE (V ) in m 3 Figure 4.15. Relationship between maximum expectable magnitude and typical magnitude stratified into weathering-limited and transport-limited basins c t 10" U UJ Q. X 3 x < 2 107 10 5 U 10" 1 — I T M i l l I I 1 1 1 1 1 1 10' 10d 104 10s 10* TYPICAL MAGNITUDE (V ) in m3 94 since in this case maximum magnitude values would have to closely match typical magnitudes. However, in this figure typical magnitudes for weathering-limited basins are at least an order of magnitude smaller than respective maximum expected magnitudes. The most likely explanation for this discrepancy is that the erodibility index requires better calibration that can only be achieved by monitoring active debris flow channels with regard to differential erosion along selected channel reaches. Particularly in colluvial channels, where incision has acted over centuries or millennia, the likelihood that the channel is completely filled during a maximum-magnitude debris flow may considerably exceed the life-time of a structure. Thus, a return period of this magnitude is irrelevant for planning purposes suggesting that some adjustment should be made with regard to the maximum discharge by using methods proposed earlier in this chapter. Despite this limitation, the investigation of maximum expected magnitudes has revealed some interesting sediment availability patterns which may find use in future debris flow magnitude studies. 4.4.6 Summary Debris flow volumes can be determined in three ways. First, by directly measuring the depths and areas of deposits, an approach which was used whenever a well preserved recent deposit allowed volume estimates. This method was complemented by dendrochronological evidence to delineate the margins of debris flows, and by dating adventitious roots and measuring their depth to determine both the age and volume of deposits. However, the latter procedure is very laborious, and sometimes dangerous and has therefore not found much application in this study. The second approach uses an empirical relationship between peak discharge and total volume. Since peak discharge for at least the last debris flow can usually be determined in the field, order-of-magnitude approximations can be made for total volume. In channels lacking distinct bends, peak discharge can also be determined using an empirical relationship relating peak discharge to cross-sectional area. The third method for estimating debris volume is based on the potential erodibility of colluvial channels. This requires channels to be surveyed over as much of their length as terrain constraints permit. A mathematical function fitted to the volume distribution within the surveyed 95 channel allows extrapolation over the remaining channel reaches. Unfortunately, erodibility indices are insufficiently well calibrated to allow the construction of meaningful correlations between mean magnitudes and maximum expected magnitudes. Colluvial channels known to have high debris flow frequencies (e.g. Boundary Creek, No Good Creek, Turbid Creek) could profitably be instrumented with scour chains along the channel bottom and the sidewalls, and resurveyed after each event along several reaches to derive more appropriate indices of erodibility. 96 CHAPTER 5. PHYSICAL CHARACTERISTICS OF DEBRIS FLOW BASINS 5.1 INTRODUCTION In the first chapter it was noted that to adequately characterize basin attributes related to debris flows, a combination of morphometric and geotechnical factors is appropriate. In this chapter, I discuss the considerations used in the selection of the various morphometric and geotechnical parameters used in this study. First, it is appropriate to provide some background discussion of this topic. Chorley et al. (1957, p. 138) defined geomorphometry, a subdiscipline of geomorphology, as the science " which treats the geometry of the landscape". Mark (1975) defined it as the science that "attempts to describe quantitatively the form of the land surface". Morphometric variables were selected according to the following criteria: 1. Sufficient evidence from existing studies that the variables influence the frequency and magnitude of debris flows. 2. The variables are readily measured. 3. Since many geomorphometric parameters are interrelated, only those of maximum relevance and having a minimum of information replication were chosen. That is data redundancies are kept to minimum. 4. The need for expert judgment in parameter measurement is minimal to avoid widely differing interpretations of the same terrain features. 5. The selected variables include a wide variety of morphometric and geological conditions to ensure a high degree of transfer of the resulting model into other mountain areas. 5.2 USE OF MORPHOMETRY IN DEBRIS FLOW STUDIES Although there have been a number of studies on the frequency and magnitude of debris flows, only a few studies have attempted to predict the size or frequency of debris flows based on basin variables. Much of this work has already been reviewed in Chapter 2. 97 Table 5.1. Use of morphometry in debris flow studies Author i Objectives Original variables Variables retained Harnpel (1977) 1 prediction of magnitude, j factor of safety basin area, fan slope basin area, fan slope Ikeya(1981a) | prediction of magnitude channel length and width channel length and width Ikeya (1981b) j prediction of magnitude i normalized by area basin area basin area Watanabe(1981) 1 prediction of magnitude i andfrequency basin area basin area Okubo and Mizuyama (1981) I prediction of factor of 1 safety channel length, width, depth of scour channel length, width, depth of scour Takahashi (1981) 1 prediction of magnitude i normalized by area basin area basin area Ikeya and Mizuyama (1982) j prediction of magnitude channel length channel length Mizuyama (1982) | p r e d i c t i ° n ° f m a g n i t u d e i normalized by area basin area basin area Kronfellner-Kraus (1983) I prediction of magnitude mean slope, basin area mean slope, basin area Thurber (1983) 1 prediction of magnitude basin area basin area Hungr et al. (1984) 1 prediction of magnitude basin area basin area VanDine (1985) 1 prediction of magnitude basin area basin area Jackson (1987) j process discrimination 1 (fluvial/debris flow) basin relief, mean slope, basin area. Melton's ruggedness number mean slope, Melton's ruggedness number, basin area Johnson et al. (1991) j prediction of frequency j and magnitude hypsometry, relief ratio, drainage density, total stream length, bifurcation ratio, basin elongation ratio hypsometry, relief ratio, Mark (1992) j prediction of frequency slope slope Ellen (1993) 1 prediction of frequency slope slope Table 5.1 includes data from VanDine (1985), Table 4. 98 Most studies have considered only one dependent variable (frequency or magnitude) which is then correlated with one morphometrical independent variable (e.g. Hampel, 1977; Ikeya, 1981b; Watanabe, 1981; Hungr et al. 1984). Results from these bivariate correlations are generally poor, which is hardly surprising, given the variety of possible factors that can influence the triggering and volume of debris flows. Nevertheless, some bivariate correlations have proved useful in assigning limiting envelopes for debris flow magnitude. Figure 5.1 demonstrates the relationship between basin area and debris flow volume of all study sites which can be useful in assigning a limiting envelope to determine whether a basin seems to be susceptible to debris flows activity. Although the obvious scatter of points does not allow the construction of a regression curve with predictive capabilities, an envelope can be drawn that provides a limiting value of volume for a given basin area. This envelope can provide a first approximation of minimum magnitudes for the respective basin size. Hungr et al. (1984) and VanDine (1985) also related debris flow magnitudes to basin area and found that debris flows were most likely in basins with areas less than 5 km2 along the eastern shore of Howe Sound, north of Vancouver. The reason for this is that an increase in basin size is generally related to a decrease in the mean gradient, until eventually debris flows that originate in the steeper part of the basin are unlikely to reach the fan. In summary, morphometric variables have been used in landslide analysis to discriminate between stable and unstable areas. Studies which attempted to predict debris flow frequency or magnitude based on morphometry focused on univariate correlations which at best can provide a limiting envelope for the independent variable, but are insufficient for predictive purposes. A notable exception is the study by Johnson et al. (1991) who integrated several morphometric factors in their model. Despite these efforts, little is known about the causal effects of individual morphometric or geotechnical variables, or combinations thereof, on the occurrence of debris flows. For this reason, variable choice is the first important step in building an appropriate predictive model. 99 Figure 5.1 Relationship between basin area and debris flow volume 5.3 CHOICE OF MORPHOMETRIC PARAMETERS 5.3.1 Basin Area and Percent Active Area Basin area is an essential parameter controlling debris flow occurrence for several reasons. It subsumes morphometric parameters such as gradient as well as process-linked parameters such as runoff and sediment yield. An increase in basin size also increases the amount of potentially transportable sediment, since larger basins usually contain more potential debris flow source areas, and the longer channel networks mean more stored sediment. However, as noted earlier, there is a finite upper basin size beyond which it is unlikely that debris flows will reach the fan. At the other end of the scale, very small basins contain very steep slopes and limited runoff are rockfall dominated, and a cone rather than a debris flow fan develops at the basin outlet. However, it is clearly unwise to use basin area as the sole criterion to discriminate between debris flow and non-100 debris flow basins. For example, in this study, several debris flow basins with areas exceeding 10 km2 are documented, although according to Thurber Engineering Ltd. (1983) this situation is considered quite rare in the Howe Sound area. The discrepancy is perhaps explained by the fact that the sample examined by Thurber was mainly bouldery debris flows of high frictional strength from predominantly granitic source areas. Bouldery debris flows tend to have higher viscosity, lower mobility, and steeper stopping angle than those from volcanic source areas, which are typically muddy and highly mobile. This suggests that an adjustment of the area-debris flow occurrence envelope is appropriate according to debris flow material type and mobility. In some cases, even bouldery debris flows can occur in large watersheds, particularly in hanging valleys where there is an abrupt increase of gradient in the lower part of the channel. Alternatively, temporary channel blockage and debris release during the rupture of landslide dams or extremely high jokulhlaup discharges, can both produce debris flows in basins with low overall gradients. A recent debris flow at Pierce Creek, which drains a 6.5 km2 large watershed in Chilliwack Valley, serves as an example. Here, a hanging valley with an overall gradient of approximately 14 degrees initiated a debris flow in a 42° steep old landslide deposit close.to the basin outlet and commenced its path down a 20-25° degree channel. Although basin area is a useful predictor variable due to its correlations with runoff, sediment yield, and slope, it does not provide any information on the basin area actually feeding sediment in the debris flow channels. For this reason, percent active area (%ACT) was introduced. Percent active area is a measure of the proportion of the basin that is actively shedding material to the debris flow channel. This variable is also deemed important because it provides information on the amount of available material, even if the thickness of the mapped unit cannot readily be determined. Percent active area was determined from air photograph analysis while basin area was determined on 1:50,000 topographic maps by digital planimetry. 5.3.2 Slope Slope angle controls the gravitational force driving most geomorphic work and thus both debris flow initiation and debris transport. Strahler (1956) used slope sine as a surrogate for the 101 downslope component of the acceleration of gravity. Tangent slope has also been used, and is the first derivative of altitude with respect to distance. In this study, slope is contained as a tangent function in the relief ratio (RELRAT). Mark (1974) pointed out that, unlike those geomorphic parameters which are defined for finite sub-areas of a surface, slope is defined at every point as the gradient of a plane tangent to the surface at that point. In practice, however, slope is calculated from contour maps and was measured from the highest to lowest point in the basin in this study. Modern geographic information systems (GIS) offer a variety of slope measures which are easily computed if topographic information is available in digital form. GIS was not applied for two reasons. First, this study aims for simple data acquisition with little technical requirements. However, most basins were not available as digital files in 1:20,000 scale. Second, basin characteristics that are routinely offered by GIS do not provide any more meaningful information than that obtained by manual data gathering. 5.3.3 Hypsometric Integral Hypsometry has been defined by Clark (1966) as "the measurement of the interrelationships of area and altitude". These measures describe the distribution of landmass with elevation and are based on the hypsometric curve; which plots the relative cumulative basin area against relative elevation. Strahler (1952) and Langbein (1947) pointed out that the hypsometric curve might be useful in hydrologic and erosional studies. In this study, hypsometric integral was used because, if all other factors are constant, convex or straight basins with high hypsometric integrals should be more active in producing debris flows than basins with concave long profiles and low hypsometric integrals. The reason is that debris deposition rather than mobilization tends to occur in local concavities due to a decrease in slope. In reality, slopes are highly variable within a basin and convex reaches tend to alternate with concave reaches. Even minor changes in local slope may significantly change the depositional behaviour of a debris flow. The hypsometric curve can therefore only be an approximation of the actual basin curvature at the scale of analysis used in this study. 102 The "hypsometric integral" is derived from the hypsometric curve. Strahler (1952, p. 1121) defines the hypsometric integral (HI) as: i [5.1] HI=Ja(h)dh 0 where a(h) is the hypsometric function and h is the relative height defined by the difference between the actual elevation z and the lowest elevation divided by the total relief. Haan and Johnson (1966) used the elevation of randomly located points to construct the hypsometric curve, which would reduce the time required for obtaining the curve as opposed to the conventional method by a factor of between 4 and 10. Finally, Pike and Wilson (1971) showed mathematically, that the elevation-relief ratio E (Wood and Snell, 1960) equals the hypsometric integral: „ Z — Zmin [5.2] E Z max — Zmin with z being the mean elevation. Pike and Wilson (1971) used point sampling to calculate the elevation-relief ratio. They claim that this method reduces the time required to obtain the hypsometric integral by a factor of 3. The three numbers required can easily be obtained from a contour map. Minimum and maximum elevation can quickly be read off the map. Mean elevation is obtained by a sample of points evenly distributed over the drainage basin which can be accomplished by overlaying a transparent grid of regularly spaced points. Comparing results from hypsometric integrals obtained from 10 basins using both methods, Pike and Wilson (1971) showed that the absolute difference between the two data sets amounts to 1%. They claim that a sample size of 40 to 50 points would yield values of E accurate to 0.01. Using different sized computer-generated grid overlays for the same basins in this study, it was found that 30 -35 points are sufficient to achieve this same level of accuracy. Pike and Wilson's method was chosen because of its accuracy, speed and ease of use. To further improve efficiency of the method, several overlays with different grid point spacing were produced to accommodate varying basin size, and the one with the number of sample points closest to 35 was used. The average time to 103 obtain the hypsometric integral using this method averages between 15 and 20 minutes, compared to 45 minutes on average for the graphical solution. Additional elevations were marked onto contour lines to avoid map reading errors. This is particularly important when contours are very narrowly spaced. Elevations were read to the closest 1/2 contour (50 feet or 20 m). 5.3.4 Basin Relief and Relief Ratio Basin relief is defined as the vertical difference between the highest and lowest point in a basin, and is therefore an expression of the potential energy in a basin. In this study the lowest point was determined as the apex of the debris fan because this point is best defined in most cases. Relief is closely related to basin slope (e.g. Salisbury, 1962) and basin size, and may therefore be normalized by dividing it by other linear dimensions of the basin. Relief ratio is defined as the ratio of relief and horizontal distance between the highest and lowest point of the basin. It describes the vertical dimension or amplitude of topography (Mark, 1974). Relief ratio was introduced as a geomorphometric measure by Partsch (1911), who coined the term "Reliefenergie". Schumm (1954, 1956, 1963) employed relief quantitatively in several studies. He showed that sediment yield was closely related to the ratio of basin relief to basin diameter for small basins in the southwestern United States. Ahnert (1970, 1972) adopted the same measure of basin relief for correlation with denudation rates, subsequently integrated into theoretical models of slope evolution. Relief ratio incorporates slope as already stated and for this reason, slope will not be included in the statistical analysis in Chapter 6 to avoid data redundancy and problem of multicollinearity. A correlation between relief ratio and the occurrence of debris flow events might be expected since the steeper a basin is, the closer the removable sediment stored in the channels will be to its angle of repose. Furthermore, steeper channels are more likely to convey all debris to the area of deposition with little intermittent storage. However, basins with steep overall slope, but pronounced local concavities can promote debris flow deposition and prevent debris flows from reaching the fan. Angel Creek in the Mount Meager volcanic complex, which has not been included in the sample because of a lack of recent debris flow deposits on the fan, serves as an 104 example for this basin type. At this location, a late Holocene rock avalanche occurred which left a scar in a cirque-like basin, which was subsequently scoured by ice during neoglacial glacier advances. These ice re-advances have produced a half-moon shaped moraine with an approximate height of 10 metres. Frequent debris flows from adjacent talus slopes lose their momentum at the bottom of the cirque or run up against the moraine (Figure 5.2 ). This example shows that the measures basin relief and relief ratio alone can not provide conclusive answers about the frequency and magnitude characteristics of debris flow basins, and other variables must be considered if specific basin characteristics are to be included in a predictive model. Figure 5.2. Fresh debris flow deposits at the bottom of Angel Creek cirque, Mount Meager 105 5.3.5 Roughness and Drainage Density Roughness refers to the irregularity of a topographic surface (Mark, 1974). Roughness or ruggedness has proven difficult to define numerically. It is an important variable in this study because it reflects the true surface area developed by dissection. Hobson (1972) pointed out that there are as many roughness definitions as there are roughness studies which is in accordance with the suggestion by Stone and Dugundji (1965), Hobson (1967) and Mark (1974) that any measure of roughness should be consistently defined by several parameters - a principle which is followed in this study. Stone and Dugundji included five, and Hobson nine different roughness measures in their studies. The terms texture and grain are generally used to describe the significant "wavelength" of topographic detail. Texture is defined as the shortest, and grain the longest, significant wavelength in the topography. Wood and Snell (1960) described grain as "the size of area over which the other factors are to be measured. It is dependent on the spacing of major ridges and valleys and thus indicates texture of topography." Smith (1950) suggested a texture ratio: T=N/P, with N being the number of crenulations on the selected contour and P, the length of the perimeter of the basin. Smith chose the contour displaying the most crenulations. He discovered that the texture ratio is closely related to drainage density by the power function: Dd = 1.658 T 1 1 1 5 . As Mark (1974) pointed out, this correlation is not surprising because the inverse of T is related to the average distance between contour crenulations along the selected contour. Since each crenulation represents a perennial or ephemeral channel, the inverse of T must be related to the mean distance between channels, which is the inverse of drainage density. Another variable describing basin roughness is drainage density Dd Drainage density is defined as the total length of stream channels per unit area by Horton (1945, p. 283). Dd has been used successfully in several studies and close correlations have been found with mean stream discharge (e.g. Carlston, 1963), mean annual precipitation (Chorley and Morgan, 1962) and sediment yield (Abrahams, 1972). Roberts and Klingman (1972) showed that there is a good relation between instantaneous stream discharge and the total length of flowing channels at a particular time. Mark (1974) summarized several methods of sampling for drainage density. In 106 his study drainage density was calculated by measuring the total length of "blue lines" on the topographic map and dividing them by contributing area. He found an inverse relationship between Ddand the year of map publication. In this study, drainage density was sampled to obtain a quantitative measure of linear debris-delivering systems. It was interpreted and sampled in two ways. The conventional method measures "blue line" streams from topographic maps at the scale 1:50,000 divided by basin area. Although easy to obtain, Dd by this method was found to be of little use since most of the 34 debris flow basins showed neither continuous nor even stippled blue lines, suggesting no perennial or ephemeral stream flow, which is not the case. Particularly in basins with dense forest canopy, this probably reflects the cartographer's inability to identify and assign a channel. Secondly, the onset of blue line status for a channel is a very subjective matter and prone to large variation. For these reasons, and the premise that as much information as possible should be taken off direct images of reality such as air photographs, a different measure of Dd was used. On air photographs with an approximate scale of 1:15,000, drainage pathways incised more than five metres were traced. These pathways were termed zero-order drainages, and the derived morphometric variable was termed 0-order drainage density 'DD-0', because most of the channels showed signs of only occasional water flow draining into first-order channels. Another parameter describing basin roughness, which is closely related to drainage density, is the ruggedness number (RUGNR) originally developed by Melton (1958), and is defined as the product of drainage density and basin relief. It was shown to be useful by Patton and Baker (1976), who related morphometry to flood peak discharge, and by Jackson et al. (1987) who found a good relationship between the slope of debris flow fans and Melton's ruggedness number in their study of debris flow basins in the Canadian Rocky Mountains. However, Jackson et al. used the ruggedness number, R, defined by Melton (1965) as R = HbAb-°-5 with H b basin height and Ab basin area. The original version of Melton's ruggedness number was used in this study. Since the blue-line drainage density proved to be an inadequate measure of basin dissection, DD-0 was used to replace Dj. For situations where no air photographs were available, and to provide another measure of basin roughness, a ruggedness number (RUG-0) was quantified by measuring the length of 107 contours (lc) within a given vertical interval of a basin, divided by the length of the shortest distance between the points of contour intersection with the basin perimeter (dc). A mean of these numbers over several vertical intervals yields a measure of dissection over n equally spaced contours (Figure 5.3): y lc [5.3] RUG-0 ^ n This variable is similar to Smith's (1950) texture ratio, as outlined above, but instead of using a mean of contour crenulations suggested in equation [5.3], he used the contour with the maximum number of crenulations. Figure 5.3 Construction of the RUG-0 variable 108 This procedure is particularly meaningful in small catchments which lack a pronounced thalweg. The greater the ratio between the length of the main channel to the average width of the basin, the more bias will be introduced because the length of the contours is then not a function of the basin dissection, but reflects more the amount of basin elongation. Basins with abnormally long thalwegs were flagged to facilitate outlier diagnostics during statistical analysis. This section has identified and developed morphometric variables which are believed to exert some control on debris flow activity. In the following section, source materials will be investigated which are considered to reduce the unexplained variability of debris flow activity. 5.4 GEOTECHNICAL CLASSIFICATION OF SOURCE MATERIALS Basin geometry alone cannot adequately describe magnitude and frequency characteristics of debris flows, because it does not contain any information on the nature and availability of erodable and transportable materials in the source area, which are likely to be more influencing factors. Even the most favourable combination of morphometric conditions cannot account for debris flow occurrence if, for example, the entire basin consists of massive unweathered rock. Although this scenario is fairly uncommon, it shows that the characteristics of erodible sediments in the basin are at least as important as morphometric factors. To account for additional explained variance in any statistical model, a variety of geotechnical factors needs to be developed to describe and quantify aspects such as the susceptibility to comminution and in situ erosion, as well as the amount of material available. In this context, it is important to consider the spatial limitations of the methods used. Data must be easily and replicably obtainable over small areas. In the following, engineering and geomorphological applications of cohesive and non-cohesive natural materials are reviewed to develop appropriate methods to characterize the source materials of debris flows which can subsequently be applied for basin-scale stability problems. 109 5.4.1 Rock Mass Classifications A n assessment of the effect of lithology on the delivery of sediment and thus geomorphic activity of a basin, requires a description of the rock mass in a manner relevant for geomorphologists. This calls for a quantification of the geotechnical properties of the source material. Typically, geological maps are based on the age and type of rocks rather than their strength characteristics. Geotechnical engineers and engineering geologists have long recognized the inadequacy of geological maps for geotechnical applications and have developed several rock mass classifications during the past three decades. These classifications are based on parameters such as uniaxial compressive strength, persistence (continuity) of joints, their separation ("aperture"), joint roughness, joint filling (gouge), degree of weathering, drill-core quality, rock unit orientation, spacing of discontinuities, the presence and behaviour of ground water, as well as a variety of field tests. These field-based observations and measurements are commonly complemented by laboratory analysis, including uniaxial and triaxial compression tests, point-load tests as well as density, porosity, water content, and swelling tests (Bieniawski, 1989). For geomorphological purposes many of these tests and procedures are impractical since they either require heavy and expensive equipment and assume vehicular access, or are only applicable in site specific rather than basin-scale studies. The following example illustrates the inadequacy of geologic maps with regard to rock stability considerations. Hotsprings Creek, a tributary of Meager Creek in upper Lillooet River basin, is situated within the western coast belt consisting primarily of quartz diorite (Monger and Journeay, 1994). Quartz-diorite rock forms sheer cliffs and very stable rock faces in many parts of the Coast Plutonic Complex; at Hotsprings Creek, however, it is highly jointed and prone to continuous raveling (Figure A-7). One reason for this behaviour might be the recent activity of the Mount Meager volcanic complex, immediately north of Hotsprings Creek. It is likely that stresses caused by magma rising into the nearby complex, coupled with volcano-seismic activity during eruptions, could have altered the original structural properties of the plutonic rock. It is also evident that completely different rock types can behave similarly in a geotechnical context. For example, folded and faulted sedimentary rock of the Cayoosh Assemblage east of the town of 110 Lillooet, showed similar rock-quality properties to deeply weathered, and hydrothermally altered pyroclastic rock as determined by point load tests (cf. Section 5.4.2). To account for differences in rock-quality, Selby (1980) modified engineering rock mass classification schemes for geomorphological purposes. He used only those parameters that could readily be obtained in the field with a minimum of instrumentation. In his classification he suggested measuring intact rock strength via Schmidt hammer testing, weathering indices, spacing of discontinuities, joint orientation, joint width, fracture continuity, and groundwater discharge factors. This scheme proved useful in process studies directly linked to bedrock disintegration. In the present study, however, many debris flows are neither initiated by rock avalanches or rockfall, nor do the source materials necessarily consist of bedrock derived material. Failures also occur in talus slopes, basal till blankets, by rejuvenation of old debris flow deposits, slumps into the recent channels, and erosion of morainal material. Any existing rock-mass classification would therefore have to be modified to include failures from sources other than disintegrating bedrock. The other problem inherent in Selby's approach is its lack of applicability to basin scale studies. A thorough documentation of all parameters along a 50 m transect can take several hours, particularly if conditions change significantly over its length. On a basin-wide scale this transect would fail to provide information on the variability of bedrock behaviour across the basin. Depending on the total area of bedrock contributing sediment, a very large number of transects would have to be investigated. Given the inaccessibility and extremely hazardous working conditions in the initiation zones of debris flow basins, this task proved to be too dangerous and time consuming to justify its use in a basin-scale field investigation. The problem of geologic maps being inadequate for stability considerations has also been encountered in hazard evaluations. During investigations for a project to predict the consequences of hazardous geologic processes in the San Mateo County (Brabb, 1995), 41 bedrock units were combined to produce 21 units with similar physical properties and inferred engineering characteristics. Physical characteristics included composition, hardness, fracture spacing, bedding character, and stratigraphic thickness of bedrock as well as texture of the soil cover. Since Brabb's project assessed many different types of instability, it focused on different characteristics from those examined in the present study. This points to the difficulty of designing geologic or I l l geomorphic slope stability maps which could find widespread application among geotechnical engineers and geomorphologists. There is sufficient reason to conclude that each type of study requires process-dependent, custom-tailored assessment of the most appropriate parameters. 5.4.2 Point Load Tests as a Surrogate for Compressive Strength of Basin Rocks As discussed in the previous section, the degree of fracturing is believed to influence rock stability and thus debris supply to debris flow systems as explained above. Similarly, the geotechnical behaviour of unfractured rocks is known to play a role in debris supply to the basins. In this study, point load strength is used as a surrogate for erodibility of exposed bedrock. It is based on the assumption that there is a correlation between the strength characteristics of intact clasts and the susceptibility to physical and chemical weathering mechanisms. Rocks with little strength are believed to disintegrate faster and thus supply material at a higher rate to the debris flow system than rocks with higher strength. Experiments carried out by Moss (1972) indicate that compressive strength is directly associated with fracturing and breakage. Point-load strength tests are also highly correlated with uniaxial compressive strength as shown by Broch and Franklin (1972). The advantage of point load tests over uniaxial compressive strength tests is that they can be applied to irregularly shaped clasts (Selby, 1982). The underlying assumption in this study is that the rocks collected on the debris flow fans are representative of the lithologic assemblages within the contributing area of the basin. This assumption is necessary since a random sample of rock clasts in the debris flow source areas is either impossible due to inaccessibility, or is too hazardous to justify the attempt. It is recognized that very weak clasts might have disintegrated along their path to the fan to a clast size that was not readily sampled; therefore the results are biased towards the more competent clasts. Typically, 25 clasts of approximately fist size were collected to represent the proportions of source rocks in the contributing part of the basin. According to ISRM (1985) recommendations, particles with c-axis dimension less than 30% larger than the b-axis dimension were excluded from the sample. Fractured clasts were also avoided to avoid breakage along pre-determined failure planes. Over 800 specimens were collected and subjected to standard test procedures as described 112 by ISRM (1985). Clasts were inserted between the cone-shaped metal platens of the point load tester, and load was applied continuously until failure occurred. The point load strength index, Is, (MPa) at failure is obtained by: [5.4] IS = P/A with A (m2) denoting the minimum cross-sectional area measured through the loading points and P (kN) the applied load at failure. Pressure P is measured by a hydraulic gauge linked to the loading device. Brook (1985) showed that the point load required to break a clast is proportional to its volume, which calls for a size correction to compare rock strengths over specimens of varying sizes. In this study, cross-sectional area was used as a surrogate for size. It was determined by tracing the failure planes, which typically represent the minimum cross-sectional area, onto graph paper. Subsequently, the area of each cross-section was measured by digital planimetry. The mean point-load strength index was then used in the statistical analysis. Figure 5.4 shows a frequency plot of mean strength values for all basins. The strength indices are roughly normally distributed, and the results do not allow a clear separation into lithologic groups. A stratification into volcanic, intrusive and other rock groups proved to be inconclusive, because of the large variability of strength indices in each group. Generally, igneous intrusive rocks tested much higher than volcanic rocks, and indices ranged from 0.5 MPa (Pothole Creek, volcanic tuff) to 14.2 MPa (Rainy Creek, igneous intrusive granitic). However, there were several notable exceptions. Intrusive rocks in the Lytton area (Kaboose Creek, Snake Gully, Fool's Gold basins) showed significantly lower strength than rocks of similar composition in the adjacent Coast Mountains; the differences probably reflect the greater age and more advanced weathering of the Mount Lytton Batholith materials. Significant variation was also found in the strength of plutonic rocks of approximately the same age. Granitic rocks from debris flow basins in the Hope region have generally lower strengths than granitic rocks in the Pemberton area. Mount Ludwig Creek showed an exceptionally low value for plutonic rock (3.3 MPa). Several rocks in this sample disintegrated when placed between the metal platens of the point load tester. 113 One reason for this abnormal behaviour might be the location of the gully in what appears to be a fault zone. Motion along the fault may have induced micro-scale fracturing of these rocks. Figure 5.4. Frequency plot of mean point load strength indices as derived from point load tests on approximately 800 intact, fist-size clasts collected on debris fans compressive strength frequency plot 7 f — i 1 1 1 1 1 1 1 1—q 0 1.5 3 4.5 6 7.5 9 11 12 14 15 Point Load Strength Index (I in MPa) S 5.4.3 Terrain Maps and Slope Stability Mapping Existing geological maps as well as rock mass classifications are only of limited value to this study because their resolution was found to be inappropriate for the requirements of this study, since detailed structural characterizations of rock masses were clearly non-feasible at the basin-scale. For this reason, a review was conducted of other procedures for hillslope assessment, such as terrain- and slope stability mapping. Terrain maps are a specific type of surficial geology map showing the distribution of surficial deposits and related landforms. They provide information on recent geomorphological processes, which is not usually part of conventional surficial geological maps (Resources Inventory Committee, 1995). Fulton introduced the present terrain mapping scheme to British 114 Columbia in the early 1970s (Boydell, 1992; Resources Inventory Committee, 1995). In 1988, Howes and Kenk published the revised edition or the Terrain Classification System for British Columbia. From the 1980s to the present, terrain maps and their derivatives such as geological hazard maps and slope stability maps were widely applied at local and regional scales. Table 5.2 shows a comparison between existing mapping schemes. This table will be used to test the usefulness of conventional slope stability maps for the requirements of this study. Slope stability mapping involves the complete mapping of the basins with no areas or polygons left blank, a characteristic which is not required in the present study. Slope classes are only of limited application since very steep slopes can still be relatively stable in terms of major debris sources, and slopes much flatter than the angle of repose of their material usually do not contribute much sediment to the debris flow system over the time scale considered in this study. Surficial materials indicated on terrain maps (e.g. sand, gravel, boulders) were not specified in detail, since debris flows are capable of transporting all grain sizes from clay to large boulders, and there is no information on the relative influence of grain size on debris flow initiation. Finally, presently active processes were considered insofar as they were delivering sediment to the debris flow system. In summary, it appears that conventional slope stability maps include more information than required for the present study. Instead of attempting to modify the existing mapping scheme, a set of new stability criteria were developed which are presented in the next section. Table 5.2. Comparison between interpretive maps and the requirements of this study. Features Slope stability map derived from terrain maps Requirement for this study (see Table 5.2.) spatial coverage complete only contributing areas closed polygons yes yes slope classes yes not considered soil drainage yes notconsidered materials yes not considered landform yes yes active processes yes some slope stability class yes yes 115 5.4.4 Stability Mapping in Debris Flow Basins Considering the limitations inherent in bedrock maps, geotechnical rock-classification schemes, and slope stability maps, as derived from terrain maps, basin-scale stability assessments were tailored to the purpose of this study. A first step was to partition the total basin area into stable and unstable areas. Only those areas considered capable of contributing sediment to debris flow channels over a 100-200 year time frame were considered. Contributing areas were identified using the criteria listed in Table 5.2. Implicit in the stability classification is the existence of stable areas, which are not presently contributing sediment to debris flow channels. A three-fold stability classification was applied to both non-cohesive (NC) and cohesive (C) materials in each debris flow basin: Class 1 very unstable, Class 2 unstable, Class 3 moderately unstable (Table 5.2). This subjective, ordinal assessment was based on air photograph analysis, supplemented and confirmed by field mapping. The division into non-cohesive and cohesive material classes was deemed important, since embankments of non-cohesive materials were observed to be constantly failing by raveling and sliding, whereas the greater strength of cohesive formations rendered them more susceptible to discrete episodes of instability. Air photographs were digitized using the software program "MDSD" (Jordforsk, 1992) which requires only a two-dimensional digitizer. This program eliminates photograph parallax and creates a crude digital elevation model (DEM) from control points which then allow the measurement of distances adjusted for slope. The digitizer used in this study yielded photo measurements of 100 micron accuracy, whereas analytical plotters offer accuracy around 1-20 microns. On a 1:20,000 scale air photograph, the resolution of the MDSD is therefore around 2 metres, which is well within the range acceptable for this study. Considering the degree of resolution that is obtainable from stereoscopic analysis, and given that the polygon boundaries lack precision due to the changing stability conditions in the basin over time, a planimetric error in the order of 50 metres was considered acceptable. It should be noted that the scale of the map used to obtain control points also introduces error since 1:50,000 scale maps were used to obtain control points. Maps of this scale digitized with 0.5 mm accuracy produce a planimetric error of 25 metres. 116 This three-fold stability classification requires air photograph analysis with large-scale (preferably 1:20,000 or larger scale) air photographs. Unstable areas are classified according to the criteria outlined in Table 5.3, and their area measured by digital planimetry using the program MDSD. The outcome is a slope stability map superimposed on the air photograph which shows all areas of potential instability that can contribute sediment to the debris flow system (Figure 5.5). 5.4.4.1 Weighted Stability Number and Active Area Ratio Since Class 1 polygons are more active in producing sediment to channels, the percentage active area was transformed to a Weighted Stability Number (WSN) by summing the percentage areas in each stability class, multiplying them by their respective stability class numbers, then dividing by 100 to yield a class WSN. The basin WSN is then simply the sum of the WSN across all classes. An example is given in Table 5.4; the air photograph of McLeod Creek (Figure 5.5) illustrates the computational method. McLeod Creek basin comprises a total area of 7.2 km2, has four NC polygons in Class 1 and one polygon in Class 2, which together cover 26% of the total area contributing material to the debris flow channels. In the cohesive (C) category, three polygons are in Class 1, three in Class 2, and one in Class 3, together covering 74% of the contributing area. This procedure indicates that the areas actually contributing debris to the channel system are given proper ordinal weighting. In this example, the basin WSN for McLeod Creek is 1.6. Given the total range of WSN from 1.0 to 2.5 across all 34 basins, the computed WSN classifies McLeod Creek as moderately unstable in this particular mountain region. The WSN is then normalized as an Area-Activity Ratio (AAR), defined as [5.5] AAR = (Ac * WSN) / A t where A c is the total contributing area (i.e. the sum of the partial areas in Classes 1, 2, and 3), and A t is total basin area. For McLeod basin, with a basin area of 7.2 km2, a total contributing area of 1.7 km 2 and a WSN of 1.6, the AAR is 0.38. The higher the ratio of contributing area to total basin area and the higher the WSN, the higher the AAR. Thus AAR provides a good estimate of the degree of instability, normalized by basin area. In this study, AAR values range from 0.001 to 117 2.263 with a mean of 0.633. Very low AAR values are typical of completely forested basins, in which debris is supplied primarily from channel sidewalls comprising only a small fraction of the entire basin area Table 5.3. Stability criteria of different sediments and bedrock moderately unstable unstable very unstable stability class 3 2 1 coarse, cohesionless materials (CL): partially vegetated, often some stable parts, signs of continuous (talus slopes, moraines, ablation till low angle slopes with though most areas activity, very fresh channel fills) infrequent signs of actively shedding appearance activity, channel sidewalls debris partly vegetated cohesive materials (C): some signs of instability frequent signs of major scarps and basal till, volcanic tuffs, weak instability (raveling. continuous activity sediments, massive crystalline rock rockfall, slumping) (raveling, rockfall, slumping) Table 5.4. Calculation of weighted stability number (WSN)for McLeod Creek category/stability class total area in class % of total contributing weighted stability number (km2) area (SC*%) CL-1 0.3 20 (0.20) 0.20 CL-2 0.1 6 (0.06) 0.12 CL-3 0.0 0 (0.00) 0.00 C-l 0.5 31(0.31) 0.31 C-2 0.6 35 (0.35) 0.70 C-3 0.2 8 (0.08) 0.24 sum 1.57 118 Figure 5.5 Air photograph of McLeod Creek showing stability polygons. Stability classes are derived from Table 5.3. WSN is calculated in Table 5.4. Scale approx. 1:15,000 119 5.4.4.2. Ground checking of stability mapping Ground checking of stability mapping from air photographs was carried out during August 1995. Ground checking as part of the field work was necessary to verify the original mapping from air photographs, and to identify and map features which had not been previously recognized. Many debris flow source areas inaccessible by foot required the use of helicopters. Unfortunately, very bad weather conditions throughout August 1995 made it impossible to visit the source areas of all 34 research basins. An additional obstacle to successful fieldwork at high-elevation sites was the fact that rockfall hazards, insurmountable rock faces and crevassed glaciers did not allow a complete ground-check of all contributing areas in some basins. These limitations imply that the level of detail of ground checking varied among basins, so that work carried out in great detail in some areas is not be comparable between all sites. For this reason, adjustments to stability classifications were made only when it was obvious, from even a cursory examination of a site, that a polygon had been misclassified. Despite these hindrances, ground checking proved to be very helpful in confirming the consistency of the general classification scheme. Fifteen basins were visited and subjected to more detail ground investigations (marked with '*' in Tables 5.4 and 5.5). Approximately 95% of all stability polygons in those basins seemed to be reasonably well classified, which increased confidence in the method. A typical source for erroneously grouped polygons was very unstable colluvial material (CL-1), which turned out to be bedrock controlled and only covered with a thin discontinuous veneer of colluvial material derived from unstable rockslopes higher in the basin. The principal problem inherent in the stability approach used here is the high degree of subjectivity that is required from the mapper. Table 5.2 can aid in the process of correctly classifying polygons of unstable landforms, but does not replace experience. 5.5 D I S C U S S I O N Figure 5.6 is designed to illustrate schematically the linkages and controls of morphometric and geotechnical parameters on factors which are believed to be the most important one in determining debris flow frequency. Exact causative explanations of variable interaction are not possible in most 120 cases because of a lack of models describing these mechanisms. Most hypothesized factor interactions have been discussed in this chapter and will therefore not be repeated here. Figure 5.6 Influence and control of morphometric and geotechnical parameters on factors determining frequency and magnitude of debris flows BASIN DEBRIS FLOW ACTIVITY The left-hand side of Figure 5.6 indicates that climatic events are a primary control of debris flow initiation. However, these have not been included in the study due to severe data limitations. Not only is there uncertainty concerning climatic events responsible for recent debris flow cycles, but there also exists uncertainty about the climatic history of the study area. Prediction of debris flow frequency is therefore believed to be the least reliable part of this scheme. The remainder of Figure 5.6 indicates the intercorrelations between factors which control debris flow magnitude. The dominant factor which influences debris flow magnitude is believed to be sediment availability. Sediment availability controls the degree to which a debris flow basin is 121 prone to produce a debris flow and provides information on potential debris flow volumes. It is primarily determined by ruggedness (RUG-0 and DD-0) as well as CONTAR. The weighted stability number (WSN), compressive strength (COMPST), and the active area ratio (AAR) are the geotechnical variables developed in this chapter which all determine the characteristics of the debris flow source materials as described above. A conventional framework for viewing sediment availability and delivery involves stratification of debris flow basins into initiation zone, transportation zone, and deposition zone (e.g. Forest Practices Code, Gully Assessment Guidebooks, 1995). Figure 5.6 adopts this convention and indicates the variables which are considered to have some impact on debris flow activity in these zones. An change in ruggedness is believed to change the likelihood of debris flow initiation whereas slope, expressed as relief ratio, is considered the dominant variable controlling debris transport and deposition. The steeper the channel gradient the more material will be delivered to the fan, thereby determining debris flow magnitude as defined in Chapter 4. Debris flow initiation is also believed to be controlled by the contributing area (CONTAR), and the active area ratio (AAR) since it incorporates the area-normalized stability of the source material. Lower channel gradients and slope concavities, which are represented by hypsometry (HYPIN), promote intermittent channel debris deposition, particularly of smaller flows, which then feeds back to debris flow frequency. Relative relief (RELRAT) will determine debris deposition along the channel, and thus ultimately influences the volume of material which will be deposited on the debris flow fan. Tables 5.5 and 5.6 summarize information on frequency and magnitude of debris flows as well as the geotechnical and morphometric variables for weathering-limited and transport-limited basins. Means, standard deviations, and ranges of values were determined for each variable to highlight the distinct differences inherent in the two basin subsets. For example, debris flows occur almost twice as frequently in transport-limited basins as in weathering-limited basins. It has been mentioned before that this is most likely due to the time needed for post-event sediment recharge required in weathering-limited basins, whereas transport-limited basins respond more directly to precipitation intensity thresholds. Debris flow volumes are significantly lower in weathering-limited basins, which again is due to the supply limitations in those basins. This difference is less pronounced for peak discharge, which indicates that flow duration which is 122 dependent on sediment availability as suggested in section 4.4.2, is an important control of debris flow volume. It is also notable that the standard deviation of debris flow volume in transport-limited basins is approximately five times higher than that of weathering-limited basins. A higher standard deviation reflects the higher range of magnitudes for transport-limited basins, and a similar range of magnitudes for weathering-limited basins, which supports the concept of typical debris flow magnitude in those basins. The considerably higher debris flow volumes in transport-limited basins can also be explained by the variety of mass movement processes such as debris avalanches, large rockfalls, and rotational slumps which can trigger debris flows in these basins. There exists a pronounced difference in basin ruggedness between the two basin groups, which is shown by the mean values of the variables RUGNR, and DD-O, but only a slightly higher roughness of the RUG-0 variable. The percent active area in transport-limited basins is on average about 2.5 times larger, and the contributing area 2 times larger than in weathering-limited basins, although average total basins size is smaller in the transport-limited group. This result was expected, considering that the availability of transportable sediment was used as a criterion to distinguish between weathering-limited and transport-limited basins in Chapter 4. As a consequence, WSN is lower , and AAR is higher in transport-limited basins, indicating that the active areas are more susceptible to erosion than in weathering-limited basins. This conclusion is further strengthened by the generally lower compressive strength values found in transport-limited basins. All other variables show only minor variations between the two basin groups, suggesting their limited use as variables to discriminate between weathering-limited and transport-limited basins. This topic is explored further in Chapter 6. 123 Table 5.5. Geotechnical and morphometric variables for weathering-limited basins BASIN FREQ CrnaK VOL DRA RELRAT BASREL RUGNR RUG-0 AREA DOO HYPINT %ACT AAR SLOPE WSN CONTAR COMST ev/tyr m 3 /s m3 m DOO km 2 km ton2 MPa PETERSEN 0.278 200 20,000 5600 0.33 1556 1556 1.90 8.6 1.0 0.67 21 0.524 18 2.49 1.81 8.8 NO LAW 0.195 460 25,400 5000 0.41 1784 2854 1.54 3.2 1.6 0.59 26 0.410 22 1.82 0.72 10.2 MNT.CURRIE * 0.231 240 12,000 2800 0.76 1983 9518 1.34 1.7 4.8 0.53 63 1.003 37 1.64 1.04 11.8 TERMINAL 0.092 480 26,600 2400 0.41 2028 4462 2.10 8.9 2.2 0.57 45 0.943 22 2.10 4.00 4.0 MNT. LUDWIG 0.077 sso 30,000 2300 0.74 1196 7415 1.35 0.6 6.2 0.68 11 0.233 37 2.00 0.07 3.3 EUREKA 0.178 150 10,300 1800 0.40 1940 1940 2.01 10.1 1.0 0.59 1 0.001 22 1.00 0.01 7.2 CANYON* 0.158 220 10,700 1700 0.42 1834 2751 1.96 3.9 1.5 0.55 28 0.458 23 1.16 1.54 4.3 2-MILE 0.089 270 13,700 1200 0.62 1762 1586 1.41 0.8 0.9 0.53 24 0.500 32 2.00 0.20 7.5 COLLIS 0.115 210 10,000 1200 0.27 1388 1804 1.30 10 1.3 0.78 15 0.239 15 1.56 1.53 8.0 ROGER'S 0.083 100 13,400 1100 0.38 1784 5174 1.76 4.9 2.9 0.67 38 0.755 21 2.00 1.85 8.5 NIGHTMARE* 0.054 280 14,300 800 0.61 1891 5106 1.31 1.6 2.7 0.53 51 1.190 31 1.78 1.07 9.6 ENDURANCE 0.082 120 5,400 440 0.53 1906 1906 1.67 9.5 1.0 0.68 25 0.629 28 2.51 2.38 11.2 RAINY 0.090 110 4,900 440 0.39 1876 2439 1.68 6.3 1.3 0.62 17 0.313 21 1.84 1.07 14.2 CLEARWATER 0.068 130 6,000 410 0.71 1769 6192 1.48 2.6 3.5 0.55 92 2.263 35 2.51 2.38 8.0 DEEPA 0.071 110 4,900 350 0.67 1312 4067 1.20 2.2 3.1 0.63 27 0.273 34 1.00 0.60 9.5 19 Mile Ck. 0.081 75 3,200 260 0.41 1632 2774 1.48 4.9 1.7 0.51 3 0.074 22 2.00 0.18 9.3 PATTERSON 0.044 120 5,400 240 0.55 1408 845 1.60 6.1 0.6 0.40 1 0.010 29 1.00 0.06 6.2 LAST DAY 0.076 70 3,000 230 0.44 1601 2562 1.24 2.5 1.6 0.68 10 0.232 24 2.32 0.25 11.5 HOWE* 0.073 30 1,100 80 0.50 1412 6636 1.26 0.8 4.7 0.55 25 0.443 27 1.77 0.20 2.6 WILDCAT * 0.065 30 1,100 70 0.59 1327 3092 2.19 3.1 2.3 0.57 7 0.064 31 1.00 0.20 6.5 SNAKE * 0.046 15 500 20 0.45 930 1860 1.87 3.1 2.0 0.57 10 0.131 24 1.40 0.29 2.2 KABOOSE* 0.050 5 150 8 0.67 732 4978 1.15 0.4 6.8 0.53 62 0.600 34 1.00 0.24 3.0 MEAN 0.107 180 10,100 1300 0.51 1593 3705 1.58 4.6 2.5 0.59 27 0.513 27 1.72 0.99 7.6 STAND. DEV. 0.217 150 8,800 1500 0.14 346 2238 0.31 3.2 1.7 0.08 23 0.508 6 0.52 1.02 3.3 RANGE 0.044- 5- 150- 8- 0.27- 732- 345- 1.15- 0.4- 0.6- 0.4- 1- 0.001- 15- 1.0- 0.01- 27-0.278 550 30,000 5600 0.76 2028 9518 2.19 10.1 6.8 0.78 92 2.263 37 2.51 4.00 14.2 Table 5.6. Geotechnical and morphometric variables for transport-limited basins BASIN FREQ Qmax VOL DFIA RELRAT BASREL RUGNR RUG-0 AREA DDO HYPINT %ACT AAR SLOPE WSN CONTAR COMST ev/yr m 3 /s m 3 m DDO ton2 km km 2 MPa TURBID 0.273 500 160,000 43,700 0.35 2028 5678 2.42 8.5 2.8 0.46 58 0.642 19 1.10 4.96 5.7 BOUNDARY * 0.571 100 33,100 18,900 0.51 1296 4925 2.03 2.3 3.8 0.43 52 0.688 27 1.33 1.19 3.5 CAPRICORN 0.078 1600 100,000 7,800 0.38 2068 4550 2.20 14.4 2.2 0.56 67 0.980 21 1.47 9.60 6.2 FERGUSSON * 0.171 120 39,800 6,800 0.67 1071 3320 1.16 1.6 3.1 0.48 48 0.575 34 1.21 0.76 5.9 NOGOOD* 0.600 30 10,000 6,000 0.53 1388 8328 1.68 2.1 6.0 0.40 71 0.498 28 1.34 0.78 4.5 HOTSPRINGS 0.107 400 21,600 2,311 0.37 1717 4808 2.11 6.2 2.8 0.63 54 0.883 20 1.63 3.36 7.4 GUNBARREL 2 * 0.110 60 20,000 2,200 0.71 1205 6507 1.11 1.1 5.4 0.59 87 1.075 35 1.24 0.91 6.8 GUNBARREL 3 * 0.103 50 16,700 1,700 0.75 1175 8695 1.06 0.4 7.4 0.58 95 0.946 37 1.00 0.35 6.8 FOOLS'GOLD* 0.086 140 6,500 559 0.41 1357 3393 1.26 1.5 2.5 0.53 62 0.937 22 1.48 0.95 4.5 POTHOLE 0.028 110 36,500 1,000 0.61 1186 6523 1.29 0.5 5.5 0.48 65 0.660 31 1.00 0.33 0.5 MCLEOD * 0.048 340 17,900 900 0.52 1891 3971 1.14 7.2 2.1 0.51 23 0.369 27 1.60 1.66 9.5 GUNBARREL 1 * 0.085 10 3,400 300 0.75 1205 4820 1.08 0.3 4.0 0.59 95 1.983 37 2.12 0.29 6.8 MEAN 0.188 290 38,800 7,700 0.54 1466 5460 1.65 3.8 4.0 0.52 65 0.853 28 1.38 2.10 5.2 STAND. DEV. 0.196 440 45,900 12,500 0.15 360 1758 0.51 4.4 1.7 0.07 21 0.436 7 0.32 2.74 2.0 RANGE 0.028- 10- 3400- 300- 0.38- 1071- 3328- 1.06- 0.3- 2.1- 0.4- 23- 0.369- 19- 1.0- 0.29- 0.5-0.600 1600 160,000 43,700 0.75 2068 8695 2.42 14.4 7.4 0.63 95 1.983 37 2.12 9.60 7.4 note: basins are listed in order from highest to lowest debris flow activity (DFIA) basins marked with * were examined in detail in the field a summary table of variable definitions is provided on the following page (Table 5.7) Table 5.7. Definitions of morphometric and geotechnical variables 124 Variable Definition FREQ Frequency of debris flow events (1/return interval) [events/year] Qmax Maximum instantaneous debris flow peak discharge [m3/year] VOL Total debris flow volume which has traveled beyond the fan apex [m3] DFIA Debris flow index of activity, product of FREQ and VOL RELRAT Relief ratio, ratio of relief and horizontal distance between highest point of basin and fan apex BASREL Basin relief, vertical difference between highest and lowest point in a basin [m]. RUGNR Ruggedness number (after Melton, 1958), product of drainage density and drainage relief RUG-0 Ruggedness number, quantifying basin dissection (Figure 5.4 and Eqn 5.3). AREA Total area of a debris flow basin [km2] DD-0 Zero-order drainage density, total length of stream channels per unit area [km/km2] HYPINT Hypsometric integral, distribution of area with elevation (Eqn 5.1 and 5.2) %ACT Percent active area, proportion of basin area actively shedding debris to the channels AAR Active Area Ratio, area-normalized estimate of the degree of instability within a basin (Eqn. 5.5) SLOPE Average basin slope from highest point of basin to fan apex [degrees] WSN Weighted Stability Number, weighted average of source material stability characteristics (Table 5.4) CONTAR Contributing area, total area which is actively shedding debris to the channels [km2] COMPST Compressive strength, point load strength index as a surrogate for compressive strength [MPa] 5.6 Summary Several morphometric variables were examined for their effectiveness as debris flow predictors, and include measures of the basin relief, curvature, ruggedness, area, and slope. Bedrock maps, geotechnical rock classifications, and interpretive maps derived from terrain base maps were reviewed to find a suitable procedure for quantifying basin stability and debris flow potential. Conventional bedrock maps based primarily on rock type and age do not provide much information relevant to this investigation. Geotechnical rock classifications provide detailed procedures suitable for engineering design at the site level, but are considered unworkable at the basin scale. Slope stability maps derived from terrain base maps are useful in regional studies, but include variables that are of limited importance in this study. For these reasons, a stability classification was developed involving only those areas actually contributing sediment to the debris 125 flow system. Their stability was then assessed and stability numbers were assigned to each slope stability class. With this system, it was possible to quantify the stability of a basin with regard to sediment availability and delivery. Two variables derived from this information were the Weighted Stability Number, which provides information on the degree of stability in the most active parts of a basin, and the Area Activity Ratio, which assists in characterizing the overall stability of the basin. The influence of structural controls on debris flow occurrence are still poorly understood and requires a great deal of additional work to include specific rock mass characteristics in a statistical model. This section indicated the main limitations for this type of analysis and suggested the measurement of additional geotechnical properties such as joint spacing and joint orientation. Bridging of rock-mass engineering and geomorphological applications was suggested as an avenue for future research. The computation of mean values, standard deviations, and range of values for the dependent and predictor variables in weathering-limited and transport-limited basins has emphasized the differences in the two sub-groups of basins, and has highlighted those variables which would be most useful in a numerical separation of these basin types. This is discussed under the heading of discriminant analysis in Chapter 6. 126 CHAPTER 6. STATISTICAL METHODS FOR PREDICTING DEBRIS FLOW ACTIVITY 6.1 I N T R O D U C T I O N The major objective of this thesis is to develop a means of estimating the average probable activity of debris flow basins for which no prior information exists on the magnitude and frequency of events. In Chapter 2 it was concluded that no single parameter would likely be adequate to characterize the frequency or magnitude variables. Accordingly, to improve predictive power, a multivariate model is proposed in which several morphometric and geotechnical parameters are regressed against the dependent variables peak discharge (Qmax). which comprises an alternative measure of debris flow magnitude, average debris flow volume (VOL), average frequency of debris flows (FREQ) and debris flow index of activity (DFIA), defined as the product of mean frequency and mean magnitude. The chapter is structured as follows: In section 6.2, discriminant function analysis is used to develop an objective means of classifying basins as either weathering-limited or transport-limited, a division that is considered indispensable to improving the predictive power of the regression model. The rationale for this classification is that each basin type should exhibit a different frequency-magnitude behaviour. An a priori classification into weathering-limited and transport-limited basins is performed by applying the same criteria that were used to stratify the basins for volume calculations using peak discharge (Chapter 4). The stratified regression model is then tested using a sample of 12 control basins which were not included in the original regression analysis. A multivariate regression model is developed in Section 6.3 which incorporates basin attributes to predict the above four response variables. The model results are presented as a set of regression equations that can be applied to basins where the four response variables are unknown. As will be demonstrated, the predictive power of an unstratified regression model is unsatisfactory. In particular frequency does not seem to be predictable, which is perhaps not surprising since it is strongly controlled by hydroclimatic factors which, as already noted, cannot be readily quantified for each basin (cf. Figure 5.6 and related discussion). 127 6.2 A PRIORI IDENTIFICATION OF BASIN TYPE In Section 4.2.2, transport-limited and weathering-limited basins were identified according to qualitative criteria involving the estimated availability of sediment in the debris flow source areas. In this section, a set of variables is extracted via discriminant analysis to obtain more objective measures of debris availability, and thereby to provide numerical a working method for recognizing transport-limited and weathering-limited basins in this area. 6.2.1 Discriminant-Function Analysis Discriminant analysis seeks to classify individuals or objects into mutually exclusive groups on the basis of a set of predictor variables measured on both groups. The analysis proceeds by finding a linear function that maximizes the ratio of between-group to within-group variability within multi-dimensional space (Dillon and Goldstein, 1984). The discriminant function is thus a set of coefficients for computing a discriminant score for each member of each group of objects. This score becomes a classification score that determines the group to which each object most likely belongs. This is reducible to a probability of membership, based on the proximity of the case score to the group mean score. The morphometric and geotechnical variables used as independent variables in the discriminant analysis are identical to those employed later in the multiple regression analysis. The first step in the discriminant analysis is to use the a priori classified cases to run a stepwise analysis on the 34 original basins to obtain a function that minimizes the misclassification of basins. Stepwise discriminant analysis selects variables for inclusion based on the magnitude of their additional, or partial, contributions to the discrimination process. The choice of the variable is based on the highest F-value, computed from the ratio of between-group to within-group variance of discriminant scores at each step in the analysis. Variables are then discarded from the model if the computed F-value is less than the user-specified F-to-remove value. (The F-to-remove value is defined as the F value of the partial Wilks' lambda associated with the unique contribution of the respective variable to the discriminatory power of the model). This procedure resulted in the 128 retention of three variables, listed in order in Table 6.1. The F-value was chosen so that statistical significance at a reasonable a-value (0.10) is guaranteed, whilst at the same time retaining as few variables as possible to avoid unnecessary model complexity. This procedure was implemented by the statistical software package STATISTICA (Statsoft, 1994). Table 6.1 Summary of stepwise discriminant analysis for classification of transport-limited and weathering-limited basins Variables F-value degrees of freedom p-level Lambda %ACT 19.97 32 0.00009 0.616 AAR 10.62 31 0.00271 0.459 DD-0 3.83 30 0.05961 0.407 The variables %ACT, AAR, and DD-0 that were extracted by the model have physical as well as statistical significance. The larger the active area of a basin, the more sediment will be available for transport with regard to the total basin area. AAR contains the degree of stability normalized for basin area, and provides a meaningful measure of the erodibility of available materials. The DD-0 is an index for basin roughness or basin dissection. An increase in the degree of dissection means more area exposed to denudational processes, thus increasing the amount of erodible sediment and creating more potential failure sites. This in turn increases the probability for frequent debris flow initiation. Equations 6.1 and 6.2 present the two discriminant functions for weathering and transport-limited systems: [6.1] CS w l = -1.838 - 0.051* %ACT + 3.566*AAR + 0.986*DD-0 [6.2] CS ti = -6.508 + 0.243* %ACT - 6.322*AAR + 0.046*DD-0 Using these equations, only three of 34 basins are misclassified. The squared Mahalanobis distance (Table 6.2) indicates the distances from the group centroids (means) of each basin, and therefore provides a numerical measure of misclassification. 129 Table 6.2 Squared Mahalanobis distances form group centroids Basin observed classification weathering-limited (p=0.676) transport-Umited (p=0.324) 19-Mile weathering-limited 1.337 12.097 2-Mile weamering-limited 1.189 6.534 Eureka weamering-limited 1.619 10.793 Wildcat weathering-limited 1.140 10.536 Herrling weamering-limited 15.652 33.314 Clearwater weamering-limited 13.458 18.612 Turbid transport-Umited 9.484 1.236 Terminal weamering-limited 0.988 5.202 Endurance weathering-limited 0.935 8.432 Gunbarrel 1 * transport-limited 9.581 8.375 Gunbarrel 2 transport-limited 13.233 1.401 Gunbarrel 3 transport-Umited 20.711 5.388 Snake Gully weathering-limited 0.722 9.060 Fool's Gold transport-limited 6.328 1.001 Kaboose * weathering-limited 8.177 4.266 Peterson weathering-limited 0.800 8.570 Deepa weathering-limited 0.943 4.171 Last Day weamering-limited 0.703 10.287 Rainy weathering-limited 0.661 7.172 No Law weathering-limited 0.932 4.638 McLeod* transport-limited 0.227 5.824 Mnt. Currie weathering-limited 2.848 2.562 Nightmare weathering-limited 2.057 8.572 Ferguson transport-Umited 3.7% 0.661 Howe weathering-limited 3.821 14.593 Pothole transport-Umited 7.980 1.051 Capricorn transport-Umited 9.934 1.957 Canyon weamering-limited 1.131 4.423 Hotsprings transport-Umited 2.511 1.380 No Good transport-Umited 16.359 3.642 Boundary transport-Umited 2.731 0.798 Big Bear weamering-limited 1.952 10.554 Collis weamering-limited 0.844 7.066 Rogers weathering-limited 0.246 6.170 Misclassified basins are marked with * , and p denotes the relative proportion of basin type in the data set. 130 According to this table the three misclassified basins are Gunbarrel I, Kaboose Creek and McLeod Creek basins. An examination of Table 6.2 indicates that Gunbarrel I is misclassified only by a very small margin. According to the discriminant function, Gunbarrel I and McLeod Creek should both be classified as weathering-limited basins. McLeod Creek has a low %ACT (23%), which is most likely the reason for its misclassification. However, this 23% active area consists of very steep, highly unstable talus slopes that contain large amounts of unconsolidated material. The talus deposits are rapidly re-supplied with fresh material from toppling rock slopes above. According to these observations, the a priori classification of McLeod Creek was not revised. Kaboose Creek was probably misclassified as a transport-limited basin because of its high DD-0 (6.8) and high %ACT (62%). It was originally classified as weathering-limited because most of the available sediment is derived from rock face denudation, which necessitates system recharge after a debris flow has scoured the channel. Based on this observation, Kaboose Creek was not re-classified despite its unusually high percentage of contributing area. This discussion demonstrates that the identification of weathering-limitation versus transport-limitation is sensitive to the decision of what constitutes an "active area", as discussed in Chapter 5. The Mahalanobis distance table also identifies those basins which are intermittant between tranport-limited or weathering-limited. Marginal classifications were assigned to Gunbarrel I, Clearwater Creek, Mount Currie Creek, and Hotsprings Creek. Table 5.4 shows that both Clearwater Creek and Mount Currie Creek have a high %ACT and high DD-0 that explains their marginal position in the discriminatory space. Both basins are characterized by steep, unvegetated bedrock faces which points towards the difficulty in deciding what percentage of the bedrock controlled area is actively shedding material to the debris flow channel. Gunbarrel I basin is characterized by a high %ACT, and an average DD-0, but a high AAR which is atypical for transport-limited basins. Hotsprings Creek basin on the other hand displays average AAR, but a low DD-0 and below average %ACT for transport-limited basins which accounts for its marginal classification as transport-limited. It is unlikely that all 34 basins would be classified correctly according to the post-hoc prediction model, which does not reflect erroneous a priori classification, but rather suggests that not all discrimination factors were included in the analysis. It is also to be expected that not all basins can be unambiguously assigned to either group, because a sample of 34 basins is likely to 131 contain the entire spectrum of sediment availability. However, an encouraging result is that 91% of all cases were correctly classified, which indicates good performance of a model based on just three discriminatory variables. %ACT, AAR and DD-0 are all independent of any particular geologic setting or climatic regime. It is expected that the discriminant function can be transferred to regions outside southwestern British Columbia, without losing discriminatory power. To validate the model, 12 control basins were pre-classified as weathering-limited or transport-limited types. Equations 6.1 and 6.2 were then used to calculate two classification scores for each of the test basins: CSwi and CSti which can be used to assess the probability of membership in the weathering-limited and transport-limited groups. A case is classified as belonging to the group for which is has the higher classification score. Table 6.3 shows that all control basins were classified correctly. An interesting case, however, is Cheekye River basin which, based on the original qualitative criteria, would be classified as a typical transport-limited basin since it contains huge quantities of friable Quaternary volcanic debris. However, its classification score, indicates a transport-limited case, but only by a very small margin (0.049). Several model test runs were carried out to investigate the influence of %ACT on this classification score. It was found that a change of only +5% would have produced a re-classification of this particular basin more definitively in one or other group, again emphasizing the important control of the percent active area variable. Transitional basins which are roughly equidistant from each group centroid, require special consideration for management purposes, since small disturbances would change their status from weathering-limited to transport-limited. This would effectively change the frequency and magnitude characteristics of debris flows. The other basins were considerably less sensitive to minor changes in the percent active area, suggesting that discrimination with the chosen set of variables can be used with some degree of confidence. A final set of equations was then developed by combining the 12 test basins and the 34 original basins to investigate the robustness of the original model as well as providing a regionally valid means for classifying debris flow basins according to sediment availability on a regional basis. A final check showed that the model was able to correctly classify 96% of all basins, confirming the robustness of the classification procedure. 132 Table 6.3 Classification scores for test basins basin w-limited score (Sw|) t-limited score (Sti) classification Agassiz Mountain Creek 1.748 -2.684 weathering-limited Charles Creek 0.963 1.888 transport-hmited Alberta Creek -0.998 -3.820 weathering-limited Magnesia Creek 0.401 -3.213 weamering-limited M-Creek 0.570 -2.820 weamering-limited Turpin Creek 3.788 -2.164 weathering-limited Newman Creek 2.136 -1.552 weathering-limited Hope Creek 3.882 -0.410 weamering-limited AlexanderCreek 0.757 -6.747 weathering-limited ThackerCreek 0.596 -0.384 weathering-limited Ted Creek 0.405 -3.743 weathering-limited Cheekye River 0.733 0.782 tramport-limited [6.3] C S W ) = -2.027 - 0.067*%ACT + 4.331*AAR + 1.139*DD-0 [6.4] CSu = -7.095 + 0.297* % A C T - 7.347* A A R - 0.169*DD-0 The addition of other variables to the discriminant function does not significantly improve the power of the model. The final set of functions [Equation 6.3 and 6.4] based on the combined data set are useful for differentiating between weathering-limited and transport-limited basins; they also aid in selecting the appropriate function for determining total debris flow volume from peak discharge measurements (cf. Chapter 4.4.2). It should be noted the regression coefficients have not changed significantly from equations 6.1 and 6.2 to 6.3 and 6.4. The largest change has occurred for DD-0 for which also the sign has changed. Changes in sign do not necessarily indicate positive or negative relationships between the predictors and the dependent variables (cf. section 6.3.4.2). This notation should be kept in mind for geomorphic interpretations of these equations. The disproportionally large change in the coefficient for DD-0, indicates a larger variability of this variable compared to A A R and % A C T . 6.2.2 Summary The objective identification of sediment availability, as expressed by the concept of weathering- and transport-limitation, is important for a better understanding of the frequency and 133 magnitude characteristics of debris flow basins. Discriminant function analysis also provides important indications as to the most relevant variables for inclusion in the regression analysis. Having suggested the importance of sediment availability for improving predictive capabilities of a statistical model, the next section addresses the multiple regression model for predicting the four debris flow response variables Qmax. VOL, FREQ, and DFIA (debris flow index of activity). 6.3 LINEAR-REGRESSION MODEL Multiple regression is the primary statistical tool used in empirical prediction of dependent variables from a linear combination of predictor variables. Although intuitive causal mechanisms are the foundation of the choice for predictor variables, regression techniques themselves do not provide any information about causal mechanisms, and can only be used to confirm intuitive best-judgments as to the major variables of physical significance. Regression equations were developed for each of the four dependent variables (Qmax. VOL, FREQ, and DFIA). Figure 6.1 summarizes the steps that were followed in building the multiple-regression model. The linear regression model can be expressed in terms of X variables with normally and identically distributed independent error terms ej as: [6.5] Yi = bo + biXii + b2Xi2 + ... + bpX i ( P + q where: bo, bi,...bp are computed regression coefficients measured for p predictor variables Xu, .... X i > p are predictor variables, measure on n cases (i = 1,..., n) e, are error terms (residuals from regression) This model with p predictor variables does not explicitly include interaction effects between the variables. However, in complex systems, some interactions are to be expected and can rarely be completely avoided. To account for interactions, one can either deliberately include interaction terms in the regression model, or try to avoid unwanted interactions. Some predictor variables in this study are themselves functions of other variables, and therefore techniques are used to avoid data redundancies reflected in between-variable interactions. 134 As is evident form the term "linear", it is assumed that relations between dependent and predictor variables are in constant proportion to one another. In practice this assumption is difficult to confirm, but there are several procedures to test whether linearity holds true. If obvious deviations from the model are detected, variables can be linearized by fitting non-linear components to the regression model. Bivariate scatterplots were produced to detect deviations from linearity, but indicated that linearity assumption was not grossly violated. Another prerequisite for regression analysis is the assumption of normality of the dependent variable. In cases where this assumption is violated, variables can be transformed to approximate normal sample distributions. Frequency plots of the predictor and dependent variables indicated non-normality for the FREQ, Qmax. VOL, and DFIA. Normality was further tested by a Chi-square goodness-of-fit test at a = 0.05. The results of this test also suggested non-normality for these same variables. Non-normal variables were accordingly log-transformed to remove skewness and achieve the desired normality (Figure 6.2). BASREL, AREA, RUG-0 and CONTAR were also found to be non-normal. Predictor variables in regression models are considered fixed and therefore are not required to be normal. However, log-transformation of these variables allowed an improved least-square fit, which warranted this transformation. The next step indicated in Figure 6.1 is to diagnose relations and interactions among the predictor variables to ensure stable correlation coefficients and thus replicability of the analysis. The techniques used are outlined in the following section, and include an examination of the correlation matrix, forward-stepwise regression, and the computation of variance inflation factors. Variables were then reduced to find the "best" set of predictor variables so as to minimize model complexity. The "tentative model" consists of a summary of preliminary regression equations that are based on the original dataset. Residual and outlier diagnostics are then performed to detect outlying cases which strongly bias the regression equations. To validate the statistical model, 12 basins with known frequency and magnitude characteristics were selected, and the predictor variables contained in the original equations were measured for these test basins. The original model was then extended by including the test basins, and the residual and outlier analysis was repeated for the combined dataset. A final collinearity check was performed to guarantee a minimum of variable interaction in the final regression equations. 135 Figure 6.1 Development of the final regression model D A T A C O L L E C T I O N J TESTS FOR NORMALITY, AND LINEARITY OF ^ Y-VARIABLE y ASSUMPTIONS VIOLATED DIAGNOSTICS FOR RELATIONSHIPS AND INTERACTIONS . OF X-VARIABLES ^ ASSUMPTIONS VIOLATED 1 VARIABLE REDUCTION I C TENTATIVE MODEL J*-T RESIDUALS ANALYSIS AND .OUTLIER DIAGNOSTICS. (^M I MODEL VALIDATION W RESIDUAL ANALYSIS OUTLIER DIAGNOSTICS .MULTICOLLINEARITY CHECK REMEDIES FINAL " X ( R E G R E S S I O N ) \ A / I O D E I _ ^ / Figure 6.2. Log-transformation of the dependent variables FREQ, Qmax, VOL and DFIA FREQUENCY DISTRIBUTION LOG-TRANSFORM (FREQ) 12 10 8 6 4 2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 frequency (events/yr) -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 frequency (log) PEAK DISCHARGE DISTRIBUTION 1200 Q _ (mVs) 1600 2000 LOG-TRANSFORM (QMAX) 12 10 8 c § 6 o 4 2 0, 33 2.5 (log) 3.5 VOLUME DISTRIBUTION LOG-TRANSFORM (VOL) 30 60 90 volume * 103 (m3) 120 DFIA DISTRIBUTION 20 p — i 1 1 1 1 1 1 r -15 LOG-TRANSFORM (DFIA) 10 20 30 DFIA *10 3 _j i i i & i 50 1 2 3 DFIA (log) 137 6.3.1 Multicollinearity and Variable Selection of the Regression Model In Chapter 5 it was shown that some basin variables are significantly intercorrelated, which introduces data redundancies (Table 6.4). Past research (e.g. Melton, 1958; Mark, 1975) has shown that variables associated with basin morphometry are both causally and mathematically interrelated. Removing all interrelated variables is therefore neither possible nor desirable. Excessive variable interrelation, however, can cause variations in the estimated regression coefficients among samples to the extent that prediction becomes non-replicable, and exhibits severe intra -and inter-regional instabilities. Perfectly correlated variables yield a number of possible response functions with different fitted values for the independent variables, rendering predictions useless. On the other side of the intercorrelation spectrum are uncorrelated predictor variables, for which regression coefficients remain the same, regardless of which other predictors are included in the model. This behaviour is desirable since it allows the analysis of predictions with only one variable changed, while all others are fixed. A formal collinearity check consists of two steps: (1) detect the presence of collinear relations among the variables, and (2) assess the extent to which these relations have degraded regression parameter estimates to decide whether omission of a variable or other remedies are necessary (Belsley et al., 1980). One widely used method to diagnose multicollinearity is to simply correlate pairs of predictor variables. Another approach is to delete one or more variables from the linear model to reduce collinearity, with the effect of reducing the standard errors of the estimated regression coefficients for the predictor variables remaining in the model. There are two main problems associated with this procedure. One is that no direct information is obtained about the discarded variables. The second has already been stated, and is that the values of the regression coefficient for independent variables in the model are affected by the discarded variables (Neter et al, 1989). Forward-stepwise regression, discussed in the next section, was chosen to overcome some of the problems associated with multicollinearity. 138 Table 6.4. Pearson correlation matrix of dependent and predictor variables IRELRATSBASREL iRUGNR : RUG-0 • AREA j D&O i HYFIN S %ACT } AAR WSN iCONTAR iCOMST FREQ Q M A X V O L DHA R E L R A T 1 00 -0.43 0.46 -0.59 -0.68 0.64 -0.23 0.46 0.39 -0.13 -0.29 -0.10 -0.18 -0.34 -0.15 -0.19 B A S R E L 100 -0.10 0.44 0.60 -0.55 0.09 -0.15 0.09 0.39 0.39 0.54 0.24 0.72 0.50 0.48 RUGNR 100 -0.20 -0.41 0.83 -0.17 0.70 0.50 -0.11 0.25 -0.24 0.16 -0.05 0.21 0.23 RUG-0 1 00 0.64 -0.41 -0.09 -0.22 -0.18 -0.05 0.29 -0.11 0.33 0.35 0.27 0.34 A R E A 1.00 -0.64 0.27 -0.53 -0.36 0.21 0.29 0.30 0.16 0.63 0.34 0.32 D r > 0 1 00 -0.20 0.64 0.32 -0.29 -0.09 -0.44 0.03 -0.38 -0.05 -0.03 HYPIN 100 -0.26 -0.05 0.45 0.08 0.39 -0.19 0.11 -0.10 -0.15 % A C T 1.00 0.84 -0.08 0.41 -0.20 0.16 -0.13 0.24 0.25 A A R 1.00 0.33 0.44 0.01 0.04 -0.02 0.17 0.15 WSN 100 0.26 0.43 0.00 0.18 0.00 0.00 C O N T A R 1.00 0.17 0.23 0.42 0.44 0.43 C O M S T 1 00 0.10 0.24 0.05 0.08 FREQ 1.00 0.18 0.42 0.70 Q M A X 1 00 0.80 0.69 V O L 1.00 0.94 DFIA ; . 1.00 note: correlations significant at a=0.05arebold-faced 6.3.1.1 Forward-stepwise regression Forward-stepwise regression was developed to economize computational efforts for studies with large numbers of predictor variables. Its principal difference to the all-possible-regressions approach is the fact that the "best" subset of predictor variables is found (Neter et al., 1989). Stepwise-selection procedures allow the decision to include a predictor to be reversed, which is not possible with sequential methods. Forward-stepwise regression enters predictor variables until the "optimal" regression model is found. This is achieved by computing the partial correlation between each predictor variable and the dependent variable. The predictor variable with the highest simple correlation is entered first, followed successively by variables with the largest partial correlation with the variables already included in the model. At each step, an F-statistic is computed from the ratio of the mean square due to regression to the error (residual) mean square. This is continued until an F-value associated with the last entered variable is greater than the 139 specified critical F-to-enter value. The variable entered first is now checked for possible removal by computing the F-value as if it were the last variable entered. If this variable is highly correlated with a subsequently entered predictor, and if its F value does not exceed a pre-specified F-to-remove value, it will be removed from the model. In other words, the F-to-remove value determines how "insignificant" the contribution of a variable in the regression equation has to be in order for it to be removed from the regression equation. This procedure continues until no included variable can be added according to the specified critical F-to-enter value (Dillon and Goldstein, 1984). The distinct advantage of the stepwise procedure is that by allowing the removal of highly correlated variables, it avoids typical multicollinearity problems. Despite this advantage, it remains difficult to assess the effects of individual predictors on the dependent variable when strong relations exist between the predictor variables. As a remedy for serious multicollinearity, the STATISTICA (Statsoft, 1994) algorithm includes a condition index, which pre-screens the data for strong intercorrelations and issues an appropriate warning (e.g. Dillon and Goldstein, 1984; Belsley, et al., 1980). From the above, it is obvious that the selection of the "best" set of predictor variables is dependent on the F-to-enter and F-to-remove values. The approach taken in this study was to specify the F-value in a form that the rule "number of predictor variables, NVAR, < number of cases, N, times six" was not violated. This rule was designed to avoid the problem of "overfitting", which occurs when more predictor variables are included than are necessary to obtain a robust model. In addition, such an overfilled regression would be unlikely to be replicable in slightly different field areas. The general rule that NVAR < N/6 was suggested by Neter et al. (1989). Other authors have suggested a variable-to-observation ratio of between 5 and 10 (Draper and Smith, 1981). In this study, 34 observations are available in the original data set, which implies that, according to the rule stated above, no more than 5 variables should be used in the final regression model. In most instances this rule could be met with an F-to-enter of 1.0 and F-to-remove of 0. The analysis was repeated if the addition of a new predictor variable did not significantly improve the coefficient of determination, R2. This procedure guaranteed that a minimum number of variables explained a maximum amount of variance of the dependent variable with the least model complexity. 140 Table 6.5 summarizes the regression equations of the initial model. Comparing the Revalues shows that the stratification into weathering-limited and transport-limited subgroups has a higher predictive power compared with the unstratified dataset. Caution must be exercised, however, in comparing two regression models with the same dependent variables but differing numbers of predictor variables. The higher R2-value does not necessarily reflect an improved least-squares fit, and thus represent a "better model" (Dillon and Goldstein, 1984). To guarantee comparability of the two models, it is better to use an adjusted multiple coefficient of determination (R 2), since this takes into account the number of predictors present in the model. The relationship between R2 and R 2 is expressed as: — i , n-l [6.6] R2 = l-(l-R2) n-p with n denoting the number of cases, and p the number of predictor variables retained in the model. Equation [6.6] shows that for p > 1, R2 < R2, implying that as the number of predictors increases, R2 becomes increasingly less than R2. The stratification into weathering-limited and transport-limited basins proves to be particularly effective for the prediction of FREQ, VOL and DFIA, and Qmax in transport-limited basins. The capability of the model to predict frequency differs significantly between the weathering-limited ( R2= 0.31) and transport-limited basins ( R2= 0.75). This striking difference is most likely due to the different abilities of the two basin types to respond to hydroclimatic events. Debris flows in transport-limited basins should respond directly to the exceedance of a precipitation threshold, and therefore are more predictable than debris flows in weathering-limited basins where recharge rates, which could not be integrated in the model, are an important frequency-controlling factor. The comparison of the ^-statistics should be interpreted with care, because the sums of squares of the dependent variable (SSY) are different for the two data sets. For this reason, the standard error of the estimate (SE), which is a measure of dispersion of the observed values about the regression line, was examined as an additional means to compare the regression equations. Table 6.5 shows that the standard error decreases between the original and transport-limited 141 regression lines, which shows that minor differences in SSY do not significantly influence the adjusted coefficient of multiple correlation. Table 6.5. Summary of preliminary regression model D A T A S E T N EQUATION R2 R2 SE ORIGINAL 34 logFREO = -1.83 + 0.33*RUG-0+0.78*%ACT+0.03*COMST-0.31*AAR .27 .20 .27 34 loeOm?n = -9.01 + 3.78*logBASREL+0.58*IogAREA-0.06*OOMST-0.47*RUG-0 .66 .63 .33 34 logVOL = -8.76 + 3.92*logBASREL+1.87*%ACT-0.76*AAR .45 .41 .48 34 logDFIA = -13.28 + 5.01*logBASREL+2.65*%ACT-1.13*AAR .47 .44 .60 ORIGINAL 22 logFREO = -4.91 + 1.10*logBASREL+ 0.61HYPIN .38 .31 .20 WEATH-LIM 22 logOma* = -6.64 + 2.88* lo gBASREL+0.50* logAREA-0.47* RUG-0 .62 .60 .33 22 logVOL = -6.52 + 3.34*loeBASREL+0.60*logAREA-0.45*RUG-0 .63 .59 .38 22 logDFIA = -14.13 + 4.87*logBASREL+2.22*HYPIN .61 .59 .49 ORIGINAL 12 logFREQ = 1.39 - 6.20*HYPIN-0.18*OOMST .80 .78 .18 TRANS-LIM 12 logQmax = 2.44+ 1.18*logOONTAR-0.08*OOMST .83 .81 .27 12 logVOL = 5.54 + 0.67*logOONTAR-0.88*WSN .80 .78 .23 12 logDFIA = 3.28 + 0.91*RUG-0-0.88*WSN .75 .73 .35 note: underlined variables are significant at a = 0.05 — 2 R denotes the adjusted coefficient of multiple determination, calculated via equation [6.6] Table 6.6 lists the coefficients of determination (R2) for univariate and multivariate relations between the dependent and predictor variables. As proposed in Chapter 1, the combination of several morphometric and geotechnical variables yields far better predictive results than any univariate model, although the multivariate models are by no means cumbersome and, furthermore, involve physically realistic variables. Table 6.6. Comparison between univariate and multivariate (MV) models using R2-statistics (original dataset) RELRAT BASREL RUGNR RUGO AREA DD-0 HYP1N MACT AAR SLOPE WSN CONTAR COMST HV FREQ 0.03 0 .0 6 0.00 0.11 0.00 0.00 0.03 0.00 0.00 0.04 0.00 0.05 0.00 0.2 7 Qmax 0.12 0 .52 0.00 0.12 0.39 0.15 0.00 0.00 0.00 0.12 0.03 0.18 0.06 0 .66 VOL 0.00 0 .2 5 0.04 0.07 0.11 0.00 0.00 0.06 0.00 0.00 0.00 0.19 0.00 0.4 5 DFIA 0.04 0 .2 3 0.05 0.11 0.10 0.00 0.00 0.06 0.00 0.04 0.00 0.18 0.00 0 .47 R 2 values for BASREL are bold-faced to indicate best univariate model. 142 6.3.2 Model Refinement: Residual Analysis and Outlier Diagnostics Having developed the initial regression equations, further analysis was conducted to identify outlying cases via residual analysis. The purpose was to assess the effect of outliers on the regression models, and to assess whether the outlier should remain in the model or should be discarded. However, outliers are not necessarily a sign for inappropriate model design, but often contain interesting and important sample information. In the following, a variety of residual analyses will be presented that help to identify outliers and aid in an assessment of the magnitude of their impact on the regression equations. Outlying cases are then discussed according to their unusual position, and a decision is made as to whether to retain the case or exclude it from further analysis. Plots of residuals were investigated with respect to outlying cases and if a case fell outside the +3 standard deviation of residuals, it was considered an extreme outlier. Once an outlier had been identified, it was checked whether errors could have occurred during data collection or data transfer. Another statistic which proved useful for outlier detection was a plot of deleted residual versus residuals. Deleted residuals are the standardized residuals for their respective cases that one would obtain if the cases had been excluded from the analysis. The deleted residual, is A A calculated as di = Yi - Yno where Yno is the predicted value andF, the observed value for the /th case omission. Pronounced outliers will have the tendency to pull the regression line towards their locations in space, causing a biased least-squares estimates. This method is complimentary to ordinary residual analysis but has the advantage of identifying outlying Y values that were not detected by ordinary residual plots. An example for the use of this statistic is illustrated in Figures 6.4 and 6.5. Mahalanobis' and Cook's distances are other frequently used measures to detect outliers. Mahalanobis1 distance, D 2 , measures the distance of an observation from its group centroid,?i, with i denoting the group in question. Yi are obtained by applying the vector of regression coefficients to the mean score vector in each group (Yi = b' x(). For a two-group example, this measure is calculated as: 143 [6.7] D2 = &Yi = b'x,-bxz with A Yt = y, - Y2. D 2 can then be used to determine whether or not between-groups differences are statistically significant. In this study, Mahalanobis' distances were plotted ranked by magnitude, and extreme outlying cases were flagged for comparison with other outlier diagnostics. Cook's distance measure, Dj, is derived from the confidence region for all regression coefficients, b ,^ simultaneously. Dj can be expressed as: where Y is the vector of the fitted values when all n cases are used for the regression fit, and Y^ is the vector of the fitted values when the ith case is deleted. pMSE is the mean square error. The magnitude of Dj can be assessed by referring to the corresponding F distribution F(p, n-p). Determining the percentile of this distribution for the highest Dj will then indicate the magnitude of its influence on the overall regression fit. Following the same graphical approach used for the Mahalanobis' distance analysis, all Dj values were plotted ranked by magnitude and extreme outlying Dj values were identified by their position and flagged for further analysis. Figure 6.3 illustrates the graphic outlier diagnoses for the combined dataset with VOL as the predictor variable, clearly identifying case 15 (Kaboose Creek) as an outlying case. The five outlier diagnostic techniques were compared with respect to the consistency of results. Mildly outlying data points were not discarded, but the reason for the existence of an outlier was investigated. Only when there was evidence supported by several of the diagnostic methods described above that the outlying point strongly altered the regression estimate, was the outlying case excluded and the analysis repeated. For ambiguous cases, the regression analysis was repeated with and without the outlying cases, and the intercept value and regression coefficients were compared. Following this procedure, the FREQ estimate for the original weathering-limited and transport-limited datasets were recalculated after deleting Kaboose Creek as an outlying case. Kaboose Creek had already been identified as a somewhat outlying point in the A A A A [6.8] D. = ( r - r ( 0 ) ( r - r ( 0 ) pMSE, 144 discriminant function analysis. The reason for its outlying position is very difficult to assess since the control of individual basin parameters on debris flow frequency is speculative, given that both BASREL and HYPIN are both very poorly correlated with FREQ. The Q m a x and VOL estimates for the transport-limited data set were recalculated after omission of Pothole Creek and Gunbarrel I, respectively. Omission of these cases improved the R value by only 4% for the Qmax and VOL estimates, but had the desirable effect of stabilizing the regression coefficients. The Q m a x regression model was further simplified by retaining only CONTAR in the regression model. Pothole Creek was identified as an outlying case due to its very low compressive strength of 0.5 MPa. Average debris flow volume for Gunbarrel I is somewhat too low given the size of its contributing area in the transport-limited basin category, which may explain its outlying position. Having identified serious outliers and having excluded them from the model, the next step was the validation of the regression model, with a completely different set of debris flow basins. Figure 6.3 Cook's distance for the volume variable (original dataset) 0) JO E 3 C ra .a 0.1 0.2 0.3 0.4 Cook's distance 0.5 0.6 145 6.3.3 Model Validation Neter et al. (1989) list three different methods for validating regression models. (i) Comparison of results with theoretical expectations, earlier empirical results, and simulation results. (ii) Splitting of the existing data set into two similarly-sized groups, one of which is used to test the model which was developed from the other data set. (iii) Collection of a new data, to test the model. Comparison with theoretical expectations or simulation, as suggested in point 1, was not possible because of a lack of theory and physical models on which such expectations could be based. Data splitting, as suggested in (ii), is only meaningful when enough observations exist to develop a reliable model in which the number of observations is at least 6 to 10 times the number of variables. Splitting the data set in two sets of 17 observations, and then stratifying these sets further into weathering and transport-limited basins would have resulted in too few variables to retain, which would have adversely influenced the predictive power of the model. For these reasons, new data were collected to check the regression model. Few basins are available in which frequency, magnitude and peak discharge of debris flows are all reasonably well known. The majority of suitable basins where these factors have been determined to some degree of accuracy are located along the east shore of Howe Sound. Basin characteristics and associated debris flows in this area have been addressed in several reports and scientific publications and therefore will not be repeated here (Thurber Engineering Ltd., 1983; VanDine, 1985; Hungr et al. 1984; Church, 1985; Bovis and Dagg 1987,1988; Church and Miles, 1987). The basins used for validation are listed in table 6.7. Thurber Engineering Ltd. (1983) compiled a comprehensive document on frequency and magnitude of debris flows based on Field observations, media reports and eye-witness accounts. Data on frequency, magnitude and peak discharge were extracted from this report. Due to urban development and logging on the debris fans, dendrochronologic analysis of the depositional area and along the channels was not attempted. Other test basins include Cheekye River basin, located some 20 km north of Squamish. This basin was selected because its frequency and magnitude characteristics have been well studied 146 recently because of proposed development plans on Cheekye River debris fan (Thurber Engineering Ltd. and Golder Associates Ltd., 1993). Two more basins in the upper Fraser Valley were added to the test sample. One basin is located near Agassiz Mountain (Martin et al., 1984); the other (Ted Creek) is located approximately 2 km east of Peterson Creek and 2 km west of Mount Ludwig Creek, between Chilliwack and Hope. The highly destructive Ted Creek debris flows have been described by Evans and Lister (1984) and Slaymaker et al. (1987). Finally, two basins on the north side of Hope Mountain, south of the town of Hope, were added to the sample as a result of two debris flows that occurred in November, December 1995 at Hope Creek and Alexander Creek. It should be emphasized that due to the relatively short observation period, the record of debris flows along Howe Sound, as well as the Hope area, is limited, and errors associated with the estimation of mean volumes and frequency must be expected. The magnitude of these errors is impossible to quantify precisely because there are neither theoretical models nor sufficient information on which to base such error analysis. Errors associated with volume determinations are typically due to a poor knowledge of flow thickness and poor preservation of the flow deposit. In cases where several natural incisions exist flow thickness can usually be estimated with an accuracy of +50%. Errors associated with debris flow frequency estimation depend on the length of record, the recurrence interval, and on the method used for debris flow dating, and are therefore extremely difficult to quantify. Longer records with more frequent flows are less sensitive to a single event unrecognized event than are short records where only few debris flow events have been recorded. Radiocarbon dating, for example, will most likely record very large events, since it is only those events which tend to deposit large amounts of organic debris, which increases the likelihood that a datable sample will be found. However, dendrochronologic methods yield dates on debris flows with a large spectrum of volumes but are unlikely to record small, confined flows, as discussed in Chapter 4. Morphometric and geotechnical data were collected in the sample of test basins according to the methods outlined in Chapter 5. The test data set was then subjected to the same residual analyses and outlier diagnostics as the original dataset. The twelve test basins were stratified into weathering-limited and transport-limited sub-types according to the results obtained in the 147 discriminant analysis in Section 6.1. Only Charles Creek and Cheekye River were identified as transport-limited. Following this classification, Qmax, VOL, DFIA and FREQ were calculated for each basin and compared to the predicted results presented in Table 6.7. In general, the model tends to underestimate peak discharge and volume, and overestimates both the debris flow index of activity and frequency. It is difficult to assign definitive reasons for the lack of predictive ability of the model in each basin because, unlike the experimental situations where the dependent variables are commonly error-free, Qmax, VOL, DFIA and FREQ in the test data set are estimated from available field evidence which usually contains some error as discussed above. A possible explanation for the underestimation of the VOL and Qm ax variables could be the fact that only large and destructive events have been recorded along Howe Sound, as well as at Agassiz Mountain and Cheekye River. Frequency data are probably more reliable in the original data set because of the availability of a fairly complete tree-ring record at these sites. About one half of the magnitude estimates and about one third of the entire dataset were predicted with reasonable accuracy. Particularly good estimates were achieved for magnitude at M-Creek, and Turpi n Creek. The overestimation of DFIA for Cheekye River is a resultant of the very high RUG-0 value, which is due in turn to the high density of zero-order and first-order channels in the upper part of the watershed. Cheekye basin is very unusual with respect to ruggedness compared to other basins in Quaternary volcanic complexes of the Garibaldi volcanic belt (see Table 5.4). The severe underestimation of the Hope Creek debris flow volume and peak discharge may be due to intensive scour of the colluvial channel and the unique structural geology of the basin. Hope Creek is situated close to an extension of the Fraser Fault zone, which may have severely weakened the rock by tectonic activity. The unusually large fan that extends towards Highway 3 is indicative of high rates of geomorphic activity. This unique characteristic of Hope Creek is not incorporated as a predictor variable in the regression model. Magnitude estimation at Hope Creek is based on only three recorded flows, indicating that the assumed "observed" volume may in fact be too much influenced by the the 1995 flow with a magnitude of approximately 55,000 m 3 (Jakob et al., 1996). Severe underestimation of total volume and peak discharge in Ted Creek may be due to the same problem of basing average volume on two recent debris flow events, which might have been 148 unusually large. Another reason for the low predicted value in this basin is a possible underestimation of the percent active area. Most initiation zones of the 1983 debris flow were located in areas initially classified as non-active on older air photographs, which was verified by field checking. A final problem of determining average debris flow volume at Ted Creek is that 75% of the watershed area was logged prior to 1966, and slope instabilities caused by clearcutting may have contributed to the size of the 1968 and 1983 events. Ted Creek was identified as an outlier for volume and DFIA calculation in the combined weathering-limited data set and therefore was excluded. Although it is interesting to examine the reasons for both under- and overestimations of debris flow magnitudes derived from this model, planning and engineering considerations require that particular emphasis is given to severe underestimations. Future research should therefore attempt to investigate the reasons for these anomalies that will ultimately lead to a refinement of the model presented here. Debris flow volume estimates are based on fragmentary records of recent flows and can not be interpreted as error-free arithmetic means. Volume estimates may include errors up to 100% as noted earlier which explains discrepancies between observed and predicted volumes by a factor of two. After testing the original models on the test data set, the two data sets were combined and the outlier statistics were repeated to check for anomalies that might have resulted due to the combination of the two sets of data. Cheekye River basin had to be excluded from the combined model since it was identified as an extreme outlier which severely affected the least-squares estimate for peak discharge prediction. Figure 6.4 and 6.5 are plots of deleted residuals against residuals with and without Cheekye River basin. This graphical technique was verified by comparing the regression equations with and without this case (equations 6.9 and 6.10). The changes in the regression intercept and the regression coefficients for the BASREL and CONTAR variables are minor. The change from 0.36 to 0.49 in the AREA coefficient is due to the comparatively large size of Cheekye basin. This change, however, is minor compared to the change in RUG-O from 0.06 to 0.33. As noted earlier, the degree of dissection in upper Cheekye basin is considerably higher than that found in other basin. This is reflected in the 9.87 RUG-0 149 value; this is a factor of 4.1 higher than the highest values in all other basins. This analysis led to the omission of Cheekye River basin from the final regression model. The overall perfomance of the model on the test data set is disappointing given that only about one third of all predictions could be regarded as reasonably close to the observed value. However, it must be emphasized that the test data set is based on a generally very short and incomplete record. Therefore it can not be concluded that the model presented here is inadequate to predict debris flow frequency-magnitude characteristics, but rather highlights the need for continuing monitoring to obtain more reliable frequency and magnitude estimates to refine and verify the model. Preliminary air photograph analysis is of great importance to determine the appropriate choice of the different regression models summarized in Table 6.9. Where doubt exists as to the sediment availability classification, the transport-limited equation, which yields conservatively higher values, should be used. The equation for the combined data set generally yields higher values for both Q m a x and VOL, but predictive power is lower due to both the significantly lower R2-value and the standard error of the estimate compared with that of the original data set. Each regression model was checked with regard to the linearity assumption by producing scatterplots of predicted values versus residuals of the dependent variables. Homogenous scatter around the center line indicated linearity, thus confirming the preliminary bivariate correlation test conducted at the beginning of this section. The pattern of scatter also indicated the existence of predominantly linear relations between the dependent and predictor variables. 6.3.4 Final Regression Model The preceding section has shown that model validation is associated with several problems that may be due in part to the quality of the test data set as well as the number of cases used for comparison. Comparison of regression results did not lead to firm conclusions as to the validity of the original model. The final regression model is based on a combination of the original and the test data sets, which were analysed according to the methods described above. To ensure proper use of the regression equations, the basins were again stratified in weathering -or transport limited 150 Table 6.7. Observed versus predicted magnitude and frequency for the 12 test basins. Variables are defined in table 5.7. Qmax Qmax VOL VOL DFIA DFIA FREQ FREQ B A S I N (obs.) (pred.) (obs.) (pred.) (obs.) (pred.) (obs.) (pred.) Agassiz Ck. 220 30 11000 , 1000 670 80 0.06 0.05 Charles Ck. 270 90 15000 11000 840 1180 0.06 0.06 Alberta Ck. 110 80 5000 3000 110 260 0.02 0.07 Magnesia Ck. unknown 180 unknown 10000 unknown 1200 0.03 0.11 M-Ck. 210 200 10000 10000 110 680 0.01 0.09 Turpin Ck. 40 50 2000 2000 20 170 0.01 0.06 Newman Ck. 160 90 8000 4000 340 300 0.05 0.07 Hope Ck. 270 60 15000 2000 1820 260 0.18 0.07 Alexander Ck. 120 170 5000 9000 380 970 0.03 0.11 Thacker Ck. 250 170 23000 8000 1820 710 0.18 0.10 TedCk. 400 40 42000 1000 3100 40 0.07 0.05 Cheekye R. 300 1000 100,000 180,000 9300 1.6-101 1 0.09 0.05 note: Bold-faced numbers were calculated using the equation for the original transport-limited data set All other figures were determined from the original weathering-limited data set (Table 6.7) Table 6.7 (cont.) Basin year(s) of debris flow(s) volume(s) source Agassiz Ck. 1962, '82 11,000 ('82) Martin etal. (1984) Charles Ck. 1969, 72, '72, '81 10,000-20,000 Thurber (1983) Alberta Ck. 1982, '83 450, 10,000 Thurber (1983) Magnesia Ck. 1962, '81 unknown Thurber (1983), VanDine (1985) M-Ck. 1981 20,000 Thurber (1983) Turpin Ck. 1983 1,500 Thurber (1983) Newman Ck. 1969, '81, '83 7,500 Thurber (1983) Hope Ck. ca. 1970, '84, '95 3,000*. 6,400, 55,000 Thurber (1984), Jakob et al. (19%) Alexander Ck. 1984, '95 5,400# ('95) own observation Thacker Ck. 1983, '95 11,200, 35,000 Thurber, 1983, own observation TedCk. 1967, '83 60,000; 24,000* Thurber, 1985, Slaymaker, et al. (1987) Cheekye R. 1920's, '58, '93, several approx. 100,000 Thurber and Golder (1989) other events (poor resolution) Dubek (pers. comm.) •volume estimate based on deposit area and assumed depth # volume estimated empirically via cross-section and peak discharge-total volume relationship (Fig. 4.8) 151 Figure 6.4 Deleted residuals versus residuals for Qmax of the combined data set 2.6 w 3 •g tn a> DC & <D Q 1.4 0.8 0.2 -0.4 o M C H E E K Y E RIVER -0.8 -0.6 -0.4 -0.2 0 0.2 Residuals 0.4 0.6 0.8 1 Regression 95% confid. [6.9] logQmax =-3.36 + 1.72«logBASREL+0.36«logAREA-0.06«RUG-O+0.13*logCONTAR (R2=0.43) Figure 6.5. Deleted residuals versus residuals for Qmax after omission of Cheekye River basin Residuals vs. Deleted Residuals Dependent variable: QMAX 2.6 1.4 TO in 0.8 •o S 0.2 -0.4 -1 • • • • • ' • - • • . . , 1 _ , , . -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 Residuals Regression 95% confid. [6.10] logQmax = -3.14 + 1.77«logBASREL+0.49«logAREA-0.33«RUG-O+0.14«CONTAR (R2=0.45) 152 according to the procedures outlined in section 6.1. Before accepting the final regression model, a final test for multicollinearity of the model variables was conducted to guarantee replicability of the results. 6.3.4.1 Multicollinearity Check The test for multicollinearity in the predictor variables ensures that including interrelated variables will not destabilize the regression estimates. It has been claimed above that complete avoidance of variable collinearities in this type of study is neither possible nor desirable due to the complex response mechanisms acting in natural basins (Schumm, 1973). Minor multicollinearity is therefore acceptable if the regression coefficients are not seriously affected. The degree to which regression coefficients are affected can be verified by choosing predictors from any given basin and calculating the values of the dependent variables to compare coefficients. A decision can then be made as to whether the deviations are within an acceptable margin of error. An alternative is the analysis of variance inflation factors (VIF). Although multicollinearity was largely avoided by applying forward-stepwise regression, an examination of the correlation matrix (Table 6.3) of variables retained in the model suggests some remaining intercorrelation. A final inspection was therefore conducted with regard to excessive variable intercorrelation to avoid model bias and thus guarantee model replicability. The VIF measures the degree to which the variances of the estimated regression coefficients are inflated, as compared to the case where the independent variables are not linearly related. According to Neter et al. (1989), VIF can be calculated as: [6.11] (VIF)k = (1-Ric2)-1 with k = 1, 2 , p - 1 and Rk2 is the coefficient of multiple determination when a predictor variable, Xk, is regressed on the other X variables in the model. The variance inflation factor is equal to 1 when Rk2 = 0, which implies that Xk shows no correlation with any of the other X variables. For Rk2 > 0, VIF will be greater than 1, indicating inflated variances. Table 6.8 shows the calculated variance inflation factors for the final regression equations. A maximum VIF of 10 is generally taken to indicate serious influence on the least-square estimate (Neter et al., 1989). As expected from an examination of the correlation matrix, the variables % ACT and AAR display high VIF values compared to other variables retained in the model (Table 6.8). Although high in comparison, Table 6.8 suggests that the critical VIF value of 10 is not exceeded, which suggests that multicollinearity does not warrant further attention, Table 6.8. Variance Inflation Factors (VIF) for final regression equations C O M B I N E D D A T A S E T N=46 F R E Q Qmax V O L D F I A variable VIF variable VIF variable VIF variable VIF %ACT 5.65 BASREL 1.58 AREA 2.22 AREA 2.26 A R E A 1.22 RUG-0 1.94 %ACT 2.39 %ACT 2.39 AAR 5.97 AREA 2.25 CONTAR 2.24 WSN 1.06 WSN 1.74 CONTAR 1.24 CONTAR 2.51 ORIGINAL D A T A S E T N=34 RUG-0 1.08 BASREL 2.29 BASREL 1.23 BASREL 1.23 %ACT 1.11 RUG-0 2.37 %ACT 4.12 %ACT 4.12 COMST 1.07 AREA 2.34 AAR 4.06 AAR 4.06 COMST 1.95 TEST D A T A S E T N = 12 HYPIN 1.04 HYPIN 1.02 RELRAT 1.61 HYPIN 1.20 CONTAR 1.04 CONTAR 1.02 %ACT 1.61 WSN 1.20 ORIGINAL WEATHERING-LIMITED N=22 BASREL 1.01 BASREL 0.80 BASREL 1.19 BASREL 1.00 HYPIN 1.01 AREA 0.60 AREA 1.61 HYPIN 1.00 RUG-0 0.70 RUG-0 1.46 ORIGINAL TRANSPORT-LIMITED N=12 HYPIN 1.05 CONTAR 0.00 WSN 1.40 RUG-0 1.01 RUG-0 1.05 CONTAR 1.40 WSN 1.01 C O M B I N E D W E A T H E R I N G - L I M I T E D N=32 BASREL 1.01 R E L R A T 1.36 BASREL 1.01 BASREL 1.01 HYPIN 1.01 BASREL 1.09 HYPIN 1.01 HYPIN 1.01 HYPIN 1.32 C O M B I N E D T R A N S P O R T - L I M I T E D N=14 HYPIN 1.14 CONTAR 1.01 CONTAR 1.00 CONTAR 1.00 %ACT 1.14 RUGNR 1.01 WSN 1.00 HYPIN 1.00 154 Table 6.9 Summary of regression results of the final models D A T A S E T N EQUATION R2 R2 S E ORIGINAL 34 logFREO = -1.83 + 0.33*RUG-0 + 0.78*%ACT + 0.03*COMST - 0.31*AAR .27 .20 .27 34 l o p O m i l x = -9.01 + 3.78*loeBASREL + 0.58*IogAREA - 0.06*COMST - 0.47*RUG-0 .66 .63 .33 34 logVOL = -8.76 + 3.92*IogBASREL + 1.87*%ACT - 0.76*AAR .45 .41 .48 34 logDFIA = -13.28 + 5.01*logBASREL+2.65*%ACT-1.13*AAR .47 .44 .60 TEST 12 logFREQ = no correlation 12 l o g Q m a x = 1.52 + 0.28«IogCONTAR + 1.50.HYPIN .32 .17 .24 12 logVOL = 6.34 - 4.73*RELRAT + 2.12*%ACT .66 .58 .29 12 logDFIA = -1.65 + 5.20*HYPIN + 0.85*WSN .36 .07 .62 COMBINED 46 logFREO = -1.76 + 1.44»%ACT + 0.26*logAREA - 0.63*AAR + 0.26*WSN .25 .17 .32 45 l o P ° m a \ =-314 + 1.77*logBASREL + 0.49*loeAREA - 0.33.RUG-0 + 0.14»logCONTAR .45 .40 .37 46 logVOL = 3.46 + 0.11*iOgCONTAR + 0.64*loeAREA + 0.80*%ACT .33 .28 .48 46 logDFIA = 2.19 + 0.05*logCONTAR - 0.11 *WSN + 1.40*%ACT + 0.96*AREA .37 .29 .65 ORIGINAL 21 logFREO = -5.90 + 1.39*IoeBASREL+ 0.70HYPIN .38 .31 .20 WEATH-LIM 22 l o g O m a x =-6.64 + 2.88*IogBASREL+ 0.50*logAREA - 0.47*RUG-0 .62 .55 .33 22 logVOL = -6.52 + 3.34«logBASREL + 0.60*logAREA - 0.45*RUG-0 .63 .57 .38 22 logDFIA = -14.13 + 4.87*logBASREL+2.22*HYPIN .61 .57 .49 ORIGINAL 12 logFREQ = 1.39 - 6.20*HYPIN - 0.18*COMST .42 .78 .18 TRANS-L IM 11 logOm^ = 1.97 + 1.24*logCONTAR .86 .85 .24 11 logVOL = 6.31 + 0.91*loeCONTAR - 1.52*WSN .83 .79 .22 12 logDFIA = 3.28 + 0.91*RUG-0-0.88*WSN .75 .73 .35 TEST 10 logFREQ = no correlation WEATH-LIM 10 l o g Q m a x = 0.12 + 2.45*HYPIN + 0.47*WSN .37 .19 .25 10 loeVOL = 5.02 - 1.83*RELRAT .38 .30 .30 10 logDFIA = -1.57 + 5.05*HYPIN + 0.84*WSN .25 .04 .66 COMBINED 32 logFREQ =-4.74 + 0.93*logBASREL+1.10*HYPIN .14 .08 .33 WEATH-LIM 32 loeO m i , x =-7.16 + 2.45*loeBASREL+ 1.93.HYPIN + 0.66*RELRAT .35 .28 .38 31 logVOL = -7.69 + 3.27*loeBASREL + 1.82*HYPIN .49 .45 .38 31 logDFIA = -12.41 + 4.21*logBASREL + 2.84*HYPIN .46 .42 .54 COMBINED 14 loeFREO = 0.61 - 3.60*HYPIN + 0.55«%ACT .40 .29 .32 TRANS-LIM 14 l o g O m i r t = 5.75 + 0.94*loeCONTAR - 0.99*logRUGNR .76 .71 .30 13 loeVOL = 5.54 + 0.67*i o gCONTAR - 0.89*WSN .80 .76 .22 14 logDFIA = 5.90 + 0.74*IogCONTAR - 4.84*HYP1N .66 .59 .40 note: underlined variables are significant at a = 0.05 R denotes the adjusted multiple coefficient of determination calculated via equation [6.5]. 155 6.3.4.2 Discussion Having produced the final set of regression equations, it is appropriate to discuss the regression coefficients and the physical meaning of the variables included in the models. It is important to note that the magnitude of each coefficient should be assessed with care since the variables were not standardized, and variables are not entirely independent of one another. This means that if a given coefficient, bj, is greater than bj this does not necessarily imply that the associated predictor variable, Xj_ is greater or more important than Xj. In other words, the weighting of x n is dependent on the scale of the measured variable, and on dependencies amongst the predictor variables. Standardization of variables was not conducted because similar linear measures were used and variables with largely different scales were log-transformed, with the effect of reducing scale differences. In addition, the influence of the regression coefficients on the y-variable assumes that all other factors are kept constant which, despite all efforts to avoid excessive interaction, is not completely assured. Before discussing the physical meaning of predictor variables it should be noted that the positive or negative signs of some coefficients do not necessarily indicate a positive or negative relationship between the predictors and the dependent variables. This is due to the fact that adding or dropping correlated predictor variables can change the sign of regression coefficients. However, strong correlations with positive or negative signs, which is indicated in Table 6.4, did not experience a change in sign. Peak discharge ( Q m a x ) and volume (VOL) in the original data set are best predicted by B A S R E L in combination with A R E A and % A C T , which indicate the potential energy in a basin and the amount of erodible sediment. Qmax is further controlled by COMST and RUG-0. The prediction of V O L and DFIA, which includes V O L as a product term, are also influenced by A A R . A A R indicates the stability of erodible sediment normalized by basin area. More unstable material in a proportionally large contributing area should be conducive to higher magnitudes and therefore to higher debris flow activity. This argument is further strengthened by the inclusion of C O N T A R that is also an important predictor variable for magnitude ( Q m a x a ° d VOL) as well as DFIA prediction. Clearly, it was only possible to identify areas as belonging to one of the three classes that showed some potential for instability over the period of record (approximately 200 years). It 156 is possible that some areas classified as "stable" on this time scale may be unstable on a longer time scale. However, this cannot be assessed here because of a lack of field evidence. Therefore, areas classed as unstable according to AAR and WSN likely only provide minimum volume estimates over longer time periods. In weathering-limited basins, HYPIN is an important predictor for VOL and Qmax- The hypsometric integral reflects curvature and distribution of mass in a basin. Higher HYPIN suggests basins with convex-profiles which can easily convey mobilized material and encourage additional sediment entainment through channel erosion. RUG-0 has also been identified as a variable influencing Q m a x and VOL. Changes in RUG-0 are expected to cause a decrease or increase in sediment availability because an increase in basin ruggedness will increase the true area which is exposed to weathering processes. This link is supported by the relatively high correlation coefficient between AREA and RUG-0 of 0.64 (cf. Table 6.4). Frequency is controlled by a variety of variables with fewer consistencies than magnitude prediction. HYPIN, %ACT, COMST and BASREL are the only variables that occur repeatedly. Following the above argument, higher HYPIN values encourage easy debris delivery and may therefore influence the frequency of debris flows. The prediction of FREQ for the original transport-limited basins is particularly encouraging. With only two variables (HYPIN and COMST), 78% of the total variance can be explained. Lower values of COMPST mean higher rates of erosion of rocks on rock faces and in the channel. Low compressive strength could therefore indicate more material recharge, and hence influence debris flow frequency. An investigation of the variables involved in the regression model shows that variables have a clear physical meaning, which supports most of the assertions made in Chapter 5. Some variables show coefficients of opposite sign to those expected, probably due to the correlations among predictor variables that can be expected in a situation where complex and little understood system responses make it impossible to evaluate the influence of a change of a single variable on other variables while controlling all other model parameters. A more extensive discussion of the meaning of retained variables was not attempted to avoid erroneous interpretation for similar reasons. 157 6.4 S U M M A R Y In this chapter, a statistical model was developed with the objective of predicting debris flow peak discharge, volume, frequency, and activity from physical basin attributes. Prediction is based on the assumption that the values of the predictor variables will not significantly change, at least within the time frame of design discharge calculations. It was shown that models including several variables perform significantly better than bivariate regressions. A classification of basins into weathering-limited and transport-limited types generally yields improved least-squares estimates because the two basin types display distinctly different frequency-magnitude characteristics. Caution should be exercised with basins that are transitory between the two basin types. Small changes in the value of one or more discriminatory variables may cause a change from weathering-limited conditions to transport-limited conditions and vice versa. For engineering applications, the more conservative transport-limited estimates should be chosen. Frequencies in transport-limited basins are more predictable than those in weathering-limited basins, which probably reflects their direct response to the exceedance of climatic thresholds. The combined models for weathering-limited and transport-limited basins do not improve the predictive power of the original model based on the original data set. This is probably a result of the introduction of extreme variability, as well as differences in quality between the two sets of data which stem from the different methods of event reconstruction. Debris flow frequency estimates for weathering-limited basins should be interpreted very cautiously because of the generally low coefficients of multiple determination (R2). The low R 2 values are probably due to the fact that debris flow frequency is controlled largely by climatic factors that were not included in the models. This point is further confirmed by comparing model fit in the cases of weathering-limited and transport-limited basins. In the latter case, predictability is much higher, which was expected considering that virtually unlimited amounts of sediment in a basin respond much more readily to particular hydroclimatic effects. In weathering-limited basins, recharge rates, also not included in the model, are a major control of frequency. The independence of the selected morphometric and geotechnical variables to geologic, morphologic and climatic conditions should allow transfer of the equations developed here to other 158 regions. It is important to note that reliable estimates can only be obtained for data that lie within the range of data points for the significant variables. Testing of this model in other regions forms an important avenue of future research in this field. It is important to emphasize that the models presented here are not suitable for estimating maximum expected magnitudes, as defined earlier. Presently, there is no physical or theoretical basis to confidently estimate maximum expected magnitudes for debris flows. This is due to the poor quality of volumetric data of older debris flows, which does not allow the construction of reliable frequency-magnitude relations. It is also due to an insufficient understanding of the erodibility of material in colluvial channels and of sediment storage and recharge times. A final problem is that the size and number of triggering failures in bedrock or colluvial material cannot reliably be estimated for older debris flows because of the tempo of geomorphic change, which has obliterated or eroded most of the evidence. As pointed out in Chapter 4, section 4.4.3, average magnitudes of debris flows in weathering-limited basins are more likely to be close to maximum expected magnitudes in cases where the majority of debris flow material is entrained from the channel. The mean magnitude in transport-limited basins is very sensitive to the number of events for which magnitudes have been reconstructed since deviations from the mean are larger than in weathering-limited basins. The equations presented here provide a reasonable approximation to average magnitudes. For practical applications, this implies lower factors-of-safety for the construction of mitigative structures at weathering-limited basins and higher factors-of-safety in transport-limited basins. As already stated, in cases where the discriminatory scores are similar, and there is doubt as to the basin type, the transport-limited model should be chosen to avoid gross underestimation of debris flow volume or peak discharge. 159 CHAPTER 7. CONCLUSIONS Data on the frequency and magnitude of debris flows were obtained from 34 basins in southwestern British Columbia. Although the quality and quantity of this data set is one of the most comprehensive compiled worldwide to date, the nature of the data does generally not allow the formal application of extreme value frequency analysis as commonly used for perennial streams in hydrology (cf. Chapter 2). Unlike most rivers, debris flow channels exhibit episodic rather than continuous behaviour and, in any case, are not usually monitored. Partial duration series analysis is based on several statistical assumptions which, in the case of debris flows, could only be met in some transport-limited basins, but a lack of several reliable magnitude estimates precludes further analysis using these methods. Most data on debris flow phenomena must be collected by labour-and time intensive field investigations that rely heavily on the indirect documentation of events by depositional and botanical evidence. Much of this evidence is removed over time either by erosion, fire, logging or through burial by subsequent events. At the high-magnitude end of the scale, events generally occur outside of the the time window available for this study. These limitations warrant the adoption of a different approach in this study in which the frequency and magnitude of debris flows are related to morphometrical and geotechnical characteristics of the debris flow basins. In order to test the hypotheses outlined in Chapter 1, data on the frequencies and magnitudes of debris flows were collected by a variety of field and laboratory techniques and complemented with historical accounts and air photograph analysis. Dendrochronological methods, such as the dating of scarred trees and the determination of the onset of reaction wood, proved to be the most reliable methods to establish debris flow frequencies on a regional-scale in this particular mountain environment. In a number of cases, volumetric measurements of deposits were made. In other cases, an empirical approach was used by relating peak discharge to total flow volume based on data collected in the study area. Another approach, based on the relation between channel cross-section and peak discharge, was developed for sites where velocity estimates could not be made because of a lack of channel bends suitable for super-elevation 160 calculations. The principal problem encountered with magnitude determination was the lack of volumetric or peak discharge information from the older debris flow events. Field evidence was used to determine a "typical" magnitude of debris flows from weathering-limited basins, based on the assertion that supply-limitations and mobility thresholds determine the volume of debris flows such that very small or very large flows are unlikely to occur. In weathering-limited basins, the arithmetic mean of documented flows may therefore approximate a typical magnitude for a given basin. In transport-limited basins, where the range of volumes is much more variable, available field information was used to approximate magnitude. In this context, the results in Tables 5.5 and 5.6 show that the deviations of debris flow magnitudes from the mean in transport-limited basins are likely to be considerably larger than those of weathering-limited basins; the concept of typical magnitude is therefore more difficult to apply. Maximum expected magnitudes were computed to obtain a measure of an upper limit for debris flow volume based on the assumption that channels will not change significantly during this time period. There are two fundamental problems associated with the calculation of maximum expected magnitudes. First, many debris flow channels are accessible in only some reaches, and extrapolations of debris volumes along the channel proved to be unreliable. Second, a recent debris flow near the town of Hope, in which the pre-existing channel was incised to an unexpectedly great depth and width, indicates our lack of understanding of erosion, particularly in colluvial debris flow channels. The uncertainty involved in the calculation of maximum expected magnitudes is probably the reason for generally poor correlations between typical magnitudes and maximum expected magnitudes as discussed in Chapter 4. Thus maximum expected magnitude was not included in the statistical model. Despite this uncertainty, maximum expected magnitudes calculated by means of erodibility indices are useful in evaluating the distribution of erodible sediments along debris flow channels, and provide a method for comparison between channels. Past studies have attempted to correlate and regress frequency and magnitude characteristics of debris flows with one or two basin attributes and have yielded disappointing results. This is not surprising considering the complex response mechanisms that are acting in geomorphologically active basins. An additional disadvantage of these univariate and bivariate morphometric models is a general neglect of the geotechnical variables which control debris supply and hence debris flow 161 magnitude. These considerations led to the development of a statistical model that incorporates morphometric as well as geotechnical variables to predict frequency and magnitude of debris flows. Morphometric parameters were selected according to knowledge acquired from previous studies that have attempted correlations between basin parameters with measures of erosion. Adjustments had to be made for variables that were unsuitable for this particular study in their conventional form. For example, drainage density was determined from zero-order and first-order channels identified on air photographs, since conventional "blue-line" drainage on maps was found to be a poor representation of the actual channel system. A measure of ruggedness or basin dissection was developed based on the intensity of contour crenulations (RUG-0). Geotechnical variables were developed ad hoc because existing material classification systems were found to be inappropriate for the scale and purpose of this study. The Weighted Stability Number (WSN) measures the degree of stability of the active parts of a basin, and the Active Area Ratio (AAR) describes the overall stability of the basin. The advantage of the morphometric and geotechnical data, which were selected as input parameters to the statistical model, is their relative ease of measurement from air photographs and topographic maps, supported by a minimum of field work, at least in preliminary studies. The identification of unstable areas, as required for the calculation of the Active Area Ratio and the Weighted Stability Number, calls for expert judgment, but is aided by a number of criteria that facilitate the correct classification of these areas. The preliminary regression models involving three or four variables were somewhat disappointing in their predictive power, but still performed considerably better than previously used univariate and bivariate models. A stratification of all basins into weathering-limited and transport-limited basins based on discriminant analysis was conducted to test the second hypothesis, which stated that these sub-groups would significantly enhance the predictive power of the statistical model. Explained variance in the regression models improved by up to 38%, suggesting that stratification of basins according to their sediment supply characteristics is an important prerequisite for obtaining a better prediction of debris flow frequency and magnitude. Particularly successful equations were developed from the original data set for the prediction of frequency, peak discharge, total volume, and the debris flow index of activity in transport-limited 162 basins. Explained variances were 78%, 87% and 81%, and 73% respectively. The same variables yielded figures of 35%, 74%, 78%, and 63% respectively from the combined data set comprising the original 34 basins and 12 test basins. The success of these predictions confirms the assumption that mean debris flow frequency and magnitude in transport-limited basins should be better predictable than those of weathering-limited basins. In general, debris flow frequency was the least predictable variable, which is probably due to its climatic controls, which were not included in the model. The fact that predictability of transport-limited basins is higher than that of weathering-limited basins is considered to reflect the virtually unlimited amounts of sediment in the former basin type, making it more responsive to hydroclimatic events. Despite their comparatively high coefficients of multiple determination, the regression equations for transport-limited basins should be used with care since they were developed from a relatively small data set of only 11 to 14 basins. A larger sample of transport-limited basins is needed to validate the robustness of these functions. The regression model was tested by an independent sample of 12 basins for which frequency and magnitude had previously be determined. In some cases, the equations provide reasonable estimates, but in most instances the predicted and observed values diverge considerably. This divergence might be partly explained by the differences in quality between the two data sets, and partly due to the fact that some predictor variables in the test sample fell outside the range of the variables included in the original model. Presently there is no dataset available in British Columbia, which would allow an unbiased sample for model validation. Although the regression results using the test sample of basins were somewhat disappointing, all test basins were correctly classified according to the discriminant function used to stratify the original sample of basins into weathering-limited and transport-limited types. In its present form, the regression model is not confined to specific geologic or climatic regions because neither bedrock geology nor climate was explicitly addressed in the selection of predictor variables. Geology was represented by the geotechnical variables WSN, AAR and COMST, which can be determined from any type of bedrock or surficial materials. It is therefore expected that the model could be used in other mountain regions with similar morphometric and 163 geotechnical characteristics with no more work involved than the remeasurement of variables that were identified as significant in this study. Interpretation of the controls exerted by individual variables on the frequency and magnitude relations of debris flows confirmed most of the assertions made in Chapter 5. The results presented here should be regarded as guidelines rather than as final statements. If more precise frequency and magnitude characteristics are the objective of a study, detailed field work is still required. In order to improve confidence in the regression and discriminant function, the highest research priority must be to monitor all researched basins with regard to the recurrence of debris flows. In a mountainous province such as British Columbia, in which development is opening new transportation corridors through hazardous terrain, and where development is beginning to encroach upon steep hillsides, a process as destructive and as threatening as debris flow must be met with increased awareness and intensified research. Only long-term monitoring of the recurrence and volume of debris flows will improve our numerical predictive capabilities. A n increase in record length of high quality data is desirable to improve the "true" mean of dependent variables such as volume, peak discharge, and frequency, which can then be adjusted after each recorded event. Another obvious advantage of long-term monitoring is that other important correlations, such as the relation between mean and maximum magnitude in weathering-limited and transport-limited basins, can be investigated to improve confidence in the regression model. Ultimately, a data set of this quality will allow the detection of activity fluctuations in response to changes in land use or climate. The second research priority is to add more basins to the existing sample for which debris flow frequencies and magnitudes are known. Including additional basins would allow further stratification according to either climate or geologic controls without the problem of having to discard predictor variables because of inadequate sample size. A n increase in sample size would also allow the incorporation of other predictor variables with the aim of further reducing unexplained variance in the key dependent variables. Apart from increasing the sample size and improving the resolution of the frequency and magnitude estimates, new variables describing basin morphometry or geotechnical characteristics, or hydroclimatic indices should be sought. In particular the description of geotechnical characteristics requires refinement. This can only be accomplished by finding a suitable method 164 which is able to identify important variables quantifying rock mass stability as well as the relative stability of noncohesive materials appropriate for basin-scale investigations. Ground level photogrammetry could prove helpful in overcoming the problem of inaccessibility and hazardous work conditions, and would yield important data such as joint spacing, and joint orientation. Exact prediction of the frequency and magnitude of debris flows is still unattainable. However, this study has advanced our knowledge in several areas of frequency and magnitude characteristics of debris flows in southwestern British Columbia. Specifically, the following contributions to science have been made: When compared with previous studies in this region, the record of debris flow frequency developed in this study is unparalleled. Certainly, the quality of the data record was adequate for exploring the correlations between frequency and controlling terrain variables. Based on this record it was successfully shown that multivariate models are clearly superior to uni -or bivariate models that try to predict frequency and magnitude of debris flows. The adoption of the concept of weathering-limited and transport-1 imitated debris flow basins provided a framework which was effective for predicting debris flow behaviour, and greatly enhanced predictive power of the statistical model. A geotechnical system for mapping basin stability was introduced and proved helpful in quantifying material supply to debris flow channels. It also added considerably to the explained variance of frequency and magnitude characteristics of debris flows. In addition, the study has contributed to the empirical prediction of debris flow frequencies and magnitudes, and the methods developed here may serve the needs of preliminary investigations. In particular, many debris flow basins have not preserved a clearly reconstructable record of previous debris flow frequency and magnitude either due to absence of long-lived trees or erosion of previous deposits. Despite these geomorphic constraints, the early phases of any engineering investigation require estimates of expected debris flow activity before detailed design work can be accomplished. This research has also re-emphasized the importance of a close, mutually beneficial interaction between the engineering and geomorphological fields within the stated limitations of the models, a refinement and periodic recalculation of computed magnitude and frequency estimates will eventually provide more reliable tools for assessing and controlling debris flow hazards. 165 REFERENCES Abrahams, A.D. 1972. Drainage densities and sediment yields in eastern Australia. Australian Geographical Studies, 10, 19-41. Association of Forest Service Employees for Environmental Ethics (AFSEEE) 1996. Report on landslides in Oregon. Ahnert, F. 1970. Functional relationships between denudation, relief, and uplift in large mid-latitude drainage basins. American Journal of Sciences, 268, 243-263. Ahnert, F. 1972. Relief and denudation - theoretical test of an empirical correlation. International Geography Congress, 1972, Vancouver, B.C. Alestalo, J. 1971. Dendrochronological interpretation of geomorphic processes. Fennia, 105, 1-140. Azimi, C. and Desvarreux, P. 1974. A study of one special type of mudflow in the French Alps. Quarterly Journal of Engineering Geology, 7,329-38. Belsley, D.A., Kuh, E. and Welsch, R.E. 1980. Regression Diagnostics: Identifying influential data and sources of collinearity. New York, John Wiley & Sons. Benda, L. and Cundy, T.W. 1990. Predicting deposition of debris flows in mountain channels. Canadian Geotechnical Journal, 27,409-417. Benda, L. and Dunne. T. 1987. Sediment routing by debris flow. In Erosion and Sedimentation in the Pacific Rim. Beschta, R.L., Blinn, T., Grant, G.E., Ice, G.G. and Swanson, F.J. (eds.) International Association of Hydrological Sciences, Publication no. 165, 213-223. Bienieawski, Z.T. 1989. Engineering rock mass classification in rock engineering. In Bieniawski, Z.T. (ed.). Exploration for rock engineering, Proceedings of the Symposium, Vol. 1, 97-106. Birks, H.J.B. 1980. The present flora and vegetation of the moraines of the Klutlan Glacier, Yukon Territory, Canada: A study in plant succession. Quaternary Research, 14,60-86. Bovis, M.J. 1982. Uphill-facing (antislope) scarps in the Coast Mountains. Geological Society of America Bulletin, 93,804-812. Bovis, M.J. 1985. Earthflows in the interior plateau, southwest British Columbia. Canadian Geotechnical Journal, 22,313-334. Bovis, M.J. and Dagg, B.R. 1987. Mechanism of debris supply to steep channels along Howe Sound, southwest British Columbia. IAHS Publication no. 165, 191-200. Bovis, M.J. and Dagg, B.R. 1988. A model for debris accumulation and mobilization in steep mountain channels: Hydrological Sciences Journal, 33,589-604. Bovis, M.J. and Dagg, B.R. 1992. Debris flow triggering by impulsive loading: mechanical modelling and case studies. Canadian Geotechnical Journal, 29,345-352. Bovis, M.J. and Evans, S.G. 1995. Rock slope movement along the Mount Currie "fault scarp", southern Coast Mountains, British Columbia. Canadian Journal of Earth Science 32, 2015-2020. 166 Bovis, M.J. and Evans, S.G. 1996. Extensive deformation of rock slopes in the southern Coast Mountains, southwestern British Columbia, Canada. Engineering Geology. In press. Boydell, A.N. 1992. Surficial geology and land use, report to the surficial geology task group. Unpublished report for the Earth Sciences Task Force Research Inventory Committee. Victoria Brabb, E. 1995. The San Mateo County California GIS project for predicting the consequences of hazardous geologic processes. Carrara, A. and Guzzetti, F. (eds.). Geographical Information Systems in assessing natural hazards. Kluwer Academic Publishers, p. 299-334. Broch, E. and Franklin, J.A. 1972. The point load strength test. International Journal of Rock Mechanics and Mining Science, 9, 669-697. Brook, N. 1985. The equivalent core diameter method of size and shape correction in point load testing. International Journal of Rock Mechanics, Mining Science and Geomechanics Abstracts, 22, 61-70. Brooks, G.R. and Hickin, E.J. 1991. Debris avalanche impoundments of Squamish River, Mount Cayley area, southwestern British Columbia. Canadian Journal of Earth Sciences, 28: 1375-1385. Broscoe, A.J. and Thomson, S. 1969. Observations on an alpine mudflow, Steele Creek, Yukon. Canadian Journal of Earth Sciences, 6,219-229. Buchanan, P. and Savigny, K.W. 1990. Factors controlling debris avalanche initiation. Canadian Geotechnical Journal, 27, 659-675. Buchanan, P., Savigny, K.W., DeVries, J. 1990. A method for modeling water tables at debris avalanche headscarps. Journal of Hydrology, 113, 61-88. Caine, N. 1980. The rainfall intensity duration control of shallow landslides and debris flows. Geografiska Annaler 62A, 1/2, 23-27. Campbell., R.H. 1975. Soil slips, debris flows, and rainstorms in the Santa Monica Mountains and vicinity, southern California: U.S. Geological Survey Professional Paper 851,51 p. Cannon, S.H. 1989. An evaluation of the travel-distance potential of debris flows. Utah, Geological and Mineral Survey Miscellaneous Publication 89-2,35p. Cannon, S.H. 1993. An empirical model for the volume-change behavior of debris flows. Hydraulic Engineering '93, Proceedings, Vol. 2, 1768-1773. Cannon, S.H. and Ellen, S.D. 1985. Rainfall conditions for abundant debris avalanches in the San Francisco Bay region, California. California Geology 38, no. 12, 267-272. Cannon, S.H. and Ellen, S.D. 1988. Rainfall that resulted in abundant debris flow activity during the storm. In Landslides, floods, and marine effects of the storm of January 3-5, 1982, in the San Franscisco Bay Region, California, U.S. Geological Survey, Professional Paper 1434, p. 27-34. Cannon, S.H. and Savage, W.Z. 1988. A mass change model for the estimation of debris flow runout. Journal of Geology, 96, 221-227. Carlston, C.W. 1963. Drainage density and streamflow. U.S. Geological Survey Professional Paper 422-C, 8 p. 167 Carter, N.M. 1931. Exploration in the Lillooet River watershed. Canadian Alpine Journal, 21, 8-18. Chorley, R.J. Malm, D.E.C. and Pogorzelski, H.A. 1957. A new standard for measuring drainage basin shape. American Journal of Science, 255, 138-141. Chorley, R.J. and Morgan, M.A. 1962. Comparison of morphometric features. Unaka Mountains, Tennessee and North Carolina and Dartmour, England. Geological Society of America Bulletin, 73, 17-34. Chow, V.T. 1959. Open channel hydraulics. McGraw Hill, New York, 680 p. Church, M. 1985. Debris torrents and natural hazards of steep mountain channels; east shore of Howe Sound. In Jackson, L.E., Church , M., Clague, J.J., and Eisbacher, G.M.(eds.). Slope hazards in the southern Coast Mountains of British Columbia: Geological Society of America Cordilleran Section Annual Meeting, Vancouver, British Columbia, May 6-10, Field Trip 4, Guidebook, 4.2-4.22. Church, M. and Ryder, J.M. 1972. Paraglacial sedimentation: a consideration of fluvial processes conditioned by glaciation. Geological Society of America, Bulletin, 83,3059-3072. Church, M. and Miles, M.J. 1987. Meteorological antecedents to debris flow in southwestern British Columbia; some case studies. In Debris Flows/Avalanches: Process, Recognition, and Mitigation. Geological Society of America, Reviews in Engineering Geology, VII: 63-79. Clague, J.J. and Souther, J.G. 1982. The Dusty Creek landslide on Mount Cayley, British Columbia. Canadian Journal of Earth Sciences, 19, 524-539. Clague, J.J., Evans, S.G., and Blown, I.G. 1985. A debris flow triggered by the breaching of moraine dammed lake, Klattasine Creek, British Columbia. Canadian Journal of Earth Sciences, 22, 1492-1502. Clark, J.I. 1966. Morphometry from maps. In Dury, G.H., (ed.) Essays in Geomorphology, Heinemann, London, 235-274. Costa, J.E. 1984. Physical geomorphology of debris flows. In Developments and applications of geomorphology. Costa, J.E. and Fleisher, P.J. (eds.). Springer Verlag, 268-317. Costa, J.E. 1988. Rheologic, geomorphic, and sedimentologic differentiation of water floods, hyperconcentrated flows, and debris flows. In Baker, V.R., Kochel, R.C., and Patton, P.C. (eds.). Flood geomorphology. John Wiley and Sons, 113-122. Costa, J.E. 1996. Personal communication. Costa, J.E. and Jarrett, R.D. 1981. Debris flows in small mountain stream channels of Colorado and their hydrologic implication. Bulletin of the Association of Engineering Geologists, 18, 309-322. Crandell, D.R. 1971. Postglacial lahars from Mount Rainier volcano, Washington. U.S. Geological Survey, Professional Paper 677,75 p. Crandell, D.R., Mullineaux, D.R., and Miller, C D . 1979. Vocanic-hazard studies in the Cascade Range of the western United States. In: Volcanic Activity and Human Ecology. Sheets, P.D. and Grayson, D.K. (eds.) Academic Press, New York, 195-219. Croft, S.A.S. 1983. Stability Assessment of the Capricorn Creek Valley, British Columbia. B.Sc. thesis, Department of Geological Sciences, University of British Columbia. 168 Cruden, D.M. and Lu, Z-Y. 1992. The rock slide and debris flow from Mount Cayley, British Columbia in June, 1984, Canadian Geotechnical Journal, 29, 614-626. Dillon, W.R. and Goldstein, M. 1984. Multivariate Analysis - Methods and Applications. John Wiley & Sons, New York, 587 p. Draper, N.R. and Smith, H. 1981. Applied regression analysis. Second edition. John Wiley & Sons. New York. Douglass, A.E. 1919. Climatic cycles and tree-growth. A study of the annual rings in trees in relation to climate and solar activity. Car. Inst. Washington, 1,1-127. Dubek, T.H. 1996. Personal communication. Duffell, S. and McTaggart, K.C. 1952. Ashcroft map-area. British Columbia. Geological Survey of Canada Memoir 262, 122 p. Eisbacher, G.H. and Clague, J.J. 1984. Destructive mass movements in high mountains: Hazard and management. Geological Survey of Canada. Paper 84-16,230 p. Ellen, S.D., Mark, R.K., Cannon, S.H., Knifong, D.L. 1993. Map of debris flow hazard in the Honolulu District of Oahu, Hawaii. United States Geological Survey Open-File Report, 93-213. Ellen, S.D. and Mark, R.K. 1993. Mapping debris flow hazard in Honolulu using a DEM. Hydraulic Engineering '93, Proceedings, Vol. 2, 1774-1779. Evans, S.G. 1986. Landslide damming in the Cordillera of western Canada. In Schuster R.L. (ed.) Landslide Dams: Process, Risk and Mitigation. Schuster. American Society of Civil Engineering, Special Publication, 3,111-130. Evans, S.G. 1987. A rock avalanche from the peak of Mount Meager, British Columbia. Geological Survey of Canada, Paper 87-1 A, 929-934. Evans, S.G. 1990. Massive debris avalanches from volcanoes in the Garibaldi Volcanic Belt, British Columbia. Program with Abstracts, 15, Geological Association of Canada/Mineralogical Association of Canada, p. A38. Evans. S.G. 1992. Landslides and river damming events associated with the Plinth Peak volcanic eruption, southwestern British Columbia, GeoHazards '92, Vancouver, 405-412. Evans, S.G. and Lister, D.R. 1984. The geomorphic effects of the July 1983 rainstorms in the southern Cordillera and their impact on transportation facilities. Geological Survey of Canada Paper 84-IB, 223-235. Evans, S.G. and Brooks, G.R. 1991. Prehistoric debris avalanches from Mount Cayley volcano, British Columbia. Canadian Journal of Earth Sciences, 28,1365-1374. Fairchild, L.H. 1985. Lahars at Mount St. Helens, Washington. Unpublished Ph.D. thesis, University of Washington, Seattle, 374 p. Fannin, R.J. and Rollerson, T.P. 1993. Debris flows: some physical characteristics and behaviour. Canadian Geotechnical Journal, 30,71-81. Fannin, R.J. and Wise, M.P. 1995. A method for calculation of debris flow travel distance. 48th Canadian Geotechnical Conference, Vol. 2,643-650. Fitzharris, B.B. 1975. Snow accumulation and deposition on a west coast midlatitude mountain. Ph.D. Thesis, University of British Columbia, Vancouver, B.C. 169 Forest Practices Code of British Columbia - Regulations. 1995. B.C. Environment, Victoria, B.C. Forest Practices Code of British Columbia. 1995. Section 17(2). Operational Planning Regulation. Gully Assessment Procedure Guidebook. 40 p. Gilbert, R. 1972. Observations on lacustrine sedimentation at Lillooet Delta, British Columbia, 187-194. In O. Slaymaker and H.J. McPherson, (eds.): Moutain Geomorphology. Tantalus Research, Vancouver, 274 p. Gilbert, R. 1973. Observations of lacustrine sedimentation at Lillooet Lake, British Columbia. Unpublished Ph.D. thesis, University of British Columbia, 193 p. Gilbert, R. 1975. Sedimentation in Lillooet Lake, British Columbia. Canadian Journal of Earth Sciences, 12, 1696-1711. Gontscharov, E.P. 1962. Sjels in Tadshykistan. Proceedings of the Fifth Symposium on Sjels, Baku, 109-114. Govi, M. and Sorzana, P.F. 1980. Landslide susceptibiity as a function of critical rainfall amount in Piedmont basins (North-Western Italy): Studia Geomorphologica Carpatho-Balcanica, 14, 43-61. Green, N.L., Armstron, R.L., Harakal, J.E., Souther, J.G., and Read, P.B. 1988. Eruptive history and K-Ar geochronology of the late Cenozoic Garibaldi volcanic belt, southwestern British Columbia. Geological Society of America Bulletin, 100,563-579. Haeberli, W. Rickenmann, D., Roesli, U., Zimmermann, M. 1990. Investigation of the 1987 debris flows in the Swiss Alps: General concept and geophysical soundings. In: Hydrology in Mountainous Regions, part II. IAHS Publication no. 194, 303-310. Haan, C.T. and Johnson, H.P. 1966. Rapid determination of hypsometric curves. Geological Society of America Bulletin, 77,123-126. Hampel, R. 1977. Geschiebewirtschaft in Wildbachen. Zeitschrift des Vereins der Diplomingenieure der Wildbach und Lawineneverbauung Osterreichs, 4,3-34,53-144. Hart, J.S. 1979. Clastic sediment sources and suspended sediment yield in a Coast Mountains watershed, British Columbia. Unpublished M.A. thesis, Department of Geography, University of British Columbia. Hawley, P.M. 1989. Identification, evaluation and mitigation of debris flow hazards in British Columbia. In Proceedings of the XX Conference, International Erosion Control Association, Vancouver, B.C., 471p. Heginbottom, J.A. 1972. Surficial geology of Taseko Lakes map area (92-0). Geological Survey of Canada, Paper 72-14. Henscher, S.R. 1987. The implications of joints and structures for slope stability. In Slope Stability. Anderson, M.G. and Richards, K.S. (eds.). John Wiley and Sons Ltd., p. 145-186. Hobson, R.D. 1967. FORTRAN IV programs to determine surface roughness in topography for the CDC 3400 computer. Contribution 14, State Geological Survey, University of Kansas, Lawrence, 28 p. Hobson, R.D. 1972. Surface roughness in topography: a quantitiative approach. In Chorley, R.J. (ed.): Spational Analysis in Geomorphology, Methuen & Co. Ltd. London, 221-246. 170 Howes, D.E. and Kenk, E. 1988. Terrain classification system for British Columbia (revised edition). Manual 10, B.C. Ministry of Environment, Victoria. Hungr, O., Morgan, G.C. and Kellerhals, R. 1984. Quantitative analysis of debris torrent hazards for design of remedial measures. Canadian Geotechnical Journal, 21,663-677. Hungr, O., Morgan, G.C, VanDine, D.F., and Lister, D.R. 1987. In Costa, J.E. and Wieczorek, G.F. (eds.): Debris flow defenses in British Columbia. In Debris Flows/Avalanches: Process, Recognition, and Mitigation. Geological Society of America, Reviews in Engineering Geology, VII: 201-222. Hungr, O. and Rawlings, G. 1995. Assessment of terrain hazard for planning purposes: Cheekye Fan, British Columbia. 48th Canadian Geotechnical Conference, Vancouver, B.C. Sptember 25-27, 509-517. Hupp, C.R. 1984. Dendrogeomorphic evidence of debris flows in frequency and magnitude at Mount Shasta, CA. Environmental Geology and Water Sciences, 6, 121-128. Ikeya, H. 1981a. A method of designation for area in danger of debris flow. Erosion and sediment transport in Pacific Rim steeplands. International Association of Hydrological Sciences (Christchurch, New Zealand), 132, 576-587. Ikeya, H. 1981b. Investigation for debris flow countermeasure. Translated from Civil Engineering Journal, 23(7), (original in Japanese). Ikeya, H. and Mizuyama, T. 1982. Flow and deposit properties of debris flow. Public Works Research Institute, Tsukuba, Japan. Report 157-2, 88-153 (in Japanese). Innes, J.L. 1983. Lichenometric dating of debris-flow deposits on alpine colluvial fans in southwest Norway. Earth Surface Processes and Landforms, 10, 519-524. Innes, J.L. 1985. Magnitude - frequency relations of debris flows in northwest Europe. Geografiska Annaler, 67 A (1-2), 23-32. International Society of Rock Mechanics (ISRM). Commission on Testing Methods. 1985. Suggested method for determining point load strength. International Journal of Rock Mechanics, Mining Science, and Geomechanical Abstracts, 22,53-60. Jackson, L.E. 1977. Dating and recurrence frequency of prehistoric mudflows near Big Sur, Monterey County, California. United States Geological Survey Journal of Research 5, 17-32. Jackson, L.E. 1979. A catastrophic outburst flood (jokulhlaup) mechanism for debris flow generation at the Spiral Tunnels, Kicking Horse River basin, British Columbia. Canadian Geotechnical Journal, 16, 806-813. Jackson, L.E. 1987. Debris flow hazard in the Canadian Rocky Mountains. Geological Survey of Canada. Paper 86-11, 20p. Jackson, L.E., Church, M., Clague, J.J., Eisbacher, G.H. 1985. Slope hazards in the southern Coast Mountains of B.C. Field Trip Guide. Jackson, L.E., Kostaschuk, R.A., and MacDonald, G.M. 1987. Identification of debris flow hazard on alluvial fans in the Canadian Rocky Mountains. In Costa, J.E. and Wieczorek, G.F. (eds.): Debris flow defenses in British Columbia. In Debris Flows/Avalanches: Process, Recognition, and Mitigation. Geological Society of America, Reviews in Engineering Geology, VII, 115-124. 171 Jackson, L.E., Hungr, O., Gardner, J.S. and Mackay, C. 1989. Cathedral Mountain debris flows. Bulletin of the International Association of Engineering Geology, 40,35-54. Jakob, M. and Bovis, M.J. 1996. Morphometrical and geotechnical controls of debris flow activity, southern Coast Mountains, British Columbia, Canada. Zeitschrift fiir Geomorphologie Neue Folge, Supplementband 104, 13-26. Jakob, M., Hungr, O., Thomson, B. 1996. Two debris flows with anomalously high magnitude. Abstract submitted to the First International Conference on Debris Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, San Francisco, August 1997. Johnson, A.M. and Rodine, J.R. 1984. Debris flow. In Brunsden, D. and Prior, D.B. (eds.). Slope instability, John Wiley & Sons, 257-361. Johnson, P.A, McCuen, R.H., Hromadga, T.v. 1991. Magnitude and frequency of debris flows. Journal of Hydrology, 123, 68-82. Jordan, P. 1986. Meager Creek Hotsprings. Hazard study and recreation site plan. British Columbia Ministry of Forests, Squamish District. Jordan, P. 1987. Impacts of mass movement events on rivers in the southern Coast Mountains, British Columbia: Summary report. Canada, Environment, Water Resources Branch, Inland Waters Directorate, IWD-HQ-WRB-SS-87-3,62P. Jordan, P. and Slaymaker, O. 1991. Holocene sediment production in Lillooet River basin, British Columbia: A sediment budget approach. Geographie physique et Quaternaire, 45,45-57. Jordan, P. 1994. Debris flows in the southern Coast Mountains, British Columbia: Dynamic behaviour and physical properties. Unpublished Ph.D. thesis. The University of British Columbia, 258 p. Jordforsk. 1992. MDSD Guide. Center of Soil and Environmental Research, As, Norway. Josza, L. 1988. Increment core sampling techniques for high quality cores. Forintek Canada Corporation. Special Publication SP-30. Journeay, J.M. and Mahoney, J.B. 1994. Cayoosh assemblage: regional correlations and implications for terrane linkages in the southern Coast Belt, British Columbia. In Current Research, Part A, Geological Survey of Canada, 92-1 A, 225-235. Keefer, D.K., Wilson, R.C., Mark, R.K., Brabb, E.E., Brown, W.M., Ellen, S.D., Harp, E.L., Wieczorek, G.F., Alger, C.S., Zatkin, R.S. 1987. Real-time landslide warning during heavy rainfall. Science, 238, 921-925. Kemmerikh, A.O. 1978. Mudflows in the Pamir and Pamir-Alay. Soviet Hydrology, 17, 307-309. Kobashi, S. and Suzuki, M. 1987. The critical rainfall (danger index) for disasters caused by debris flows and slope failure. Proceedings of the Corvallis Symposium on Erosion and Sedimentation in the Pacific Ring, IAHS Publication 165,201-211. Kotarba, A. 1992. High energy geomorphic events in the Polish Tatra Mountains. Geografiska Annaler, 74 A (2-3), 123-131. Kronfellner-Kraus, G. 1983. Torrent erosion and its control in Europe and some research activities in this field in Austria. SABO-The Erosion-Control Engineering Society-Japan, 3(126), 33-44. 172 Langbein, W.B. 1947. Topographic characteristics of drainage basins. U.S. Geological Survey., Water Supply Paper 968-C, 99-114. Langbein, W.B. 1960. Hydrologic data networks and methods of extrapolating or extending available hydrologic data. Hydrologic Networks and Methods. Rood Control Series 15. United Nations Economics Commission for Asia and the Far East, Bankok, Thailand. Larson, S. 1982. Effects on the slopes of Longyear Valley, Spitsbergen, after a heavy rainstorm in July 1972. Geografiska Annaler, 64 A, 105-125. Leopold, L.B. and Maddock, T.Jr. 1953. The hydrologic geometry of stream channels and some physiographic implications. U.S.Geological Survey Professional Paper 282-B, 39-85. LeRoy, O.E. 1908. Preliminary report on a portion of the main coast of British Columbia and adjacent islands included in New Westminster and Nanaimo districts. Geological Survey of Canada, Report 996, Vancouver. Lips, E.W. and Wieczorek, G.F. 1990. Recurrence of debris flows on an alluvial fan in Central Utah. Hydraulics/Hydrology of Arid Lands. Proceedings of the International Symposium, San Diego, California, 555-560. Lu, Z.Y. 1988. Rock avalanches on Mount Cayley, British Columbia. Unpublished M.Sc. thesis, University of Alberta, Edmonton, Alberta. Lu, Z.Y. 1992. Prehistoric debris avalanches from Mount Cayley volcano, British Columbia: Discussion. Canadian Journal of Earth Sciences, 29, 1342-1343. Lu, Z.Y. 1993. Landslides and geotechnical properties of volcanic tuff on Mount Cayley, British Columbia. Unpublished Ph.D. thesis. University of Alberta, Edmonton, 263 p. Lu, Z.Y. and Cruden, D.M. 1996. Two debris flow modes on Mount Cayley, British Columbia. Canadian Geotechnical Journal, 33, 123-139. Major, J.J. and Voight, B. 1986. Sedimentology and clast orientation of the 18 May 1980 southwest flank lahars, Mount St. Helens, Washington. Journal of Sedimentary Petrology, 56, 691-705. Mark, D.M. 1974. A comparison of computer-based terrain storage methods with respect to the evaluation of certain geomorphometric measures. Unpublished MSc. thesis. The University of British Columbia, 171 pp. Mark, D.M. 1975. Geomorphometric parameters: a review and evaluation. Geografiska Annaler, 57 A, 165-177. Mark, R.K. and Newman, E.B. 1988. Rainfall totals before and during the storm. Distribution and correlation with damaging landslides. In Landslides, floods, and marine effects of the storm of January 3-5, 1982, in the San Franscisco Bay Region, California, U.S. Geological Survey, Professional Paper 1434, 8-17. Martin, D.C. , Piteau, D.R., Pearce, R.A., and Hawley, P.M. 1984. Remedial measures for debris flows at the Agassiz Mountain Institution, British Columbia. Canadian Geotechnical Journal, 21, 505-517. Martin, D.C. 1989. Remedial measures for debris flows in British Columbia. Bulletin of the Association of Engineering Geologists, 25, 505-517. Mathews, W.H. 1958. Geology of the Mount Garibalide map-area, southwestern British Columbia, Canada. Bulletin of the Geological Society of America, 69,161-198. 173 Melton, M.A. 1958. Correlation structure of morphometric properties of drainage systems and their controlling agents. Journal of Geology, 66,442-460. Melton, M.A. 1965. The geomorphic and paleoclimatic significance of alluvial deposits in southern Arizona. Journal of Geology, 73, 1-38. Miles, M.J. 1984. Hydrometeorological conditions associated with debris torrents. The Canadian Association of Geographers, Annual Meeting, Nanaimo, B.C., 21-24. Miles, M.J. and Kellerhals, R. 1981. Some enigneering aspects of debris torrents. In Proceedings of the Canadian Society for Civil Engineering, 5th Canadian Hydrotechnical Conference, Fredericton, New Brunswick, 395-421. Millard, T. 1993. Sediment in Forested and Logged Gullies, coastal British Columbia. Unpublished M.Sc. thesis, University of British Columbia, Vancouver, 202p. Mizuyama, T. 1982. Analysis of sediment yield and transport data for erosion control works. In Recent developments in the explanation and prediction of erosion and sediment yield (Proceedings of the Exeter Symposium, July 1982). International Association of Hydraulics Research, 137, 177-182. Mizuyama, T., Kobashi, S. and Ou, G. 1992. Prediction of debris flow peak discharge. Interpraevent. Internationales Symposium. Bern, Switzerland. Tagespublikation, Band 4, 99-108. Mokievsky-Zubok, O. 1977. Glacier-caused slide near Pylon Peak. British Columbia. Canadian Journal of Earth Sciences, 15, 1039-1052. Monger, J.W.H. 1989. Geology of Hope and Ashcroft map-areas; Geological Survey of Canada, maps, 41-1989, 42-1989, (scale 1:250,000). Monger, J.W.H. and McMillan, W.J. 1984. Bedrock geology of Ashcroft (92-1) map area. Geological Survey of Canada, Open File 980. Monger, J.W.H. and Journeay, J.M. 1994. Guide to the geology and tectonic evolution of the southern Coast Mountains. Geological Survey of Canada. Open File, 2490. Moore, D.P. and Mathews, W.H. 1978. The Rubble Creek landslide, southwestern British Columbia. Canadian Journal of Earth Sciences, 15, 1039-1052. Morgan, G.C., Rawlings, G.E., and Sobkowicz, J.C. 1992. Evaluating total risk to communities from large debris flows. In Geotechnique and Natural Hazards, Symposium. Vancouver Geotechnical Society and Canadian Geotechnical Society, Vancouver, B.C., 225-236. Morton, D.M. and Campbell, R.M. 1974. Spring mudflows at Wrightwood, southern California. Quarterly Journal of Engineering Geology, 7,377-384. Moss, A.J. 1972. Technique for assessment of particle breakage in natural and artificial environments. Journal of Sedimentary Petrology,,42,725-728. Muhs, D, R., Thorson, R.M., Clague, J.J., Mathews, W.H., McDowell, P.F., Kelsey, H.M. 1987. Pacific coast and mountain systems. In Graf, W.L. (ed.). Geomorphic Systems of North America. The Geological Society of America, Centennial Special Volume 2,517-582. Nasmith, H.W. and Mercer, A.G. 1979. Design of dykes to protect against debris flows at Port Alice, British Columbia. Canadian Geotechnical Journal, 16,748-757. Neter, J., Wasserman, W. and Kutner, M.H. 1989. Applied Linear Regression Models. 2nd edition. Irwin Inc. Homewood, 667 p. 174 Nilson, T.H., Taylor, F.A. and Dean, R.M. 1976. Natural conditions that control landsliding in the San Francisco Bay region - an analysis based on data from the 1968-69 and 1972-73 rainy seasons. U.S. Geological Survey Bulletin 1424,35 p. Niyazov, B.S. and Degovets, A.S. 1975. Estimation of the parameters of catastrophic mudflows in the basins of the Lesser and Greater Almatinka Rivers. Soviet Hydrology, 2,75-80. Nyberg, R. 1985. Debris flows and slush avalanches in northwestern Swedish Lappland, distribution and geomorphological significance. Meddelanden fran Lunds Universitets Geografiska Institution Avhandlinger 97,222p. Oden, M.E. 1994. Debris recharge rates in torrented gullies on the Queen Charlotte Islands. Unpublished M.Sc. University of British Columbia. 100 p. Ohmori, H. and Hirano, M. 1988. Magnitude, frequency and geomorphological significance of rocky mud flows, landcreep and the collapse of steep slopes. Zeitschrift fur Geomorphologie, Neue Folge, Supplementband 67, 55-65. Okubo, S. and Mizuyama, T. 1981. Planning countermeasures against debris flow. Translated from Civil Engineering Journal, 23(9) (original in Japanese). Okuda, S., Suwa, H., Okunishi, K., Yokoyama, K., Nakano, M. 1980a. Observations on the motion of a debris flow and its geomorphological effects. Zeitschrift fur Geomorphologie Supplementband, 35, 142-163. Okuda, S., Okunishi, K. and Suwa, H. 1980b. Observation of debris flow at Kamikamihori Valley of Mt. Yakedake. In: Okuda, S. (ed.): Excursion guide book, 3rd meeting of IGU Commission on Field Experiments in Geomorphology. Kyoto, 116-139. O'Loughlin, C.L. 1972. A preliminary study of landslides in the Coast Mountains of southwestern British Columbia. In Slaymaker, H.O. and McPherson, H.J. (eds.). Mountain Geomorphology. Tantalus Research Ltd. Vancouver, 101-111. Osterkamp, W.R., Hupp, C.R.., Bloodroot,. 1986. Magnitude and frequency of debris flows, and areas of hazard on Mount Shasta, Northern California. U.S. Geological Survey, Professional Paper 1396-C, 1-20. Owens, I.F. 1974. Interrelationships between climatic and material characteristic in the occurrence of alpine mudflows in the Nigel Pass area, Canadian Rocky Mountains. Abhandlungen der Akademie der Wissenschaften in Gottingen, Mathematisch Physische Klasse, 29,313-318. Partsch, J. 1911. Schlesien, eine Landeskunde fur das deutsche Volk. II, Breslau. Patton , F.D. 1976. The Devastation Glacier slide, Pemberton , B.C. (abstract). In Geomorphology of the Canadian Cordillera and its bearing on mineral deposits. Geological Association of Andad, Cordilleran Section, Vancouver, B.C., p.26. Patton, P.C. and Baker, V.R. 1976. Morphometry and floods in small drainage basins subject to diverse hydrogeomorphic controls. Water Resources Research, 12, 941-952. Phipps, R.L. 1985. Collecting, preparing, crossdating, and measuring tree increment cores. U.S. Geological Survey Water Resources Investigation Report 85-4148. Pierson, T.C. 1985. Initiation and flow behavior or the 1980 Pine Creek and Muddy River lahars, Mount St. Helens, Washington. Geological Society of American Bulletin, 96, 1056-1069. Pierson, T.C. 1986. Flow behavior of channelized debris flows, Mount St. Helens, Washington. In Hillslope Processes. Edited by Abrahams, A.D. Allen & Unwin, Boston, 269-297. 175 Pierson, T.C. and Scott, K.M. 1985. Downstream dilution of a lahar: transition from debris flow to hyperconcentrated streamflow. Water Resources Research, 21,1511-1524. Pike, R.J. and Wilson, S.E. 1971. Elevation-relief ratio, hypsometric integral, and geomorphic area-altitude analysis. Geological Society of America Bulletin, 82,1079-1084. Podor, A. 1993. Recent debris flow frequency and magnitude at West Wilson Creek, Banff National Park, Alberta, M.Sc. thesis, University of Waterloo, 211 p. Rapp, A. 1960. Recent development of Mountain Slopes in Karkevagge and surroundings, northern Scandinavia. Geografiska Annaler, 42,65-200. Rapp, A. and Stromquist, L. 1979. Field experiments on mass movements in the Scandinavian mountains with special reference to Karkevagge, Swedish Lappland. Studia Geomorphologica Carpatho-Balcanica, 13,23-38. Read, P.B. 1978. Geology, Meager Creek geothermal area, British Columbia. Geological Survey of Canada, Open File 603. Read, P.B. 1990. Mount Meager complex, Garibaldi Belt, southwestern British Columbia. Geoscience Canada, 17, 167-170. Resources Inventory Committee. 1995. Guidelines and standards for terrain mapping in British Columbia, 123 p. Roberts, M.C. and Klingman, P.C. 1972. The relationship of drainage net fluctuation and discharge. International Geography 1972, W.P. Adams and F.M. Helleiner (eds.), 22nd International Geography Congress, Montreal. Roddick, J.A. 1966. Coast Crystalline Belt of British Columbia. In Tectonic history and mineral deposits of the western Cordillera. H.C. Gunning (ed.). Canadian Institute of Mining and Metallury, Special Volume, 8,73-82. Roddick, J.A. and Hutchinson, W.W. 1973. Pemberton (east half) and mainland parts of Alberni, Geological Survey of Canada, Paper 73-17. Rood, KM. 1984. An aerial photograph inventory of the frequency and yield of mass wasting on the Queen Charlotte Islands, British Columbia. British Columbia Ministry of Forests, Land Management Report 34. Rood, K.M. 1990. Site characteristics and landsliding in forested and clear-cut terrain, Queen Charlotte Islands, B.C. British Columbia Ministry of Forests, Land Management Report, No 64. Russel, S.O. 1972. Behaviour of steep creeks in a large flood. In Mountain Geomorphology. Edited by Slaymaker, H.O. and McPherson, H.J. Tantalus Research Ltd., Vancouver, 223-227. Ryder, J.M. 1971. The stratigraphy and morphology of para-glacial alluvial fans in south-central British Columbia. Canadian Journal of Earth Sciences, 8,279-298. Ryder, J.M. 1981a. Terrain inventory and Quaternary geology, Lytton, British Columbia. Geological Survey of Canada, Paper 79-25. Ryder, J.M. 1981b. Geomorphology of the southern part of the Coast Mountains of British Columbia. Zeitschrift fur Geomorphologie, 37,120-147. Ryder, J.M. and Church, M. 1985. The Lillooet terraces of Fraser River: a palaeoenvironmental enquiry. Canadian Journal of Earth Sciences, 23, 869-884. 176 SABO 1986. An index to the measures against debris flow (guideline). SABO (Erosion Control) Division SABO Department. Public Works Institute. Ministry of Construction. SABO 1987. Measures against debris flow on roads. SABO (Erosion Control) Division SABO Department. Public Works Institute. Ministry of Construction. Technical Memorandum 2492. Salisbury, N.E. 1962. Relief: slope relationships in glaciated terrain. Annals of the Association of American Geographers, 52, 358-359. Savigny, K.W. and Rinne, N.F. 1991. Assessment of landslide hazard along a pipeline corridor in mountainous terrain of southwestern British Columbia. 44th Canadian Geotechnical Conference, Calgary, 19p. Schaefer, D.G. 1980. The rainstorm of December 13-18, 1979 over southwestern British Columbia; Scientific Services Division, Atmospheric Environment Service, Environment Canada, Vancouver, unpublished, 4 p. Schaefer, D.G. (1981): Meteorological conditions associated with flooding in British Columbia, December 1980; Atmospheric Environment Service, Environment Canada, Vancouver, unpulished, 2 p. Schaefer, D.G. 1982. Meteorological conditions associated with slides and floods in the Lower Mainland of British Columbia in later October 1981, Scientific Services Division, Atmospheric Environment Service, Environment Canada, Vancouver, B.C. unpublished, 47 p. Schaefer, D.G. 1983. Meteorological conditions associated with the Alberta Creek debris torrent of February 11, 1983, Canada Department of Environment, Atmospheric Environment Service, Pacific Central Office, Vancouver, 5p. Schumm, S.A. 1954. The relation of drainage basin relief to sediment loss. Publication International Association of Hydrology, IUGG, 10th General Assembly, Rome, 1, 216-219. Schumm, S.A. 1956. Evolution of drainage systems and slopes in badlands at Perth Amboy, New Jersey. Geological Society of America Bulletin, 67,597-646. Schumm, S.A. 1963. The disparity between present rates of denudation and orogeny. United States Geological Survey Professional Paper 454-H, 13 p. Schumm, S.A. 1973. Geomorphic thresholds and complex response of drainage systems. In Morisawa, M. (ed.): Fluvial geomorphology, 299-310. Schweingruber, F.H. 1987. Beziehungen zwischen Wachstumsphasen in Jahrringabfolgen und Umweltbedingungen in der Schweiz. In: Gesellschaft fiir Strahlen- und Umweltforschung Munchen. Proceedings Symposium Klima und Witterung in Zusammenhang mit den neuartigen Waldschaden, Munchen, 165-174. Schweingruber, F.H., Eckstein, D., Serre-Bachet, F., Braker, O.U. 1990. Identification, presentation and interpretation of event years and pointer years in dendrochronology. Dendrochronologia, 8, 8-39. Schweingruber, F.H. , Wehrli, U., Aellen-Rumo, K., Aellen, M. 1991. Weiserjahre als Zeiger extremer Standorteinflusse. Schweizerische Zeitschrift fiir Forstwesen, 142,33-52. Scott, K.M. 1988. Origins, behaviour, and sedimentology of lahars and lahar-runout flows in the Toutle Cowlitz River system. U. S. Geological Survey Professional Paper 1447-A, 74 p. 177 Scott, K.M. 1989. Magnitude and frequency of lahars and lahar runout flows in the Toutle-Cowlitz River system. U.S. Geological Survey Open-file Report 90-385,106 p. Scott, K.M., Pringle, P.T. and Vallance, J.W. 1992. Sedimentology, behavior, and hazards of debris flows at Mount Rainier, Washington. U. S. Geological Survey, Open-file Report 90-385, 106p. Scott, K.M., Vallance, Pringle, P.T. 1995. Sedimentology, behaviour, and hazards of debris flows at Mount Rainier, Washington. U.S. Geological Survey Professional Paper 1547. 56 P-Selby, M.J. 1980. A rock mass strength classification for geomorphic purposes: with tests from Antarctica and New Zealand. Zeitschrift ftir Geomorphologie, 24,31-51. Selby, M.J. 1982. Hillslope materials and processes. Oxford University Press, Oxford, 264p. Sharpe, R.P. and Nobles, L.H. 1953. Mudflow of 1941 at Wrightwood, southern California. Bulletin of the Geological Society of America 64,547-560. Sharpe, D.R. 1974. Mudflows in the San Juan Mountains, Colorado: How constraints, frequency and erosional effectiveness. Unpublished M.Sc. thesis. University of Colorado. Shroder, J.F. 1980. Dendrogeomorphology: Review and new techniques of tree ring dating. Progress in Physical Geography, 4, 161-188. Shroder, J.F. and Butler, D.R. 1986. Tree-ring analysis in the earth-sciences. Proceedings of the International Symposium on Ecological Aspects of Tree-Ring Analysis, New York, 186-212. Sidle, R.C., Pearce, A.J. and O'Loughlin, C.L. 1985. Hillslope stability and land use. Water Resources Monograph 11, American Geophysical Union, 140 p. Slaymaker, O. and Gilbert, R. E. 1972. Geomorphic process and land use changes in the Coast Mountains of British Columbia, a case study, 269-279. In P.Macar and A. Pissart (eds.) Processus p r^iglaciaires Studies sur le terrain. University de Liege. Slaymaker, O. and McPherson, H.J. (eds.) 1972. Mountain Geomorphology. Tantalus Research, Vancouver, 274 p. Slaymaker, O., Hungr, O., Desloges, J., Lister, D., Miles, M., VanDine, D. 1987. Debris torrent and debris flood hazards, lower Fraser, Niculum and Coquihalla Valleys, B.C. I.U.G.G. XIX General Assembly, Excursion Guidebook. Slaymaker, O. 1988. The distinctive attributes of debris torrents. Hydrological Sciences Journal, 33, 567-573. Sobkowicz, J., Hungr, O. and Morgan, G. 1995. Probabilistic mapping of a debris flow hazard area: Cheekye Fan, British Columbia. 48th Canadian Geotechnical Conference, Vancouver, B.C. Sptember 25-27, 519-529. Smith, K.G. 1950. Standards of grading texture of erosional topography. American Journal of Science, 248, 655-668. Stone, R.O. and Dugundji, J. 1965. A study of microrelief - its mapping, classification and quantification by means of a Fourier analysis. Engineering Geology, 1,89-187. Stokes, M.A. and Smiley, T.L. 1968. An introduction to tree ring dating. The University of Chicago Press, 73p. 178 Strahler, A.N. 1952. Hypsometric (area-altitude) analysis of erosional topography. Geological Society of America Bulletin, 63,1117-1142. Strunk, H. 1991. Dendrogeomorphologische Methoden zur Ermittelung der Murfrequenz und Beispiele ihrer Anwendung. Habilitationsschrift Universitat Regensburg, 218 p. Sulebak, J.R. 1969. Mudflow in the low alpine region. Norsk Geografisk Tidsskrift, 23, 15-33. Suzuki, M. 1979 The artificial rainfall for the disasters caused by debris movement. Shin-Sabo 110, 1-7. Suzuki, M. and Kobashi, S. 1981. The critical rainfall for the disasters by slope failures. Shin-Sabo 121, 16-26. Swanston, D.N. 1974. Slope stability problems associated with timber harvesting in mountainous regions of the western United States. U.S. Forest Service General Technical Report PNW-21. Swanston, D.N. and Swanson, F.J. 1976. Timber harvesting, mass erosion, and steepland forest geomorphology in the Pacific Northwest. In: Coates, D.R. (ed.): Geomorphology and Engineering, Dowden, Hutchinson and Ros, Inc., 199-221. Swetnam, T.W., Thompson, M.A., Sutherland, E.K. 1985. Using dendrochronology to measure radial growth of defoliated trees. Spruce budworm handbook. United States Department of Agriculture. Handbook 639. Takahashi, T. 1981. Estimation of potential debris flows and their hazardous zones; soft countermeasures for a disaster. Journal of Natural Disaster Science, 3, 57-89. Thurber Consultants Ltd. 1983. Debris torrent and flooding hazards, Highway 99, Howe Sound. Report to Ministry of Transportation and Highways, British Columbia, Vancouver. Thurber Consultants Ltd. 1985. Debris torrent assessment Wahleach and Floods, Highway 1 Hope to Boston Bar Creek Summit, Coquihalla Highway. Report to British Columbia Ministry of Transportation and Highways, Vancouver. Thurber Consultants Ltd. and Golder Associates Ltd. 1992. Cheekye Fan study. Report to the Ministry of Environment, Land, and Parks, Vancouver. Tipper, H.W. 1971. Glacial geomorphology and Pleistocene history of central British Columbia. Geological Society of Canada, Bulletin 196. VanDine, D.F. 1985. Debris Flows and debris torrents in the Southern Canadian Cordillera. Canadian Geotechnical Journal, 22,44-67, Victoria. VanDine, D.F. 1996. Debris flow control structures for forest engineering. Report to the Ministry of Forests, Research Branch, 68 p. Ward, W.E. 1989. Hydrology of floods in Canada: A guide to planning and design. National Research Council Canada, Ottawa, 245 p. Watanabe, M. 1981. Debris flows and associated disasters. Translated from Civil Engineering Journal, 23(6) (original in Japanese). Waylen, P. and Woo, M.-K. 1983. Annual floods in southwestern British Columbia, Canada. Journal of Hydrology, 62,95-105. Wells, W.G. 1987. The effects of fire on the generation of debris flows in southern California. In Debris Flows/Avalanches: Process, Recognition, and Mitigation. Geological Society of America, Reviews in Engineering Geology, VII: 105-114. 179 Wharton, G. 1995. The channel-geometry method: guidelines and applications. Earth Surface Processes and Landforms, 20, 649-660. Wheeler, J.O. and Gabrielse, H. 1972. The Cordilleran structural province. In Variantions in tectonic styles in Canada. Price, R.A. and Douglas, R.J.W. (eds.). Geological Association of Canada Special Paper, 2, 2-81. Wieczorek, G.F. 1987. Effect of rainfall intensity and duration on debris flows in central Santa Cruz Mountains, California. In Debris Flows/Avalanches: Process, Recognition, and Mitigation. Geological Society of America, Reviews in Engineering Geology, VII: 93-104. Wiezcorek, G.F. and Sarmiento, J. 1988. Rainfall, piezometric levels, and debris flows near LaHonda, California, in storms between 1975 and 1983. In Landslides, floods, and marine effects of the storm of January 3-5, 1982, in the San Franscisco Bay Region, California, U.S. Geological Survey, Professional Paper, 1434, 43-62. Wigmosta, M.S. 1983. Rheology and flow dynamics of the Toutle debris flow from Mount St. Helens. Unpublished M.Sc. thesis, University of Washington, 184 p. Wilford, D.J. and Schwab, J.W. 1981. Soil mass movements in the Rennel Sound area, Queen Charlotte Islands, British Columbia. Unpublished report, British Columbia Ministry of Forests, Victoria. Wilson, R.C. and Wieczorek, G.F. 1995. Rainfall thresholds for the initiation of debris flows at La Honda, California. Environmental and Engineering Geoscience, 1,11-27. Wolman, M.G. and Miller, J.P. 1960. Magnitude and frequency of forces in geomorphic processes. Journal of Geology, 68, 54-74. Wolman, M.G. and Gerson, R. 1978. Relative scales of time and effectiveness of climate in watershed geomorphology. Earth Surface Processes, 3, 189-208. Wood, W.F. and Snell, J.B. 1960. A quantitative system for classifying landforms. Quatermaster Research and Engineering. Command, U.S. Army, Technical Report EP-124. Yamachui, D.K. 1990. A simple method for cross-dating increment cores from living trees. Canadian Journal of Forestry Research, 21,414-416. 180 APPENDIX A. DEBRIS FLOW FREQUENCY DATA This appendix summarizes all available data on debris flow frequency. The following table includes data collected from dendrochronologic evidence, historical documentation, eyewitness accounts, and published sources. The first column contains the creek names that is deciphered in the legend below. The second and third columns indicate the dendrochronologic evidence that was acquired from dating tree scars and reaction wood, respectively. If the debris flow occurred during the growing period, an effort was made to determine the approximate months in which the event most likely occurred (see Chapter 4). Columns 5 and 6 provide other debris flow dates and the source of the date information. Column 7 summarizes all the available information and indicates the most precise date. Column 8 indicates the likelihood that the debris flow actually occurred. "Very likely" is assigned to events with a 100% chance of occurrence when the debris flow was observed or when at least 2 unambiguous tree scars yielded the same date. "Likely" is assigned to debris flows that were dated by only one tree scar. Although it is very difficult to assign proper probability statements, these events are documented as having occurred because the available evidence strongly suggests debris flow impact (e.g. rock clast embedded in cambium). The final column documents the number of replications of each date. Two numbers (e.g. 2+4) indicate that the event was confirmed by two different dating methods. The first number marks dates derived from dating reaction wood, published dates, or eye-witness accounts; the second number represents dates obtained from dating tree scars. n = number of debris flows recorded (100 yrs) = length of period for which frequency was calculated F = frequency RI = return interval (F"1) DS = debris flow occurred during the dormant season ? = evidence does not allow unequivocal determination of month or season J/J = debris flow occurred probably in June or July own obs. = own observations, either witnessed the event or visited the site immediately after the event 181 APPENDIX A. DEBRIS FLOW FREQUENCY DATA Creek Name Scardate Season Reduction Other Dates Source Final dates Likelihood Total Month Date Rep. m COLLIS 1985/86 DS 1985 1985 very likely 3 1981/82 DS 1981/82 likely 1 n=14 (121 yrs) 1978/79 DS 1978/79 likely 1 F=0.115 1950/51 M/J 1950 1950 very likely 1 Rl=9 1967/68 DS 1967/68 very likely 3 1937/38 DS 1931/32 1937/38 1931/32 likely likely 1 2 1935/36 DS 1935/36 likely 1 1921/22 DS 1921 1921 very likely 1 1910/11 DS 1910/11 very likely 2 1894/95 DS 1894/95 very likely 2 1885/86 DS 1872/73 1885/86 1872/73 likely likely 1 2 #2 FERGUSON Aug 1993 own obs. 1993 very likely 1 1986/87 DS 1986/87 very likely 4 n=19 (111 yrs) 1970 J/J 1970 likely 1 F=0.171 1965 J/J 1965 very likely 4 Rl=6yrs 1960 J/J 1960 very likely 2 1958 J/J 1958 likely 1 1954/55 ? 1954/55 very likely 2 1952 M/J 1952 likely 1 1947 J/J 1947 likely 1 1942/43 DS 1942/43 likely 1 1939 J/J 1935/36 1939 1935/36 likely very likely 1 2 1932 M/J 1932 very likely 5 1929/30 DS 1929/30 very likely 2 1927/28 ? 1927/28 likely 1 1918 J/J 1918 very likely 2 1915 M/J 1915 very likely 2 1903/04 ? 1903/04 likely 1 1883/84 DS 1985 1883/84 very likely 2 #3 HOWE 1983/84 DS 1983 1983 very likely 3 1977/78 DS 1977 1977 very likely 2 n=8 (109 yrs) 1945/46 1945/46 very likely 2 F=0.073 1939/40 ? 1939/40 very likely 3 Rl=14 1936/37 ? 1936 1912/13 1936 1912/13 very likely very likely 2 4 1920/21 DS 1920/21 likely 1 1885/86 DS 1885/86 very likely 2 #4 •WILDCAT' 1989/90 ? 1989/90 likely 1 1983/84 DS 1984/85 1984 very likely 844 n=7 (124 yrs) 1955/56 ? 1955/56 1955/56 very likely 1+4 F=0.056 1941/42 DS 1941/42 very likely 4 Rl=18 1917/18 ? 1896/97 1917/18 1896/97 likely likely 1 2 1870/71 ? 1870/71 very likely 4 #5 TWO HILE 1986/87 1983/84 DS DS 1984 1984 Slaymaker et al. '87 1986/87 1984 very likely very likely 2 244 n=6 (67 yrs) F=0.089 Rl=11 1981/82 1962/63 1949/50 1927/28 DS DS DS DS 1962/63 1950 1981/82 1962/63 1949/50 1927/28 likely very likely very likely very likely 1 1 3+2 2+1 182 APPENDIX A. (continued) DEBRIS FLOW FREQUENCY DATA #6 PETERSEN 1986/87 DS 1986/87 very likely 3 1983/84 DS Slaymaker et al. '87 1983/84 very likely 4 n=5 (112 yrs) 1960/61 ? 1960/61 likely 1 F=0.044 1924/25 DS 1924/25 likely 1 Rl=22.4 1882 1882 likely 1 #7 EUREKA 1990/91 DS 1990/91 very likely 2 1983/84 DS 1983/84 very likely 2 n=7 (45 yrs) 1977/78 DS 1977/78 very likely 4 F=0.155 1970/71 DS 1970/71 likely 1 Rl=6 1966/67 ? 1965/66 1966 very likely 1+1 1962/63 DS 1962/63 likely 1 1949/50 DS 1949/50 very likely 2 #8 MNT. LUDWIG 1993 M/J 1993 very likely 3 1985 M/J 1985 likely 1 n=9 (I17yrs) 1971/72 DS 1971/72 likely 1 F=0.077 1968/69 DS 1968/69 likely 1 Rl=13 1960/61 ? 1960/61 very likely 2 1956/57 DS 1955/56 1956 very likely 2+2 1954/55 DS 1954/55 very likely 3 1917/18 ? 1917/18 likely 1 1877/78 ? 1877/78 likely 1 #9 19 Mite Creek 1979/80 DS 1979/80 very likely 9 1973/74 DS 1974/75 1974 very likely 3+3 n=8 (98 yrs) 1971/72 DS 1971/72 very likely 6 F=0.081 1947/48 ? 1947/48 likely 1 Rl=12 1944/45 DS 1944/45 very likely 3 ? 1926/27 1926/27 very likely 2 1915/16 DS 1915/16 likely 1 1896/97 DS 1896/97 likely 1 #10 'Gunbarrel 1" 1991/92 DS 1992 own obs. 1992 very likely 1 1973/74 DS 1973/74 likely 1 n=10(185yrs) 1945/46 DS 1945/46 likely 1 F=0.085 1942/43 ? 1942743 likely 1 Rl=12 1938/39 DS 1938/39 likely 1 ? 1921/22 1921/22 very likely 5 1905/06 ? 1905/06 likely 1 1898/99 ? 1898/99 very likely 3 ? 1883/84 1883/84 likely 1 1876/77 ? 1876/77 likely 1 1859/60 ? 1859/60 likely 1 1853/54 DS 1853/54 likely 1 #11 'Gunbarrel II' 1991/92 A/S 1991/92 very likely 3 ? 1984/85 1984/85 very likely 5 n=17 (154 yrs) 1981 A-J 1981 likely 1 F=0.110 1969/70 ? 1969/70 likely 1 Rl=9 ? 1958/59 1958/59 very likely 7 ? 1948/49 1948/49 very likely 3 1939/40 ? 1939/40 likely 1 1933/34 DS 1933/34 very likely 4 ? 1917/18 1917/18 very likely 4 1902/03 DS 1902/03 very likely 2 1883/84 A/S 1883/84 very likely 2 1873/74 ? 1873/74 likely 1 1860/61 DS 1860/61 likely 1 1867/68 A-J 1867/68 likely 1 1854/55 DS 1854/55 likely 1 ? 1848/49 1848/49 very likely 4 1840/41 DS 1839/40 1840 very likely 1+5 183 APPENDIX A . (continued) DEBRIS FLOW FREQUENCY D A T A _ •Gunbarrel III' 1991/92 A/S 1991/92 very likely 2 1990/91 DS 1990/91 likely 1 n=8 (78 yrs) 1983/84 DS 1983/84 very likely 3 F=0.103 1980/81 M-J 1980/81 likely 1 Rl=10 1964/65 DS 1964/65 very likely 2 1925/26 ? 1925/26 very likely 3 1919/20 1919 very likely 8 1916/17 ? 1916/17 likely 1 #13 1KABOOSE' 1991/92 DS 1991/92 likely 1 1982/83 DS 1982/83 likely 1 n=10 (105 yrs) 1976/77 DS 1976/77 likely 1 F=0.050 1961/62 DS 1961/62 very likely 4 Rl=11 1955/56 ? 1955/56 likely 1 ? 1942/43 1942/43 very likely 4 1927 J/J 1927 likely 1 1914/15 DS 1914/15 very likely 3 1900/01 A/S 1900/01 likely 1 1889/90 DS 1889/90 likely 1 # 14 'SNAKE GULLY' 1979/80 DS 1979/80 very likely 4 n=4 (86 yrs) 1953/54 DS 1953/54 likely 1 F=0.046 ? 1923/24 1923/24 very likely 5 Rl=22 1908/09 DS 1908/09 very likely 2 #15 'FOOL'S GOLD' 1984/85 DS 1984/85 likely 1 1978/79 DS 1978/79 likely 1 n=7 (81 yrs) 1974/75 DS 1974/75 likely 1 F=0.086 1965/66 DS 1965/66 very likely 2 Rl=12 1961/62 DS 1961/62 very likely 6 1952/53 DS 1952/53 very likely 4 1938/39 DS 1938/39 likely 1 1925 1925 very likely 3 1914/15 DS 1913/14 likely 1 #16 •CLEARWATER' 1988/89 7 1988/89 likely 1 1982/83 DS 1982/83 likely 1 n=7 (52 yrs) 1977/78 DS 1977/78 very likely 6 F=0.134 1973/74 ? 1973/74 very likely 2 Rl=7 1969/70 DS 1969/70 very likely 2 1958/59 DS 1958/59 likely 1 1942/43 DS 1942/43 likely 1 1915/16 7 1915/16 very likely 3 1884/85 ? 1884/85 likely 1 1848/49 7 1848/49 very likely 1+1 #17 •POTHOLE' 1985/86 ? 1985/86 likely 1 ? 1973/74 1973/74 very likely 5 n=6 (214 yrs) ? 1924/25 1924/25 very likely 3 F=0.028 ? 1892/93 1892/93 very likely 5 Rl=36 7 1850/51 1850/51 very likely 4 1780/81 ? 1780/81 likely 1 #18 HOTSPRINGS 1991/92 DS 1991/92 likely 1 1989/90 DS 1989/90 1989/90 very likely 3+5 n=27 (202 yrs) 1984 ? 1984 very likely 2 F=0.133 1981/82 DS 1981/82 likely 1 Rl=7 ? 1978/79 Jordan'94 1978/79 very likely 2 1975/76 DS 1975/76 likely 1 1966/67 DS 1965/66 1966 very likely 5+3 1957/58 ? 1959 1958 very likely 2+3 1953/54 1953/54 likely 1 1950/51 DS 1950/51 likely 1 184 APPENDIX A . (continued) DEBRIS FLOW FREQUENCY D A T A #18 1946/47 DS 1946/47 likely 1 HOTSPRINGS 1944/45 ? 1943/44 1944 very likely 1+4 (continued) 1942/43 ? 1942/43 likely 1 1935/36 DS 1935/36 likely 1 1931/32 DS 1931/32 likely 1 1926/27 DS 1926/27 likely 1 1917/18 ? 1917/18 likely 1 1910/11 ? 1910/11 likely 1 1902/03 ? 1902/03 likely 1 1892/93 ? 1892/93 very likely 2 1889/90 ? 1889/90 likely 1 1859/60 ? 1859/60 very likely 2 1848/49 ? 1848/49 likely 1 1844/45 DS 1844/45 likely 1 1829/30 DS 1829/30 likely 1 1811/12 ? 1811/12 likely 1 1791/92 ? 1791/92 likely 1 1733/34 ? 1733/34 likely 1 #19 CAPRICORN 1991/92 DS 1991/92 very likely 3 1988/89 DS 1988/89 likely 1 n=12 (154 yrs) 1972/73 DS 1972/73 1972/73 very likely 4 F=0.078 ? 1968/69 1968/69 very likely 5 Rl=13 1958/59 DS 1958/59 likely 1 1944/45 ? 1944/45 very likely 3 1933/34 DS 1933/34 very likely 2 ? 1909+2 Jordan '94 1909 likely 1 1903/04 DS 1903/04 likely 1 1873/74 DS 1873/74 likely 1 1850/51 DS 1949/50 1850 likely 1 1841/42 ? 1841/42 very likely 3 1669/70 ? 1670 1669/70 likely 1 #20 CANYON 1990 DS 1990 very likely 5 1987 Jordan *94 1987 very likely 1 n=9 (57 yrs) 1983/84 DS 1983/84 very likely 1 F=0.158 1977/78 DS 1977/78 very likely 2 Rl=6 1966/67 1962/63 DS 1966/67 1962/63 very likely very likely 2 3 1950/51 DS 1950/51 very likely 2 1946/47 DS 1946/47 very likely 2 1937/38 ? 1858/59 1937/38 1858/59 very likely likely 2 1770 1770 likely 1 #21 NO GOOD 1995 1993/94 1990 own obs. own obs. Jordan'94 1995 1993/94 1990 very likely very likely very likely ] n=10 (16 yrs) 1989 Jordan'94 1989 very likely 1 F=0.625 1988 Jordan'94 1988 very likely 1 Rl=1.6 1987 1983 1982 1980 1979 Jordan'94 Jordan'94 Jordan'94 Jordan'94 Croft '83 1987 1983 1982 1980 1979 very likely very likely very likely very likely very likely | #22 BOUNDARY 1995 own obs. 1995 very likely 1 n=5 (9yrs) 1993/94 own obs. 1993/94 very likely 1 F=0.556 1989 Jordan'94 1989 very likely 1 Rl=2 1988 1987 Jordan'94 Jordan'94 1988 1987 very likely very likely ! 185 APPENDIX A . (continued) DEBRIS FLOW FREQUENCY D A T A #23 'NIGHTMARE* 1987/88 DS 1987/88 likely 1 1984/85 ? 1984 Jordan '87 1984 very likely 2 n=8 (68 yrs) 1975/76 ? 1975 Hart 79 1975 very likely 2 F=0.118 1971/72 ? 1971/72 likely 1 Rl=9 1960/61 DS 1960/61 likely 1 1950/51 ? 1950/51 likely 1 1931/32 ? 1931/32 very likely 4 1927/28 7 1927/28 very likely 6 ? 1891/92 1891/92 likely 2 1865/66 ? 1865/66 likely 1 1842/43 ? 1842/43 likely 1 #24 MCLEOD 1992/93 1991/92 1992/93 1991/92 very likely very likely 2 1+2 n=10 (210 yrs) 1989/90 DS 1989/90 likely 1 F=0.048 1974/75 DS 1974/75 very likely 2 Rl=21 1933/34 ? 1933/34 very likely 2 1896/97 DS 1896/97 very likely 6 1835/36 DS 1823/24 1803/04 1785/86 1835/36 1823/24 1803/04 1785/86 likely very likely very likely very likely 1 6 4 3 #25 PETERSEN 1991/92 M-J 1991/92 very likely 2 1989 M-J 1989 likely 1 n=15 (54 yrs) 1987 M/J 1987 very likely 3 F=0.278 1983/84 DS 1983/84 very likely 2 Rl=4 1984/85 DS 1984/85 very likely 2 1978/79 DS 1978/79 very likely 2 1974/75 J-S 1974/75 very likely 4 1972/73 1972/73 likely 1 1970/71 DS 1970/71 very likely 2 1965/66 ? 1965/66 very likely 2 1960/61 ? 1960/61 likely 1 1957/58 ? 1957/58 likely 1 1950/51 M 1950/51 likely 1 1944/45 DS 1942/43 1943 very likely 2+7 1940/41 DS 1940/41 likely 1 1918/19 DS 1916/17 1917 very likely 1+3 1915/16 7 1915/16 likely 1 1875/76 ? 1875/76 likely 1 1863/64 7 1850/51 1863/64 1850/51 likely likely ! #26 TURBID 4-Aug 1995 Interfor 1995 very likely 1 29-Jul 1993 1991 own obs. Weldwood 1993 1991 very likely very likely ] n=9 (33 yrs) 1987 Weldwood 1987 very likely i F=0.273 1984/85 13-Jun 1984 Weldwood 1984 very likely i Rl=4 1978 1972 Weldwood Weldwood 1978 1972 very likely very likely ! 1967/68 1967 1963 Weldwood Clague& Souther'83 1967 1963 very likely very likely 4+1 186 APPENDIX A . (continued) DEBRIS FLOW FREQUENCY D A T A #27 TERMINAL 1990/91 DS 1990/91 very likely 1+2 1984/85 DS 1984/85 very likely 1+3 n=11 (119 yrs) 1981/82 DS 1981/82 very likely 3 F=0.092 1967/68 DS 1967/68 very likely 2 Rl=11 1965/66 DS 1965/66 very likely 2 1958/59 • ? 1958/59 likely 1 ? 1942/43 1942/43 very likely 3 ? 1935/36 1935/36 very likely 3 1916/17 ? 1916/17 very likely 2 1909/10 ? 1909/10 very likely 3 1875/76 ? 1875/76 very likely 3 #28 ENDURANCE 1991/92 DS 1990/91 1991 very likely 3+5 1984/85 DS 1984/85 very likely 3 n=7 (85 yrs) 1980/81 DS 1980/81 very likely 4 F=0.082 1962/63 DS 1961/62 1962 very likely 1+1 Rl=12 ? 1934/35 1934/35 very likely 3 1920/21 DS 1920/21 likely 1 ? 1909/10 1909/10 very likely 3 #29 'NO LAW 1991/92 DS 1991/92 very likely 2 1990/91 DS 1990/91 likely 1 n=15 (77 yrs) 1988/89 DS 1988/89 likely 1 F=0.195 1985/86 DS 1985/86 likely 1 Rl=5 1980/81 DS 1980/81 very likely 2 1974/75 DS 1975/76 1975 very likely 2+5 1969/70 DS 1968/69 1969 very likely 1+7 1967 M/J 1967 likely 1 1965/66 DS 1965/66 very likely 2 1961/62 ? 1961/62 very likely 3 1949/50 ? 1949/50 1949/50 very likely 2+4 1937/38 DS 1937/38 very likely 2 1931/32 ? 1931/32 very likely 2 1924/25 DS 1924/25 likely 1 1917/18 M/J 1916/17 1917 very likely 3 #30 •RAINY" 1991/92 DS 1991/92 very likely 2 1976/77 DS 1975/76 1976 very likely 3 n=13 (144 yrs) 1972 M/J 1972 likely 1 F=0.090 1958/59 DS 1958/59 very likely 3+4 Rl=11 1944/45 DS 1944/45 likely 1 1941/42 DS 1941/42 very likely 2 1931/32 ? 1931/32 very likely 1+4 1924/25 ? 1924/25 very likely 2 1917/18 ? 1915/16 1917/18 very likely 3 1887/88 DS 1887/88 likely 1 1876/77 ? 1876/77 likely 1 1854/55 DS 1854/55 likely 1 1850/51 ? 1850/51 1850/51 very likely 3 #31 ROGER'S 1990/91 DS 1990/91 likely 1 1964/65 DS 1964/65 1964/65 very likely 2+1 n=5 (60yrs) 1956/57 ? 1956/57 likely 1 F=0.083 ? 1945/46 1945/46 very likely 3 Rl=12 1932/33 DS 1932/33 very likely 2 1844/45 ? 1844/45 very likely 2 1833/34 ? 1833/34 very likely 9+2 187 APPENDIX A . (continued) DEBRIS FLOW FREQUENCY D A T A #32 •DEEPA' ? 1988/89 1988/89 very likely 5 1975/76 DS 1975/76 very likely 5 n=10 (140yrs) 1967/68 DS 1967/68 very likely 4 F=0.071 1946/47 ? 1946/47 very likely 2 RI=14 1940/41 ? 1940/41 1940/41 very likely 3 1925/26 DS 1925/26 very likely 3+5 1910/11 DS 1910/11 very likely 4 1896/97 ? 1896/97 very likely 2 1884/85 ? 1884/85 very likely 2 1854/55 ? 1854/55 likely 1+1 1776/77 DS 1776/77 likely 1 #33 •MOT. CURRIE" 1994/95 own obs. 1994/95 very likely 1 1989/90 DS 1989/90 1989/90 very likely 3 1986/87 DS 1986/87 likely 1 1981/82 DS 1981/82 likely 1 n=9 (40 yrs) 1978/79 DS 1978/79 very likely 3 F=0.225 1971/72 DS 1971/72 very likely 5+1 Rl=4 ? 1963/64 1963/64 very likely 1+1 1961/62 ? 1960/61 1961 very likely 2 1956/57 DS 1956/57 likely 1 #34 •LAST DAY' 1991/92 DS 1991/92 very likely 2 1981/82 DS 1981/82 likely 2 n=10 (105 yrs) 1974/75 DS 1974/75 very likely 1+1 F=0.148 1968/69 DS 1968/69 likely 1 Rl=7 1960/61 DS 1960/61 very likely 2 1953/54 DS 1953/54 very likely 2 1945/46 DS 1945/46 likely 1 1939/40 ? 1939/40 likely 1 1916/17 ? 1916/17 very likely 2 1890/91 DS 1890/91 likely 1 APPENDIX B. DETAILED DESCRIPTIONS OF THE STUDY BASINS 188 MOUNT MEAGER AREA Boundary Creek and No Good Creek These two adjacent basins are located on the south flank of the volcanic complex (Figure 3.1). Both creeks have been studied intensively by Jordan (1994) and some of the information presented here is derived from his work. Boundary and No Good creeks are discussed jointly because of close geomorphologie and geological similarities. The basins are underlain primarily by early Pleistocene dacitic and rhyolitic lava and pyroclastic deposits which have been hydrothermally altered (Read, 1978). Accordingly, debris flow materials consist mainly of hydrothermally altered volcanic-source material. Similar to other basins on the south flank of the Complex, Boundary and No Good creeks are underlain at greater depth by Mesozoic plutonic basement rocks with some Mesozoic metasedimentary and high-grade metamorphic rocks which crop out in the lower reaches of the basins. The upper basin of Boundary Creek is deeply dissected by steep gullies filled primarily with Fine grained unconsolidated sediments. At an elevation of approximately 1800 m these gullies are continued upwards by chutes bordered by highly unstable lava, volcanic breccias and pyroclastics (Figure A-l). In late August 1995 fieldwork was conducted in the upper parts of the basins. During the two days of fieldwork, rockfall was audible throughout the day indicating very loose bedrock in the upper basin. The ridge between the two basins is double and triple-crested with signs of recent movement indicating slow sackung-like displacement (Figure A-2). Water seepage is visible along a narrow band about 25 m downslope of the ridge on the No Good Creek side, suggesting a perched water table. It is conceivable that slices of the ridge will detach and slide into No Good or Boundary Creek, thereby mobilizing an initial debris mass in the order of 100,000-300,000 m 3 . This volume was estimated on the basis of length and width measurements of the possible slide mass, delineated by the sackung trenches. The thickness of the mass was inferred by extrapolating a semicircular failure surface from the seepage zone to the uppermost sackung trench. Figure A-l. Debris flow source area in upper Boundary creek 189 Upper No Good Creek basin below the debris shedding rock faces consists mainly of a continuous 36° steep talus and debris slope (Figure A-3). During a traverse of the upper basin it was found that the silt and sand size sediment that covers most of the slope is hardened to an extent that indentation by boots was impossible. As in the Boundary Creek basin, the talus slope is increasingly dissected at lower elevations. Field evidence suggests that some debris flows start as surficial runoff mobilizing small debris within the hard, quasi-impermeable layer. At locations where this layer is disturbed, rapid erosion in the underlying coarse sediment can mobilize large amounts of easily erodible material (Figure A-4). Both basins have recorded the highest frequency of debris flows known in the study area. Jordan (1994) described a large debris flow that descended Boundary Creek in August of 1994 which diverted Meager Creek to the other side of its valley, possibly blocking the channel for a 190 short period of time. According to Jordan, air photograph analysis indicated that this debris flow was the largest recorded during the past 15 years. Frequency data are based entirely on reports from logging companies (data provided by Jordan, 1994) and own observations. Unfortunately, the fans of both creeks have been logged and partially slash burnt which rendered them useless for dendrochronological frequency analysis. Figure A-2. Sackungen on ridge between Boundary Creek and No Good Creek, Mount Meager Volcanic Complex. Note yellow tent for scale. 191 In addition, both channels are deeply incised into debris flow diamicton, so that only very large debris flows are able to spill out of the channel before reaching the mouth at Meager Creek. Dendrochronolgic analysis of impact scars or reaction wood along the flow path would be highly biased toward low frequency high magnitude events. Volumetric analysis of debris flow deposits is severely obstructed by the fact that both creeks discharge into Meager Creek, a stream with an estimated annual flood discharge of approximately 100 m3/s (Jordan, 1994). Debris flows tend to occur during rainstorms which often coincide with high discharge of Meager Creek. Debris flows reaching Meager Creek channel are rapidly eroded during snowmelt and fall rainstorms. This leaves an unrepresentative fraction of the original sediment volume at the confluence. For this reason, the relation between Q m a x and total volume, introduced in Chapter 4, was used to infer debris flow magnitudes. Figure A-3. Smooth talus slope and erosion caused by debris flow in upper No Good Creek basin. Note small levees. Channel width is approximately 0.4 m. 192 This observation is in accordance with theoretical considerations of infiltrability and the promotion of overland flow. The rate of overland flow increases with increasing slope and decreases with an increase in roughness, indicating that smooth and steep slopes are favored locations for the occurrence of overland flow. No Good Creek basin as well as Boundary Creek basin have stored so much unconsolidated material that they are classified as supply- or transport limited. Figure A-5 shows the air photographs of the two basins with stability polygons and morphologic features detrimental to slope stability. Figure A-4. Debris flow failure scarp in talus slope, upper No Good Creek basin. Channel width and depth is approximately 5 m and 3 m, respectively 193 Figure A-5. Boundary Creek and No Good Creek basins, on the north flank of Mount Meager. Scale approx. 1:15,000 (airphotograph: 30BCB90053 No.14). 194 Canyon Creek This creek is situated about 8 km to the east of No Good Creek (Figure 3.1). Its mouth is located at the logging bridge traversing Meager Creek. In this drainage, the contact between plutonic basement rocks and the volcanic rocks ranges between 1500 and 2000 metres, which is reflected in the high proportion of basement rock in debris flow sediments. In contrast to No Good Creek, and Boundary Creek, the upper Canyon Creek basin is presently ice-covered. Rapid glacial retreat since the last neoglacial maximum has exposed poorly consolidated pyroclastic rock and debutressed adjacent rock faces of poor rock mass quality. Canyon Creek produced a large debris flow in 1990 which destroyed a logging camp near the mouth of the creek. One large landslide scar is visible on the air photographs on the north side of the basin (Figure A-6). The landslide scar coincides with a linear bench-like feature resembling antislope scarps which are abundant around the Mount Meager volcanic complex (Bovis, 1982). This linear feature extends contour parallel approximately 150 meters to a deeply incised gully. Canyon Creek was classified as weathering-limited because the channel is bedrock controlled over most of its length, and transportable debris is delivered from materials with some degree of cohesion. Capricorn Creek This basin is the largest one studied in the Mount Meager volcanic complex. Capricorn Creek serves as the major drainage of a steep glacial valley incised into andesitic flows and altered quartz diorite basement rock (Read, 1978). Croft (1983) noted closely spaced continuous joint systems within the quartz dioritic basement rock. He observed that the degree of alteration and the fracture intensity increased towards the centre of the Meager volcanic complex. Morainal material mantles the lower slopes and is found in some parts of the valley bottom where it is usually obscured by ample amounts of mass movement debris. Extensive lateral moraines of Neoglacial age parallel the stream on both sides and are presently eroding rapidly. Very large debris flows have descended Capricorn Creek in the past 25 years, some causing temporary impoundment of Meager Creek (Jordan, 1987). Capricorn Creek enters Meager Creek at a right angle, making it prone to channel impoundment. Stratified sand deposits and a 195 Figure A-6. Prominent landslide scar in upper Canyon Creek basin pronounced terrace upstream from Capricorn Creek are indicative of this process. About 2500 m upstream from the confluence with Meager Creek, a large landslide scar extends approximately 700 meter along the east ridge of Mount Meager. The exposed bedrock consists of steeply slope-parallel dipping altered quartz diorite with clay gouge fillings (Croft, 1983). Croft suggested that this landslide with an estimated volume of 10 million m 3 descended the creek in the 1920s (Croft, 1983). Extensive dendrochronologic studies carried out on the fan do not support this date. The date closest to that suggested by Croft is 1933/34. East of the landslide scar, a series of contour parallel lineaments can be found which were interpreted by Croft as tension cracks, but are more likely uphill facing scarps that developed as a consequence to rock slope deformation (Bovis and Evans, 1996). Some of these scarps can also be found on the west side of the landslide scar. Although antislope scarps are known for their slow creeping motion, it is conceivable that they were a contributing factor to the slope instability at this site. Another major 196 source of instability is caused by the deboutressing of unstable rock masses on both sides of the valley which followed glacial retreat after the last neoglacial maximum. There are three deeply incised gullies on the north side of Capricorn Creek which are all capable of producing debris flows in the order of several thousand cubic metres. Debris flows from these gullies can impound Capricorn Creek and cause a debris flow or debris flood which could then descend Capricorn Creek to its mouth and impound Meager Creek. Coarse unstratified deposits on the fan of Capricorn Creek indicate the presence of both debris flows and debris floods. It is very likely that debris flows, debris floods and landslides will reoccur in this basin. Damage can be caused by direct impact of debris against the logging road that traverses the fan, or indirectly by impounding Meager Creek, causing flooding upstream and downstream as a consequence of landslide dam failure (Jordan, 1987). Capricorn Creek was classified as transport-limited because of the existence of several sediment delivering side channels, large amounts of unconsolidated and unvegetated morainal debris, and the large landslide mass in the upper part of the basin. "Pothole Creek" "Pothole Creek" drains a small basin on the north slope of Plinth Peak., and owes its unofficial name to collapsed subsurface pipes in fine volcanic sediments which create strings of crater-like features oriented downslope. The channel itself runs over more than half of its length in unconsolidated material. According to Read (1978), bedrock of Pothole Creek basin belongs to the Plinth Assemblage, consisting primarily of light to medium gray porphyritic rhyodacite flows as well as porphyritic rhyodacite breccia and ash. A striking feature of this basin is its extremely brittle rock which might in part be due to the vicinity of the basin to the volcanic vent of the 2350 yrs. BP eruption, which produced the Bridge River tephra. Clasts struck by a rock hammer disintegrate to pebble and coarse size clasts. The debris fan is very gently inclined (2-3°), indicating debris flows with very low viscosity (Jordan, 1994). Since the basin also produces large snow avalanches that reach part of the fan, only trees on the distal part of the fan were sampled for dendrochronologic analysis. Pothole Creek was classified as transport-limited 197 because the basin consists of extremely brittle easily erodable volcanic rock which is manteled with a thick layer of volcanic debris. Hotsprings Creek Although Hotsprings Creek is not part of the Meager volcanic complex, it will be discussed here because of its vicinity to the complex which seems to have influenced the geotechnical properties of its bedrock. Hotsprings Creek drains a partially ice-covered basin southeast of the volcanic complex (Figure 3.1). The basin is underlain primarily by quartz diorite with minor amounts of quartz monzonite in its uppermost part (Read, 1978). The central part of the basin is a highly dissected bowl which produces large amounts of rockfall which are then funneled into Hotsprings Creek channel originating at the bottom of the steep (35-45°) rockface (Figure A-8). This rock bowl might have been formed initially by a post-glacial landslide which is suggested by the irregular hummocky surface on the western side of the fan as far as Meager Creek. Antislope scarps parallel contours to the north of the basin and are truncated by the bowl. Similarly to Capricorn Creek, these scarps could have contributed to the instability of the rock slope. In contrast to most other sites where Coast Mountain plutonic rock forms the contributing rock in debris flow basins, Hotsprings Creek rock is highly fractured and jointed (Figure A-7) which might be due in part to the recent volcanic activity of Mount Meager. In October 1984, a large debris flow (estimated volume: 50,000 m3) descended Hotsprings Creek and destroyed several vehicles parked at the Hotsprings Creek recreation site. Debris flows from Hotsprings Creek are the dominant natural hazard that threatens this recreation site (Jordan, 1986, 1987). Hotsprings Creek was assigned transport-limited status due to the extremely rapid recharge of sediment from the steep debris chutes, and the existence of extensive talus slopes in the upper basin. 198 MOUNT CAYLEY AREA Terminal Creek Terminal Creek flows into Squamish River approximately 50 km upstream of the town of Squamish. It is one of three basins draining the western flank of the Mount Cayley volcanic edifice. The northern part of the basin consists of andesite, dacite and minor rhyodacite flows, tephras and domes of the Mount Cayley Complex (Green et al., 1988). A broad ridge extending in southwesterly direction in the upper basin is a composite dome with peripheral ice-contact features consisting mainly of porphyritic andesite (Green et al., 1988). Terminal Creek basin illustrates the influence of changing lithology and associated topography on debris production. Most of the sediment supplied to the creek originates from volcanic sources which comprise 55% of the total basin area, indicating that young volcanic rock is a more effective debris supply source than the adjacent plutonic rock. The channel of Terminal Creek is accessible over only one third of its length. A deeply incised canyon stretches from elevation 330 m to about 460 m. Immediately upstream of the entrance of the canyon, the channel widens to about 60 metres. At this location approximately 40,000 m 3 of coarse sediment are stored. Several steep, gullies enter the main creek from the south between 1200 m and 1800 m supplying large amounts of mobilizable sediment. The reason for this asymmetrical sediment supply is fact that the south side of the creek is significantly steeper than the north side (30-40° vs. 15-25°) which is probably due to recent volcanic activity. On the 1:50,000 mapsheet 92 J3, the channel of Terminal Creek dissects the fan through its center. Recently, it has changed to the northern fan boundary. Squamish River at this location is about 100 metres wide and banked by a 5 m high terrace on the western side, so channel blockage by debris flows at this location seems very unlikely. Terminal Creek is classified as a weathering-limited system due to mostly bedrock controlled debris supply mechanism. The large amounts of sediment currently stored in sediment delivering channel point towards an intermediate position of this basin on the weathering-limited, transport-limited spectrum. 199 Figure A-7. Deeply weathered and intensely fractured granitic rock, Hotsprings Creek basin 200 Figure A-8. Air photograph showing the lower, active part of Hotsprings Creek basin. Scale approx. 1:15,000 (airphotograph: BC 7550No. 83) 201 Turbid Creek Turbid Creek is notorious for its frequent and large debris flows and debris avalanches and has been studied to great detail by several researchers (Clague and Souther, 1982; Brooks and Hickin, 1991; Evans and Brooks, 1991; Cruden and Lu, 1993; Lu, 1993; Lu and Cruden, 1996). Turbid Creek drains a steep basin on the southwest side of Mount Cayley. The lower channel is incised in a fan of Holocene origin consisting primarily of debris flow deposits underlain by glacial till (Evans and Brooks, 1991). At least eight debris flows have descended Turbid Creek in the last 25 years, one of which was observed by the author. On July 29, 1993 at 3:45 PM my field assistant and I were working at Terminal Creek, about 3 km north of Turbid Creek, when a logging truck stopped and the truck driver informed us of a "flood" in Turbid Creek. At the time of our arrival, Turbid Creek channel was filled with a rust-coloured, fast flowing sediment slurry, filling the entire channel to a width of about 15 m and a depth of approximately 2.5 m (Figure A-9). Velocity was determined by throwing pieces of wood into the moving slurry and measuring their travel time between two points. Average velocity about 50 m upstream of the culvert was approximately 5 m/s. The debris flow continued for almost 30 minutes discharging about 300,000 m 3 of sediment into the Squamish River which was confined to a channel only 15 metres in width between the quickly developing fan and the rock face on the west bank of the river. The debris arrived in regular surge intervals spaced 25 to 35 seconds apart. The surge waves were moving at a level about 0.3 m to 1 m above the surrounding debris. Boulders of up to 0.5 m diameter and up to 15 m long logs were transported in the slurry. The 2.5 m diameter road was not blocked on this occasion, but the first surge reached the road surface. After about 25 minutes, the debris flow changed to a flow phase resembling hyperconcentrated flow with rapidly declining discharge at the time we left the site. The site was re-visited 2 days later at which time the deposit had not drained, indicating a high clay content and slow consolidation of the debris matrix. According to Evans and Brooks (1991) and Lu (1993), the 1963 and 1984 landslides, as well as several debris flows and rock slides consisted mainly of volcanic tuffs. Lu (1993) investigated several geotechnical aspects of these tuffs. His tests results show low dry densities (13.6 kN/m3), high porosity (35.8%) and very low slake-durability indices (25%). The uniaxial compressive 202 Figure A-9. Debris flow in Turbid Creek, July 29, 1993 strength of dry tuff was determined from uniaxial tests as 1.5-2.4 MPa (average 2.1 MPa) which is significantly different to the 5.4 MPa which were determined by employing the point load method. This difference can be explained by the fact that sampling on the fan was not limited to tuffs, but incorporated andesitic and rhyolitic clasts with significantly higher strength indices (cf. chapter 5). Slaking tests indicate that the tuff rapidly disintegrates into fine particles which easily mix with water forming the slurry required to transport larger particles in suspension (Lu, 1993). Rock slides in upper Turbid Creek are known to have triggered large debris flows. Lu (1993) documented the 1984 and 1963 rock slides. The 1984 rock slide descended Avalanche Creek, a tributary of Turbid Creek causing temporarily damming followed by a catastrophic outflow which triggered a debris flow with an approximate magnitude of 1,000,000 m3 (Jordan, 1987). The 1963 rock slide, investigated by Clague and Souther (1982), was initiated in upper Dusty Creek basin. Temporary damming of Turbid Creek and subsequent outburst caused this 203 debris flow. The possibility of impoundment of the Squamish River, which is confined between a vertical rock face its west bank and the debris flow fan on its east bank, is the main hazard associated with Turbid Creek. Brooks and Hickin (1991) used backwater deposits to date seven or eight prehistoric channel blockages at 4800 BP, 3200 BP and 500 BP, as well as several probably smaller events between 1100 BP and 1955 AD. The 4800 BP and 500 yr BP dates were confirmed by Evans and Brooks (1991) from radiocarbon dates on Turbid Creek fan. The combination of basin morphometry and the poor geotechnical characteristics of the volcanic tuff are responsible for the basin's susceptibility to large landslides and debris flows and strongly indicate transport-limitation of available sediment. Endurance Creek This creek is not part of the Mount Cayley volcanic complex, but will be described here because of its vicinity to the last two sites. Endurance Creek basin is located on the west side of the Squamish River, opposite of Terminal Creek basin (Figure 3.1). The only ground access is by boat across the Squamish River. Endurance Creek drains a small receding icefield on the north-east side of Icecap Peak. The basin consists almost entirely of early middle Jurassic to late Jurassic quartz diorite basement rock as part of the Cloudburst Pluton (Monger and Journeay, 1994). The bedrock in this basin seems to be very competent apart from a small outcrop of loose metamorphic rock and intensely jointed granitic rock approximately 1.5 km upstream from the mouth of the creek. This outcrop is possibly a continuation of a much larger occurrence of Cretaceous metamorphics between here and the Princess Louisa Pluton about 25 km to the northwest. According to Monger and Journeay (1994), these rocks consist of undifferentiated garnet-biotite, kyanite, sillimanite schist, local amphibolite as well as felsic and metavolcanic rocks. The metamorphic rock daylights symmetrical on both sides of the main stream. On the west side, a steeply jointed rock mass shows signs of frequent rockfall and rockslides, whereas the east side is dominated by a series of rotational slumps. It is conceivable that slumps and slides from the rock faces on both sides of the stream can transfer at least several hundred cubic metres of fragmented rock into the creek. The thalweg of Endurance Creek trends south-north with the eastern side of the main channel being 204 much steeper than the western part (35° vs. 20°) because of a steeply dipping joint set on the eastern side. Some debris is supplied from talus slopes beneath rock faces that have recently be uncovered by glacial retreat, but it seems that the majority of sediment is derived from the weak metamorphics rocks and slumps on the east side of the channel. Debris flows in this basin are of moderate mean size (5,400 m3) which suggests a low probability of channel impoundment of the Squamish River at this location. There is evidence for large snow avalanches that discharge on the fan which makes frequency analysis via dendrochronological methods impossible, because scars left by avalanche debris and debris flows are not easily distinguished. However, most debris flows seem to have changed their path to a more easterly direction upon arrival at the fan which is outside the reach of avalanches and therefore suited for dendrochronological analysis. Endurance Creek was classified as weathering-limited because of the comparatively small active area and bedrock control of sediment delivery. UPPER LILLOOET RIVER VALLEY AND ADJACENT SITES Clearwater Creek Clearwater Creek drains a small basin on the north side of Lillooet River valley, near the confluence of Meager Creek and Lillooet River (Figure 3.1). Clearwater Creek is the northernmost of three steep granitic basins with strikingly similar morphometric and geologic characteristics. The dominant bedrock consists of mid Cretaceous quartz diorite (Monger and Journeay, 1994). The creek seems to follow a structurally controlled line of weakness as indicated by the unusually deep incision and very straight creek axis. On the east side of the basin at an elevation between about 1100 m and 1500 m there is a series of about 15 parallel uphill facing scarps which are truncated by the Clearwater Creek incision. The channel itself is filled with several metres of coarse, angular debris at places which thickens in lobe form at places suggesting intermittent debris flow deposition (cf. Chapter 4). Although we were able to ascend the creek only to an elevation of approximately 1200 m, inspection with binoculars indicates that bedrock is 205 more closely jointed at higher elevations where most sediment is supplied to Clearwater Creek. Clearwater Creek was assigned weathering-limited status due to the strong bedrock control of sediment delivery. Peterson Creek The mouth of this creek is located in the upper Ryan River 20 km upstream of its confluence with Lillooet River (Figure 3.1). Peterson Creek is underlain by the middle Jurassic Ryan River Pluton consisting mainly of grano-dibrite rocks (Monger and Journeay, 1994). The creek drains a large (8.6 km2) south-facing basin which is partly ice-covered. Approximately 1.2 km upstream from its confluence with Ryan River, Peterson Creek splits into two glacier fed streams. The presence of large lateral moraines shows that both glaciers have significantly retreated since the last neoglacial maximum, exposing ample amounts of unconsolidated morainal material. The western of the two streams drains a small proglacial lake which acts as a sediment trap for material eroded from the pronounced left lateral moraine. Debris flows and floods have created a large fan which has pushed Ryan River to the northern side of its floodplain. Air photographs suggest that large debris flows reach Ryan River. Upstream of the fan, the channel pattern changes from channelized to meandering in a wider floodplain indicating at least episodic aggradation at present. The western one-third of the fan has been logged making it impossible to apply dendrochronologic analysis. However, analysis of an air photograph chronosequence for this location revealed that the channel has occupied the eastern and central part of the fan, at least for the 50 years of air photograph coverage which supports the coherency of dates collected on this part of the fan. Nightmare Creek Nightmare Creek joins Ryan River just above its confluence with Lillooet River. Nightmare Creek is typical of many small and steep basins in granitic bedrock in the study area. The coarse, bouldery debris consists of quartz diorite with little fines. Debris flows from this basin have been described previously by Jordan (1987) with regard to their potential and actual 206 blockage of Ryan River. The 1975 debris flow event was witnessed by Hart (1979). The danger of channel blockage is further amplified by the fact that Ryan River channel is confined between the debris flow fan and a bedrock bluff. Jordan (1987) observed that Ryan River has a low gradient upstream of the site (0.6°), promoting considerable water volumes to collect behind a potential debris flow dam. Since most debris is removed by Ryan River this site is not suitable for direct measurement of debris flow volume. In the upper reaches of Nightmare Creek basin, poorly preserved late Holocene moraines contribute little sediment to the channel. Most sediment is supplied from the heavily jointed, steep north face of Mount Ross. Large wedge failures and sheet jointing were observed during a site visit to the upper basin in July 1994. A series of north-east striking deeply incised gullies further entrain sediment episodically and deliver it to the main gully. Nightmare Creek is located somewhere in the middle of the continuum between transport-limited and weathering-limited basins, because of the availability of sediment stored in the form of moraines and talus slopes in the upper basin. However, this sediment is not prone to debris flow initiation because of its block-sized clasts, implying either a very high precipitation threshold to trigger a debris flow event, or time for communition of coarser sediment by weathering and erosion. McLeod Creek McLeod Creek basin has been discussed previously in the context of the stability analysis methods in Chapter 5. The upper basin consists of several elongated, steep and active talus slopes which merge into one channel at an elevation of approximately 1400 m. The basin displays an unusual lithologic assemblage for this part of the Coast Mountains. While the lower part of the basin is underlain by early Cretaceous quartz diorite, at elevations of between 400 m to 1000 m bedrock consists of Gambier Group rocks. The main sediment-contributing area of the basin above 1000 m consists of pillowed and massive greenstone, greenstone breccia and tuff of the Triassic Pioneer Formation, which comprises the lower part of the Cadwallader Group (Monger and Journeay, 1994; Rusmore, 1987). McLeod Creek basin is traversed by the Owl Creek fault which forms a steep north-north-west striking reverse fault, that separates Triassic volcanic and 207 related sedimentary rocks to the east from Cretaceous volcanic and sedimentary rocks of the Gambier Assemblage to the west The fault is characterized by a 2 km wide zone of rock breakage with interspersed with slivers of altered Gambier assemblage and Cadwaller terrane volcanics (Monger and Journeay, 1994). A system of conjugate shears and quartz-filled extension fractures are included within the fault zone. A series of parallel linears along the summit ridge of Mount Barbour may be interpreted as being controlled by these shears or extension fractures. Although difficult to prove, it is possible that the original shears are now being remobilized as slow gravitational rock-mass movement, with motion vectors perpendicular to the contour lines (Figure A-10). Several of these linears are truncated by the headscarps within the basin, and there is evidence that rock toppling contributes significant amounts of sediment to the debris flow system. Although no such event has been recorded to date, very large debris flows descending McLeod Creek have the potential of blocking Lillooet River which has shifted towards the southwest as a result of the McLeod Creek debris flow fan. The abundance of large active talus slopes in the upper parts of this basin suggests transport-limitation as discussed in detail in Chapter 6. Mount Currie Creek Mount Currie is located about 6 km southeast of the town of Pemberton. Its northwest face is dissected by three main gullies with fans that abut against with the Green River flood plain. Although all three basins show signs of active debris flows, the western and central gullies were found to be inappropriate for this study, because frequent snow avalanches do not allow the growth of coniferous trees; also, existing deciduous trees and shrubs are scarred by recent flow impacts from those due to snow avalanches. As indicated previously, it is difficult to distinguish impact scars with regard to their scarring process if both avalanches and debris flows occur on the fan. The eastern gully is also affected by avalanches, but the latter do not reach the lower parts of the fan which as a result supports the mature growth of Douglas firs and yellow cedars. Dendrochronologic analysis was therefore confined to the lowest parts of the fan. 208 Figure A-10. Toppling of heavily jointed bedrock at the head of McLeod Creek basin The Mount Currie northeast ridge, which forms the upper limit of this basin, is a steep, asymmetric glacial arete underlain by strongly foliated intrusives of the Pemberton Dioritic Complex (Roddick and Hutchison, 1973; Price et al. 1985; Journeay and Mahoney, 1994). The gully pattern on the northwest face of the ridge are controlled by intersecting northeast and southwest striking joints (Figure A-l 1). Bovis and Evans (1995) monitored rock slope movement along a distinctly visible linear running oblique to the axis of the ridge, consisting of a bedrock scarp and adjacent trench. Linear trenches, tension cracks, grabens and uphill-facing scarps are all signs of active rock slope movement along the ridge. The northwest rock face immediately adjacent to the scarp displays a high degree of jointing. Locally, jointed bedrock has been overturned as much as 90 degrees by toppling (Figure A-12). Rockfall along this steep face delivers large amounts of sediment to the bowl-shaped embayments which funnel into the debris flow gullies. Unlike the two basins to the west, most debris is delivered directly from the rock 209 face to the debris flow gullies without intermittent storage in the form of talus slopes. Mount Currie basin was classified as weathering-limited due to strong bedrock controls in the upper basin, but is located closer to the middle of the sediment availability spectrum because of relatively large amounts of sediments in the lower part of the main channel. "No Law" Creek This creek is one of three creeks producing debris flows on the south slopes of Ipsoot Mountain, located in the upper Rutherford Creek drainage (Figure 3.1). The basin is underlain by rocks of the Lillooet River Intrusion, consisting mainly of middle Jurassic granodiorite. A small glacier in the upper basin is remnant of a neoglacial ice stream that extended several hundred metres further down valley, as indicated by glacial trimlines on air photographs. As a result, ample amounts of glacial sediment can be found on both valley sides above tree line. A pronounced change in slope angle that separates the glacial trough from the rest of the channel suggests that most debris flows are initiated in the steep lower part of the channel. Most sediment is supplied from channel sidewalls consisting of poorly consolidated glacial till (Figure A-13). Alders and willows are abundant in the channel and may play a significant role in stabilizing the channel sidewalls. Once a debris flow with a high instantaneous peak discharge has eroded the shrub vegetation, the channel sidewalls remain unstable until re-colonized. An increase in sidewall erosion will then cause a change in sediment supply to the channel which in turn may lead to a change in debris flow frequency and magnitude characteristics. Channels like No Law Creek may therefore be controlled by positive feedback mechanisms that can be responsible for cycles of higher debris flow activity, although it is unclear to what degree these mechanisms affect debris flow frequency and magnitude. The last debris flow in 1993 stalled within several metres of the left bank of Rutherford Creek, indicating that larger and more mobile debris flows are capable of blocking Rutherford Creek, which at this location is only about six meters wide and confined between the debris flow fan and a steep opposite slope. Catastrophic floods released by dam breakage of a debris flow deposit in Rutherford Creek could affect Highway 99 and the B.C. Rail tracks which traverse Rutherford Creek near its confluence with Green River. 210 Figure All. Air photograph of Mount Currie northwest face. Scale approx. 1:15,000 (air photograph: BC 7478, No. 276) 211 Figure A-12. Overturned toppling bedrock at Mount Currie "Rainy Creek" Rainy Creek drains the basin immediately to the east of No Law Creek (Figure 3.1). Similar to No Law basin, Rainy Creek basin is located within the Lillooet granodiorite intrusion. The channel of Rainy Creek was traversed by a low flying fixed-wing plane in August 1995. It is almost twice as large as No Law Creek basin, but significantly less active. Approximately 50% of the basin are covered by a rapidly receding glacier. Fresh lateral moraines from the last neoglacial 212 Figure A-13. Till failures along No Law Creek advance are visible on the air photographs. Glacial retreat has increased the area that actively contributes sediment to the debris flow system by at least a factor of two. One major difference between N o Law Creek and Rainy Creek, however, is the activity of the channel downstream from the local convexity caused by the glacial trough. Similar to N o Law Creek, channel sidewalls are well vegetated by slide alders, wil lows and small coniferous trees, thus protecting the more unstable channel sidewalls. N o active failure scarps where detected during the overflight indicating a period of quiescence with regard to debris flow activity. However, undercut talus slopes and 213 glacial sediments in the middle sections of the channel point to the potential for slope failures that could initiate a cycle of instability. Neither Rainy Creek nor No Law Creek show significant changes in debris flow frequencies over the past 80 years, suggesting that debris flows occur episodically and that changes in channel stability are reflected by an increase in debris flow magnitudes. Although there is local evidence for this type of behaviour, magnitude data are insufficient to draw firm conclusions. In addition, glacial advances and recessions complicate this model by increasing or decreasing the amount of actively contributing area through changes in glacial coverage. Changes in runoff associated with changes in glacial melt are probably of lesser importance since most debris flows in this area occur during autumnal rainstorms. It is interesting to note that although Rainy Creek and No Law basins are adjacent, the majority of debris flows do not appear to be synchronized. Assuming that precipitation patterns are comparable, this fact suggests a weathering-limited status for these two basins. Rutherford Creek has truncated older debris flow deposits suggesting channel blockage in the past. The existence of very coarse bouldery deposits near the confluence of Rutherford Creek and Green River indicates that large floods, which might have been triggered by the collapse of a debris flow dam, have in fact occurred in the past. "Last Day" Creek The mouth of Last Day Creek is located approximately 5 km east of Rainy Creek (Figure 3.1). With 2.5 km2 area, it is the smallest of the three debris flow basins investigated in Rutherford Creek watershed. Geologically, the drainage is still within the primarily granodiorite rocks of the Lillooet intrusion. However, unlike the other two basins, Last Day Creek is not glacier fed. Sediments for the debris flow system are derived entirely from within the bedrock channel indicating weathering-limitation to material recharge. The channel has no significant changes in slope and the basin has the highest hypsometric integral of the three basins. As pointed out in Chapter 5, high hypsometric integrals are indicative of very steep and little dissected basins with overall convex hypsometric curves, implying a direct sediment transportation along the channel with no intermittent storage. Even small debris flows in Last Day Creek are likely to reach Rutherford Creek, since a continuously steep channel slope angle inhibits debris accumulation. 214 Although the mean magnitude of Rainy Creek debris flows is comparatively small (3000 m3), debris flows are likely to temporarily block Rutherford Creek, resulting in outburst floods as suggested for No Law Creek and Rainy Creek. "Deepa Creek" Deepa Creek is located northwest of Mount Gardiner in the Cayoosh Range (Figure 3.1). It lies within the Spetch Creek Pluton, consisting primarily of middle Cretaceous granodiorite rock. The basin is dissected by two orthogonal joint systems with northwesterly and southwesterly strike. A system of parallel scarps which are a continuation from the southwesterly striking joint set is clearly delineated by a chain of lakes on the plateau northeast of Deepa Creek basin. Large sheet joints along the northwest facing rockwalls with a north-to northeasterly dip of about 50° produce slabs of up to 2 metre in thickness and up to several metres in length. These slabs whose length is controlled by joints that run orthogonal and oblique to the main joint set, detach and disintegrate into smaller particles at impact with the channel. Debris in Deepa Creek is very coarse and even large debris flows are unlikely to travel beyond the powerline corridor. Because of the obvious bedrock source of much of the debris, Deepa Creek is classed as a weathering-limited basins. STUDY SITES IN THE FRASER VALLEY BETWEEN LILLOOET AND LYTTON The study sites between Lytton and Lillooet along the Fraser River show some different topographic and surficial geology characteristics compared to other study sites in the Coast Mountains. Aspects of Quaternary history of the Lytton and Lillooet area have been described by Tipper (1971), Heginbottom (1972), Ryder (1981a), Bovis (1985) and Ryder and Church (1985). In the Lillooet area, the Fraser River has cut deeply along the fault zone which separates Fountain Ridge and the Clear Range of the Interior from the Coast Mountains (Duffell and McTaggart, 1952). Following deglaciation, the Fraser River became entrenched in drift, bedrock and occasional rock slide debris along the valley (Ryder and Church, 1985). Holocene terraces and truncated large debris flow fans which developed in response to deglaciation are found abundantly 215 on both sides of the valley. Glacial drift, early Holocene debris flow fans as well as bedrock, are the determinant sediment inputs for debris flows in this area. "Gunbarrel I, II, III" These three small basins are discussed together here because of their morphometric and geotechnical similarities. The three basins are located approximately 3 km east of the town of Lillooet on the east side of the Fraser River. The rocks in these three basins consist primarily of argillites of the early Cretaceous Jackass Mountain Group and middle Jurassic Relay Group. Interspersed are sandstones, siltstones and shales as well as occasionally conglomerates. The near vertical rock faces above the fan are intensely fractured creating a typical grain size of debris flow deposits of pebble and cobble size. Sediment derived from the basins coalesced to an extensive composite fan overlying early Holocene debris flow deposits and fluvial gravels. Late Pleistocene and early Holocene fans in this region were studied previously by Ryder (1971). Jordan (1994) studied debris flow dynamics of Gunbarrel debris flows and concluded that their matrix resembled volcanic source-area materials. The matrix content of the debris is 32% with clay content of 9% (Jordan, 1994). Clay in this basin is probably an alteration product of shale and argillite . communition. A recent debris flow in 1993 had a runout distance of 300 m on a slope of 3°. The deposit covered several hundred square metres with a fairly uniform thickness of sediment, supporting Jordan's observation that debris flows at this site can be compared to volcano-genetic debris flows with similar depostional characteristic. Pronounced levees consisting of coarser sediments than the material deposited on fans can be found along all debris flow paths in this area. Jordan (1994) suggests that this depositional behaviour might be due to the relatively low water content of debris flows. This fan is an excellent site for the application of dendrochronological method to date debris flows since trees within the depositional area are damaged but not destroyed, and since partial stem obliteration allows the dating of reaction wood. Older trees (primarily Ponderosa Pine) on the fan have been selectively logged on the fan, but well preserved stumps allowed the dating of older scars by cross-dating with living trees. 216 Gunbarrel I basin has developed by incision through stratified talus slopes, but the reason for the initiation is unknown. Presently, backward and lateral erosion in the upper bowl-shaped basin is exposing unconsolidated fine stratified sediment, thus effectively increasing the contributing area, and creating a positive feedback mechanism which may result in a change in the frequency and magnitude behaviour of the basin if the exposed slopes are not self-stabilizing. The large amounts of erodable sediment in the source area and along the channel strongly suggest transport-limitations for this basin. Figure A-14. Debris flow diamicton and talus slope at Gunbarrel 11 Gunbarrel II and III are deeply incised into older debris flow deposits, probably dating from the early Holocene, when glacial deboutressing and uncovering of unconsolidated sediment caused a period of increased mass movements subsequently referred to as the paraglacial cycle (Church and Ryder, 1972). The causes of the incision and reactivation of these debris flow 217 deposits are speculative, but there is evidence that an increase in water discharge in these basin could have caused an initial disturbance that was amplified in the manner described above. The small ephemeral stream draining a steep bedrock gully above Gunbarrel II has eroded the old debris flow diamicton in the upper part of the basin to such an extent, that there is now an undercut about one hundred metre in length of a talus slope (Figure A-14), which as a result has increased its toe slope angle from 34° to 36°. This talus slope is typical of talus found along Fountain Ridge. It consists of a surficial layer of boulder size angular clasts underlain by weakly stratified loose sediment of pebble, sand and silt size. The initial disturbance and subsequent removal of the coarse surficial layer by undercutting is followed by rapid erosion of the underlying material creating transport-limited conditions in this basin. The active talus slopes above Gunbarrel II are constantly raveling, indicating a high degree of geomorphic instability and rapid additional sediment contribution into the main channel. The onset of talus slope undercutting cannot be determined directly, because it predates the earliest air photographs. Black and white photographs displayed in the Lillooet Museum show that the talus slope was eroding in the 1930s. Snake Gully Snake Gully is a small ephemeral creek on the western slope of the Clear Range, located between Lochore Creek and McGillivray Creek (Figure 3.1). Bedrock in the source area of Snake Gully consists mainly of highly altered (zeolitized and fractured) granitic and dioritic rocks of the Mount Lytton Complex. These rocks have very low strength, as determined from point load tests of fist size clasts (cf. Chapter 5). In this region, faults are splayed off into four main slices in northwesterly direction away from the Fraser fault system (Monger, 1989). Just north of the basin, the main strand of the Fraser Fault leaves the Fraser Canyon, now trending in a northerly direction through the Fountain-Cinquefoil valley to rejoin the Fraser Valley 9 km north of Lillooet (Monger and Journeay, 1994). Parts of the basin are strongly dissected by dry debris chutes which are fed by disintegrating bedrock which suggests weathering-limited conditions in this basin. Small undercut talus slopes contribute fine sediments to the channel by frequent raveling. Figure A-15 illustrates the relatively small source area of Snake Gully, which comprises only 10% of the total basin area, 218 and helps to explain the low volumes discharged during typical debris flow events. Debris flows in Snake Gully are confined in a narrow channel and have discharged onto Highway 99, which traverses the fan. "Fool's Gold Creek" Fool's Gold Creek is situated between McGillivray Creek and Laluwissen Creek. The basin which is dominated by dioritic and volcanic rocks is part on the west flank of the Clear Range. There is a sharp transition between the altered granitic and dioritic rocks of the Mount Lytton Complex and the yellowish overlying Spence Bridge volcanics marked by a fault contact in the upper basin (Figure A-16). Clasts of both units were sampled to test rock strength. The point load strength index of the lower unit is rather low (4.0 MPa), and the strength of Spences Bridge volcanics could not be assessed by point load tests, because clasts disintegrated while being placed in the point load tester. Debris recharge rates to the channel seem to be rapid which can partially be explained by the low rock strength in the basin. Repeated site visits have documented the notable increase in size of several small sediment fans that extend from talus slopes into the channel. In contrast to Snake Gully, the greater part of the basin contributes sediment to the debris flow system (62% active area), and although its basin area is about half the area of the Snake Gully, the annual debris yield exceeds that of Snake Gully by at least a factor of 20. This underlines the importance of the debris contributing area factor. The upper basin of Fool's Gold Creek can be easily accessed via a logging road that follows McGillvray Creek and then forks off to climb close to the ridge between McGillvray and Fool's Gold Creek. Access of the middle reaches of the channel by following the creek upwards is hindered by a 15 metre high waterfall 200 metres upstream from the fan apex. Post-glacial incision of the Fraser River has left a marked terrace along its eastern banks. This river terrace and overlying colluvial deposits are deeply incised by a series of gullies which are probably remnant of older debris flow channels. A catchment dam was constructed above this incision to avoid road obliteration by debris flows. This catchment basin allows direct calculation of future debris flow volumes. At the time of writing, the dam was found to be filled with debris 219 Figure A-15. Snake Gully debris flow basin. Scale approx. 1:15,000 (air photograph: 30BCC 92010, No. 8) 220 to within three metres below of its crest, leaving space for only about 8000 m 3 of additional debris. Low rock strength and abundant fine-grained talus slopes in the upper basin suggest transport-limited conditions in this basin. Figure A-16. Contact between Spetices Bridge Volcanics and granitic rocks of the Mount Lytton Complex at Fool's Gold Creek "Kaboose Creek" Kaboose Creek is located about 8 km north of the town of Lytton, and is the central basin of three equally sized basins that drain an amphitheater-shaped bowl in south-southwesterly direction. The basin is wedged between the Babine Fault to the east and the Fountain Fault to the west. Bedrock consists primarily of highly altered granitic and dioritic rock of the Mount Lytton Complex. The granitic and dioritic rocks are unstable to the degree that even weak hammer blows cause complete clast disintegration into coarse to medium sized sand. Grusification of bedrock is observed in many areas. The dominating joint set is crossed by 5-6 randomly oriented cross-joint 221 sets. These steeply dipping sheet joints in the upper basin deliver large slabs of material to the channel and are the primary sediment delivery mechanism indicating weathering-limited conditions (Figure A-17). Characteristic soccerball-sized clasts that were rolled off the rim of the basin, not only descended more than half of the length of the basin, but also mobilized large amounts of additional sediment. This observation indicates that during heavy summer rainfall even small initial failures are able to mobilize substantial amounts of debris before the resulting debris flow becomes channelized. Figure A-17. Steeply dipping joints at Kaboose Creek basin 222 STUDY SITES IN THE HOPE-CHILLIWACK AREA Pattersen Creek Patterson Creek drains a comparatively large (6.1 km2), steep, forested basin with a local relief of 1500 metres on the north flank of Four Brothers Mountain. The creek's mouth is located approximately 1 km west of the Wahleach power house (Figure 3.1). The basin is located within the Mount Barr Pluton, which consists of Miocene granodiorite (Monger and Journeay, 1994). Unlike many small basins along the northern slopes of the Cascade Mountains in this area, the lack of a deep linear incision suggests that Pattersen Creek is not structurally controlled. Pattersen Creek forks off into a western branch at an elevation of 400 m draining three small lakes in the upper basin, and an eastern branch which extends to an elevation of approximately 800 m. A debris flow of approximately 30,000 m3 descended Pattersen Creek in July 1983 obliterating the Trans Canada Highway and the CN Rail tracks (Evans and Lister, 1984). Logging in the upper watershed occurred in the 1960s and 1970s, and air photographs taken immediately after the event indicate that at least three failure scars were associated with overgrown logging roads. A closer examination on the ground revealed that logging road fill slopes had failed and the fill material entered Patterson Creek at oblique angles, and mobilized accumulated saturated sediment in the main channel (Bovis and Dagg, 1988). Although most of the debris flow volume was derived from channel storage which suggest weathering-limited conditions, the failed road fills seem to have initiated the debris flow which then scoured the channel. Two small unnamed creeks immediately to the west of Patterson Creek contain similar failures. In response to the 1984 debris flow, a catchment basin with a capacity of approximately 40,000 m3 was constructed, 200 metres upstream of the highway. Mount Ludwig Creek The fan of Mount Ludwig Creek is located approximately 3 km west of Cheam View along the Trans Canada Highway (Figure 3.1). The creek follows the geologic contact between Cretaceous Slollicum Schist on the north east side, and Miocene granodiorite of the Mount Barr Batholith on the south west side of the basin. The geologic contact is defined by a straight 223 northwest-southeast trending gully which continues through a deeply incised notch approximately 200 metres northeast of Mount Ludwig (Figure A-18). The furrow then continues on the southeast side of Mount Ludwig as a tributary of Wahleach Creek (Figure A-19). This type of northwest-southeast striking linear is very typical of the mountain front between Cheam Peak and the town of Hope. Monger (personal communication, 1996) suggested that major faults in the area have been active well into the Tertiary period, and concurs that many linear features may be associated with faulting, although evidence of shearing is difficult to find. Faulting, particularly along geologic contacts is likely to weaken the rock structure along the plane contact. Mount Ludwig Creek itself is extremely steep (37°), and filled with very coarse debris. Observations during a helicopter flight revealed large slab failures and rock toppling in the upper channel indicating weathering-limitations for this basin. The coarseness of stored debris and the small basin area, which limits the available water supply discharge, suggests a high threshold for debris flow triggering. Given the thickness and coarseness of the accumulated debris in the upper channel, it is unlikely that debris flows are triggered by critical water discharge. The only conceivable initiation mechanism is the detachment of a failed rockmass from the adjacent rockslopes, and impact into the channel leading to impulsive loading and debris mobilization. Gully deepening by fluvial erosion has likely decreased slope stability in the basin due to undercutting of the fault-broken rock. In 1992 a debris flow of approximately 20,000-25,000 m3 descended Mount Ludwig Creek, obliterating the Trans Canada Highway. This debris flow was deflected by a natural dam across the old channel, approximately 400 m upstream from the highway. This deflection caused a change in direction of about 100 degrees and directed the debris flow towards the road, about 500 metres east of the old channel. At least 10,000 m 3 sediment accumulated on the Highway and blocked the traffic (Jones, personal communication, 1995). 224 Figure A-18. Mount Ludwig debris flow channel along pronounced linear furrow defined by fault zone 225 Figure A-19. Air photograph of Mount Ludwig Creek. Scale approx. 1:20,000 (air photograph: 30BC 83017, No. 152) 226 Wildcat Creek Wildcat Creek is located in Silverhope Valley, 2 km upstream of the village of Silverhope. The basin is underlain by Miocene granodiorite of the Chilliwack Pluton. Wildcat Creek has a small contributing area and extensive talus slopes in the upper basin appear to be inactive at present indicating that the basin is presently weathering-limited. The upper basin is traversed by the Hope Fault striking north-south. This has created a furrow which acts as a debris trap for the main channel which crosses the fault at a planimetric angle of 43° at an elevation of 700 m. Mass movements from Wildcat Creek have created a fan of considerable size (~30,000 m2) indicating the geomorphic activity of the basin. An important feature in Wildcat Creek is an old silver mine which was operated in the 1930s at an elevation of about 500 m (Hope Museum archives). Mine tailings with a total volume exceeding of several thousand cubic metres were discharged into the stream, thus adding considerable amounts of pebble to boulder size material to the system. The mine tailings have since been undercut by Wildcat Creek, but it seems that the majority of mined sediment is still in place. A storm in December 1995 which caused several other debris flows in the Hope area, did not initiate a debris flow in this basin, but caused high water discharge and bedload transport which partially destroyed the culverts and damaged the road surface. A recent failure scar is visible on the south slope of Wildcat Creek at an elevation of approximately 400 metres. At this location a shallow translational rock slide detached along prominent, steeply dipping granitic sheet joints sometime between 1983 and 1992, as is evident on air photographs. The debris mass entered the main channel at a 90° angle, but does not appear to have initiated a debris flow since the rockslide material still rests in the channel. Water is flowing through the coarse debris and no signs temporary impoundment are visible. Eureka Creek Eureka Creek drains a large (10 km2) elongated hanging valley approximately 600 metres south of Wildcat Creek (Figure 3.1). Large parts of the watershed were clearcut in the 1980s, and poor logging and road building practices seem to have introduced additional sediment to the channel. The middle part of the basin has a low gradient (<10°), which is anomalous for debris 227 flow channels since such slope angles usually promote the deposition rather than transport of coarse debris. Debris flow initiation processes in Eureka Creek are therefore most likely based on mobilization of debris in the steeper lower part of the channel (20-25°) by either debris slides or by exceedance of a critical discharge (VanDine, 1985). A large blocky debris flow deposit of approximately 5000 m3, laid down on both sides of the road, probably represents only a fraction of the original debris which was discharge into Silverhope Creek and removed by fluvial erosion. Silverhope Creek is confined in a narrow bedrock channel at its confluence with Eureka Creek which renders this site susceptible to channel impoundment. Channel blockage, and subsequent outbreak of water through breaching of the natural dam, could have devastating consequences for the village of Silverhope located on the floodplain several kilometres downstream of the stated confluence. Eureka Creek was visited shortly after the December 1995 rainstorm, and a debris flow of an estimated volume of 5000-10,000 m 3 was noted to have passed under the road bridge without causing significant damage. Field evidence suggests that the debris mass then blocked at least parts of Silverhope Creek and diverted its flood waters into the road embankment which subsequently undermined the road surface causing it to collapse. Eureka Creek was assigned weathering-limited status due to a lack of well defined source areas with abundant sediment supply. Two-Mile-Creek Two-Mile Creek is one of four steep debris flow producing creeks descending the north flank of Hope Mountain (1838 m) (Figure, 3.1). Highway 3 crosses the creek about 1 km east of Hope. The local bedrock consists of early Tertiary granodiorite and monzogranite, and on the west side of the basin, metamorphic rock belonging to the Custer Gneiss Assemblage. Two Mile basin was examined by Bovis and Dagg (1992) with regard to the mechanism that led to the initiation of the January 1984 debris flow. According to Slaymaker et al. (1987) a small event in the upper west branch of Two Mile Creek in 1983 entered the main channel at an oblique angle (55°) thereby transmitting energy directly down channel. This caused a debris surge 228 that traveled almost one kilometre, but deposited in the channel near apex of the debris cone. In January 1984,40,000 m 3 of rock slid into the upper western tributary, and half of the material was deposited at slopes between 34° and 38°. The other half of the mass descended the main channel, triggering two debris flow surges which flowed the entire length of the channel. These events damaged the main beams of the highway bridge and discharged into Coquihalla River. Large, steeply dipping joints in the upper basin control the failure of individual slabs and larger rock masses. The rock mass that initiated the 1984 debris flow entered the channel at a very low angle, which directed considerable thrust down channel (Bovis and Dagg, 1992). The lack of fines of the landslide materials causes high flow resistance, so that large amounts of water or momentum transfer into accumulated channel debris are required to initiate debris flows in this environment. It is notable that during the two storm sequence in November/December 1995, 2-Mile Creek was the only creek on the north slope of Hope Mountain that did not produce a debris flow. Two Mile Creek remains a very hazardous debris torrent channel with the potential to damage the highway and the capability of temporarily blocking Coquihalla River. This basin is a classical example for weathering-limitations because of strongly bedrock controlled sediment supply mechanisms. Outram Creek Outram Creek drains a 4.9 km2 large basin at the entrance of Manning Park (Figure 3.1). Bedrock in the upper part of the basin consists of granodiorite and monzogranite of the early Tertiary Mount Outram Pluton which has intruded the underlying Carboniferous to middle Jurassic Hozameen Complex consisting of radiolarian cherts, pelite, mafic volcanic rocks, minor carbonate, and gabbro and associated ultramafic rocks (Monger and Journeay, 1994). The lower channel incised through a thick glacial till blanket. The upper channel branches off into several small tributaries which drain the southwestern slope of Mount Outram. These upper channels all show signs of frequent snow avalanche activity, as indicated by well developed runout paths overgrown by shrubs and slide alders. It is interesting to note that the occurrence of snow avalanches seems to have a mitigative effect on debris flows, at least along those channel reaches affected by avalanches. Snow avalanches suppress the growth of mature coniferous trees and promote the 229 development of small dense shrubs such as alders and willows. In the paths of less frequented avalanches, small coniferous trees grow in very dense stands. These trees have a krummholz-like appearance and are very resistant to disturbance. Their dense stands successfully suppress local sidewall erosion. There is no obvious sediment link between the upper talus slopes and the adjacent channels. The majority of sediment seems to be derived from the channel sidewalls of the lower channel which is reflected in the rock composition on the fan and the creek bed, and suggests weathering-limited conditions for this basin. A well developed fan extends to the highway, where a small berm protects traffic from debris flows. A certain risk is produced by the fact that the debris flow deflection berm directs flows towards a parking and picnic area at the entrance of Manning Park. STUDY SITES AT OTHER LOCATIONS Collis Creek Collis Creek basin is located on the east side of Hurley River, approximately 20 km south of the town of Goldbridge (Figure 3.1). Several northwest-southeast striking thrust faults have been mapped south of the basin (Monger and Journeay, 1994). The area to the east of the main channel is underlain primarily by phyllites, sandstones, siltstones and volcanic tuffs, basaltic flows and volcanic breccias of the Jurassic Cayoosh Assemblage. The western side of the basin is underlain by undifferentiated conglomerates, sandstones, siltstones, shales and volcanic flows of the Cretaceous Taylor Group. The basin is atypical for most debris flow basins in the study area with regard to size (10 km2) and the overall slope (21°) which emphasizes that these measures alone cannot adequately describe the susceptibility of a basin to debris flow occurrence. The apparent activity of Collis Creek basin is largely explained by the proximity of a steep (30°) sub-basin close to the basin outlet, 85% of whose area is contributing sediment. In the remainder of the basin only 15% of the total area is actively contributing sediment to the channels. Slope instability in the southern part of the basin is caused by phyllite rock dipping at 45° towards the main channel, which promotes frequent rock slides. The southeastern comer of the active sub-basin has experienced a recent rockslope failure which was identified from the air photograph 230 (BCC 93096 No. 193) and confirmed in the field. Extensive talus slopes below the headwall of this failure are feeding material into the channel which suggests an intermediate position in the sediment availability spectrum, but still within a weathering-limited environment. Although Collis Creek receives sediment from other sources along its main stem channel, the sediment input from the steep sub-basin seems to be the most active. Well developed levees along its channels suggest frequent debris flows. Unlike debris flows from sub-basins further upstream Collis Creek, those from the lower sub-basin are more likely to maintain mobility due to the steepness of its channel and the oblique angle of confluence with the Collis Creek main stem, which promotes continuing mobility and entrainment of water in the moving debris mass. Roger's Creek Roger's Creek is a large (4.9 km2) basin with an average slope of 21° (Figure 3.1). A saddle about 800 m west of the peak follows the contact between the greenstone and chert successions of the Bridge River Complex to the southeast and laminated siltstones and turbidites of the Cayoosh Assemblage to the northwest. The contact dips to the southeast and is structurally inverted along the overturned limb of a southwest-verging syncline (Monger and Journeay, 1994). This syncline is cut by a system of southwest-directed thrust faults. The basin slope might suggest only moderate debris flow hazard, but this is misleading because a large proportion of debris flow material is derived from the very steep (35°- 45°) and highly dissected south face of an unnamed peak 2.5 km north of the creek mouth. Another major source of sediment is an outcrop of massive glacio fluvial sediments approximately 1.5 km upstream from the fan of Roger's Creek. Continued downcutting of the creek at this location causes instability and the addition of sediment to the channel. The Duffy Lake road which connects the towns of Lillooet with Pemberton, traverses the debris flow fan and is exposed to a potential hazard from debris flows, as is a Forest Service campground on the north side of Cayoosh Creek. Past debris flows have reached Cayoosh Creek, as indicated by debris flow diamicton in the undercut north bank of the creek. The Ministry of Highways has recognized the problem of potential highway obliteration and an 8 m high deflection dam was built about 400 m upstream of the road which now conveys debris into Cayoosh Creek 231 100 m downstream from the highway bridge. Roger's Creek was assigned weathering-limited status because of recharge control from the dissected south face and the glacio fluvial sediments. Fergusson Creek Fergusson Creek basin is located approximately 5 km north of the town of Bralorne (Figure 3.1). The Creek drains a steep 34° basin consisting primarily of undifferentiated cherts, gabbro, pelite and mafic volcanic rocks and talc-carbonate schists of the Bridge River Complex. The upper basin contains unstable talus slopes that are fed by rockfall from highly jointed bedrock dipping at (50-60°) towards the east (i.e. toward the main channel). The east side of the basin truncates a series of parallel uphill facing scarps extending several hundred metres to the east. Toppling associated with these scarps is contributing additional boulder-size material to the debris flow source area. Frequent snow avalanches descend the south face of Mount Fergusson. This fact did not seem to be a significant problem because Fergusson Creek is deeply incised into the upper fan with the result that debris flows travel beyond the reaches of avalanches. To avoid confusion with tree scars caused by avalanches, dendrochronology was limited to the lower parts of the fan which are not affected by avalanches. The low runout angle of the debris flows and the thin layer of debris flow lobes with uniform thickness, both indicate highly mobile debris flows from this basin. Fergusson Creek basin classifies as transport-limited due to the large amounts of transportable sediments in the upper part of the basin (Figure A-20). Howe Creek Howe Creek drains a small (0.8 km2) basin on south side of Carpenter Lake and approximately 15 km east of the town of Gold Bridge (Figure 3.1). The basin is located within the Bridge River complex with a distribution of rock types similar to those found in Fergusson Creek basin, located 15 km south of Howe Creek. Siltstones are the dominant rock type in the upper basin and are randomly jointed with narrow joint spacing and deeply weathered. In the lower section of the actively eroding part of the basin, a dominating joint set dips approximately 50° to the north and promotes failures along the bedding planes. Rock creep is suggested by a series of 232 uphill-facing scarps just to the west of the basin, many of which are truncated by the Howe Creek basin headwalls. Clasts with 50 cm diameter that were rolled from the escarpment completely disintegrated to less than fist size particles during saltation in the channel indicating the rapid communiuon process during rock fall. Howe Creek was classified as weathering-limited due to its relatively small active area (25%) and bedrock controlled debris supply mechanisms. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share