Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Identification of natural and logging-related landslide in the Capilano River basin (coastal British… Brardinoni, Francesco 2001

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

Item Metadata

Download

Media
831-ubc_2001-0330_fixed.pdf [ 12.15MB ]
Metadata
JSON: 831-1.0090429.json
JSON-LD: 831-1.0090429-ld.json
RDF/XML (Pretty): 831-1.0090429-rdf.xml
RDF/JSON: 831-1.0090429-rdf.json
Turtle: 831-1.0090429-turtle.txt
N-Triples: 831-1.0090429-rdf-ntriples.txt
Original Record: 831-1.0090429-source.json
Full Text
831-1.0090429-fulltext.txt
Citation
831-1.0090429.ris

Full Text

IDENTIFICATION OF NATURAL AND LOGGING-RELATED LANDSLIDE IN THE CAPILANO RIVER BASIN (COASTAL BRITISH COLUMBIA): A COMPARISON BETWEEN REMOTELY SENSED SURVEY AND FIELD SURVEY by Francesco Brardinorii Diploma di Laurea, Universita Ca Foscari di Venezia, 1999 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master of Science in THE FACULTY OF GRADUATE STUDIES (Department of Geography) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA August 2001 © Francesco Brardinoni, 2001 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 The University of British Columbia Vancouver, Canada • Date Qjl} AUfrDST 2)001 ABSTRACT In the Pacific Northwest landslide inventories are routinely compiled by means of aerial photo interpretation. When examining photo pairs the forest canopy, notably in old-growth forest, hides a population of "not visible" landslides. The present study attempts to estimate how important is the contribution of landslides not detectable from aerial photographs, to the global mass of sediment production from mass failures on forested terrain of the Capilano basin. To achieve this, aerial photo interpretation has been coupled with intensive fieldwork for identification and measurement of all landslides. In order to minimise bias in the comparison and integration of field-collected and air photo-collected data it was decided to define a 30-year time window. Incidentally, it has been possible to prove how landslide scars that appear on a single photo set would date further back than 30 years. Results show that "not visible" landslides can represent up to 85 percent of the total number of failures and can account for up to 30 percent the total volume of debris mobilised. Rates of sediment production differ greatly (one order of magnitude) between two sub-basins of the study area, suggesting that such figures should be generalised with care within a physiographic region. The difference in denudation rate is explained qualitatively by GIS-based analysis of slope frequency distributions, drainage density and spatial distribution of surficial materials. Fieldwork has demonstrated that gully-related failures have a greater importance than one could expect from air photo interpretation. ANOVA and nonparametric tests indicate that careful logging in East Cap Creek has produced no detectable effects on mass wasting. Similarly, Sisters Creek, where timber harvesting stopped about 20 years before the start of our 30-year time window, has apparently recovered from the signs of past extensive logging. The existence of "not visible" events affected in a minor way conclusions about the impact of logging on slope stability in terms of land use (management) effects. It had a major impact on the nature of landslide magnitude-frequency relations and, finally is demonstrated to have implications for British Columbia Terrain Stability Classification from the terrain sensitivity point of view. u CONTENTS ABSTRACT ii List of Tables v List of Figures vii List of Photos ix Acknowledgements x Chapter 1 INTRODUCTION 1 1.1 Aims 1 1.2 Hypotheses 3 1.3 Previous Studies on the Main Topics 4 1.3.1 Sediment Production from Landsliding and Impact of Logging 6 1.3.2 Landslide Visibility through Time 12 1.3.3 Landslide Magnitude-Frequency 14 Chapter 2 METHODOLOGY 16 2.1 Statistical Analysis 20 2.2 GIS-Based Analysis 22 Chapter 3 STUDY AREA 26 3.1 Physiography, Bedrock Geology and Geomorphology of the Watershed 28 3.2 Overview of Regional Climate 32 Chapter 4 RESULTS 37 4.1 Landslide visibility from aerial photographs 37 4.2 Landslide frequency and associated sediment production initiation 40 4.2.1 Descriptive Statistics 40 in 4.2.1 ANOVA and Nonparametric Tests 45 4.3 The Impact of Forest Management on Mass Wasting 49 4.4 GIS as a Tool for Justifying Inter-basin Variability of Denudation Rates 56 4.5 Landslide Magnitude-Frequency Relations 64 Chapter 5 DISCUSSION 74 5.1 Landslide Visibility from Aerial Photographs 74 5.2 Landslide frequency and associated sediment production 80 5.3 The impact of timber harvesting on landslide initiation 84 5.4 GIS-based analysis: a Tool for Explaining Patterns of Sediment Production from Mass Wasting 88 5.5 Landslide Magnitude-Frequency Relations 92 Chapter 6 CONCLUSIONS 96 BIBLIOGRAPHY 103 Appendix A. Terrain Stability Assessment in British Columbia 112 Appendix B. Types of Mass Movement Related to Timber Harvesting 121 Appendix C. Fieldwork Checklists 122 Appendix D. Precipitation Data Tables 124 IV List of Tables 1.3.1 Impact of timber harvesting on landslide density and related rate of denudation (modified after Robison et al., 1999) 7 2.1 Principal failure attributes 19 2.2 Dependent variables and factors of Location-Survey analysis 21 2.3 Dependent variables and factors of Location-Stability-Land use analysis 22 3.1 Physical characteristics of the study areas 28 3.2.1 Climate stations in or near the GVRD Watersheds 33 3.2.2 Water Survey of Canada gauging stations in the Capilano Watershed 34 4.2.1 Denudation rates from landsliding and landslide densities during a ~30-yr window (1968-2000) as obtained from air photo interpretation coupled with intensive ground checking 42 4.2.2 Denudation rates from landsliding during a ~ 30-yr window (1968-2000) as obtained from air photo interpretation 42 4.2.3 Denudation and landslide density ratios 43 4.2.4 Sediment delivery to streams (stream connection) 44 4.2.5 One-way ANOVA-like analyses of the entire database (Survey) 45 4.2.6 Two-way ANOVA-like analyses of the entire database (Survey-Location) 46 4.2.7 One-way ANOVA-like analyses: Survey within Locations 47 4.3.1 Management effects on annual rates of landsliding as obtained from fieldwork-coupled air photo-based landslide inventory 49 4.3.2 Management effects on annual denudation rates as obtained from fieldwork-, coupled air photo-based landslide inventory 50 ,. 4.3.3 Management effects on rates of landsliding and denudation: comparison between air photo-based and field-coupled landslide inventories 51 4.3.4 Two-way ANOVA of the entire database (Location-Stability-Land use) 52 4.3.5 One-way ANOVA-like analyses of the entire database (Location-Stability-Land use) 53 v 4.3.6 One-way ANOVA: Land use within Locations 53 4.3.7 One-way ANOVA: Undisturbed Forest between Locations 53 4.4.1 Elevation and Slope Gradient as extracted from a 25-metre DEM 56 4.4.2 Relation between Surficial Materials and Aspect, Capilano Watershed 59 4.4.3 Drainage density of the field-surveyed area as extracted from a 25-metre Digital Elevation Model 61 4.5.1 Logged areas in the QCI, the Capilano and the San Juan basins 70 4.5.2 Kolmogorov-Smirnov tests 73 5.2.1 Land use extensions in East Cap and Sisters Creeks 81 A. 1 An example of terrain stability class criteria, Subjective Rating Analysis 115 A.2 Example of terrain stability class criteria: Probabilistic Univariate Analysis 116 A.3 Reconnaissance terrain stability classification 118 A.4 Detailed terrain stability classification 118 A.5 Comparison of various terrain stability mapping systems used in BC 119 D.l Snow course in or near the GVRD Watersheds 124 D.2 Rainfall total-duration-frequency at selected AES stations 125 VI List of Figures 1.3.1 Illustration of air photo canopy displacement and visible portion of a landslide (from Pyles and Froehlich) 8 1.3.2 Failure scar on till (from Swanston, 1969) 10 1.3.3 Failure scar on colluvium (from Swanston, 1969) 10 2.1 The 30-year window framework adopted to obtain the quantitative database 18 3.1 Map of the Capilano watershed indicating the location of the ground-survey areas 27 4.1.1. Landslide frequency distributions as obtained from interpretation of a single photo set and from five sequential photo sets 37 4.1.2 Landslide visibility on recently harvested cut-blocks 38 4.1.3 Landslide visibility on forests older than 50 years 38 4.1.4 Landslide visibility through time due to vegetation re-growth 39 4.2.1 Cumulative percent volume distributions (East Cap & Sisters Creeks) 41 4.2.2 Survey-Location interaction of landslide density 48 4.2.3 Survey-Location interaction of mobilised debris 48 4.3.1 Box-Whisker plot of the medians of mobilised volumes: Old-growth forest vs. Old Logging 54 4.3.2 Box-Whisker plot of the medians of landslide density: Old-growth forest vs. Old Logging 54 4.4.1 Elevation frequency distributions: Sisters Creek vs. East Cap Creek 57 4.4.2 Slope frequency distributions: Sisters Creek vs. East Cap Creek 58 4.4.3 Slope aspect as extracted from a 25-metre DEM 59 4.4.4 Landslide distribution by aspect for the Capilano watershed (from GVRD, 1999) 60 vii 4.4.5 Distributions of surficial materials in Sisters Creek 62 4.4.6 Distributions of surficial materials in East Cap Creek 62 4.5.1 Landslide volume distribution as obtained from air photo interpretation coupled with intensive ground survey 64 4.5.2 (A) Histogram showing the magnitude range of landslides "not visible" from air photo interpretation; (B) Particular of events not greater than 2000 m3. 65 4.5.3 Landslide volume distribution as obtained from air photo interpretation (scale 1.15,000) 66 4.5.4 Landslide Magnitude-Frequency (volume categorical bins: 200-m step) 66 4.5 Flowchart showing how was obtained the Capilano predicted landslide volume distribution. 67 4.5.5 Landslide volume distribution (volume categorical bins: 200-m3 step) 68 4.5.6 Landslide Magnitude-Frequency (volume categorical bins logarithmically equally spaced) 68 4.5.7 Landslide Magnitude-Frequency (volume categorical bins: 400-m step) 69 4.5.8 Landslide volume distributions from API (Capilano basin, San Juan basin, Queen Charlotte Islands) 71 4.5.9 Landslide Magnitude Frequency from API: Capilano basin, San Juan basin, Queen Charlotte Islands 72 4.5.10 Cumulative percent volume distributions: Capilano basin, San Juan basin, Queen Charlotte Islands 72 5.1.1 Total vegetative cover as a function of age for landslide scar (after Smith et al., 1986) 78 5.1.2 Factors affecting landslide visibility 79 A. 1 Example of BC surficial material mapping 119 V l l l List of Photos 1.1 Not visible debris slide in Cameron Creek 2 1.2 Not visible debris slide in Strachan Creek 2 1.3 Tree standing in the middle of a recent debris flow track 11 3.1 Granodioritic bedrock in the East Cap basin 29 3.2 Basal till exposed on a debris slide sidescarp 31 5.1 Shallow debris slide in a recently harvested cut-block of the East Cap basin 75 IX Acknowledgements The list of persons who provided substantial help during the development of this research project is fairly rich. This reflects my philosophy of conducting research that is founded in collaboration, exchange of ideas and techniques. I am grateful in primis to Olav Slaymaker, Michael Church and June Ryder who supported my ideas from the beginning, encouraged me to continue the research when things seemed not to go for the better, and gave constructive comments to the completion of the thesis. Antal Kozak kindly acted as basic reference for the statistical analysis. Thanks to Roy Sidle and Dan Hogan who helped defining research methodologies during the pre-fieldwork season. The help of GVRD officials, in particular Derek Bonin and Dave Dunkley, to supply me data and to facilitate the access to the GVRD Watersheds reserve is gratefully acknowledged. Special thanks to Dave Dunkley, who has always been willing to help me through the logistical problems of the fieldwork season. Brian Klinkenberg and Erik Schiefer provided help with GIS. Much appreciation for the help received from Russell White, both for sharing his field experience and intuitions. Thanks to Liz-Anne Strik and Stephanie Sork for sharing uncomfortable rainy days out in the field. Michael Bovis, Marwan Hassan, Terry Rollerson and Frederick Swanson provided constructive comments that helped refining the data analysis layout. The first year of Master study was funded through a scholarship granted by the Universita Ca Foscari Di Venezia; the second year was financially covered by a Government of Canada Award. I am finally grateful to my family and all my close friends who have always been by my side during the difficulties I encountered while adjusting to the North American "environment". FRANCESCO BRARDINONI The University of British Columbia August 2001 x Chapter 1 INTRODUCTION 1.1 Aims Landslides, defined as the sudden downward movement of a mass of rock, debris or earth, are important agents of soil erosion in steep mountains that require inventory in relation to land use. Air photo interpretation is a convenient and efficient way to appraise landslide occurrence in steep terrain with difficult access. When comparing rates of landsliding between undisturbed and harvested terrain by means of air photo-based landslide inventories one is faced with a bias, that is the forest canopy hides scars of "small" landslides (cf. Photos 1.1 and 1.2) generally smaller than 200m2 (Rood, 1984) in area. The main purpose of this study is to investigate how large a proportion of failures is missed, both in terms of number of events and volume of mobilised debris, at a given air photo scale. Two questions follow directly from the general objective: 1. Is the proportion of "missed" events constant throughout a physiographic region (i.e. Pacific Ranges of the Coast Mountains) at a given air photo scale? 2. Can the existence of "not visible" events affect: A) Conclusions about the impact of logging on slope stability in terms of land use (management) effects? B) The nature of landslide1 magnitude-frequency distributions? C) Practical application of British Columbia Terrain Stability Classification (cf. App. A)? In this thesis the term landslide includes both debris slides and debris flows/torrents. 1 Photo 1.1. Not visible debris slide in Cameron Creek. Photo 1.2. Not visible debris slide in Strachan Creek. 2 1.2 Hypotheses Hoi: In coastal British Columbia, the time window for detection of landslides from A/P is empirically estimated to be 30 years. It is hypothesised that no event detected on a photo set is older than 30 years (the testing of such hypothesis has to be conducted on episodic events, cf. Section 1.3 for the definition of episodic events). H02: Two areas of the Capilano watershed, East Cap and Sisters Creeks have been selected as study sites. Being characterised by similar geological setting, climate, and type of vegetation cover, it is hypothesised they have no significantly different rate of sediment production from mass wasting processes. Confounding factors are considered minimised by comparing polygons of "equal" land use and stability (according to the BC terrain stability classification). H03: Timber harvesting, when isolated from other confounding factors (through delineation of BC terrain stability classification polygons), is not responsible for the acceleration of the natural pattern of mass wasting on forested terrain. Old-growth vs. old extensive logging are compared in East Cap basin; old-growth vs. recent careful logging are compared in Sisters basin. H04: Landslide volume distributions obtained respectively from air-photo interpretation and from air-photo interpretation coupled with field survey are equal. In fact, the forest canopy would hide a population of mass wasting events; these are not going to significantly change the shape of the air-photo based volume distribution. 3 1.3 Previous Studies on the Main Topics Mass failures are generally regarded as dominant mechanisms of sediment production throughout the Pacific Northwest. A number of studies have well documented how forestry practices, a dominant anthropic activity in this part of North America, can increase production of sediment (Swanson et al, 1987). Types of shallow rapid movements that are directly affected by timber harvesting are debris avalanches, debris slides {episodic events), and debris flows/torrents {recurring events). This simple classification considers the frequency of occurrence of an event at a given location. Recurring events are defined as those that occur on a decennial time scale. They are gully related. We know that a number of contingencies have to occur for the triggering of such failures, including weather, surface condition, and the history of gully sediment recharging (Bovis and Dagg, 1992). Episodic landslides are sporadic events that occur at irregular time intervals, they have a recurrence period greater than 100 years at a given location. Their occurrence is not connected to the sediment recharging process of gullies (cf. Section 5.4), but is rather related to the more uniformitarian process of regolith evolution. In the footsteps of Caine's comprehensive study (1980) researchers have inquired into the meteorological conditions associated with hillslope failures, aiming to evaluate rainfall (intensity and duration) thresholds for sliding (Church and Miles, 1987; Hogan and Schwab, 1992). Particularly in British Columbia, great attention has been paid to the role of gullies, as preferential pathways of sediment delivery to streams (Swanson and Swanston, 1976; Jordan, 1994; Bovis et al., 1998). Due to their high degree of steepness2 and moisture, gullies, are to 2 It is worth to remember that slope gradient (P) is probably the most important antecedent factor affecting landslide occurrence (cf. Mohr-Coulomb equation in section 2.2). 4 be seen as the most unstable, hence sensitive, sites in this forested-geomrphological context (Rollerson, 1992). The impact of logging on mass wasting processes has been the central theme of many studies in British Columbia (O'Loughlin, 1972; Schwab, 1983, 1988; Rood, 1984, 1990; Krag et al., 1986; Sauder et al., 1987; Gimbarzevsky, 1988; Rollerson, 1992). These studies have emphasised acceleration of landslide rates as a management effect (i.e. road construction and timber harvesting). However, the complexity of the topic goes well beyond the crude comparison of landslide frequencies and sediment yields and it is reflected in the ways in which research has been conducted in the Pacific Northwest: from different perspectives (scales) and with different goals. Timber harvesting takes effect on at least two levels: 1. It removes the live tree component - mechanical level; 2. It modifies water interception and conveyance - hydrologic level. Preliminary results from independent studies carried by Grainger and Jordan (B.C. FSRB, in prep.) have suggested how coastal BC might be dominated by root-strength (in-clearcut) failures while water-diversion related (road-induced) ones would be more common in the Interior. Such indications are based on statistical analysis comparing relative number of road-related slides to in-cutblock ones and need further analysis. Detailed studies on the impact of road construction in forested terrain have been carried out in Oregon and Washington (Reid and Dunne, 1984; Montgomery, 1994; Wemple et al., 1996; Luce and Wemple, 2001). Roads are generally held to be the most significant contributors to accelerated sediment production in logged areas. 5 1.3.1 Sediment Production from Landsliding and Impact of Logging Compilations of landslide studies conducted in the Pacific Northwest (Sidle et al., 1985; Meehan et al., 1991) have exerted strong influence on decision making in forest policy and ecosystem management. The high degree of variability of the impact of logging (cf. ratios on table 1.3.1) on the rates of landsliding and sediment production demonstrate the complexity of the issue, due to the high number of factors involved (for a comprehensive list of factors affecting slope stability cf. Sidle et al., 1985). The majority of these studies rely (completely or heavily) on air-photo interpretation (cf. Table 1.3.1). How appropriate are these air photo-based comparative studies to detect differences in landslide densities and erosion rates between recently harvested cut-blocks and old-growth forest? The issue of slide identification in the field as opposed to in air photos has been addressed in different ways, empirical (Rood, 1984; Schwab, 1988; NHC, 1997; Rollerson et al., 2001) and strictly theoretical ones (Pyles and Froehlich, 1987). Empirical approaches try to overcome the bias by excluding landslides below a certain area size (cut off) that is considered still to be assuredly detectable in recent harvested cut-blocks and in primary forest. Such threshold seems to have increased through time in the literature, O'Loughlin (1972) considered debris slides and avalanches larger than 1000 ft2 (93 m2); Rood (1984), events larger than 200 m2; Schwab (1988), larger than 400 m2; Rollerson et al. (2001) larger than 500 m . When comparing rates of landsliding between logged and unlogged terrain from air photos, the effect of corrections (cut-off) in removing the bias of different visibility leaves the comparison incomplete, because it neglects all "events smaller than". In the presentation of the results this specification should always be stated clearly. 6 Table 1.3.1. Impact of timber harvesting on landslide density and related rate of denudation (modified after Robison et al., 1999). Reference Amaranthus et al., 1985 Dyrness, 1967 Fiksdal, 1974 Hicks, 1982 Hughes and Edwards, 1978 Ketcheson and Froehlich, 1978 Lyons, 1982 Marion, 1981 Morrison, 1975 Robison et al., 1999 Rood, 1984 Schwab, 1983 Swanson and Dyrness, 1975 Swanson and Grant, 1982 Swanston and Swanson, 1976 O'Loughlin, 1972 Brardinoni et al., 2001 Site Siskiyou Mtn., Oregon Oregon Cascades, H.J. Andrews Olympic Pen. Wash. Sequaleho Cr. Oregon Cascades, Middle Santiam Oregon Cascades, Umpqua Basin Oregon Coast Range, Mapleton Area Oregon Cascades, 1959-67 Oregon Cascades, 1967-72 Oregon Cascades, Blue River Oregon Cascades, Alder Creek Oregon Cascades, near Vida Oregon Cascades, Elk Creek Oregon Coast Range, Mapleton Oregon Coast Range, Scottsburg B.C. Graham and Moresby Islands B.C. Queen Charlotte Islands Oreg. Case, H.J. Andrews Unstable Oregon Cascades, WNF Mod Stable Oregon Cascades, WNF Stable N.W. Coastal British Columbia S.W. Coastal British Columbia S.W. B.C. Coast Mnts, Howe Sound B.C. Cascades, Center Creek B.C. Cascades, Nesakwatch Creek S.W. B.C. Coast Mnts, Chapman Cr. S.W. Vancouver Island, San Juan R. Measurement Type* Air Air/Size Air/Field Visit Air/Field Visit Ground Ground Air/Field Visit Air/Field Visit Air/Field Visit Air/Size Ground Ground Ground Ground Air/Size Mixed Mixed Mixed Mixed Air/Size Mixed Air/Size Air/Size Air/Size Air/Size Air/Size Recently Ratio LS Density 19.0 9.8 0.0 3.6 8.0 2.2 22.8 6.8 10.0 13.5 1.4 0.8 1.9 5.2 30.0 17.0 3.2 3.0 7.0 5.0 3.6 10.3 6.9 29.5 36.4 15.9 harvested Ratio LS Denudation 6.8 5.0 0.0 3.4 10.0 3.4 29.5 10.0 9.0 2.6 3.2 0.3 1.5 2.6 31.2 5.0 2.8 2.5 5.0 2.2 2.3 40 10 47 28 24 * Measurement Type: 1. Air: study based entirely on air photo interpretation. 2.Air/Size: study based on air photos with a minimum landslide size used to mitigate bias between undisturbed forest and recently harvested cut-blocks. 3. Mixed: combination of more than one method of surveys (e.g. air photo-based survey to detect landslides in clearcuts and a ground-based sample in older forest). 4. Air/Field Visit: air-based sample with non-systematic field visits used to get some indications on the percentage of landslides missed. 5. Ground: study that detects failures on a systematic sampling of landslides using the channel network and/or the slope contours as a search path. 7 Pyles and Froehlich (1987) tackle the question from a more formal/theoretical point of view, focusing the discussion on the limitations of air photo-based landslide inventories. They list a series of factors that affect photo detection capability, these are shadow, season, canopy height, ground slope, radial distance and direction from the principal point of a photograph (cf. Figure 1.3.1). Figure 1.3.1. Illustration of air photo canopy displacement and visible portion of a landslide (from Pyles and Froehlich, 1987). An important question is how these factors vary from one physiographic area to another. For example, in the Oregon Coast Range slides are small (approx 50-100 m3), the terrain highly dissected, forest cover dense, so slides are commonly obscured and hard to see on photos. In contrast, in the Cascade Mountains slides are bigger and may have longer runout zones, so may be more detectable. Any analysis of photo detection capability using field case studies needs to define the properties of the events and the context: forest stature and cover, topography as it affects shading and runout, etc. (F. J. Swanson, pers. comm., 2001) - basically the "texture" of the landscape. 8 In addition, it should be noted that, the measurement of landslide magnitude (width and length) from air photos contains relevant errors, hence the identification of landslides itself is not the only problem. It is anticipated that better precision can be achieved in estimating landslide size and distinguishing between initiation-transportation and deposition zones on recent cut-blocks than in old growth forest. This question is an important one, particularly when one integrates field and air photo measurements (cf. Section 5). In particular, failures occurring on till mantled terrain tend to have a linear shape whose width is difficult to estimate from air photos. Furthermore, it is not uncommon to see single trees that have "survived" the catastrophic flow standing in the middle of the scar, hence increasing the degree of uncertainty from air photo interpretation, cf. Photo 1.3). Swanston (1969), in a study conducted in southeast coastal Alaska, provides a careful landslide scar morphological distinction, later confirmed by Howes (1981) for northwestern Vancouver Island, between bedrock-colluvial soil slides and till soil slides. The former leave behind a wedge-shaped scar expanding downslope, the latter tend to have a spoon-shaped initiation zone and a typically confined, narrow transportation zone down to the valley floor (cf. Figures 1.3.2 and 1.3.3). Similar morphological characteristics were found while conducting ground survey in the Capilano watershed. The issue of "not visible" landslides became a controversial one after the 1996 floods in Oregon. Some environmental groups did a flyover inspection and reported high slide frequency in cut and roaded areas, while others were concerned that many slides were missed in the forested areas, so a good reference for evaluating management impact was missing. This issue also affected the Siuslaw National Forest assessment of storm impact (F. J. 9 Figure 1.3.2. Failure scar on till (from Swanston, 1969). s Figure 1.3.3. Failure scar on colluvium (from Swanston, 1969) 10 after the 1996 storm events a large project (Robison et al., 1999), to date the only study that has systematically compared ground based landslide inventories with air photo based ones. Photo 1.3. Tree standing in the middle of a recent debris flow track (note signs on the bark). 11 About half of the landslides in recent plantations were visible in photos at the 1:6,000 scale. In mature and old growth forest, the air photos revealed 5%, at most, of the ground surveyed slides. Such percentages should not surprise the reader, in fact landslide size in the Oregon Coast Range is typically smaller than any of the air-photo thresholds we mentioned earlier. When commenting on the results of Robison et al., it is essential to bear in mind two operational specifications that undoubtedly have affected their data. These are: a) they counted and measured only those events that impacted stream channels b) the study focused on areas considered to have experienced the most severe impacts from the 1996 storms (results are expected to be well beyond the average forestland response to these individual storms). 1.3.2 Landslide Visibility through Time Clearly the issue of landslide visibility is closely related to the main topic of this research project. I selected a few references from the Pacific Northwest to better delineate the problem. A study (Reid and Dunne, 1996) conducted in coastal Washington indicates that many of the small landslides are no longer visible after 10 to 20 years, while failure scars of the sizes that dominate sediment input (> 800 m3) remain easily identifiable until they are 30 to 50 years old. Apparently, up to date no experimental study in British Columbia has dealt with the issue of landslide visibility through time from aerial photos, and no project has tried to compare/integrate landslides visible on air photos with those hidden by the forest canopy. Mass failures observable on aerial photographs in forested terrain can date as far back as 100 years (Rood, 1984). However, the exponential temporal trend of scar percent area 12 covered by colonising vegetation makes it impossible to collect all qualitative and quantitative landslide attributes (cf. Appendix C). In particular, on landslides older than 40-60 years one would not be able to distinguish between initiation and transportation zones (Smith et al., 1983). Smith et al. visited a sample of 50 debris slides in the Queen Charlotte Islands with average size of 23000 m (failure size includes initiation, transportation and deposition zones). However, they are relatively large events and, as such, landslide size is expected to affect the pattern of vegetation re-growth. Field measurement of 68 slides in the Capilano Watershed indicates that identification on the air photos (approximate scale 1:15,000) of landslide tracks older than 30 years is uncommon, rare in the case of failures older than 40 years. Thus 30 years was established as the upper limit of ages of all mass failures visible on a single photo set (GVRD, 1999). Field experience suggests that "small" landslides can "disappear" on logged terrain in less than 30 years, because the difference in canopy heights and appearance is not very significant. In natural forests the canopy on landslide scars tends to stand out across the adjacent forest cover (marked differences in terms of height and texture/colour on the air photos), so one can often see landslides that are 50 years old or older (T. P. Rollerson, pers. comm., 2001). Snyder (2000) conducted a test for visibility through interpretation of sequential air photo sets on 24 debris flows, which occurred in 1965. Every event that was detected on the air photos was classified for visibility and vegetation re-growth through time. Visibility was assumed to be a function of the relative difference in brightness, contrast, and vegetative 3 Field investigation conducted by JMRATA Ltd. for the GVRD. 13 recolonisation between the failure scar and surrounding forest cover. Accordingly, three categories of visibility were defined: highly visible (absent or very little vegetation on the scar), moderately visible (approximately 50% of the scar is covered by vegetation),. and scarcely visible (almost the entire scar was covered with vegetative re-growth). About 13% of the monitored failures (3 out of 24) were still moderately visible 31 years after they occurred. 1.3.3 Landslide Magnitude-Frequency Landslide magnitude-frequency relations appear to have a power-law scaling, either if one defines landslide magnitude as scar area (Hovius et al., 1997, 2000) or the total area disturbed including the deposition zone (Pelletier et al., 1997), or if one uses volume. Bak et al. (1987, 1988), borrowing the term from the thermodynamics literature, have suggested the concept of self-organised criticality as the theoretical explanation for the power-law magnitude frequency distribution that one can see in failures resulting from the overcoming of critical thresholds. By definition self-organised criticality characterises those systems that maintain themselves far from equilibrium by rearranging their configuration through internal feedback to a state in which a system parameter fluctuates with power-law distributed adjustments around a marginally stable critical value (Turcotte, 1992). A power-law scaling for landslide magnitude-frequency has implications and useful applications in the definition of hazard assessment (Hungr et al., 1999). From an exclusively theoretical stand point, long-term landscape evolution modelling can incorporate such power-law relationships, which in turn can be employed to identify increase in the fractal dimensions of the topography of montane terrain (Dadson, 2000). 14 More pertinent to the issue of landslide identification in densely forested areas, the existence of a power-law relation for shallow rapid mass failures would sensibly improve precision of rapid evaluation of sediment budgets. Of particular practical relevance is to verify whether the power-law that one commonly observes for air photo detectable failures still remains scale-invariant for "not visible" landslides. 15 Chapter 2 METHODOLOGY In this chapter are included the operational statements of measurements made in the research project. Initially a drainage basin that had been partly subjected to timber harvesting was selected. The study entailed the following sequence of actions: 1) air photo interpretation and compilation of landslide inventory. The scale of aerial photography available ranged from 1:12,000 to 1:15,000; 2) intensive fieldwork with the double purpose of: A) ground-checking the classification4 of terrain polygons both in terms of surficial material and terrain stability; and B) recognition and measurement of every distinguishable failure within each traversed polygon. In every study-polygon we walked along the entire gully network and covered almost the whole area of the polygon, hence we can assert with good confidence that we were able to detect all landslides present. The inventories have been carried out with precise operational specification: A) Inspection of all ground classified in stability classes III, IV, and V. We can assume that no mass failure has occurred on stability classes I and II; B) Systematic measurement of all traversed "not visible" events; 4 JMRATA Ltd. following BC Terrain Stability Classification guidelines of the BC Forest Practices Code, 1995, 1999) have conducted both types of mapping. The work was done mainly by air photo interpretation with only 25% of polygons ground checked at the time of map compilation. 16 B) Systematic measurement of all traversed "not visible" events; C) Road related landslides are not included. Due to the fact that continuous road maintenance has been conducted in the watershed, an indefinite number of landslide scars has been masked/rehabilitated. In order to minimise any bias in the comparison/integration of field-collected and air photo-collected data it was decided to define a time window, that is we considered only those mass wasting events that occurred during the last X years (cf. Figure 2.1). References cited earlier in section 1.3 (Smith et al., 1983, 1986; Rood, 1984; Reid and Dunne, 1996) and field logistic reasons suggested selection of a 30-year window framework. In addition, here we have the opportunity to check the reliability of the 30-year window assumption that has been used for estimating annual average volume of sediment moved by landslides in each terrain polygon of the Capilano Watershed (GVRD, 1999). Age of landslides was estimated in the field (minimum age) by means of dendrochronology, count of branch levels on saplings, and presence of litter on the scar; on aerial photos through inspection of sequential photo sets. Having distinguished between episodic and recurrent landslides; failures were then classified according to their degree of activity. There are a number of examples that classify landslides into active and dormant (Zaruba and Mencl, 1969; Erskine, 1972). This distinction is relative to the time frame one is looking at. In our case dormant (inactive or fossil) mass movements are those that, though still visible on air-photos, had occurred more than 30 years ago and, as such, they were not taken into consideration. Following these criteria, a quantitative database was obtained from the integration of fieldwork and air photo interpretation (Cf. Figure 2.1). 17 The air photo-based landslide inventory was conducted with a SOKKIA MS27 stereoscope (3x and 8x magnification). Each mass movement event, when possible, was partitioned into initiation, transportation and deposition zone. The technique adopted to evaluate the mobilised volume associated to each failure was that described by Rood (1984). LANDSLIDE TEMPORAL CLASSIFICATION • Number of Landslides \ • Disturbed Area (m') • Mobilised Volume (m ) Figure 2.1. The 30-year window framework adopted to obtain the quantitative database. In the field, linear measurements were taken with hip-chain and metric tape, slope gradients with clinometer, slope aspects and direction of traverses with compass, elevation 18 with analogue altimeter. Rugged topography of the terrain and density of forest cover have prevented us from using GPS (Global Positioning System) devices 5. This would have minimised errors in locating ground measurements such as the location of "not visible" landslides within terrain polygons, and delineation of the routes travelled while traversing. Table 2.1. Principal failure attributes. Failure Attribute Qualitative Quantitative Type Location at initiation point Slope profile Land use Stream connection Surficial material Elevation Slope gradient Slope aspect Height of headscarp Height of sidescarp Mobilised volume Categories/Units - debris slide - debris flow/torrent - open slope - headwall - side-wall - gully-channel - convex - concave - straight - complex - undisturbed (old-growth) - recent logging - old logging - fire - entered a stream - not entered a stream - entered a gully e.g. Till, colluvium, bedrock, etc. - metres a.s.l. - at initiation - at transportation (degrees) - at deposition - degrees from north (N) - metres - metres - cubic metres For each event a number of attributes were recorded (Cf. Appendix C), the most significant of which are reported in Table 2.1. Among the other quantitative attributes, the 5 GPS requires an unobstructed view of the sky. Solid obstacles such as mountains and dense tree cover can block satellite signals, make them intermittent and cause errors due to reflections. 19 computation of mobilised volume of debris deserved particular attention; it entailed the following: 1. distinction among initiation, transportation and deposition zone; 2. measurement of the initiation and transportation zone (width, length, height of side scarp) and volume estimation; In air photo interpretation the following steps were also taken. 3. Scale computation: scale = (flight height - terrain elevation)/focal length of lens; 4. Width and length, calculated at point 2, have to be multiplied for the scale (correction factor). 5. Volume = corrected width * corrected length * depth Alternatively one could have measured the volume of the deposition zone. However, this option was discarded because stream-connected failures might have their deposition zone (e.g. flow fans) readily washed out by the stream they have impacted. Field and remotely sensed data were managed and analysed using Microsoft Excel, Statistica, and SPSS. Arc View and Arclnfo were used to manage thematic maps. 2.1 Statistical Analysis We used two thematic maps, a land use (or disturbance) map and a terrain stability map. The land base is subdivided into polygons. A polygon is considered as an experimental unit, and each polygon that receives the same classification ("treatment", i.e. survey type, land use, stability class) is regarded as a replication. 6 For those failures that were only measured from aerial photos, depth was assumed to be 1.52 m in Sisters Creek, and 0.78 m in East Cap Creek (average headscarp values of the ground-measured events). 20 A "two step" statistical analysis was performed. Firstly one-way ANOVA-like tests were conducted to evaluate how the survey method (i.e. air-photo vs. air-photo coupled with intensive field investigation) affected landslide frequency per unit area (LS/ha) and its associated amount of mobilised debris (m /ha). Secondly the experimental design was made more complex (two-way ANOVA-like tests) so that the location effect (Sisters vs. East Cap) on survey capability could be incorporated. The large discrepancies in landslide densities and mobilised debris between locations (cf. Table 4.2.1) encouraged us to regard the two sub-basins as blocks. Dependent variables and factors are shown in Table 2.2. Table 2.2. Dependent Variables and Factors of Location-Survey analysis. Dependent Variables Block/Factor Factors - No. of landslides per hectare (LS/ha) - Vol. of mobilised debris per hectare (m3/ha) - LOCATION (East Cap, Sisters) - SURVEY (Air Photo, Air Photo & Field) The assumptions of ANOVA are not quite met by both dependent variables, notably normality of distributions and homogeneity of variances are not satisfied. Therefore one-way nonparametric methods such as Kruskal-Wallis and Median test were added (cf. Table 4.2.4). The last is regarded as the most appropriate of the three options presented. In fact, the Median test is particularly useful when the scale contains artificial limits and many cases fall at either extreme of the scale ("off-the scale"), in this study more than 50% of observations (polygons) have a value of zero (no landslides, hence no mobilised debris). Three main factors have been analysed (cf. Table 2.3) for statistical significance to try isolating the impact of logging on landslide initiation. Dependent variables are the same as those used in the "survey" analysis; likewise ANOVA assumptions are not quite met. It 21 follows that standard ANOVA methods have been coupled with their nonparametric equivalents. This time the analysis was conducted with a progressive simplification of the design layout. The following were applied in sequence: factorial experiment in a fixed block design (locations acting as blocks), full factorial, and 1-way ANOVA-like tests. Table 2.3 Dependent Variables and Factors of Location-Land use-Stability analysis. Dependent Variables Block/Factor Factors - No. of landslides per hectare (LS/ha) - Vol. of mobilised debris per hectare (m3/ha) - LOCATION (East Cap, Sisters) - LAND USE (Undisturbed forest, Old logging, Recent logging, Fire) - STABILITY (III, IV, V) 2.2 GIS-Based Analysis East Cap and Sisters Creeks exhibited such unexpected contrasts in terms of landslide density and denudation rate (see section 4.2), that a GIS-based analysis was explored as a way of resolving the difference. The Digital Elevation Model is derived from the Terrain Resource Information Management (TRIM) product of the Province of British Columbia, Canada, a 1:20,000 scale digital map. TRIM DEMs are produced directly from photogrammetric models designed to produce triangulated irregular networks (TINs), and are composed of elevation points and break lines (Geographic Data BC, 1992). The TIN, a format good for storage, was then converted to a Grid of 25-metre resolution, as recommended by Geographic Data BC, to facilitate analysis operations. The nature and quality of information available, ease of GIS-extraction, and power of physical explanation are the factors that determined which macroscopic variables could be 22 extracted from the 25-metre DEM. These were elevation, slope gradient, slope aspect, drainage density and spatial partition of surficial materials across stability classes. Elevation affects the amount of precipitation, rate of physical weathering (e.g. freeze-thaw activity and magnitude and persistence of snow cover) and the hydrologic regime of a basin. Elevation frequency distribution has been obtained by selecting the surveyed terrain polygons from the terrain theme (shapefile). Elevation data were then converted to a grid (polygons not selected result in "no data" cells) to allow for overlay operations via Arc View Map Calculator. The grid so-obtained is termed "terrain grid". Hence, the elevation grid of the whole basin is overlaid on the terrain grid as follows: [(terrain grid)*0] + elevation grid = clipped terrain grid Finally, elevation histograms are plotted (via Arc View Histogram button) for Sisters and East Cap Creeks separately. Slope gradient and aspect frequency distributions were obtained using the same procedure described above for elevation (slope gradient and slope aspect grids replace the elevation grid). Although generalisations are to be avoided, slope gradient is one of the most important geomorphic factors for shallow mass movement processes (Sidle et al., 1985). Slope gradient (P) appears in the Mohr-Coulomb7 equation (equation 1) and directly affects both normal (o~) and tangential (x) components of the shear stress. S = C + AC + (a - u) tan<|>' (1) 7 It is the Mohr-Coulomb equation for the case of infinite slope analysis. Such analysis is applied on terrain where the thickness of the soil mantle is small (negligible) compared with the length of the slope and where the failure plane is parallel to the slope surface, an analysis that generally fits well to «steeplands» of coastal BC. 23 Where: u = pore water pressure = MZyw cos p a = total normal stress = [(1 - M) ym + Mysat]Z cos2(3 + WT cos2p x = total tangential stress = [(1 - M) ym + Mysat]Z sinpcosp +WT sinP AC = cohesion caused by root systems C = effective soil cohesion; Z = soil depth; p = slope angle; WT= weight of vegetation; <j/= angle of shearing resistance; ym = soil unit weight ysat= saturated soil unit weight; yw = water unit weight; M = height of water table as decimal fraction of Z; Slope aspect influences terrain exposure to storm fronts, it affects fluctuations of pore water pressure (u), and alternation of weathering environment (oxidising/reducing) brought on by wet/dry and/or freeze/thaw cycles. Increase in pore water pressure alters slope stability by reducing the effective normal stress and thus soil shear strength. Drainage density (Dd) describes landscape degree of topographic dissection. It is defined by the ratio between the total length of the channel network and its drainage area. Specifically, in mountainous areas of coastal British Columbia a greater drainage density corresponds with a higher number of zero and first-order streams (gullies), ephemeral streams which naturally experience recurrent failures (i.e. debris flows/torrents) on a decennial time scale. The scale of TRIM data (1:20,000) does not capture the entire stream network. No study has yet investigated the relationship between measured (as mapped on TRIM maps) and actual drainage density (as calculated for the higher-order channel network). Mapped Dd probably underestimates actual Dd, however the associated error is 24 clipped onto the selected terrain polygons. Polygon areas and stream lengths are measured via "Polyline" arcscript, hence drainage density is obtained by dividing the total length of the clipped stream network by the area of the selected polygons. As for the spatial distribution of surficial materials across slope stability classes, special attention will be paid to glacial deposits (or till) and post-glacial deposits (or colluvium). Till is considered to be less stable, having been deposited under sub-glacial, en-glacial or supra-glacial conditions. Till is therefore still adjusting to a sub-aerial, non-glacial environment. Conversely, colluvium deposits, a product of sub-aerial mass movement are arranged in more stable configurations (J. M. Ryder, pers. comm., 2001). Here the intent is to transform the qualitative information contained in the terrain mapping into a more quantitative form. To achieve this, the area of each polygon was subdivided according to the proportion indicated on the corresponding label (cf. Appendix A on British Columbia Terrain Stability Classification). Subsequently, the computed areas of each surficial material category (i.e. M, C, R, /CM) were summed according to their B.C. terrain stability class label and the relevant histograms were obtained (cf. Figures 4.5.4 and 4.5.5). 25 Chapter 3 STUDY AREA The Capilano River basin (198 km2, Source: GVRD Watershed Management, 1997) has been selected as the study area. Several reasons justify the choice. Firstly, Capilano is considered the closest forested watershed to Vancouver that would meet our experimental requirements. The accessibility of the watershed constitutes an advantage in terms of costs, time and, above all, in terms of flexibility during the fieldwork season. Secondly, forest harvesting has been conducted with well-defined and documented techniques in the last 50 years: clearcut and partial cutting (helicopter) followed by grapple, highlead, or skyline-full suspension yarding (D. Bonin, pers.comm., 2001). Thirdly, being one of the three sources of water supply for the Greater Vancouver Area, the watershed has been intensively monitored: a wide range of thematic maps and detailed air photo coverage (through time and space) are available. The latter allows dating of failures on air photos with high precision. The existence of relevant thematic maps constitutes a unique advantage in applying GIS techniques of data analysis. Of particular interest are the land use (by GVRD) and terrain materials mapping (by J. M. Ryder & Associates). Furthermore geology is "homogeneous" (Cf. Section 3.1). Within the watershed intensive ground checking was conducted on Sisters and East Cap Creeks (cf. Figure 3.1). The former is the closest location presenting areas of old extensive logging on steep terrain (class IV and V polygons), the latter being the closest location presenting recent cut-blocks on steep terrain. In the selection of these study areas, sites underlain by lacustrine and glacio-lacustrine materials were avoided (cf. later in this 26 Figure 3.1. Map of the Capilano watershed indicating the location of the ground-surveyed areas. 27 section for an extended explanation). Summary information on the study areas is reported on table 3.1. Table 3.1. Physical characteristics of the study areas. Watershed Drainage Area (km2) Elevation (m a.s.l.) ** Topography Bedrock Geology and Surficial Material Biogeoclimatic Zones Capilano R. 198 120-1800 Sisters Cr. 22.7 (6.2*) 145-1635 (265-1195*) East Cap Cr. 40.9 (6.7*) 338-1720 (355-1320*) Rugged, slopes steeper than 35° (70%), steepness generally increasing with elevation. Primarily intrusive igneous rocks (granodiorite, quartz diorite, diorite). Bedrock is exposed at mid and upper elevations: valley floors and lower to mid slopes are mantled by till (glacial drift) and colluvium (post-glacial deposits). Coastal Western Hemlock Zone (CHW) Mountain Hemlock Zone (MH) * Ground surveyed area. ** Elevations are extracted from a 25-metre DEM. 3.1 Physiography, Bedrock Geology, and Geomorphology of the Watershed The watershed lies within the Pacific Ranges of the Coast Mountains and shares most of the physiographic characteristics of this larger region. The most striking features of the Capilano valley are rugged topography and slopes typically steeper than 35° (70%), with steepness generally increasing with elevation. The rugged landscape is the result of the combined effects of tectonic uplift, rock strength, and glacial erosion. Bedrock consists primarily of intrusive igneous rocks -granodiorite, quartz diorite, diorite, and lesser amounts of gabbro and migmatite. There are metamorphic-dominated formations in places (i.e. Gambier Group and Twin Islands Group -Roddick, 1965) which, however, cover very limited and comparable portions of Sisters and 28 East Cap Creeks. After field and air-photo surveys such areas did not stand out for greater slope instability. These rocks have great intact strength, and joints and other planes of weakness tend to be widely spaced (cf. Photo 3.1). This has allowed the formation and preservation of steep, glacially carved buttresses and gully walls. Photo 3.1. Granodioritic bedrock in the East Cap basin. 29 Though the effects of the last glaciation (Fraser Glaciation) dominate the montane landscape that we see today, during the Pleistocene epoch the southern Coast Mountains were affected by a series of glaciations. After the ice receded, erosion and landslides were likely more common on the steep valley-sides than at present. Glacial deposits were eroded and transported by debris slides, debris flows, and streams, and sediments were re-deposited on lower slopes and valley floors as colluvium and fluvial materials. Formation of colluvial (debris flow) fans, talus slopes, alluvial fans and floodplains, and erosion and enlargement of gullies and canyons has continued throughout post-glacial time (GVRD, 1999). Major bedrock-controlled landforms of glacial erosion dominate the landscapes of the Capilano Watershed. The coarse-grained rock has promoted (through weathering and mass wasting) the production of coarse-textured derivative materials, such as sandy, bouldery till and blocky colluvium. Gentle and moderate slopes, especially at mid to low elevations, are mantled by till. Till is the most extensive surficial material and constitutes the primary source of fine sediments. It is an unsorted, unstratified diamicton, and matrix-supported material. The matrix supports clasts (particles greater than 2 mm in size) that range in size from pebbles to boulders. Two main subtypes can be distinguished: basal till and ablation till. Basal till was deposited sub-glacially as a result of melting of basal ice (beneath moving glaciers) loaded with rocky debris. Typically the most cohesive and most highly consolidated of all surficial materials, basal till has very low permeability and high shear 30 strength ty = 30-40° (cf. Mohr-Coulomb equation). Clasts are rounded and large (cf. Photo 3.2). Ablation till has been deposited by ablation of ice (generally close to an ice front), causing surface concentration of clasts, and flowage of material off an "ablation surface". As a consequence, it can present lenses of stratified material. It is comparatively loose, non-compact, (due to absence of ice load) and tends to be quite permeable. Interlocking of angular (or sub-angular) clasts makes its shear strength moderately high. Photo 3.2. Basal till exposed on a debris slide sidescarp (Strachan basin). Surficial materials deposited on hillslopes in post-glacial time consist primarily of colluvium. This is a collective term used to classify surficial materials derived from a combination of slopewash and mass movement (gravitational) processes like falls, slides and flows. As such, colluvial deposits tend to flank steep slopes and characterise landforms like 31 talus slopes, avalanche cones, and debris-flow fans. Given the wide range of possible genesis, its geotechnical properties are highly site specific. Nevertheless, one can generalise on these aspects: colluvium is generally clast-supported material8 (clasts are angular due to the short travel distance from source), of loose consistency and has a high permeability (especially when compared to basal till). Surficial deposits of lacustrine and glacio-lacustrine origin cover localised portions of the watershed, namely the northwestern shore of Capilano Lake. They are well known for being particularly subject to slides and slumps (cf. Ryder and Howes, 1984; Thurber Eng. Ltd., 1996). We have avoided such deposits because they are not sufficiently widespread in the Pacific Ranges, hence not typically representative of the surficial materials in this region. 3.2 Overview of Regional Climate 9 The climate in the Pacific Ranges of the Coast Mountains consists of two main seasons. A "wetter" one starts approximately from October and continues through May, characterised by an almost uninterrupted series of large-scale oceanic storms that encroach upon the coast. A much drier regime characterises June, July, August and September, when only intense small-scale oceanic storms disrupt the persistent high-pressure domain. Accordingly, in the Capilano watershed (at Cleveland Dam station) about 80% of the annual precipitation falls during the "wetter" season. Minimum monthly totals occur in July and August, accounting for about 4% of the normal annual precipitation. Convergence in valleys and topographic uplift tend to increase precipitation abruptly north of the Lower Mainland, where fronts intersect the Pacific Ranges. Annual precipitation 8 Deposits resulting from debris flows and slumps are often matrix-supported. 9 This section is largely based on information derived from GVRD Analysis Report Watershed Management Plan #5, February 1999 (referred as GYRD, 1999). 32 is estimated to approach 5,000 mm in places at the head of passes through the mountains. Elevation also affects the snowfall proportion of total precipitation. Measurements from snow courses (cf. Table D.l) show that at elevations of around 1,000 m, the normal maximum water equivalent in the snow pack is at least 1,600 mm (GVRD, 1999). The overall pattern of precipitation variability with elevation and distance along valley floors is complicated; hence major deviations from the generally understood trends are to be expected. Table 3.2.1. Climate Stations in or near the GVRD Watersheds (modified from: GVRD, 1999). Station Period of Record Elev. (m) Climate Data1 Annual Rainfall (mm) Annual Snowfall (cm) Annual Precip. (mm) Extreme Daily (mm) Atmospheric Environment Service Stations within the Watersheds Seymour Falls Coquitlam Lake 1927-97 1924-82 244 161 3814 3468 219 148 4033 3616 314 213 Selected AES Stations near the Watersheds Mount Seymour N. Vancouver -Cleveland Dam N. Vancouver -Grouse Mtn N. Vancouver -Upper Lynn N. Vancouver -Lynn Ck Hollyburn Ridge . Climate data ai 1958-68 1968-90 1971-97 1960-80 1964-83 1954-90 e quoted a 823 157 1128 177 191 930 s norma 2450 2319 1774 2494 2576 2115 s for the pe 605 71 817 123 118 790 riod of recoi 2760 2390 2565 2612 2696 2916 -d. 174 141 148 170 163 217 Note: at Mt. Seymour the annual precipitation value does not quite match the combined annual rainfall and snowfall. At Cleveland Dam (Capilano Intake), average annual precipitation is 2,390 mm (1968-1990). Snowfall accounted for about 3% of the total. Mean annual runoff from the watershed is 3,667 mm (cf. Table 3.2.2) and, accounting for losses due to evapotranspiration, average precipitation over the watershed would have to be around 4,200 mm. This implies 33 that average annual precipitation at higher elevations, in the interior of the watershed, must be at least 4,500 mm in order to generate the mean annual runoff (GVRD, 1999). Table 3.2.2. Water Survey of Canada Gauging Stations in the Capilano Watershed (modified from: GVRD, 1999). Station Capilano River above Intake Capilano River at Canyon Capilano River (West Fork) Capilano River (East Fork) Watershed Area (km2) 172 197 69.9 41.4 Period of Record2 (to 1997) 1914-55MC;58-97RC 1929-54RC;55RS 1926MS;27-28RC; 29-30RS 1926MS;27-28RC; 29-30RS Flow Characteristics1 Mean Annual Flow (m3/s) 20 20.2 8.63 4.02 Mean Annual Runoff (mm) 3667 3234 3894 3062 1. Flow characteristics calculated from all years of record, to 1995 at active stations. 2. M = Manual; R = Recording; C = Continuous; S = Seasonal. The Water Survey of Canada (WSC) gauges, on the West and East Fork of the Capilano River (cf. Table 3.2.2) show higher annual runoff from the west side of the watershed (3894 mm). GVRD (1999) reports that the greatest annual precipitation falls in the upper Capilano Watershed, near Daniels Creek (immediately west of East Cap basin, cf. Figure 3.1). Overall, annual precipitation within the Capilano basin seems to decline to the east, with lower runoff recorded in East Capilano (3062 mm) than in West Capilano, however, Table 3.2.1 reveals a regional trend for increasing precipitation eastward (Cleveland Dam-Lynn Creek-Seymour Falls l0-Coquitlam). Schaefer and Nikleva (1973) produced a precipitation map of the Vancouver area (including GVRD watersheds) by constructing a simple model based on two topographic 10 Precipitation at Seymour Falls is higher than at Coquitlam, which is located further east. One can explain the value by observing that Seymour Falls station is located further inside the valley than Coquitlam station, hence the effect of valley channelling would be greater in this former location. 34 factors that are known to exert a strong influence on precipitation, namely elevation and valley channelling. Mountain slopes force air masses to lift; this causes cooling, possible condensation of clouds, and increase chance of precipitation. Precipitation increases also as air masses converge from the sea inland through valleys where they are subject to a further uplift. Therefore maximum precipitation values are to be found at valley heads that are facing dominant winds. According to their map, Sisters and East Cap basins have similar mean annual precipitation values. Given the focus of the project, the triggering of mass failures (and associated rates of sediment production), it is likely that scaling the analysis up to the synoptic scale, to maximum rainfall during one-day to one-week storms would be more relevant than looking at mean annual values. The way the rain is distributed through time (rainfall intensity) is more important than rainfall annual amount (considering that we are comparing two tributaries belonging to the same bio-geo-climatic zone). Studies conducted in the early 80's (Caine, 1980; Innes, 1983) associate the triggering of debris torrents (hence debris slides) with the overcoming of specific precipitation thresholds. Church and Miles (1987) conducted a study on the meteorological antecedents to debris flow along the creeks of South-eastern shore of Howe Sound. They reported (p.74) that intense storms may not trigger a debris torrent; a very minor storm may do so, occasionally They conclude by suggesting that since antecedent conditions are complex, no universal criterion seems to be adequate in establishing precipitation thresholds for the occurrence of debris torrents. The existence of an inadequate monitoring network was also an issue (and still remains today) as many intense precipitation events were probably missed. 35 At Seymour Falls", which has continuously operated since 1927, the largest annual storms typically occur in the winter months, though maximum one-day storms occasionally occur in summer. The station has substantially higher rainfall, during one to five-day storms, than any of the other stations located in the area (cf. Tables 3.2.2 and D.2). Its 100-year one-day rainfall is about 45% larger than at Cleveland Dam, and 60% larger than at Grouse Mountain or Coquitlam Lake (GVRD, 1999). The data we possess do not allow us to draw any ultimate conclusion as to spatial variability of meteorological forcing within the Capilano watershed. The two different scale-approaches leave us with a contrasting situation, the trend of annual runoff would indicate an east-to-west gradient; conversely intensity of one to five-day rainstorms does not show any east-west trend. " Seymour Falls climate station is the only one of the three (including Cleveland Dam and Coquitlam Lake) located at the intake of each GVRD watershed whose records are nowadays still reported by Atmospheric Environment Service (AES). 36 Chapter 4 RESULTS 4.1 Landslide visibility from aerial photographs The ability to locate landslides from aerial photographs varies as a function of the age and area of the slope failure. A landslide inventory was compiled, based on the interpretation of five sequential photo sets (taken in 1968, 1976, 1984, 1992 and 1996). This inventory was compared with the inventory conducted by GVRD on a single photo set (1992). In this latter case it was assumed that all visible slides in a photo pair were 30 years-old or younger. Sisters & East Cap Creeks •0 a> o O" 2 I M JI ~~i~-4-•--$-•* i~-"~ F"Mr™ H Mul'ip.o Photo-Set I I Single Photo-Set —!- * ( < J UtUL jud 1 oco 2000 3000";5v|; -./Looo"".-. . 5000; )Mf#&. 6000 Landslide;yolume;;(m3)x,: :<g::\ '"' ;>; #&•' Figure 4.1.1. Landslide frequency distributions as obtained from interpretation of a single photo set and from five sequential photo sets. Figure 4.1.1 shows that the multiple photo-set methodology gives higher accuracy. Errors were caused by the presence of snow, shade, photographic quality, etc. but mainly 37 they were due to misinterpretation of active events, that is many debris slides and debris flows classified as active in the first place, after interpretation of sequential photo-sets were found to be older than 30 years. Recent Cut-blocks co CD CD CO • D ~-\c " f~ 100 10 £, k. A \ • 1 !A X>- """: [ r c O . . . . r ° o • - 150 m 2 o 8 Not Visible-Landslide Visible Landslide ;20 20 60 100 Landslide ID 140 180 220 Figure 4.1.2 Landslide visibility on recently harvested cut-blocks. 80 Landslide ID Figure 4.1.3. Landslide visibility on forests older than 50 years. 38 In this case study better accuracy means a drastic reduction in the number of events and related mobilised volume of debris. In fact, only 34 of the 76 failures mapped by GVRD occurred over the last 30 years (or 60,000 m3 of the 142,000 m3 originally inventoried). Figure 4.1.2 shows that the maximum area of a "not visible" landslide in a recently harvested cut-block (logged less than 15 years ago) is about 150 m2. On forests older than 30 years, the largest landslide scar that was not detectable during air photo interpretation has an area of about 650 m2 (cf. Figure 4.1.3). Only one landslide smaller than this was detected. -,«s- -* .-, ; M ^ -10"= i - $?,"? ' 2 0 " . ,.., - • : '•" V-Years* since landslic 40V.:: Figure 4.1.4 Landslide visibility through time due to vegetation re-growth (cf. Section 1.3.2). Note: The time envelope accounts for the uncertainty of landslide occurrence and it is based on the time separation between consecutive photo sets (about 8 years). 7 9 Scar magnitude of the studied landslides ranged between 75 m and 15,450 m , but no significant correlation was found between initial scar size and rate of vegetation re-growth. It should be noted that the time interval between available air photographs ranged between 4 and 8 years. In gully related failures (recurring events) the pattern of vegetation re-growth is 39 so irregular — due also to the disturbance of seasonal fluvial processes — that the time interval of air photos is too long to allow documentation of the macro-pattern of vegetation colonisation. Therefore, the 30-year time window had to be tested on episodic events, whose process of vegetation re-growth is fairly uniform. A test for debris slide visibility through time (Snyder, 2000) was performed on a set of 23 events in the Capilano River basin. The test is described in section 1.3.2. In this case the age was considered to be a minimum one (time since failure first appeared on an air photograph). Four events (17 %) showed no major sign of re-vegetation for longer than 32 years (cf. Figure 4.1.4). Dotted lines indicate that we don't know when such events will become not visible from air photo, this because these scars were still visible at the end of our time window (i.e. 1996). 4.2 Landslide frequency and associated sediment production Two different types of survey of landslide density and the associated amount of mobilised volume of debris in the Capilano basin were conducted: (a) exclusively by air photo interpretation, and (b) by air photo interpretation coupled with intensive field survey. 4.2.1 Descriptive statistics By inspecting Figure 4.2.1 one can observe how in ground-surveyed areas of East Cap and Sisters Creeks "not-visible" failures do contribute significant amounts of debris to the failures already detected on air photos. Additional amounts of debris inventoried by means of fieldwork ranges among volume classes from 10% to 18%. This observation 40 deserves a more in depth analysis to see whether the two types of survey have similar relations within the two sub-areas. E O > c . ID 100 80 60 -. <D , An CD . > E 3 ' o 20 o : o -o -' i ....6--—6---"--<!>"-" P " " : y i ...A- A A .?'''yf"" ; / >; } p 6/ - . , 6 - " - 6 ' ' . ' Ar.T.,.,,r,i' -A- A ' O Air Photo & Field Work -A-- Air Photo i i O - ' o : O CM O O O o o o o o o o o o CO o o o o o o CO o .o o en o o o o o o -••';••-/'. j-r .-.-, , Landslide Volume.(m)' ; " " ' •,:;-"• Figure 4.2.1 Cumulative Percent Volume Distributions (East Cap & Sisters Creeks). Table 4.2.1 shows that the contribution of "not visible" events to the total number of landslides and, more important, to the total volume of mobilised debris differs greatly between the two tributaries of the Capilano River. In other words the lumping of the two study areas is clearly hiding the spatial variability of the two variables (number of landslides, m3 of mobilised debris). When East Cap and Sisters Creeks are analysed separately, denudation rates show a one order of magnitude difference. It would clearly be inappropriate to generalise such figures over one physiographic region (i.e. Pacific Ranges of the Coast Mountains). The question that arises is how to explain this large difference? 41 Table 4.2.1. Denudation rates from landsliding and landslide densities during a ~ 30-yr window (1968-2000) as obtained from air photo interpretation coupled with intensive ground checking. Location Field surveyed area (ha) Number of Landslides Mobilised Volume (m3) Landslide Density Landsliding Annual Rate Denudation* Denudation Annual Rate** Air photo Field Total Air photo Field Total LS/ha LS/ha/yr m3/km2 m3/km2/yr m3/ha/yr East Cap Creek 668.4 14 20 34 7869 412 8281 41.2% 58.8 % 100% 95% 5 % 100% 0.051 0.0017 1239.7 41.3 0.41 Sisters Creek 615 20 117 137 52480 21999 74479 14.6% 85.4 % 100% 70.5% 29.5% 100% 0.223 0.0074 12110.4 403.7 4.04 Sum 1283.4 34 137 171 60349 22411 82760 0.133 0.0044 6450.5 215 2.15 * Denudation = Mobilised Volume / Surveyed Area **Denudation Annual Rate = (Mobilised Volume / Surveyed Area) Time window East Cap Creek and Sisters Creek share similar geology, vegetation cover and climate. As it will be shown in section 4.3 (cf. Table 4.3.7), using both ANOVA and nonparametric tests on old-growth forest (control), volumes of mobilised debris per unit area are significantly greater in Sisters Creek than in East Cap Creek. In order to clarify these unexpected results we decided to perform a GIS-based topographic analysis (cf. Section 4.5). Table 4.2.2. Denudation rates from landsliding during a ~ 30-yr window (1968-2000) as obtained from air photo interpretation. Location Landslide Density (LS/ha) Landsliding Annual Rate (LS/ha/yr) Denudation (m3/ha) Denudation Annual Rate (m3/ha/yr) East Cap Creek 0.021 0.0007 11.7 0.39 Sisters Creek 0.033 0.0011 85.3 2.84 42 Denudation rates and landslide densities obtained from air photos only (cf. Table 4.2.2) are lower than those obtained by coupling intensive ground truthing. The ratio between the two types of survey varies with location and with the dependent variable considered, higher in Sisters Creek, lower in East Cap Creek, higher for denudation rates, lower for landslide densities (cf. Table 4.2.3). Also the Sisters-East Cap ratio increases when one couples fieldwork to the routine inventory method. The highest survey ratio is in Sisters Creek for landslide density, which is about 7, while the highest Sisters-East Cap ratios are recorded for denudation, 7.3 with air photo and nearly 10 with field-coupled remote survey. Table 4.2.3. Denudation and landslide density ratios. Ratio Landslide Density Landsliding Annual Rate Denudation Denudation Annual Rate Air photo & Ground / Air photo East Cap Cr. 2.4 1.05 Sisters Cr. 6.8 1.42 Sisters / East Cap Air photo 1.6 7.3 Air photo & Ground 4.4 9.8 Now let's see if "not visible" events contribute significantly to the sediment loading of montane stream channels or if they are mainly unconnected to the stream network. This aspect has clear management implications in terms of water quality (i.e. Capilano Reservoir) and fish habitat. According to the field data, about 58% of the mobilised volume (cf. Table 4.2.4) enters the channel network directly, 35% is delivered to gullies and most probably will be evacuated to the stream network via debris torrent in a decennial time scale (from 10 to 40 years, cf. Thurber Eng. Ltd., 1996). However, air photo interpretation indicated that only 3.2 43 % of the mobilised volumes were delivered to gullies, and here lies the highest discrepancy between air photo and field data. This is an extraordinarily important finding (a) from a process perspective and (b) from an applied perspective; it also raises a further question - why are there more gully failures in Sisters than East Cap? Table 4.2.4. Sediment delivery to streams (stream connection). Stream Connection Field Air Photo SUM Gully Stream No TOTAL Gully Stream No TOTAL Gully Stream No TOTAL No. of Landslide 52 67 18 137 3 27 4 34 55 94 22 171 % No. of Landslide 38 48.9 13.1 100 8.8 79.4 11.8 100 32.2 55 12.8 100 Volumes (m3) 7848 12927 1636 22411 1945 47139 11265 60349 9793 60066 12901 82760 % Volume 35 57.7 7.3 100 3.2 78.1 18.7 100 11.8 72.6 15.6 100 Through fieldwork coupling of the air photo inventory the percentage of gully-connected failures and the associated mobilised volume increased about four times (from 8.8 % to 32.2 and from 3.2 % to 11.8 %). The proportion of stream-connected failures decreased by nearly one-third, and the mobilised volume displayed only 5.5 % reduction. Percentage of unconnected debris slides showed minor changes both in terms of number of events and volumes. If one integrates field with air photo measurements, 72.6% of mobilised volumes were delivered to the stream channels. In terms of number of failures, about 49% of "not 44 visible" events were directly connected to stream channels and 38% to gullies. 55% of the total number of landslides detected with the field-coupled survey were delivering sediment to streams. Such percentages demonstrate how a comparison with Robison et al.'s findings, where only failures delivering sediment to streams were considered, would be misleading. 4.2.2 ANOVA and Nonparametric Tests Table 4.2.5 One-way ANOVA-like of the entire database (Survey). Factor Survey Dependent Variable Volume (rrrVha) Number (LS/ha) 1-way ANOVA Kruskal - Wallis ANOVA Median test Significance ns .002 .0003 .0000 .0001 .0000 Table 4.2.5 shows how field-coupled survey gives values for both dependent variables that are significantly different from the remotely-based one. The only exception is the one-way ANOVA for volumes of mobilised debris. Values for field-coupled air photo survey are significantly greater than solely air photo-based values (t-test). The large differences in landslide densities and mobilised volumes of debris described between the two sub-basins (locations) have prompted the writer to handle locations as fixed blocks. Blocking was justified indeed for both dependent variables, location terms being significant. A fixed block design, when blocking is used to increase precision of comparisons, assumes no block-by-treatment interaction (i.e. location-by-survey), while such interaction is of primary interest when blocking is used to broaden the scope of inference (Leon and Mee, 2000). As for the full factorial experiment, the advantage of having a factorial arrangement of treatments (or factors) over a block design lies in the fact that one can analyse interactions, 45 and in case of not meaningful interactions one can reduce the number of treatment means in the multiple comparison test. Table 4.2.6. Two-way ANOVA-like of the entire database (Survey-Location). Dependent Variable Volume (m3/ha) Number (LS/ha) Factor Survey Location Survey * Location Survey Location Survey * Location Full Factorial Experiment Factorial Exp. in a Fixed Block Design Friedman Test Significance ns .002 ns .003 .001 .011 ns .002 n/a .001 .001 n/a .000 n/a n/a n/a According to the full factorial ANOVA (cf. Table 4.2.6) the interaction is significant for one dependent variable (LS/ha). Therefore it is not possible to comment on Locations or Survey separately, instead, having detected the significant interaction, its "meaning" has to be graphically evaluated by inspecting the plot of the means (cf. Figures 4.2.2-3). Such plots display that the coupling of air photo interpretation with intensive fieldwork has sensibly increased average values of both dependent variables in Sisters Creek. In East Cap Creek such increase appears to be negligible. This difference between the two locations is particularly large in terms of landslide density (LS/ha), making the location-by-survey interaction significant. As for the volumes of debris per hectare (m /ha), they show no significant survey-by-location interaction, hence one can comment on Survey and Location separately. The former is not significant; the latter is significant (cf. Table 4.2.6), with Sisters greater than East Cap. The.nonparametric equivalent of a two-way ANOVA, the Friedman test, requires the same assumptions as the fixed block design (i.e. null survey-location interaction) and as such 46 it can be performed only on mobilised volumes of debris (cf. Table 4.2.6). The test indicates that at least one of the Survey-Location combinations is different from the others; this is the field-coupled air photo interpretation conducted in Sisters Creek. Among a variety of other multiple comparison tests, Tukey's HSD (Honest Significant Difference) ' has been selected as the proper one to be performed. It is considered to be of intermediate conservatism. The test, in line with Friedman's indication, tells us that the coupling of air photo interpretation with intensive fieldwork in Sisters Creek gives a significantly greater amount of mobilised debris per unit area (m3/ha) than any other survey-location combination does. Without a two-way non-parametric test one would not have detected that volumes per hectare obtained via field-coupled survey in Sisters are significantly greater than all other location-survey combinations. Hence, solely from air photos East Cap and Sisters would not have shown a significant discrepancy (AP/East Cap and AP/Sisters are not significantly different). This result stresses the importance of fieldwork. Table 4.2.7. One-way ANOVA-like: Survey within Locations. Factor Survey Location Sisters East Cap Dependent Variable m3/ha LS/ha mb/ha LS/ha 1-way ANOVA Kruskal - Wallis ANOVA Median test Significance ns .002 ns ns .0003 .0000 ns ns .0001 .0001 ns ns Finally, the database was split in two location-wise; hence a series of one-way ANOVA-like tests was performed. Regardless of the dependent variable examined, in East 1 In fact a generalisation of Tukey's test for unequal sample sizes is applied. 47 Cap survey types were not significantly different; conversely, in Sisters Creek field-coupled survey was significantly greater than the air photo-based one (cf. Table 4.2.7). SURVEY'LOCATION 06 0.5 h • j %L* ?:2 \ 0.4 0.3 0.2 0,1 0.0 0.1 I I I k l isters •'"'•"'-" - O - A.r Phjto - a - Air Photo & Ground -,,'. • East Cap Figure 4.2.2 Survey-Location interaction of landslide density. SURVEY'LOCATION 250 20C '50 '00 I j= 50 50 1f0 - O - Air Photo -m- Air Photo & Ground Sisters East Cap Figure 4.2.3 Survey-Location interaction of mobilised debris. 48 4.3 The Impact of Forest Management on Mass Wasting The impact of logging on slope stability was approached through analysis of annual rates of landsliding and annual rates of denudation as compared between disturbed and undisturbed forest (cf. Tables 4.3.1-2). Then, in conformity with the analytical approach adopted in section 4.2, landslide density and the related volumes of mobilised debris were selected as dependent variables and analysed via ANOVA and its nonparametric counterparts (cf. Tables 4.3.4-7). Table 4.3.1. Management effects on annual rates of landsliding* as obtained from fieldwork-coupled air photo-based landslide inventory. SUB BASIN Sisters Cr. East Cap Cr. LAND USE Undisturbed Forest Old Logging Undisturbed Forest Recent Logging Rate of landsliding (LS*ha"1*yr1) Management Effects STABILITY CLASS III 0.0079 0.0021 0.0011 0.0034 IV 0.0075 0.0051 0.0008 0.0023 V 0.0098 0.0078 0.0023 0.0059 III IV V 1.0 x0.3 x0.7 x0.8 1.0 x3.1 x2.9 x2.6 * Annual Rate of Landsliding = r£(Number of landslides) / £(Area of surveyed polygons**)! Time window **Polygons of equal Land use and Stability Class. Annual rate of landsliding spanned over an order of magnitude, from 8*10"4 to 9.8*10"3 LS*ha"'*yr"' (cf. Table 4.3.1). Annual rate of denudation varied between 7*10~3 and 7 m3*ha"1*yr"1 (three orders of magnitude, cf. Table 4.3.2). In class V landslide density displayed a slight decrease as a result of old logging (Sisters Creek) and a moderate increase (x 2.6) after recent logging (East Cap Creek); denudation recorded an acceleration factor of 9 in East Cap and of 3 in Sisters. Class IV 49 showed minor differences in density (x 0.7) and denudation (x 0.9) in Sisters. Major acceleration appeared in East Cap, though the calculations are made on a relatively small area with a low background (natural) rate. In class III recent logging increased landslide density while old logging is associated with reduced effects (probably the logging sites were chosen for assurance of stability). Denudation showed major acceleration in East Cap but, again, calculations are made on a relatively small area characterised by a low undisturbed rate. Table 4.3.2. Management effects on annual denudation rates* as obtained from fieldwork-coupled air photo-based landslide inventory. SUB BASIN Sisters Creek East Cap Creek LAND USE Undisturbed Forest Old logging Undisturbed Forest Recent logging Denudation rate (m3*ha"1*yr1) SI III 0.13 0.39 0.007 0.31 IV 0.67 0.60 0.009 0.76 Management Effects rABILITY CLASS V 4.51 7.04 0.64 5.57 III IV V 1.0 x1 .6 x0 .9 x 3 1.0 X44.3 X84.4 x8 .7 * Annual Denudation Rate = [^(Mobilised Volume) / £(Area of surveyed polygons**)! Time window **Polygons of equal Land use and Stability Class. To note that class IV polygons show somehow unexpected behaviour (they never rank between class III and V values, as one theoretically would predict), such outcome undoubtedly reflects the objective difficulty of delineating/labelling class IV (potentially unstable) polygons. In addition, one should consider the choice of logging sites, which will tend to be more conservative in higher (in)stability classes. Observed effects may be induced by "good" management of harvest. Rate of landsliding showed a fairly constant acceleration 50 factor within locations (i.e. East Cap Cr., Sisters Cr.) across stability classes. Conversely, the same factor displayed a high degree of variability for denudation rate. Background denudation rates (i.e. Undisturbed forest) are notably higher in Sisters Creek than in East Cap Creek, Sister/East Cap ratio ranging between 7 (class V) and 70 (class IV). Such numbers should be interpreted carefully because of the crude/simple style of the analysis; it is believed that more reliable results are obtained from ANOVA-like tests (cf. later in this section). Table 4.3.3. Management effects on rates of landsliding and denudation: comparison between air photo-based and field-coupled landslide inventories. SURVEY Air photo Air photo & Ground SUB BASIN Sisters East Cap Sisters East Cap LAND USE Undisturbed Forest Old Logging Undisturbed Forest Recent Logging Undisturbed Forest Old Logging Undisturbed Forest Recent Logging LS*ha"1*yr"1 0.0023 0.0005 0.0005 0.0026 0.0093 0.0052 0.0016 0.0034 m3*ha 1*yr1 1.92 0.66 0.33 1.46 3.49 2.59 0.34 1.53 Management Effects l_S*ha"1*yr"1 1.0 0.2 1.0 5.2 1.0 0.6 1.0 2.1 m3*ha"1*yr"1 1.0 0.3 1.0 4.4 1.0 0.7 1.0 4.5 The comparison of the management effect between air photo-based and field-coupled surveys (cf. Table 4.3.3) showed "minor" differences. Fieldwork had the effect of shifting the old logging-old growth ratio towards unity (no acceleration), a ratio that one can justify by saying that the old logged terrain had recovered from forestry disturbance. According to air photo derived data recent logging produced an acceleration factor of x 5.2 (in terms of 51 landsliding rate — LS/ha/yr) as opposed to x 2.1 obtained in the field-coupled survey; management effect of recent logging on denudation rates remained the same between survey methods. Table 4.3.4. Two-way ANOVA of the entire database (Location-Stability-Land use). ANOVA setting Factorial Experiment in a Fixed Block Design Full Factorial Experiment ** Dependent Variable m3/ha LS/ha m3/ha LS/ha Factor Location Stability* Land use* Location Stability* Land use* Location Stability Land use Location Stability Land use Significance ns .012 ns .026 ns ns ns .015 ns .032 ns ns Stability-Land use interactions are not significant. ** All interactions are not significant. In light of what is reported in Table 4.3.4, location blocking was justified for landslide density (p < .05) and not justified for mobilised volumes (p > 0.1), hence a full factorial approach is to be recommended for the latter variable. Therefore locations are significantly different in terms of number of landslides per unit area (Sisters having greater values than East Cap). Stability classes are significantly different in yield, post-hoc tests (i.e. Bonferroni's and Sheffes', not reported here) show how class V has significantly greater mobilised volumes of debris than both class IV and class III. The adoption of a Full Factorial Experiment (in place of the blocking) has not determined any significant improvement to the explanation of the variability. Friedman's test would be the proper nonparametric test one would apply as alternative to a Factorial Experiment in a Fixed Block Design. However, the test does not admit missing values, in 52 other words it requires every treatment to have equal (or very similar) number of replications, which is not the case for the several treatments/factors of this database. As for the Full Factorial Experiment, no nonparametric equivalent is known. This situation forced me to simplify the experimental design; hence to move from a two-way to a one-way layout, where equivalent nonparametric tests are Kruskal-Wallis ANOVA and median test (Cf. Section 4.2 for extended explanation). Table 4.3.5. One-way ANOVA-like of the entire database (Location-Stability-Land use). Factor Location (East Cap, Sisters) Stability (III, IV, V) Land use* (OG, Ho, Hr, F) *OG: old growt Dependent Variable my/ha LS/ha m3/ha LS/ha m3/ha LS/ha 1; Ho: old harve 1-way ANOVA Kruskal - Wallis ANOVA Median test Significance .009 .002 .004 ns ns ns .003 .004 .001 .010 ns ns sting; Hr: recent harvesting; F: fire. .009 .016 .009 .011 ns ns Table 4.3.6. One-way ANOVA: Land use within Locations. Location Sisters East Cap Factor Land use (OG, Ho) Land use (OG, Hr) Dependent Variable mb/ha LS/ha m3/ha LS/ha 1-way ANOVA Kruskal -Wa l l i s ANOVA Median test Significance ns ns ns ns ns .029 ns ns .046 .046 ns ns Table 4.3.7. One-way ANOVA: Undisturbed Forest between Locations. Dependent Variable ma/ha LS/ha 1-way ANOVA Kruskal -Wal l i s ANOVA Median test Significance .036 .007 .007 .004 .017 .036 53 In contrast to what was found with the Fixed Block Design, one-way layout analysis (cf. Table 4.3.5) shows that Location and Stability are significantly different for both dependent variables (Median test, p < .05). 2600-•0 2200 to 1800. Q> 1400! y,o . -* :'*•> 100.0 • ~o -?, < 03 ). Ml '*' . 1 600., 200. -200; I Min-Max I I 25%-75% • Median value 'iv«;;rS Old-growth forest ':-'- ' \ LAND USE '^ | ; | ;bld.J6ggir tg ; , •ft?. Figure 4.3.1. Box-Whisker plot of the medians of mobilised volumes: Old-growth forest vs. Old Logging. 5'."5>-r 4.5 CO w 3.5 g 2 5 Q '. CD 'CO ? -a-c-> CO 1.5 0.5 -0.5 • I Min-Max CD 25%-75% • Median value OldT'growth forest "•"•• Old logging L'A'ND USE Figure 4.3.2. Box-Whisker plot of the medians of landslide density: Old-growth forest vs. Old logging. 54 Table 4.3.6 reports the analysis results of the database restricted only to Undisturbed Forest, Recent Logging and Old Logging. This data arrangement reveals that Old Logging gives significantly lower values (Median test, p = .046) for both dependent variables than Undisturbed Forest does (cf. Figures 4.3.1 and 4.3.2). The same variables, when compared between Recently Logged polygons and Undisturbed ones, do not display significantly different values (cf. Table 4.3.6): East Cap. Despite the fact that East Cap was logged with more careful (environmentally friendly) techniques, the longer time that Old Logged terrain in Sisters had for recovering than recently harvested polygons in East Cap appears to dominate in controlling landslide density and the associate rate of denudation. In seeking to eliminate land use confounding between locations (i.e. Sisters and East Cap), the database has been restricted exclusively to terrain covered by Undisturbed Forest. Landslide density and its associated mobilised volume of debris are significantly greater in Sisters Creek's "old-growth" polygons (cf. Table 4.3.7), than in East Cap Creek ones. The pattern of outcomes is somewhat complex. The following statistical differences were found for both dependent variables: Sisters Creek significantly greater than East Cap Creek and old-growth forest in Sisters Creek significantly greater than in East Cap Creek. Class V was recognised as significantly greater than class IV and III, old logging was significantly smaller then old-growth forest (though p = 0.046 is a borderline value). Descriptive statistics and ANOVA-like tests generally produced consistent results. Exceptions are represented by the management effects due to recent logging and old logging. In the first case descriptive statistics displayed high acceleration factors (notably in terms of denudation rates, cf. Table 4.3.2) whereas ANOVA-like tests did not detect significant 55 differences between recent logging and undisturbed forest (cf. Table 4.3.6). As for old logging, descriptive statistics displayed management factors ranging between 0.9 and 3 for mobilised volumes (cf. Table 4.3.2), by contrast, the median test (not Kruskall-Wallis) indicated higher median volumes mobilised from old-growth forest than from old logging. 4.4 GIS as a Tool for Justifying Inter-basin Variability of Denudation Rates In seeking a qualitative explanation for the one order of magnitude difference in landslide denudation rates between the two study areas, a GIS-based topographic analysis was performed of the area analysed. Elevation, slope gradient, slope aspect, drainage density and spatial partition of surficial materials across stability classes are the variables under discussion. Table 4.4.1. Elevation and Slope Gradient as extracted from a 25-metre DEM. Variable Location Mean Std. Error of the Mean Std. Deviation Min Max Count Elevation East Cap Cr. 832 1.8 221 338 1320 15655 Sisters Cr. 664 1.6 181 265 1195 12865 Slope Gradient East Cap Cr. 28.1 0.06 8.3 1 59 15655 Sisters Cr. 31.1 0.08 9.1 1 62 12865 In making the final comparison of elevations and slopes between the two study areas we will inspect the average values together with the frequency distributions. This is because the average elevation/gradient in one location or the other may be inflated or depressed by the range of slope included in the analysis. 56 EAST CAP — • SISTERS I - - - - « I ! ; . , " " "? 1 I 200 400 600 800 1000 1200 1400 1600 Elevation (metres) Figure 4.4.1. Elevation frequency distributions: Sisters Creek vs. East Cap Creek. East Cap Creek is located at higher elevation than Sisters Creek, and mean slope gradients are significantly different (cf. Table 4.4.1). The inspection of the frequency distributions seems to confirm this situation (cf. Figure 4.4.1). Maximum frequency is reached at the 600-800 m category in Sisters, whereas elevation frequency peaks at the 1000-1200 m category in the other Creek. Higher elevation might have slightly contributed, as we will see later on (cf. Distribution of surficial materials across stability classes) to a faster evacuation of glacial till deposits from East Cap upper and mid slope locations (higher elevation -> higher precipitation -> faster hydrologic cycle -> faster regime of sediment transport). In contrast to what is observed for elevations, the two slope frequency distributions have similar shape (cf. Figure 4.4.2). K-S test showed that distributions were not significantly different for p < 0.1. However, average slope in Sisters Creek is significantly higher than in East Cap Creek. 57 6000 |T 5000 CD o m 4000 • a "S ° 3000 > • . o § 2000 cr o> x_ ^ 1000 *• 4500.'r 4000 •< m : '%°»?' "55 '•'•%, ^ 3000 UJ ;'. -i-\Z. 2500-° ••; -. z 2000 ' '•:| ;>15"p0' £ 1000 ,.,LL .''.'500 0 ;. "'1, 0-19r"9. 20-24.9 25:29.9 30-34.9 35-39.9, 40-44.9 v 45-49.9 • > 50, " :' " : • • , . . . * , : -, .'"V- -{'A.;.'" - <<>, ( i ' ^ j - - •^ -•> • i">: .?i\v *~V ' ' '- ' T~ ' 'Slope Gradient (degrees) '«.. " ' ? Figure 4.4.2. Slope frequency distributions: Sisters Creek vs. East Cap Creek. Martin et al. (in press.) data from the Queen Charlotte Islands indicate a critical gradient for sliding of about 25° to 30°. Landslide initiation was found very unusual below such values. Though it is difficult to generalise about a critical slope gradient for mass failing, it is reasonable to assert that terrain with slopes over 35° can be regarded as shallow mass movement-prone, when soil is present (Sidle et al. 1985). In such respect, percentages of slope steeper than 35° are significantly different (Z-test), Sisters Creek having 34% of DEM cells with slope gradient steeper than 35°, East Cap Creek only 19.5%. The influence of slope aspect on landsliding deserves particular attention; several studies have examined this issue in the study area. Approximately 75% of the landslides inventoried by O'Loughlin (1972) in Howe Sound, Capilano and Seymour watersheds exhibited a southerly aspect. He justified such landslide preferential orientation by noting that "north-facing slopes are rocky and broken, a condition which discourages landslide \ 1 • • ; • EAST CAP • • • • SISTERS i • .............L............J • 1 ; \ \ ; i ! ! : ; • : ; ; ; i I 58 formation, while south-facing slopes are relatively uniform and underlain by an extensive unweathered till substratum". This spatial arrangement of surficial materials is likely to derive from the north to south flow line that characterised the movement of Pleistocene ice sheets in the Pacific Ranges (Armstrong and Brown, 1954). Relations between surficial materials and aspect have been investigated more systematically for the entire Capilano Watershed (GVRD, 1999). The analysis focused on till, colluvium, and rocky terrain. Results reflect what O'Loughlin previously suggested. Rock dominates northerly aspects, colluvium and till are most extensive on west, south and east facing aspects (Cf. Table 4.4.2). 5000 4000 30C0 2000 1000 EAST CAP SISTERS r -• mm mm mm 1 i _ | ! \ •••• i N NE E ; SE \ ' S Slope Aspect:-. SW W NW Figure 4.4.3. Slope Aspect as extracted from a 25-metre DEM. Table 4.4.2. Relation between Surficial Materials and Aspect, Capilano Watershed (after GVRD, 1999). Aspect North South East West Colluvium (ha) 984 1519 1741 1915 Till (ha) 510 1086 1344 854 Bedrock (ha) 1587 1695 1943 2134 Area per Aspect (ha) 3122 4502 5220 5092 Colluvium Index 0.32 0.34 0.33 0.38 Till Index 0.16 0.24 0.26 0.17 Bedrock Index 0.51 0.38 0.37 0.42 59 In addition to the relative distribution of surficial materials, climate is expected to be in part responsible for making south and west-facing slopes the most landslide-prone ones. Typically in the Pacific Ranges, heaviest rains are brought by southwesterly air-flows; south and westerly facing slopes are therefore more directly exposed. Secondly, northerly exposures are likely to remain more uniformly damp (no abrupt fluctuations in pore water pressure). Lastly, alternation of the conditions in weathering environment brought on by wet/dry cycles is enhanced on southerly facing slopes. In light of the above, southerly and westerly oriented slopes are expected to be more favourable for landsliding. Figure 4.4.3 shows that south-facing slopes are of comparable extent in East Cap and Sisters Creeks, west-facing slopes are more abundant in East Cap Creek. Therefore, in terms of slope aspect, East Cap Creek should be more susceptible to landsliding than Sisters Creek. Figure 4.4.4. Landslide distribution by aspect for the Capilano watershed (from GVRD, 1999). 60 Apparently landslides do not occur at any preferential slope aspect, though they appear to be slightly more abundant on southerly and easterly orientations (cf. Figure 4.4.4). This could be explained by the relative predominance of rock on northerly and colluvium on westerly facing terrain. Thus it is unlikely that aspect plays a significant causal role for landsliding in the watershed. Drainage density (cf. Table 4.4.3) is greater in Sisters Creek than in East Cap Creek, hence partly explaining the greater denudation rate for the former sub-study area. In the extraction of the drainage density, to avoid any "edge effect" (which would result in overestimating the variable), it was decided to include in the calculation all terrain polygons adjacent to the surveyed stream channels, even when only one side of the stream was ground checked. This explains why field-surveyed areas in Table 4.4.3 are greater than those reported in Table 4.2.1. Table 4.4.3. Drainage density of the field-surveyed area as extracted from a 25-metre DEM. Location Terrain Area (km2) Total Channel Length (km) Drainage Density (km2/km) East Cap Cr. 9.8 21.1 2.2 Sisters Cr. 8.0 29.0 3.6 In terms of the spatial distribution of surficial materials throughout stability classes (cf. Figures 4.4.5 and 4.4.6) two main points can be made. Firstly, in Sisters Creek there is a dominance of class V cover, in East Cap Creek class IV covers the majority of the land. Thus, the B.C. terrain stability classes are consistent (same direction) with the outlined discrepancies in denudation rates. Secondly, if one considers only class V (unstable) 61 Sisters Creek m M • • R 1 1 C ^ /CM Figure 4.4.5. Distribution of surficial materials in Sisters Creek. East Cap Creek Figure 4.4.6. Distribution of surficial materials in East Cap Creek. 62 polygons, while colluvium (C) and bedrock (R) cover almost the same surficial extension in the two sub-areas, till deposits (M) cover in Sisters Creek an area that is more than double the till mantled terrain in East Cap Creek. In addition, there are nearly 0.2 km2 of terrain where colluvium partially covers underlying till (/CM), a configuration that can be considered as unstable as till itself. Briefly summarising, a denudation rate from landsliding that in Sisters Creek is one order of magnitude greater than in East Cap Creek can be explained by considering that Sisters Creek possesses: A. steeper slopes, 34 % steeper than 35°, whereas in East Cap Creek only 20% of the terrain is steeper than 35°; B. major till on class V slopes, where it covers a surface of 2 km versus just 0.7 km in East Cap Creek; C. a higher drainage density, 3.6 km/km2 compared to 2.2 km/km2 of East Cap Creek. From here to manage to explain quantitatively (by means of a deterministic model) rates of sediment production there is still a long way to go and I am not going to pursue such task in this study project. 63 4. s Landslide Magnitude-Frequency Relations Landslide volume frequency distribution obtained from air photo interpretation (cf. Figure 4.5.3) possesses a "gamma-like" shape. By contrast, Figure 4.5.1, that plots volume distribution of landslides detected via air-photo interpretation coupled with careful ground surveys indicates that the large number of "missed" failures transforms the distribution into an "exponential-like" one. Thus, in this study area the volume frequency distribution obtained by air photo-based landslide inventory deviates significantly from reality. The extent of the ground-surveyed area seems to be adequate to detect the entire magnitude range of "not visible" landslides. In fact, the decreasing trend for "small" events in figure 4.5.2 starts in the volume category 1600-1800 m3; accordingly the largest "not visible" event that we measured during fieldwork was estimated to be 1717 m3 (cf. Figure 4.5.2). Sisters & East Cap Creeks Air Photo & Ground 1800 *- 2600 3400 -..4200 ! ^V-tff'.,; .''-ft-" .•; Can dsl id & Volume (rr?)/itf 5000 5800 Figure 4.5.1. Landslide volume distribution (exponential-like trend) as obtained from air photo interpretation coupled with intensive ground survey. 64 Sisters & East Cap Creeks 200v;, i40P 600 800 1000 1200, .vr4,0<K\Kl600 1800 2000 .Landslide Volume (m 3) • .« Figure 4.5.2. (A). Histogram showing the magnitude range of landslides "not visible" from air photo interpretation, and hence detectable only via ground survey (B). Particular of events not greater than 2000 nr. 65 «50. 40 CJ r-<i> - i cr UL-' 30 ;.2o •10 i o o o o CM OJ O - ' o CM CM Capilano River Basin Air Photo UUJJM MhA Mr^-Mr-o ,.. CM CO O O O O CM CM o o CM CD O o CM. ,, i ' f o o o o o o o o CM CM CM CM rr» rr» o -i— Landslide Volumefm Figure 4.5.3 Landslide volume distribution (gamma-like trend) as obtained from air photo interpretation (scale 1:15,000). Capilano River \m 3 O O O O _ J O O O O ' LO <Q N O O O O K t -O O O O O O O O O O O O O O O O O O LO (D N 00 0 0 Landslide Volume (m Figure 4.5.4. Landslide Magnitude-Frequency (volume categorical bins: 200-m3 step). 66 The calculation of landslide frequency (intended as number of events per unit area per year) allows extrapolation of East Cap & Sisters Creeks findings to the whole Capilano watershed (cf. Figure 4.5.3). The kink observed for air photo detectable landslide, at about 3200 m , is highly reduced (if still existing) when one integrates remotely sensed survey with intensive ground survey. We expect to reduce the high scatter in the data around the magnitude-frequency power-law relationship by changing landslide volume categorical bins (that in Figure 4.5.4 has a value of 200 m3). _c Field-coupled A/P landslide inventory LS volume frequency distribution of East Cap and Sisters Creeks LS magnitude-frequency (#LS/km2/yr)ofEastCap and Sisters Creeks + A/P-based landslide inventory LS volume frequency distribution of the Capilano watershed LS magnitude-frequency (#LS/km2/yr) of the CaDilano watershed A/P-field coupled LS magnitude-frequency (#LS/km2/yr) of the Capilano watershed C Predicted LS volume frequency \ distribution of the Capilano watershed J Flowchart 4.5. Flowchart showing how was obtained the Capilano predicted landslide volume distribution. 67 Back-transformation of the magnitude-frequency values of figure 4.5.4 allowed obtaining (through extrapolation) the "predicted" landslide volume distribution (air photo coupled with fieldwork) for the entire Capilano river basin (cf. Figure 4.5.5). !"}:' ''}(§'','';"'' Capilano Watershed: Predicted Landslide Frequency Distribution - -.---1600 '^  - — 1400 1200 ,J000 c g 800 UJ o 600 400 200 ,-j 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 , 0 0 0 0 0 0 0 o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o ' I I D C M I O O ' t t D W l O O ' t C O C M t D O ^ C D W f f l O ' t f f l W I D O f l f f l C l t D O r ^ p i n i N n n ^ f l ^ i n n ( O t & ( D N S f f l ( D ( 0 0 ) 0 ) O O O r r t M Landslide Volume (rrT Figure 4.5.5. Landslide Frequency Distribution (volume categorical bins: 200-m3 step). ' T-S Capilano Watershed 0 001 - O - Air photo -+- Air photo & Ground : .„ -A 10 100 Landslide Volume (m. 1000 10000 Figure 4.5.6. Landslide Magnitude-Frequency (volume categorical bins are logarithmically equally spaced). 68 Figure 4.5.6 represents the "classical" way of plotting landslide magnitude-frequency relationship (Hovius et al., 1997, 2000; Stark and Hovius, 2001), that is arranging volume categorical bins as "logarithmically equally spaced". This representation, if on one hand it minimises the scatter, on the other produces very small volume bins towards the domain of "smaller" failures. The reduction of the sample population per bin has the effect of enhancing the kink around 3200 m3. The adoption of 400-m metric categorical interval (cf. Figure 4.5.7) has the effect of minimising the scattering around the power-law relation, a kink around volume 3000 m3 can also be observed. A single power-law does not seem to fit the whole range of landslide magnitude-frequency. As also indicated by Stark and Hovius (2001) a double power-law (or Pareto model fit) appears to allow the fitting of a greater range of volumes, and as such is less affected by the scatter in the high magnitude tail. Capilano Watershed , "'v0 100 t-'E*'-'6.010 ' '>.-• i f »-' 1 O •;;«- • . . ; . [' c ' • • - . . ' • • • ' " . . 13 .'r • ••cr- ».v, . ' : £ 0.001 Landslide Volume (m3) Figure 4.5.7. Landslide Magnitude-Frequency (volume categorical bins: 400-m3 step). CM CO ^ - L O C O r - C O C D O 69 In figure 4.5.9 the magnitude-frequency relation of the Capilano is plotted together with two areas located in the Insular Ranges of British Columbia: the Queen Charlotte Islands' (QCI) and San Juan River2 (SJR), this latter located on the south-western shore of Vancouver Island. In comparing/analysing the three areas one has to take into account that different mappers have compiled the landslide inventories. Then differences in land use (cf. Table 4.5.1. both recently and old harvested areas), climate, and physiography (hence bio-geo-climatic zone) are also expected to exert major influences. Both inspection of landslide volume distributions (cf. Figure 4.5.8) and cumulative percent volumes (cf. Figure 4.5.10) will help in detecting relevant differences. Table 4.5.1. Logged areas in the QCI, the Capilano and the San Juan basins. Location Watershed Area (km2) Area Logged** (km2) Capilano River Basin 198 43 (22%) Recent 5 (2.5%) Old 38 (19.5%) Queen Charlotte* Islands 351 47 (13.4%) San Juan River Basin 517 129 (25%) This study comprises 27 different watersheds. ** Recent: logged not more than 50 years ago. Old: logged more than 50 years ago. A few differences can be pointed out, the Capilano watershed has definitely a smaller proportion of large events, the largest event falling in the 10,200 m3 category; events larger than 12,200 m represent 2.6 % and 1.2% of the landslides occurring respectively in SJR and in the QCI. San Juan River displays the highest variability in landslide size (longest tail). In the QCI 83% of the landslides were smaller than 2000 m3 (cf. Figure 4.5.10). 1 Source: Rood (1984). 2 Source: Northwest Hydraulic Consultants (1997). 70 140 120 100 to S 80 > LU O 60 40 20 1 ' • : . : : : H Capilano River lllilBle 500 c CD 5 o I2U 100 80 60 40 20 n 1.1 : ; ; : ; : : ; : : : ; : ; : : • : : ; B I San Juan River 1 . 1 ; 1 ; • ; I I . I l . 11. I I . 11.11.11. 1 .11.11.E l .E l .9 . • . • .ma.M.m. tsn .m..n_._ ._ . • . . — ™ 11 o o o o o o o o o o o o o o o o o o o o 0 « « N ( 0 0 « C D CM,,CD ^ ^ T - w o j c o n n ^ ^ -o o o o o o o o o o o o O -st CO CM CD O l O LO U} CD CO h -o o o o o o o o o o o o o o o o o o o o C O M ( 0 0 » a ) I M l D O ' * N f l l ( 0 0 ) 0 ) 0 ) 0 0 r ^ o o o o o o CO CM CM T - CM CM Landslide Volume (m Figure 4.5.8. Landslide Volume Distributions. 71 0 . 0 1 . >. 0.001 -CJ <D .... , . Z3 O" CD l e - 4 < y '•... . / ! A'\ i'r"jt--s, K :H S \ A J ; : - ; ; V '•••.. i fc : : : ! : ; : : n ; : ' ' - o - Capilano -••— San Juan ••- QCI >:• \ '• i + ; + « * • . CM CO • * O O O O O O O O O O O O n 10 s o c n o O O O O O O O O O O O O O O O O O O O O O •* . in to s cooo ' Landslide Volume (fn Figure 4.5.9. Landslide Magnitude-Frequency from air photos: Capilano R., San Juan R., and Queen Charlotte Islands (volume bins are logarithmically equally spaced). 1400 3000' ; • 4600 6200 7800 -. 9400 11000": >12200 .Landslide Volume (m\) Figure 4.5.10. Cumulative Percent Volume Distributions. Values of the Kolmogorov-Smirnov (K-S) tests that were performed on the cumulative percent distributions are reported on table 4.5.2. The three areas are all significantly different from each other, though the degree of dissimilarity, as one could have 72 expected, is smallest between San Juan River and the Queen Charlotte Islands (Insular Ranges). Table 4.5.2. Kolmogorov-Smirnov tests of cumulative percent volume distributions. K-S Contrast Capilano-San Juan Capilano-QCI QCI-San Juan Significance p < .001 p < .005 p<.05 73 Chapter 5 DISCUSSION Similarly to the structure that was adopted for the results, the discussion is organised into five sub-sections: (1) landslide visibility from aerial photographs; (2) survey type, landslide density and associated sediment production; (3) the impact of timber harvesting on landslide initiation; (4) GIS-based analysis; (5) landslide magnitude-frequency distributions. 5.1. Landslide Visibility from Aerial Photographs The issue of landslide visibility was articulated here into three aspects: the maximum area of "not visible" landslides, the capability of landslide detection using a single photo set versus interpretation of multiple sequential photo sets, and landslide visibility through time (due to the pattern of vegetation re-growth on the scar). However basic the issue of estimating the maximum size of "air photo undetectable" landslide scar would appear to be, no study has faced the question systematically1. Possible explanations for such a gap in the literature are (a) the underestimation of the importance of small failures; and (b) the work of ground survey requires tenacity and personal discomfort. The value found for "not visible" failures on forest older than 50 years is larger (650 m2) than what has always been assumed in previous studies (Rood, 1984; Schwab, 1988; Rollerson et al., 2001). Such large events were located on the lower portion of very steep (between 40° and 47°) old growth forested slopes of Strachan Creek (Sisters Creek meso-' Robison et al., (1999) took up the question of not visible failures. They considered only channel-connected events and this constitutes an inherent bias in the evaluation/comparison of their findings. 74 watershed). On recently harvested cut-blocks the value dropped to about 150 m2. In "young" clear-cuts the problem in detecting slides lies in the shallow nature of the events (cf. Photo 5.1), which facilitates a very fast process of re-vegetation. Shallow landslide scars and accumulations of logging woody debris (slash) look alike from air photos, so often field visits are needed for verification. Photo 5.1. Shallow debris slide in a recently harvested cut-block of the East Cap basin. 75 In addition to the other visibility factors already encompassed, we found that specifically in narrow valleys (i.e. Strachan and Cameron Creeks, tributaries of Sisters Creek), relatively large failures are more difficult to identify on the lower portion of the slopes. Due to the progressive downhill increase of pore water pressure, clusters of debris slides are often found on the lower portion of the slope and frequently such failures had not been detected from antecedent air photo interpretation. Terrain stability mapping is routinely conducted through interpretation of one aerial photo set plus some ground checking (typically TSIL C 2). It is expected that by disposing of multiple sequential photo sets one would maximise the capability of detecting landslide scars through the rugged and densely forested terrain of coastal BC. We found that in a sample area of about 60 km2 only half of the events originally mapped (by another independent study: GVRD, 1999) as having occurred 30 years ago or later, were actually active during that time window. This aspect has important implications both in terms of class V polygons labelling (over-delineation of unstable polygons) and in terms of evaluation of sediment budgets (overestimation of denudation rates from shallow landsliding). It has to be admitted that, due to the elongated shape of channelised debris torrents, their recognition from air photos is inaccurate, nevertheless one cannot assume that every major gully experienced a debris torrent during the last 30 years. Therefore one is left with nothing but the best air photo interpretation achievable. In support to this speculation, Thurber Eng. Ltd. (1996) reported that debris torrents in the Capilano Reservoir area occurred with a recurrence time ranging between 10 and 40 years. If one considers that this statement was made on the basis of air photo coverage of 54 2 TSIL (Terrain Survey Intensity Level) refers to the proportion of polygons that are checked in the field. In the Capilano a percentage between 25% and 50% of polygons was ground checked, this corresponds to the level C. 76 years (1936-1990) it is clear that this monitoring time was not long enough for a 40-year event. Hence inherent errors are to be expected and recurrence intervals may be greater than 40 years. For testing landslide visibility through time we selected exclusively episodic mass movements, this because retrogressive headwall failures, sidewall slides, and the hydrologic regime of gullies prevent vegetation from colonising the majority of gully-related (recurring) scars. In fact, the lack of vegetation re-growth on a gully from air photo can be interpreted as if a debris flow had just occurred even when this was due to minor sediment washing. Hydraulically, gullies are classified as first or second order streams (Horton, 1945). They have a seasonal regime and occupy only a small portion of the whole landscape, however they are key sites for sediment supply and transport processes. Steep sidewalls and headwall compose the initiation portion of a gully, the major area of sediment supply. Due to their morphology and moisture condition, they are very susceptible to mass wasting events, in particular channelised debris flows. Debris flow, in particular, is the main geomorphic process that occurs along a gully (Swanston and Swanson, 1976). From aerial photos and field traversing of Sisters and East Cap Creeks, the gully network appears to have been very stable in the last 30 years. Only the formation of one new gully (scoured down to bedrock by a large debris flow) was recorded in November 1990 in Sisters Creek (as also reported in Thurber Eng. Ltd., 1991). The main debris flow activity occurs in streams which originate in the high mountainous areas, and the gully network constitutes a preferential low order pathway for sediment transport. Similar observations were reported by Thurber Eng. Ltd. (1996). According to that report, comparison of 1936 air photos with more recent ones did not appear to show significant changes in the rates of 77 denudation occurring in the area (around the Capilano reservoir) during a 50-year period. The main gullies appeared to be old features, with no new major gullies forming. Smith et al. (1986, cf. Figure 5.1.1) reported that the proportion of exposed mineral soil (or bedrock) decreases exponentially with increasing age of slides. In their study, average values show that approximately 50 % of the landslide scar area (scar moderately visible from A/P) was re-vegetated after 20 years and 75-80 % was recovered within about 50 years (probably scar is no longer visible from A/P) since landslide occurred. Figure 4.1.4 shows how the pattern of vegetation re-growth would be slower in the Charlottes (Smith et al., 1986) than in the Capilano basin. Here 50% of the events became not visible after 20 years and only 20 % of them were still visible after about 30 years. 100 S 80 ^ 60 <D *•• (0 *-• d> o> 40 > o >_ c5 20 o (/) 0 0 20 40 60 80 100 120 Age (years) Figure 5.1.1. Total vegetative cover as a function of age for landslide scars (after Smith et al., 1986). 78 If on one hand the comparison between single and multiple photo sets is undoubtedly affected by the mapper who conducts the air photo interpretation, results and speculations on the maximum area of "not visible" debris slides and the pattern of landslide visibility through time can be extended/generalised to the whole physiographic region of the Pacific Ranges. In figure 5.1.2 I summarise the factors that have proven to affect landslide visibility during air photo interpretation. Areas where one should expect not to be able to detect the presence of eventual relatively "large" landslides are lower portions of steep old growth forested slopes located in narrow valleys, as reported above. LAND-USE: OLD GROWTH /OLD LOGGING RECENT LOGGING GULLY RELATION: SIDEWALL/GULLY CHANNEL OPEN SLOPE/HEADWALL SLOPE GRADIENT: STEEP GENTLE VALLEY WIDTH: NARROW WIDE SLOPE POSITION: LOWER PORTION UPPER PORTION STREAM CONNECTION: CONNECTED NOT CONNECTED Figure 5.1.2. Factors affecting landslide visibility, they are ranked in order of importance from the most important (top) to the least (bottom). In addition, gully-related failures connected to permanent streams will be more difficult to detect. This is because stream-connected failures might have their deposition zone (e.g. flow fans) readily washed out by the stream they have impacted. 79 Factors that are manifestly related to the quality of photography itself such as nominal scale, sensor type (colour, black and white, etc.), weather conditions (snow cover, clouds) do not appear in Figure 5.1.2. 5.2 Survey Type, Landslide Density and Associated Sediment Production In section 4.2 analysis of landslide density and denudation rates was tackled by considering two main factors affecting mass failure detection: type of survey, which is external to the analysed system, and location, which is an intrinsic property of the system and can be regarded as the spatial heterogeneity of the system relative to its propensity to fail. Intensive ground checking on the study areas (Sisters and East Cap Creeks) has shown that "not visible" landslides accounted for about 27 % of the total volume of debris mobilised via mass movements during the last 30 years. Even more interesting, the two sub-watersheds exhibit very different rates of sediment production from landsliding (about one order of magnitude discrepancy), Sisters Creek being the more active. This last "unexpected" result, reported on table 4.2.1, constitutes one of the most original aspects of this research project and as such has highly influenced its further development. In fact the entire GIS-based analysis was performed in recognition of this finding. The adjective "unexpected" is motivated by the fact that, even though these two Capilano River tributaries possess almost identical biophysical characteristics they behave in strikingly different ways. Differences go beyond the large discrepancies in denudation rates; in fact field survey has allowed identification of a relevant additional number of debris slides and flows in Sisters Creek, but little change in East Cap Creek. 80 One aspect to consider in comparing fieldwork landslide detection efficiency is the relative abundance of different land-use cover, in our case the relative extension of old-growth, old logged and recently logged areas. In Sisters, the presence of a large portion of old harvested (logged more than 50 years ago) terrain and only 8.6 hectares of recently harvested cut-blocks (cf. Table 5.2.1) does not favour landslide detection from air photos. Conversely, the absence of old logging land-use category and 89.2 hectares of recently developed clear-cuts aid remote recognition of mass failures in East Cap. We consider historic fires to have no effect on landslide visibility; canopy height in fire-affected forests is comparable to that in old logged and old growth forested terrain. Thus "fire" was incorporated into old growth forest land-use category. Table 5.2.1. Land-use extensions in East Cap and Sisters Creeks. Location Extension Land-use Old growth (undisturbed) Old growth (including fire) Old harvesting Recent harvesting Fire Total surveyed area (ha) East Cap Cr. hectares 409.3 578.8 0 89.2 169.5 % 61.3 86.6 0 13.4 25.3 668 Sisters Cr. hectares 271 345.1 261.3 8.6 74.1 % 44.1 56.1 42.5 1.4 12 615 The effects of some other controlling factors that may explain the "Sisters-East Cap case" will be discussed later in the chapter. To this point we have commented on the crude results coming from the lumping of the entire database, or at best split into the two tributary domains. More reliable speculations can be made on the results coming from the more sophisticated and classic statistical 81 analyses, represented by ANOVA and its corresponding nonparametric tests. In other words we will see if the trends observed so far have any statistical significance. The power of the ANOVA approach lies in the possibility to take advantage of replications; a polygon is the so-called experimental unit and those polygons that receive the same "treatment" (i.e. survey, location) are replications. The greater the number of replications, the larger the number of degrees of freedom in the error term and the smaller the experimental error which represents the unexplained variability of the dependent variable (i.e. LS/ha, m3/ha). Two-way layouts possess a greater explanatory power (more complete explanation) than one-way analyses as they incorporate both factors over the entire database. In this way treatment interactions can be evaluated and unexplained variability (experimental error) is minimised. The overall survey-by-location picture can be summarised by saying that the same treatment (survey type) performed differently in different blocks (locations). In particular, field-coupled surveys have exhibited significantly greater landslide densities and denudations in Sisters Creek, the same variables in East Cap Creek have shown to be rather indifferent to the conduct of intensive fieldwork as a supplement to air photo interpretation (confirming the simple counts). Friedman's test showed that the coupling of air photo interpretation with intensive fieldwork in Sisters Creek gives significantly greater mobilised volumes (m3/ha) than any other survey-location combination did. This finding emphasises the importance of intensive ground checking. Solely from air photos, East Cap and Sisters Creeks would not have shown 82 a statistically significant difference in mobilised volumes per hectare (AP/East Cap and AP/Sisters are not significantly different). Apparently some forested terrains hide important numbers of "not visible" landslides, some others do not, therefore the issue of not visible events can be complex even within a single drainage basin (i.e. Capilano River basin). This finding has very important practical implications, meaning that in East Cap-like areas fieldwork is virtually unnecessary and, in perspective of sediment budget evaluation virtually no correction factor is needed to account for not visible events (they account for just 5% of the total volume mobilised). Conversely, Sisters-like areas would benefit from intensive fieldwork in order to evaluate the "invisible" volume of mobilised debris. A last aspect that confirms the benefit of conducting intensive fieldwork is the recognition that gully-related events have a greater importance than one could expect from air photo interpretation. They constitute more than one-third (both in terms of number of events and of mobilised volumes) of the "missed" events, while from air photos they accounted for just 8.8 % of the total number of failures (3.2 % of the associated mobilised volumes). This reinforces the point that gully-related failures are particularly difficult to detect from remote sensing. As for landslide sediment delivery to streams, numbers on table 4.2.4 are in line with speculations made in section 5.1. The major geomorphic work at this time scale (30-year window) takes place along the entire channel network. Gullies constitute preferential locations for sediment detachment (sidewall and headwall debris slides). Hence the whole stream channel network functions as a preferential transportation pathway for debris flows/torrents through the landscape. Unconnected failures account for only 12.8 % of the 83 total number of landslides and 15.6 % of the associated mobilised volume of debris. In this mountain environment the drainage network is extremely efficient in evacuating the sediment load. This occurs through alternation of a "normal fluvial regime", where supply-limited amounts of small material systematically are eroded from the stream channel, and of a "catastrophic regime" which entails removal of large quantities of sediment and woody debris via debris flows/torrents (Nistor, 1996). 5.3. The Impact of Forest Management on Mass Wasting In analysing the impact of logging on landslide initiation and sediment production the BC Terrain Stability Classification has served as an ideal tool for delineating terrain portions of "equal" or comparable stability, hence for disentangling the confounding between land use (external factor) and terrain intrinsic physical properties. The stability labelling can be seen as a summary of the intrinsic property of the terrain relative to its propensity to fail. In fact, this classification (cf. Appendix A) assesses terrain stability in function of selected static-geomorphic properties (e.g. presence of landslide scars, presence of gullies, dominant slope gradient, type of surficial material, etc.). By definition class III and IV polygons should not contain any landslide (hence no volumes of mobilised debris). As a result of ground survey these polygons are found to present landslide scars of events that have occurred within the last 30 years. The fact that stability classes do not rank properly in terms of mobilised volumes should not surprise; in fact the BC classification does not take into account landslide-associated mobilised volumes, but only the presence/absence of failures (regardless of their density). 84 Denudation rates in old logged terrain have lower accelerations than in recently harvested cut-blocks (cf. Table 4.3.2). In terms of landslide rates, old logging presents even a reduction when compared to old growth forest (cf. Table 4.3.1). These results can be justified by saying that old logged polygons had recovered from disturbance by the time they were ground-visited (here logging took place mainly between 1900's and 1930's, however not closer than 50 years ago). At the same time one should note that denudation rates are calculated over the total area of terrain (of equal land use and stability). The summation can mask inherent spatial variability and accelerations may be just consequential artefacts. This analytical approach lacks replications, as the experiment was conducted only once. Hence it should be regarded as valid only at the exploratory level. Any ultimate interpretation could be misleading. One of the main questions that puts air photo based landslide studies under suspicion is the relative bias introduced (due to different conditions of visibility) when one compares mass wasting activity in logged and undisturbed terrain. This case study has indicated that, management effect as perceived between air photo-based and field-coupled surveys (cf. Table 4.3.3), varies very little. Fieldwork had the effect of shifting the old logging-old growth ratio from negative accelerations to near constancy. Management effect of recent logging on denudation rates remained constant survey-wise, while landsliding ratios exhibited a 3-time decrease as a result of fieldwork coupling. This is due to the higher number of "hidden" slides in old-growth forest than in recent cut-blocks, whereas in Sisters canopies of old logged and old-growth forests have a similar effect on landslide visibility. Denudation rates in table 4.3.3 are systematically higher than those recorded by O'Loughlin (1972) on the North Shore Mountains, 0.11 n r W y r ' in undisturbed forest, 0.25 85 n r W ' y r 1 in clear-cuts, with an acceleration factor of 2.3. Reasons for lower values can be ascribed to various factors. Firstly O'Loughlin's denudation rates were calculated over an area (650 km ) - much larger than the 12.8 km of our study area - hence our values can be just part of the inherent spatial variability of the area that O'Loughlin surveyed. Secondly, a great part of the landslide inventory was conducted on l:30,000-scale aerial photographs, only 11 of the 32 years monitored were interpreted at the scale of our study (1:15,000). Lastly, the meteorological forcing between 1939 and 1970 could have been more severe than between 1969 and 2000, however we don't have any data that can support such a hypothesis. ANOVA-like analyses have generally produced results consistent with what was previously indicated by the calculation of landslide density and denudation rates over the entire area of each land-use category (descriptive statistics, cf. Tables 4.3.1-4.3.3). In fact, landslide density and denudation in recently harvested cut-blocks were not significantly different from those in undisturbed-forested polygons. They were instead significantly smaller in old logged areas than in the "control" polygons (undisturbed forest). This indicates that careful logging in East Cap Creek has produced no detectable effects on mass wasting. On the other hand, Sisters Creek by the last 30 years has apparently recovered from the signs of past extensive logging. Timber harvesting stopped about 50 years ago, 20 years before the start of our 30-year time window, a recovery period that is generally in line with that indicated by Sidle et al. (1985). They define the time interval from 3 to 9 years after logging as the period of greatest susceptibility to landslide occurrence. According to their approach, general recovery of rooting strength should take place within 20 years. Descriptive statistics displayed high acceleration factors brought about by recent logging (notably in terms of denudation rates, cf. Table 4.3.2) whereas ANOVA-like tests 86 did not detect significant differences between recent logging and undisturbed forest (cf. Table 4.3.6). Similarly, in the case of old logging, denudation rates were generally slightly higher in old logging terrain than in undisturbed forest, but the median test indicated the opposite (cf. Figure 4.3.1). We ascribe these discrepancies to the properties of the database: notably the high number of polygons presenting no landslide (zeros) and the consequent exponential-like distribution of both dependent variables. The appraisal of this inconsistency has a very important implication. It poses a question mark on the reliability of those studies that have calculated the management effect (i.e. land-use) over the entire land base without trying to break down spatial heterogeneity of terrain in meaningful components nor taking advantage of replications. For examples of the criticised approach see table 12 in Rood (1984) and table 10a in Sidle et al. (1985). Also, it appears that long-term land use may be important. Besides what has already been mentioned for land use differences, through application of ANOVA and nonparametric tests we were able to show that the following factors were statistically different for both dependent variables (LS/ha, m3/ha): (a) location: Sisters Creek significantly greater than East Cap Creek; (b) stability: class V significantly greater than class IV and III (this was proved via Bonferroni's and Sheffes' tests); (c) control between locations: old growth forest in Sisters Creek was recognised as significantly greater than in East Cap Creek. In particular, thanks to the adoption of nonparametric tests it was possible to detect several significant differences (i.e. old growth-old logging, stability) that ANOVA could not spot, because the database did not meet the required assumptions (cf. Section 2.1). 87 5.4 GIS-based Analysis: a Tool for Explaining Patterns of Sediment Production from Mass Wasting We performed the GIS-based analysis with the objective to explain the large discrepancy in denudation rate that was found between Sisters and East Cap Creeks. The analysis entailed extraction of elevation, slope gradient, slope aspect, drainage density and spatial distribution of surficial materials. Such macroscopic variables were selected in relation to the nature and quality of information available, ease of GIS-extraction, and power of physical explanation. The outcome depicted Sisters and East Cap basins as two contrasting geomorphic environments. Topography of steep slopes, one third of them steeper than 35°, and till-mantled slopes that cover about 40% of unstable (class V) polygons, are a large source of potential non-coherent material. Lastly, high drainage density allows efficient evacuation of the sediment load brought into the channel network via sidewall debris slides and wind-throw. These are all physiographic characteristics that make Sisters basin a highly landslide-prone environment. Conversely, in East Cap slopes are generally gentler (20% are steeper than 35°) till is mainly located on less steep polygons (class IV), also drainage density is significantly lower, thus imparting a lower connectivity to the system. These observations can find their justification in some classic concepts of landscape evolution literature. Accordingly, Carson and Kirkby (1972) proposed a classification that subdivides systems into weathering- and transport-limited ones (also termed supply-limited and -unlimited). Within the context of drainage basin sediment dynamics, in the first case the controlling (or limiting) factor is sediment production, in the other sediment mobilisation. In the mountainous-forested environment of coastal British Columbia transport-limited basins typically have a high density of headwater channels incised into thick glacial 88 drift or closely jointed bedrock. This ensures virtually unlimited debris supply in addition to many unstable trigger points for debris slides (hence for debris flows). Supply-limited basins denote slower recharge rates and fewer zones of instability. This is usually due to more massive bedrock or a thinner cover of glacial drift (Bovis and Jakob, 1999). In this sense East Cap and Sisters Creeks seem to be good examples of respectively supply-limited and transport-limited basins. The considerable inter-basin variability in sediment production from landsliding can find a satisfactory general explanation in the concept of geomorphic sensitivity to change (Brunsden and Thornes, 1979; Brunsden, 1980, 1993). Barriers to geomorphological change represent the system resistance to external disturbing forces and are expression of the system intrinsic control. Here we propose to classify sensitive landscapes on the basis of observable results of the interaction of the active forces. External perturbation (alias meteorological forcing) was considered to be constant between our two study basins, whereas the observable (variable) results clearly derived from the spatial interpretation of landslide density and associated denudation rate (dependent variables). Applying Brunsden's (1993) terminology to our study basins, primary resistances to geomorphological change were found to be: • Strength resistance, essentially is a function of bedrock and surficial materials strength. The availability of easily mobilisable material (i.e. till located on the steeper terrain of Sisters Creek) increases the potential sensitivity of a catchment. • Structural resistance is defined as the design of a geomorphological system. Degree of channel network development (the efficiency of slope-channel linkages) determines the relative flexibility (or rigidity) of the landscape to change. Higher drainage density in 89 Sisters Creek means decrease in structural resistance to change and increased geomorphic sensitivity. • Morphological resistance refers to the morphological configuration of the geomorphic environment. In our case this resistance is represented by slope aspect and slope gradient. The former has proven to be not significantly different between the two basins, whereas slope gradient was significantly greater in Sisters Creek, once more enhancing the sensitivity of this basin. Filter resistance is the ability to manage the way in which kinetic energy is transmitted through the system (Brunsden, 1993). In our case the forest cover functions as a shock (meteorological forcing) absorber. Its filtering role did not play a determining part in differentiating the sensitivity of Sisters and East Cap Creeks because, although in East Cap recently harvested cut-blocks were present, the area of each "carefully harvested" opening was considerably small. Briefly summarising, through the GIS-based analysis we managed to explain qualitatively the large discrepancy in rate of sediment production that was recorded between Sisters and East Cap Creeks. To go further would mean to quantify the numerical contribution of all factors involved by means of a deterministic model. In fact, a complete knowledge on how the mountainous forested landscape of coastal British Columbia functions is not yet available. Sisters Creek has been shown to be a transport-limited basin and potentially sensitive to geomorphological change. East Cap Creek presented the characteristics of a supply-limited basin and potentially resistant to geomorphological change. The issue of landscape sensitivity to geomorphic change (or stability) leads us to the concept of geomorphic hazard, which is 90 the probability of a change of a given magnitude occurring within a specified time period in a given area (Slaymaker, 1996). When compared to East Cap Creek, a history of greater landslide activity, the lower strength, structural, and morphological resistances from stability departure of Sisters Creek give to this basin a greater landslide-associated hazard. In light of'what is discussed above I think it is useful to reconsider the Terrain Stability Rating System as an index of sensitivity. Gully presence (structural resistance), dominant slope gradient (morphological resistance), type of surficial material (strength resistance), are key factors in the stability labelling of terrain polygons (cf. Appendix A) and they are all in relation to the sensitivity of the landscape to geomorphological change. In addition, the presence of past landslide activity, which is the major factor for labelling class V polygons, gives information on the past landscape sensitivity to meteorological forcing and to human impacts. If one wants to grasp information on landscape sensitivity I don't think one should look at terrain polygons singularly, rather one should consider the picture deriving from the aggregation of the polygons that form a drainage basin, as we did for East Cap and Sisters Creeks. However, this type of stability rating system is not readily suitable for the task, problems are due to the qualitative and subjective nature of the approach (geomorphic-static, cf. Appendix A). This in some way explains why we have been able to justify the differences in denudation rates between East Cap and Sisters Creeks only qualitatively. Hence, some efforts need to be spent in trying to push the system in a more engineering-deterministic direction. Assigning ranges of denudation rates to the different stability classes can constitute a promising starting point; then, within each stability class one should try to better define 91 denudation rates according to the various factors (e.g. number of gullies, type of surficial material). Clearly nobody can assure that such kind of parameterisation is going to be usefully pursued. 5.5. Landslide Magnitude-Frequency Relations The first issue to elucidate is the decreasing trend in landslide volume frequency (which corresponds to the kink observed in the magnitude-frequency relationship) that occurs for volumes smaller than about 2000 m (cf. Figure 4.5.3). There are two possible explanations for such a trend (a) poor landslide detection of "small" debris slides (also called landslide under sampling), or (b) the existence of a physical threshold for landsliding that limits the occurrence of mass failures up to a critical volume (hence weight). Intensive fieldwork has documented that virtually no critical volume has to be overcome in the forested terrain of the Capilano watershed (cf. Figure 4.5.1). We measured debris slides even smaller than 10 m . Landslide volume frequency distribution follows an exponential-like trend. Poor landslide detection from aerial photographs in densely forested terrain would seem to be the only explanation for the behaviour in question. We believe that Sisters and East Cap Creeks are representative of the watershed both in terms of land use (extrinsic factor or perturbation) — old growth, old logging and recent logging — and in terms of natural landslide activity (intrinsic factor). Most probably the two sub-basins represent the two extreme situations one can find in the watershed with the average physical behaviour of the Capilano lying somewhere in between. For this reason extrapolation of fieldwork findings to the whole watershed seemed appropriate and defendable. 92 The exponential-like trend of the volume frequency distribution (cf. Figure 4.5.5) corresponds to a power-law fit in the magnitude-frequency relationship (cf. Figure 4.5.4). The way one chooses to define volume categorical bins has been shown to heavily affect the scattering around the fitting line. Specifically, bin arrangement in a logarithmically equally spaced fashion is the classical way in which magnitude-frequency relations are plotted in the literature. Though "popular", such data categorisation produces very small volume interval towards the low magnitude-high frequency end of the landslide domain. Hence, the number of debris slides and debris flows appears to be inadequately sampled through the volume range. The effect is that of decreasing the frequency of "smallest" events, therefore enhancing the kink at about 3200 m3. A constant volume interval of 400 m3 has produced the best power-law fitting, which specifically is a double power-law fitting. Such fitting is in accordance with Stark and Hovius'(2001) characterisation of landslide3 size distributions of inventories from Taiwan and New Zealand. However, at the moment we cannot explain the kink, which still remains at about 4000 m3 (cf. Figure 4.5.7). According to Church et al., (2001), larger slides in the Capilano watershed would be constrained by slope geometry, specifically by slope length. Alternatively, this question can be approached by examining the episodic nature of large slides. The Capilano basin photographic record does not cover a time window long enough to contain a representative number of such events. According to my field experience in the watershed, I can speculate on the issue by saying that: yes larger events (i.e. channelised debris flows) do need long slopes to occur, " They were fitting large, rock-based failures. 93 hence the availability of this type of slope would be "a" controlling factor in the spatial distribution of large events. However, surficial material has been shown to play an important role (cf. Section 4.4), both in terms of landslide frequency and slide depth. Specifically, steep till mantled slopes displayed greater instability and higher landslide headscarps than rocky-colluvial ones. The original aspect of this section is that no previous study has documented with real field data the existence of a power-law landslide magnitude-frequency down to the lower end of the domain. Stark and Hovius (2001), using two air photo-sets of different nominal scale, demonstrated that the kink position was a function of air photo scale itself and stressed the need to improve mapping resolution in order to assess bulk effects such as the rate of forest disturbance. However, their speculation on the undersampling of smaller mass failures was merely qualitative. They reckoned that the number of "missed" landslides was going to affect the entire fitting but they could not say if, having accounted for all actual failures the relation would become a single power-law or would remain a double Pareto fit. Volume distribution analysis is very important for a regional model that would aim to describe and predict rate of soil denudation, sediment delivery to streams and, drainage basin sediment budget. In this context, comparisons of the three landslide inventories belonging to the Capilano River (GVRD, 1999), San Juan River (NHL, 1997) and the Queen Charlotte Islands (Rood, 1984) can illuminate the presence of similarities/dissimilarities between physiographic regions (i.e. Insular and Pacific Ranges). Besides physiographic characteristics that may influence volume frequency distributions, the comparison should always take into account differences in land use; otherwise any conclusion could be misleading from such confounding. Likewise, magnitude-frequency relations will be affected by differences in land 94 use. Departures from strictly power-law fitting could be due to land use effects. Further research is needed on the topic in order to elucidate potential connections. The Capilano (Pacific Ranges) has fairly homogeneous granodioritic geology (cf. Section 3.1). It has a large area of old logging (38 km2), and relatively small area of recently harvested cut-blocks (5 km ). San Juan and QCI (Insular Ranges) are characterised by a melange of formations, in places including schists, sedimentaries, and volcanics (more erodible materials). They have large areas of recently harvested cut-blocks, 47 km2 in the Queen Charlottes and 139 km2 in the San Juan River basin, this because the sequential photo sets cover the entire length of time when timber harvesting took place. Clearly a larger area of recently logged cut-blocks allows greater detection of "small" debris slides and debris flows. The Capilano River basin had the smallest recorded proportion of these small failure features. The most conservative conclusion one can draw from comparing the three air photo based magnitude-frequency relations is to suggest the need for a similarly intensive ground based survey for different areas in different physiographic regions. In fact, considering that air photo-based relations (cf. Figure 4.5.9) were so different from area to area, no extrapolation of the Capilano ground-based results (correction factor) would be advisable. Nevertheless, it may well prove possible to establish criteria that will allow characterisation of regional slope failure magnitude-frequency relations on the basis of comparable ground based surveys. A conclusion that more field monitoring is needed is unlikely to be welcome news. But we believe that this will be essential if sustainable strategies of land management are to be implemented. 95 Chapter 6 CONCLUSIONS The main purpose of the study was to investigate how large a proportion of failures is missed during compilation of air photo-based landslide inventories, both in terms of number of events and volume of mobilised debris, at a given air photo scale. The answer to this question has shown to be complex even within a single drainage basin of medium size (i.e. Capilano River basin). In places (Sisters Creek-like) the forest canopy does hide an important population of "not visible" landslides, in others does not (East Cap Creek-like). At the same time, the recognition of this inter-basin variability replies negatively to the question of whether the proportion of "missed" mass failures is constant throughout a physiographic region, in our case the Pacific Ranges of the Coast Mountains. It follows that, if one wants to obtain complete information for the evaluation of the sediment budget and the assessment of terrain stability, supplementary fieldwork is recommended only in Sisters Creek-like areas. Maximum area of "not visible" failures on forest older than 50 years was 650 m2, larger than what has always been assumed in previous studies (Rood, 1984; Schwab, 1988; Rollerson et al., 2001). Such large events were located on the lower portion of very steep (between 40° and 47°) old growth forested slopes. On recently harvested cut-blocks the value dropped to about 150 m2. Factors that have proven to affect landslide visibility during air photo interpretation were land use, gully relation of failure, slope gradient, valley width, slope position, and stream connection. Areas where one should expect not to be able to detect the presence of 96 eventual relatively "large" landslides are lower portions of steep old-growth forested slopes located in narrow valleys. As one could have expected, disposing of multiple sequential photo-sets during interpretation greatly improved our capability of detecting and dating landslide scars through the rugged and densely forested terrain of coastal BC. On these regards, in a sample area of about 60 km2 only half of the events originally mapped (by another independent study: GVRD, 1999) as having occurred 30 years ago or later, were actually active during that time window. As for the pattern of vegetation re-growth on landslide scars, which directly affects landslide visibility through time, 50% of the events became not visible after 20 years and only 20 % of them were still visible after about 30 years. Fieldwork has demonstrated that gully-related events have a greater importance than one could expect from air photo interpretation. They constitute more than one-third (both in terms of number of events and of mobilised volumes) of the "missed" events, while from air photos they accounted for just 8.8 % of the total number of failures (3.2 % of the associated mobilised volumes). The major geomorphic work at this time scale (30-year window) takes place along the entire channel network. Gullies constitute preferential locations for sediment detachment (sidewall and headwall debris slides). Hence the whole stream channel network functions as a preferential transportation pathway for debris flows/torrents through the landscape. One of the main questions that puts air photo-based landslide studies under suspicion is the relative bias introduced (due to different conditions of visibility) when one compares mass wasting activity in logged and undisturbed terrain. This case study has indicated that, management effect as perceived between air photo-based and field-coupled surveys vary 97 very little. Fieldwork had the effect of shifting the old logging-old growth ratio from negative accelerations to nearly constancy. Management effect of recent logging on denudation rates remained constant survey-wise, while landsliding ratios exhibited a 3-time decrease as a result of fieldwork coupling. This is due to the higher number of "hidden" slides in old-growth forest than in recent cut-blocks, whereas in Sisters canopies of old logged and old-growth forests have a similar effect on landslide visibility. As for the impact of logging on landslide initiation, according to ANOVA-like tests landslide density and denudation in recently harvested cut-blocks were not significantly different from those in undisturbed-forested polygons. They were instead significantly smaller in old logged areas than in the "control" polygons (undisturbed forest). This indicates that careful logging in East Cap Creek has produced no detectable effects on mass wasting. On the other hand, Sisters Creek by the last 30 years has apparently recovered from the signs of past extensive logging. Timber harvesting stopped about 50 years ago, 20 years before the start of our 30-year time window, a recovery period that is generally in line with what indicated by Sidle et al. (1985). They define the time interval from 3 to 9 years after logging as the period of greatest susceptibility to landslide occurrence. According to their approach, general recovery of rooting strength should take place within 20 years. ANOVA-like analyses have generally produced results consistent with what was previously indicated by descriptive statistics (cf. Tables 4.3.1-4.3.3). These methods displayed high acceleration factors brought about by recent logging (notably in terms of denudation rates, cf. Table 4.3.2) whereas ANOVA-like tests did not detect significant differences between recent logging and undisturbed forest (cf. Table 4.3.6). Similarly, in the case of old logging, denudation rates were generally slightly higher in old logging terrain 98 than in undisturbed forest, but the median test indicated the opposite, that management factors are < 1 (cf. Figure 4.3.1). We ascribe these discrepancies to the properties of the database: notably the high number of polygons presenting no landslide (zeros) and the consequent exponential-like distribution of both dependent variables. The appraisal of this inconsistency has a very important implication. It poses a question mark on the reliability of those studies that have calculated the management effect (i.e. land-use) over the entire land base without trying to break down spatial heterogeneity of terrain into meaningful components nor taking advantage of replications. Large differences between Sisters and East Cap Creeks in landslide density and denudation rates (cf. Table 4.2.1) were qualitatively explained via GIS-based analysis. The analysis entailed extraction of elevation, slope gradient, slope aspect, drainage density and spatial distribution of surficial materials. The outcome depicted Sisters and East Cap basins as two contrasting geomorphic environments. The different landslide activity between the two study areas was also justifiable through some classical concepts of geomorphology: (a) distinction between supply-limited (i.e. East Cap Creek) and transport-limited (i.e. Sisters Creek) basins (Carson and Kirkby, 1972), (b) landscape sensitivity to geomorphological change (Brunsden and Thorne, 1979; Brunsden, 1993), Sisters Creek being more sensitive than East Cap Creek. Landscape sensitivity to geomorphologic change suggests how the British Columbia Stability rating system could be usefully reconsidered in different terms. Gully presence (structural resistance), dominant slope gradient (morphological resistance), type of surficial material (strength resistance), are key factors in labelling the stability of terrain polygons and they are all in relation to the sensitivity of the landscape to geomorphological change. In 99 addition, the presence of past landslide activity, which is the major factor for labelling class V polygons, gives information on the past landscape sensitivity to meteorological forcing and to human impacts. Landscape sensitivity should be evaluated from aggregation of the polygons that form a drainage basin, as we did for East Cap and Sisters Creeks. However, this type of stability rating system is not readily suitable for the task, problems are due to the qualitative and subjective nature of the approach (geomorphic-static). Some efforts need to be spent in trying to push the system towards a more numerical-engineering direction. Intensive fieldwork has also documented that virtually no critical volume has to be overcome in the forested terrain of the Capilano watershed (cf. Figure 4.5.1). We measured debris slides even smaller than 10 m . The existence of an important population of "not visible" failures has transformed the landslide volume frequency from a "gamma-like" into an "exponential-like" one. Poor landslide detection from aerial photographs in densely forested terrain would seem to be the only explanation for the behaviour in question. The exponential-like trend of the volume frequency distribution corresponds to a power-law fit in the magnitude-frequency relationship. The way one chooses to define volume categorical bins has been shown to heavily affect the scattering around the fitting line. Specifically, bin arrangement in a logarithmically equally spaced fashion is the classical way in which magnitude-frequency relations are plotted in the literature. Though "popular", such data categorisation produces very small volume interval towards the low magnitude-high frequency end of the landslide domain. Hence, the number of debris slides and debris flows appears to be inadequately sampled through the volume range. A constant volume interval of 400 nr has produced the best power-law fitting, which specifically is a double power-law fitting. At the moment we cannot explain the kink at about 100 -5 4000 m (cf. Figure 4.5.7). One can interpret this trend by saying that larger slides would be constrained by slope geometry, specifically by slope length (Church et al., 2001). Alternatively, this question can be approached by examining the episodic nature of large slides. The Capilano basin photographic record does not cover a time window long enough to contain a representative number of such events. Personal field experience suggests that in addition to slope length, surficial material plays an important role (cf. Section 4.4), both in terms of landslide frequency and slide depth. Specifically, steep till mantled slopes displayed greater instability and higher landslide headscarps than rocky-colluvial ones. No previous study has documented with real field data the existence of a power-law landslide magnitude-frequency down to the lower end of the domain. Stark and Hovius (2001), using two air photo-sets of different nominal scale, demonstrated that the kink position was a function of air photo scale itself. They reckoned that landslide undersampling was going to affect the entire fitting but they could not determine if, having accounted for all actual failures the relation would become a single power-law or would remain a double Pareto fit. Finally, comparisons of the three landslide inventories belonging to the Capilano River, San Juan River and the Queen Charlotte Islands were conducted. Volume distribution analysis is very important for a regional model that would aim to describe and predict rate of soil denudation, sediment delivery to streams and, drainage basin sediment budget. The most conservative conclusion one can draw is to suggest the need for a similarly intensive ground based survey for different areas in different physiographic regions. In fact, considering that air photo-based relations (cf. Figure 4.5.9) were so different from area to area, no 101 extrapolation of the Capilano ground-based results (correction factor) would be advisable. Nevertheless, it may well prove possible to establish criteria that will allow characterisation of regional slope failure magnitude-frequency relations on the basis of comparable ground based surveys. A conclusion that more field monitoring is needed is unlikely to be welcome news. But we believe that this will be essential if sustainable strategies of land management are to be implemented. As to conclude, connecting with the initial aims of this research project, the existence of "not visible" events affected in a minor way conclusions about the impact of logging on slope stability in terms of land use (management) effects. It had instead a major impact on the nature of landslide magnitude-frequency relations, and finally demonstrated to have implications for British Columbia Terrain Stability Classification from the terrain sensitivity point of view. 102 BIBLIOGRAPHY Amaranthus, M.P., R.M. Rice, N.R. Barr, and R.R. Ziemer. 1985. Logging and Forest Roads to Increased Debris Slides in Southwestern Oregon. Journal of Forestry 83:229-233. Armstrong, J.E. and Brown, W.L. 1954. Late Wisconsin marine drift and associated sediments of the Lower Fraser Valley, British Columbia, Canada. Bull. Geol. Soc. Amer., 65, p. 349-364. Bak, P., Tang, C , and Weisenfeld, K. 1987. Self-organised criticality: an explanation pf 1/f noise. Physical Review Letters, 59: 381-384. Bak, P., Tang, C , and Weisenfeld, K. 1988. Self-organised criticality. Physical Review A. 38: 364-374. British Columbia Forest Service Research Branch. In prep. Proceedings of the Workshop "Terrain Stability and Forest Management in the Interior ofBC". May 23-25, 2001. Nelson, B.C. 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 Jakob, M. 1999.The role of debris supply conditions in predicting debris flow activity. Earth Surface Processes and Landforms 24: 1039-1054. Bovis, M.J., T.H. Millard, and M.E. Oden. Gully processes in Coastal British Columbia: the role of woody debris. In Carnation Creek and Queen Charlotte Islands Fish/Forestry Workshop: applying 20 years of Coast research to management solutions. Edited by D.L. Hogan, P.J. Tschaplinski, and S. Chatwin. British Columbia Ministry of Forests, Land Management Handbook 41, pp. 49-75. Brardinoni F., M. A. Hassan, and O.H. Slaymaker. (In prep.). Complex mass-wasting response of drainage basins to forest management in coastal British Columbia. Brunsden, D. 1980. Applicable models of long term landform evolution. Zeitschrift fur Geomorphologie, N.F. 36: 16-26. Brunsden, D. 1993. Barriers to geomorphological change. In: Landscape Sensitivity, edited by D.S.G. Thomas and R.J. Allison (Editors). John Wiley & Sons. 103 Brunsden, D. and J.B. Thornes. 1979. Landscape sensitivity and change. Transactions of the Institutes of British Geographers, 4: 463-484. Caine, N.1980. The rainfall intensity-duration control of shallow landslides and debris flows. GeogrAnn. 62A(l-2): 23-27. Carson, M.A. and M.J. Kirkby .1972. Hillslope form and processes, Cambridge University Press, Cambridge. Chatwin, S.C., Howes, D.E., Schwab, J.W. and Swanston, D.N. 1994. A guide for management of landslide-prone terrain in the Pacific Northwest. Second Edition. B.C. Ministry of Forests, Land Management Handbook No. 18, 220 pp. Church, M., Dadson, D., and F. Brardinoni. 2001. Statistical Analysis of Landslide Magnitude-Frequency in the Greater Vancouver Regional District Water Supply Area. Report to Greater Vancouver Regional District Watershed Planning Department. 15 pp. Church, M. and M.J. Miles. 1987. Meteorological antecedents to debris flow in sothwestern British Columbia: some case studies. Reviews in Engineering Geology, Vol.VII, Geolog. Soc. Am. Crozier, M.J. 1982. A technique for predicting the probability of mudflow and rapid landslide occurrence. In Proc. Inter. Seminar on Landslides and Mudflows, Alma-Ata USSR, October 1981, UNESCO, Paris, pp. 420-430. Dadson, S. 2000. Episodic Processes in Fluvial Landscape Evolution. M.Sc. thesis, The University of British Columbia, Vancouver, B.C., 165 pp. Dikau, R., D. Brunsden, L. Schrott, and M-L. Ibsen, (editors). 1996. Landslide recognition. Int. Assoc, of Geomorph. Publication No 5., John Wiley & Sons. Dyrness, C.T. 1967. Mass soil movements in the H.J. Andrews Experimental Forest. USDA Forest Service, Res. Pap. PNW-42, 12 pp. Erskine, C.F. 1973. Landslides in the vicinity of the Fort Randall reservoir, S.Dakota. U.S. Geol. Servey Prof. Paper 675, 64. Fell, R. 1994. Landslide risk assessment and acceptable risk. Canadian Geotechnical Journal 31:261-272. 104 Geographic Data B.C. 1992. British Columbia specifications and guidelines for geomatics. Content series. Volume 3: Digital baseline mapping at 1:20 000. Ministry of Environment, Lands and Parks Geographic Data BC, Province of British Columbia. Gimbarzevsky, P. 1988. Mass wasting on the Queen Charlotte Islands: a regional overview. B.C. Min. For. Victoria, B.C. Land Management Report No. 29. Greater Vancouver Regional District. 1999. Annex to GVRD Analysis Report Watershed Management Plan #5. February 1999. Hicks, B.A. 1982. Geology, geomorphology, and dynamics of mass movement in parts of the Middle Santiam River drainage, Western Cascades, Oregon. Masters Thesis, OSU, Corvallis, Oregon. 169 pp. Hogan, D.L., and Schwab, J.W. 1991. Meteorological conditions associated with hillslope failures on the Queen Charlotte Islands. British Columbia Ministry of Forests, Land Management Report 73. Horton, R.E. 1945. Erosional development of streams and their drainage basins: Hydrophysical approach to quantitative morphology. Bulletin, Geological Society of America, 56: 275-370. Hovius, N., Stark, C.P. and Allen, P.A. 1997. Sediment flux from a mountain belt derived by landslide mapping. Geology 25, 231-234. Hovius, N., Stark, C , Chu, H.-T., and Lin, J.-C. 2000. Sediment flux from a mountain belt derived by landslide mapping. J. Geol. 108. 73-89. Howes, D.E. 1981. Terrain inventory and geological hazards: northern Vancouver Island. BC Ministry of Environment. Terres. Studies Br., Victoria, BC Assessment Plann. Division Bull. 5. Howes, D.E. 1987. A terrain evaluation method for predicting terrain susceptible to post-logging landslide activity. BC Min. Environ. Parks. MOEP Tech. Rep. 28. Howes, D.E. and E. Kenk. 1996. Terrain classification system for British Columbia (2nd revised edition). BC Min. Environ., Victoria, BC Manual 10. Howes, D.E. and M. Sondheim. 1989. Quantitative definitions of stability classes as related to post-logging clearcut landslide occurrence (II). BC Min For. Land Management Report No. 56:167'-187. 105 Hughes, D.R. and R.V. Edwards. 1978. Granite Creek landslip survey. Unpublished Document, USFS Umpqua National Forest, Rosenburg, Oregon. 22 pp. Hungr, O., S.G. Evans, and J. Hazzard. 1999. Magnitude and frequency of rock falls and rock slides along the main transportation corridors of southwestern British Columbia. Canadian Geotechnical Journal, 36: 224-238. Hungr, O., Morgan, G.C. and Kellerhals, R. 1984. Quantitative analysis of debris flow hazards for design of remedial measures. Canadian Geotechnical Journal 22: 663-677. Hutchinson, J.N. 1992. Landslide hazard assessment. In Proc. 6th Inter. Sym. on Landslides. D.H. Bell (editor). Christchurch, N.Z. pp. 1805-1842. Innes, J.L. 1983. Debris flows. Progress in physical geography. 7: 469-501. Jordan, R.P. 1994. Debris flows in southern Coast Mountains, British Columbia: Dynamic behaviour and physical properties. Ph.D. thesis. Department Of Geography. The University of British Columbia, 258 pp. Ketcheson, G. and H.A. Froehlich. 1978. Hydrologic Factors and Environmental Impacts of Mass Soil Movements in the Oregon Coast Range. Water Resources Research Institute Bulletin 56, Oregon State University, Corvallis, Oregon. Krag, R.K., E.A. Sauder, and G.V. Welbura. 1986. A forest engineering analysis of landslides in logged areas on the Queen Charlotte Islands, British Columbia. B.C. Min. For. Victoria, B.C. Land Management Report No. 34. Leon, R.V. and R. W. Mee. 2000. Blocking Multiple Sources of Error in Small Analytic Studies. Quality Engineering, 12, No. 4. Luce, C.H. and B.C. Wemple. 2001. Hydrologic and Geomorphic Effects of Forest Roads. Spcial Issue Edited by Charles H. Luce, Beverley C. Wemple. Earth Surface Processes and Landforms. Lyons, J.K. 1982. Influence of landslides, floods, land use on channel changes of the Upper Middle Fork Willamette River, Oregon 1936-1980. M.S. Thesis, Oregon State University. Marion, D.A. 1981. Landslide occurrence in the Blue River drainage, Oregon. M.S. Thesis, Oregon State University, Corvallis. 106 Martin, Y., K. Rood, J. W. Schwab, and M. Church. 2002. Sediment Transfer by Shallow Landsliding in the Queen Charlotte Islands, British Columbia. Paper accepted for publication in: Canadian Journal of Earth Science. Meehan, W.R. 1991. Influences of forest and rangeland management on salmonid fishes and their habitats. American Fisheries Society, Special Publication 19. Montgomery, D.R. 1994. Road surface drainage, channel initiation, and slope instability. Water Resources Research. 30, 1925-1932. Morrison, P.H. 1975. Ecological and geomorphological consequences of mass movements in the Alder Creek watershed and implications for forest land management. Unpublished Bachelor Thesis, University of Oregon. 102 pp. Nistor, C.J. l996.Temporal Patterns in the Normal-Regime Fine-Sediment Cascade in Russell Creek Basin, Vancouver Island. M.Sc. thesis, The University of British Columbia, Vancouver, B.C. Northwest Hydraulic Consultants. 1997'. Forest Harvesting, Terrain Stability, Channel Morphology and Fisheries Resources: San Juan River Watershed. Final Report. O'Loughlin, C.L. 1972. An investigation of the slope stability of the steepland forest soils in the Coast Mountains, southwest British Columbia. Unpublished Ph.D thesis. University of British Columbia. 147 pp. Pelletier, J.D., B.D. Malamud, T.A. Blodgett, and D.L. Turcotte. 1997. Scale-invariance of soil moisture variability and its implications for the frequency-size distribution of landslides. Eng. Geol. 48: 254-268. Pyles, M.R. and H.A. Froehlich. 1987. Discussion of "Rates of landsliding as Impacted by timber management Activities in Northwestern California" by M. Wolfe and J. Williams. Bulletin of the Association of Engineering Geologists. Vol XXIV, No.3, pp 425-431. Reid, L.M. and T. Dunne. 1984. Sediment production from forest road surfaces. Water Resources Research 20: 1753-1761. Reid, L.M. and T. Dunne. 1996. Rapid evaluation of sediment budgets. GeoEcology paperback. Catena Verlag, p. 164. Resources Inventory Committee. 1996a. Guidelines and standards for terrain mapping in British Columbia. Res. Inv. Ctte., Victoria, BC. 107 Resources Inventory Committee. 1996b. Terrain database manual: standards for digital terrain data capture in British Columbia. Res. Inv. Ctte., Victoria, BC. Robison, E.G., K. Mills, J. Paul, L. Dent and A. Skaugset. 1999. Storm Impacts and Landslides of 1996: Final Report. Forest Practices Technical Report No. 4. Oregon Department of Forestry. Roddick, J. A. 1965. Vancouver North, Coquitlam and Pitt Lake Map Areas, British Columbia. Geological Survey of Canada, Memoir 335. RoUerson, T. 1992. Relationships between landscape attributes and landslide frequencies after logging: Skidegate Plateau, Queen Charlotte Islands. B.C. Ministry of Forests, Land Management Report No. 76. RoUerson, T., D.E. Howes and M. Sondheim. 1986. An approach to predicting post-logging slope stability in coastal British Columbia. Presented at NCASI West Coast Regional Meeting, Portland, Oregon. RoUerson, T., T.Millard, C. Jones, K.Trainor, and B. Thomson. 2001. Predicting post-logging landslide activity using terrain attributes: Coast Mountains, British Columbia. F.S.B.C. Technical Report 11. Rood, K.M. 1984. An aerial photograph inventory of the frequency and yield of mass wasting on the Queen Charlotte Islands, British Columbia. B.C. Ministry of Forests, Land Management Report No. 34, 55 pp. Rood, K.M. 1990. Site characteristics and landsliding in forested and clearcut terrain, Queen Charlotte Islands, B.C. British Columbia Ministry of Forests, Land Management Report No. 64, 46 pp. Ryder, J.M. 1981. Geomorphology of the southern part of the Coast Mountains of British Columbia. Zeitschriftfur Geomorphologie, N.F. Supplementband, 37, 120-147. Ryder, J. and D.E. Howes. 1984. Terrain information, a user's guide to terrain maps in British Columbia. BC Ministry of Environment, Victoria, BC. Ryder, J.M., G Horel and D. Maynard. 1995. The emerging role of terrain stabilty specialists in the forest industry of coastal British Columbia. BC Prof. Eng. 46:5-9. Sauder, E.A. and G.V. Welburn. 1987. Studies of yarding operations on sensitive terrain, Queen Charlotte Islands. BC Ministry of Forests. Land Management Report No.52. 108 Sauder, E.A. 1987. Logging and mass wasting in the Pacific Northwest with application to the Queen Charlotte Islands. BC Ministry of Forests. Land Management Report No.53. Schaefer D.G. and S.N. Nikleva. 1973. Mean precipitation and snowfall maps for a mountainous area of potential urban development. In Proceedings of the Forty-first Annual Western Show Conference in Denver, CO. Schumm, S.A. 1979. Geomorphic thresholds: the concept and its applications. Transactions of the Institute of British Geographers. 4: 485-515. Schwab, J.W. 1983. Mass wasting: October-November 1978 storm, Rennel Sound, Queen Charlotte Islands, British Columbia. B.C. Min. For., Victoria, B.C. Res. Note No.91. Schwab, J.W. 1988. Mass wasting impacts to forest lands: forest management implications, Queen Chalotte Timber Supply Area. In Degradation of forest land: forest soil at risk. Proc. 10th B.C. Soil Science Workshop, Feb. 1986. B.C. Min. For., Victoria B.C. Land Management Report No. 56, pp. 104-115. Sidle, R.C. 1992. A theoretical model of the effects of timber harvesting on slope stability. Water Resources Research, 28(7), 1897-1910. Sidle, R.C, A.J.Pearce and C.L.O'Loughlin. 1985. Hillslope stability and land use. Water Resources Monograph Series No. 11. Am.Geophysical Union. Washington, D.C. Sidle, R.C. and Wu, W. 1999. Simulating effects of timber harvesting on the temporal and spatial distribution of shallow landslides. Zeitschriftfur Geomorphologie, N.F. 43(2), 185-201. Slaymaker, O. 1995. Geomorphic hazards. Int. Assoc, of Geomorph. Publication No 4., John Wiley & Sons. Smith, R.B., P.R. Commandeur, and M.W. Ryan. 1983. Natural revegetation, soil development, and forest growth on the Queen Charlotte Islands. Progress report. BC Ministry of Forests, Fish/Forestry Interaction Program. Working Paper 7/83. 44 p. Smith, R.B., P.R. Commandeur, and M.W. Ryan. 1986. Soils, vegetation, and forest growth on landslides and surrounding logged and old-growth areas on the Queen Charlotte Islands. BC Ministry of Forests. Land Management Report No. 41. Snyder, K.U. 2000. Debris Flows and Flood Disturbance in Small, Mountain Watersheds. M.Sc. thesis. Oregon State University. 53 pp. 109 Stark, C. and Hovius, N. 2001. The characterization of landslide size distributions. Geophysical Research Letters. 28 (6): 1091-1094. Sterling, S.M. 1997. The influence of bedrock type on the magnitude, frequency and spatial distribution of debris torrents on northern Vancouver Island. Unpublished M.Sc. thesis. University of British Columbia. 117 pp. Swanson, F.J., L.E. Benda, S.H. Duncan, G.E. Grant, W.F. Megahan, L.M. Reid, and R.R. Ziemer. 1987. Mass failures and other processes of sediment production in Pacific Northwest forest landscape. P. 9-38. In: Streamside management: forestry and fishery interactions. University of Washington, Institute of Forest Resources, Contribution No. 57. 471 p. Swanson, F.J. and C.T. Dyrness. 1975. Impact of clearcutting and road construction on soil erosion by landslides in the Western Cascades Range, Oregon. Geology 3(7): 393-396. Swanson, F.J. and G. Grant. 1982. Rates of erosion by surface and mass erosion processes in the Willamette National Forest. U.S.D.A. Forest Service Pacific Northwest Range and Forest Experiment Station, Umpublished Document, November 22, 1982. Swanston, D.N. 1969. Mass wasting in coastal Alaska. U.S.D.A. Forest Service, Res. Pap. PNW-83. Swanston, D.N. 1974. Slope stability problems associated with timber harvesting in mountainous regions of the southwestern United States. U.S.D.A., For. Serv. Gen. Tech. Rep. PNW-021. Swanston, D.N. and F.J. Swanson, 1976. Timber harvesting, mass erosion and steepland forest geomorphology in the Pacific Northwest, p. 199-219 In: Geomorphology and Engineering, edited by D.R. Coates. Dowden, Hutchinson and Ross, Stroudsburg, PA.USA. Thurber Engineering Ltd., 1991. "Geotechnical assessment of 1990-1991 landslide events in greater Vancouver Water District watersheds". Report to GVWD. Thurber Engineering Ltd., 1993. "Assessment of turbidity-generating sediment sources and transport in the Capilano, Seymour and Coquitlam watersheds". Report to GVWD. Thurber Engineering Ltd., 1996. "Capilano Reservoir investigation of shoreline instability". Report to GVWD. 110 Tripp, D.B. and Poulin, V.A. 1992. The effects of logging and mass wasting on juvenile salmonid populations in streams on The Queen Charlotte Islands. B.C. Min. of For. Land Management Report. No. 80, 38pp. Turcotte, 1992. Fractals and chaos in geology. Cambridge University Press, Cambridge (Second Edition). Varnes, D.J. 1978. Slope movement types and processes in Landslides: Analysis and Control. Special Report. 176 pp. 11-33, Transport. Res.Board, National Academy of Science, Natl.Res.Counc, Washington, D.C. Young, S.E. 1992. Slope Stability Prediction Techniques. Unpublished M.Sc. thesis. University of British Columbia. Wemple, B. C , Jones J. A. and Grant, G. E. 1996. Channel network extension by logging roads in two basins, Western Cascades, Oregon. Water Resources Bulletin. 32, 1195-1207. Wu, W. and R.C. Sidle. 1995. A distributed slope stability model for steep forested basins. Water Resource Research. 31: 2097-2110. Zaruba, Q. and V. Mencl. 1969. Landslides and their control. Academia & Elsevier, Prague, 205. I l l Appendix A. Terrain Stability Assessment in British Columbia. Rugged topography, high precipitation, past glacial history (hence the nature of surficial material, with special reference to glacially-derived deposits) act together in characterising coastal British Columbia terrain as naturally prone to landsliding. In this part of the world timber harvesting is the main human activity that takes place on forested areas, it can increase sediment production (via triggering of landslides) and cause major effects on natural resources and public safety. This explains the development of a forestry-oriented type of terrain stability classification in British Columbia (B.C. MoF, 1996). There are two possible scenarios in which landslide hazard can be assessed: an area that has been already clear-cut, and an old-growth forest that is going to be exploited for timber production. The factors, which characterise the landslide potential, may be separated into two categories. The quasi-static variables such as geology, soil geotechnical properties, elevation, slope gradient, aspect, topographic incision, and long-term drainage patterns contribute to landslide susceptibility. The dynamic variables such as changes in soil saturation and root strength tend to induce mass movement in an area of given susceptibility. Climatic and hydrologic processes as well as human activities belong to this latter category (Wu and Sidle, 1995). Such subdivision of the landslide-governing variables is reflected in the approaches adopted for the terrain stability assessment. On one side there is what I am going to call the geomorphic-static approach, based on the analysis of configurations (structures) of the landscape. On the other stands the so-called engineering-dynamic approach, which deals with 112 processes. Hydraulic conditions, implied from geomorphology in the geomorphic-static approach, are thoroughly monitored and modelled in the engineering-dynamic approach. At the same time, the approach adopted to classify the propensity for a terrain to fail reflects both the objective (application) and the philosophical background in which the classification has been conceived. Two extremes of the range of approaches so far developed are: a) Engineering-dynamic (immanent) approach. Engineers tend to prefer the modelling at any cost. Such models are very often based on theoretical data. Strictly mathematical, such an approach is an objective and quantitative method. It shows the influence of terrain attributes and requires precise estimates of slope geometry, material strength properties and groundwater conditions. The method is difficult to use for mapping a large area and is extremely data intensive. It assumes global knowledge of the process; here lies the danger of oversimplification. b) Geomorphic-static (empirical) approach. Here geomorphologists tend to found their classification on field data — an accurate interpretation of land morphology and dynamics. Intuitive sense of important physical processes, spatial integration and synthesis are important. As a consequence, the "engineering" approach guarantees more flexibility, even though the outcomes are often statistical artefacts (effect of fitting that not always reflects what is really happening within the modelled system). The latter, somehow more empirical, tends to grant a privilege to contingencies at the expense of limiting the range of its applicability to different regions. Many times immanent predictive models are too data 113 intensive; in fact such quantity and quality of data are not really available (in terms of costs and time). In general, strictly deterministic techniques for estimating slope stability are not viable in an operational planning context (Young, 1992). Recognising that in most cases the knowledge of boundary conditions is incomplete and the required physical data (bedrock topography, detailed precipitation, etc.) may be scarce, non-existent, or highly different in quality (Rollerson et al. 1986), land managers of British Columbia have opted for essentially "index" stability assessment methods (geomorphic-static approach). The two approaches differ in respect to their applicability outside the study area where they have been originated. The geomorphic-static approach, relying heavily on the mapper's expertise clearly has a strong regional character, moreover being based on the analysis of configurations it is somewhat more rigid than the engineering-dynamic approach. This latter is based on process modelling, hence variables and constants (i.e. hydraulic conductivity, permeability, etc.) can be modified according to regional properties. Since the early 1980s, research has been carried out in coastal British Columbia on the prediction of landslides in clear-cut areas based on the study of a variety of terrain attributes. Examples include Rollerson and Sondheim (1985), Howes (1987) and Rollerson (1992). They were conceived to help refine the methods of pre-logging terrain stability mapping (subjective rating analysis). Such terrain attribute studies , based on morphologic-genetic performance, have been in part incorporated into the terrain stability assessment that is presently adopted/accepted in BC. A key step in assessing terrain stability is the development of terrain stability class criteria. During the mapping, the geomorphologist has the responsibility to consider possible 114 criteria for grouping the terrain stability into landslide hazard and/or risk classes. Classes are usually based on a combination of the terrain attributes and hazard and risk parameters such as probability of occurrence, magnitude, and intensity. Table Al. An Example of Terrain Stability Class Criteria, Subjective Rating Analysis Terrain Stability Class 1 II III IV V Example Class Criteria - Floodplains and level to undulating coastal plain areas - Most terrain with slopes <20%. Exceptions are noted in higher classes - Most gently sloping (20-40%), poorly to well drained lower slope landforms. Exceptions are noted in higher classes - Moderately sloping (40-60%), well to rapidly drained surficial deposits - Moderately sloping (40-60%), imperfectly to poorly drained surficial deposits that are not marine or lacustrine - Level to gently sloping (0-40%), imperfectly to poorly drained deep marine clays and lacustrine deposits - Moderately sloping, deeply gullied surficial deposits that are not of lacustrine or marine origin - Steeply sloping (>60%), well drained, deeply gullied surficial deposits - Steeply sloping, poorly drained surficial deposits - Moderately sloping, deeply gullied, or imperfectly to poorly drained lacustrine or marine deposits - Any areas where natural landslide scars are visible on air photographs or in the field - Very steeply sloping (>70%), imperfectly to poorly drained, deeply gullied surficial deposits (Modified from B.C. Ministry of Forests, 1995a) The method of selecting a criterion is highly dependent on the method of mapping; for some methods the selection is objective, while for others it is subjective; for some it is highly systematic and quantitative, while for others it is judgmental and qualitative. Two examples of terrain stability class criteria, one based upon a subjective rating analysis (BC Ministry of Forests, 1996a) and the other based upon a probabilistic univariate analysis (Howes, 1987), are presented in tables Al and A2. 1 Young (1992) classified them as belonging to the statistical-geographic approach. 115 Table A2. Example of Terrain Stability Class Criteria: Probabilistic Univariate Analysis. Material colluvium till; till/colluvium unconsolidated scarp colluvium till; till/colluvium till; till/colluvium colluvium fluvial till; till/colluvium colluvium rock/colluvium till fluvial colluvial fluvial till of colluvium till or colluvium fluvial over till Vlodified from He Slope (%) >36 >33 >33 >36 >33 26-33 29-36 20-33 26-33 39-36 >30 >26 >20 >36 >20 >20 variable <33 wes, 1987) Shape uniform uniform uniform uniform uniform uniform uniform uniform uniform uniform uniform uniform uniform uniform uniform uniform irregular uniform Process gullied gullied gullied gullied active flooding active fans Drainage rapid rapid-moderate rapid-moderate rapid rapid-moderate rapid-moderate rapid rapid-moderate rapid-moderate rapid rapid moderate-rapid rapid-moderate rapid-moderate rapid rapid-moderate Terrain Stability Class high high high moderate moderate moderate moderate low low low low very low very low very low very low very low very low very low In subjective cases, the criteria depend on the knowledge and experience of the mapper. However, it is specified that the mapper "should be guided by all available background data" (Ryder et al., 1995), these include slope maps, drainage maps, process inventory maps, terrain maps, terrain attribute studies and field observations. At present (BC Ministry of Forests, 1999) the Revised Forest Practices Code articulates the terrain stability mapping and assessment procedures into three stages, the framework of such methodology is the result of combining previous findings and experiences 116 accumulated during the past 20-25 years. Particularly, in light of regional results coming from probabilistic univariate analysis (Rollerson et al., 1985; Howes, 1987; Rollerson, 1992) there is the incorporation of subjective geomorphic analysis and subjective rating analysis. The three stages in increasing level of detail are: 1. Reconnaissance terrain stability mapping (cf. Table A3) identifies unstable or potentially unstable land areas from a broad perspective. It helps identify areas where more concentrated analysis is required. The method delineates 3 stability classes, areas that are presently unstable, marginally stable (potentially stable) and stable. This stage is conducted quasi-exclusively via airphoto interpretation. The distinction between potentially unstable slopes and steep but stable slopes is often difficult, involving an estimation of soil and hydrologic conditions based on the experience and judgment of the mapper without the benefit of ground observation. 2. Detailed terrain stability mapping (cf. Table A4) provides a more comprehensive assessment of terrain stability hazards. It helps to more narrowly define where terrain stability field assessment is required. The method delineates 5 stability classes (the IV and V being approximately equivalent to the unstable and marginally unstable classes of the reconnaissance terrain stability mapping, cf. Table A5). The mapping is carried out through the interpretation of air photos, later complemented by ground checking. There is no universal set of criteria, typically qualitative, for the stability classes (due to physiographic and climatic variability). The experience and preparation of the compiler play a basic part. 3. Terrain stability field assessment focuses on specific areas of concern for a proposed cutblock or road location. This final stage is applied to those areas that have been 117 previously flagged as unstable or marginally unstable (or class IV or V). After air photo interpretation, the work begins with exploratory vehicle or helicopter surveys. Subsequently the site to be assessed (both on logged and unlogged ones) are accurately appraised via foot traverse. The terrain stability specialist is expected to be familiar with the local and regional conditions that influence slope stability. Table A3. Reconnaissance Terrain Stability Classification Reconnaissance Terrain Stability Class S P U Interpretation - Stable. There is a negligible to low likelihood of landslide initiation following timber harvesting or road-building. - Potentially unstable. Expected to contain areas with a moderate likelihood of landslide initiation following timber harvesting or road construction. - Unstable. Natural landslide scars present. Expected to contain areas where there is a high likelihood of landslide initiation following timber harvesting or road construction. (Modified from BC Ministry of Forests 1995a) Table A4. Detailed Terrain Stability Classification Detailed Terrain Stability Class 1 II III IV V Interpretation - No significant stability problems exist. - There is a very low likelihood of slides following logging or road construction. - Minor slumping is expected along road cuts, especially for 1 or 2 years following construction. - Minor stability problems can develop. - Timber harvesting should not significantly reduce terrain stability. There is a low likelihood of landslide initiation following logging. - Minor slumping is expected along road cuts. There is a low likelihood of landslide initiation following road-building. - A field inspection by a terrain specialist is usually not required. - Expected to contain areas with a moderate likelihood of slide initiation following logging or road construction. Wet season construction will significantly increase the potential for road related slides. - A qualified terrain specialist prior to any development should make a field inspection of these areas, in order to assess the stability of the affected area. - Expected to contain areas with high likelihood of slide initiation following logging or road construction. Wet season construction will significantly increase the potential for road related slides. - A qualified terrain specialist prior to any development should make a field inspection of these areas, in order to assess the stability of the affected area. (Modified from BC Ministry of Forests 1995a.) 118 They may not be applicable to other climatic regions or longer time periods (generally a temporal window of 5-15 years after logging is considered). Some terrain types will have a different likelihood of failure for road building compared to timber harvesting. The extension of such classifications outside the area in which they have been created/developed probably constitutes one of the most serious limitations to this approach. Table A5. Comparison of various terrain stability mapping systems used in BC. Terrain stability class 1 II III IV V Reconnaissance stability class S S s p u ESA soil sensitivity class unclassified unclassified unclassified Es2 Es1 Figure Al. Example of BC surficial material mapping For every terrain polygon are reported in the order from the top: • Nature of surficial material and ongoing mass movement processes • drainage conditions of surficial material • dominant slope class • B.C. Stability Class 119 Terrain stability maps typically display a variety of accessory information (e.g. nature of surficial material, dominant slope class, etc.) that can help the land manager in better reading the landscape sensitivity to forestry operations. These maps are termed surficial material maps, an example has been given above (cf. Figure Al). 120 Appendix B. Types of Mass Movement Related to Timber Harvesting. debris avalanches: rapid downslope movement of disaggregated soil, rock and forest debris; they differ from debris flows in that they have a much lower water content and, unlike debris flows, they are not capable of flowing as slurries under their own weight. Debris avalanches typically begin on open slopes or within shallow hill slope depressions where groundwater is concentrated. If enough water is present debris avalanches become debris flows (Chatwin et al. 1994). debris slides: shallow failures of unsaturated, relatively unconsolidated soils and logging or forest debris by a translational or a rotational action; as regarding translational ones they have generally straight slide planes which usually develop along a boundary between soil materials of different density or permeability (Sauder, 1987). debris flows: shallow failures of water-saturated soil and debris (including organic debris) by a true flow process (Sauder, 1987). They can occur on a steep slope or in confined channel. debris torrents: special type of coarse-grained debris flows that occur in steep walled V-notch gullies (fine grained channelised debris flows are not debris torrents). They are a variant of debris flows in which much fine material has been removed by fluvial washing of sediment during the gully charging process (Sterling, 1996). 121 Appendix C. Fieldwork Checklists DAILY NOTES Date: Weather: Geographic Location (e.g. Creek): Polygon (#, stab.class): MEASUREMENTS FOR SINGLE FAILURE Time: Failure Type: Land use: Location (headwall, sidewall, open slope, etc.): Elevation at initiation point: Headscarp height: Seepage from headscarp? Sidescarp height (top or/and at Xm from top): Organic Horizon (depth): Initial scar width: Average scar width: Scar length a) initiation (till where sidescarps are visible): b) transportation: c) deposition: Scar material (?): Visual estim. of % clasts: Slope upslope from failure: Scar Slope a) initiation: b) transportation: c) deposition: Scar cross section (straight concave, convex, complex): Scar longitudinal profile (straight concave, convex, complex): Nature of slip-plane: Reached a stream or a gully? Vegetation on scar: a) quality (grass, brushes, trees) b) diameter (cores or sections) Estimated age of failure: Similar events or signs of instability? Drainage quality: Other notes: Sketch of failure: 122 TRANSECTS Date: Weather: Geographic Location (e.g. Creek): Polygon (#, stab.class): Distance (hip-chain) 0 (start) Elevation Direction Slope Gradient 123 Appendix D. Precipitation data tables (after GVRD, 1999). Table D.l Snow Courses in, or near the GVRD Watersheds Station Snow Course the GVRD Wa Orchid Lake Orchid Lake Burwell Lake Palisade Lake Palisade Lake Coquitlam Lake No. s within tersheds 3A19 3A19P 3A07 3A09 3A09P ID01 Period of Record 1972-97 1979-97 1945-84 1945-97 1993-97 1935 Lat. 49.32 49.32 49.28 49.27 49.27 49.22 Long. 123.03 123.03 123.00 123.02 123.02 122.48 Elev. (m) 1190 1190 880 880 880 300 Snow Course Normals1 April 1 Snow pack (cm) 461 -247 349 -3 April 1 Water Equiv (mm) May 1 Snow pack (cm) 2038 -1058 1567 -38 452 --330 --May 1 Water Equiv (mm) 2299 --1643 --Selected Snow Courses near the Watersheds Mount Seymour Grouse Mountain Hollyburn Cypress Lake Stave Lake 3A15 3A01 3A08 1D02 1D08 1960-89 1936-97 1945-87 1935-59 1967-97 49.22 49.23 49.24 49.22 49.35 122.57 123.05 123.11 122.13 122.19 1070 1100 1100 340 1210 357 291 374 101 399 1594 1308 1709 433 1643 355 277 --379 1812 1377 --1815 1. Snow course normals from the BC Ministry of Environment, Lands, and Parks 124 Table D.2 Rainfall Total-Duration-Frequency at Selected AES Stations Seymour Falls (1107200) 1927-1996 Duration (Days) 1 2 3 4 5 6 7 8 9 10 15 20 25 30 Rainfall Intensity (mm) Return Peiriod (yrs) 2 156.05 211.25 262.15 300.33 339.71 375.70 404.63 433.30 452.58 477.12 569.54 674.39 758.95 843.50 5 208.56 278.38 337.06 378.62 421.60 465.82 502.18 536.81 557.69 584.06 706.74 828.33 941.05 --10 243.40 322.94 386.78 430.57 475.94 525.63 566.92 605.51 627.44 655.04 797.79 930.49 --~ 25 287.34 379.12 449.47 496.09 544.47 601.06 648.56 692.13 715.40 744.54 912.61 -----50 319.92 420.79 495.97 544.68 595.29 656.99 709.11 756.37 780.63 810.91 997.76 ------100 352.33 462.22 542.21 593.00 645.83 712.62 769.32 820.26 845.50 876.92 --------Grouse Mountain (1105658) 1971-1996 Duration (Days) 1 2 3 4 5 6 7 8 9 10 15 20 25 30 Rainfall Intensity (mm) Return Period (yrs) 2 111.27 150.45 179.63 202.46 220.97 236.57 248.68 265.14 274.78 286.68 331.14 374.89 412.42 444.87 5 140.48 195.70 234.21 266.43 286.84 305.63 323.11 342.19 351.88 364.73 417.40 485.10 535.99 578.42 10 159.87 225.73 270.43 308.88 330.55 351.45 372.50 393.33 403.05 416.53 474.64 558.25 617.99 667.05 25 184.32 263.60 316.11 362.42 385.68 409.25 434.79 457.82 467.58 481.86 546.84 650.49 721.40 778.83 50 202.45 291.68 349.98 402.12 426.56 452.10 480.99 505.65 515.43 530.30 600.37 718.90 798.08 861.71 100 220.48 319.61 383.67 441.60 467.21 494.72 526.92 553.21 563.01 578.47 653.61 786.92 874.35 944.15 125 Table D.2 continued Rainfall Total-Duration-Frequency at Selected AES Stations Seymour Hatchery (110N666) 1981-1996 Duration (Days) 1 2 3 4 5 6 7 8 9 10 15 20 25 30 Rainfall Intensity (mm) Return Period (yrs) 2 149.79 204.69 254.01 300.13 345.31 373.36 406.64 435.69 460.22 481.29 593.86 703.54 822.16 909.80 5 188.51 256.68 320.79 373.03 421.95 451.15 488.83 523.28 547.13 564.30 733.17 863.09 989.09 10 214.20 291.19 365.11 421.41 472.81 502.77 543.38 581.41 604.81 619.39 825.62 968.97 25 246.60 334.70 421.00 482.42 536.95 567.88 612.17 654.71 677.55 688.87 942.21 50 270.63 366.96 462.44 527.67 584.51 616.16 663.18 709.06 731.49 740.39 100 294.52 399.05 503.66 572.67 631.82 664.17 713.91 763.12 785.13 791.63 Coquitlam Lake (1101890) 1924-1981 Duration (Days) 1 2 3 4 5 6 7 8 9 10 Rainfall Intensity (mm) Return Period (yrs) 2 128.0 196.0 232.7 262.7 296.4 329.9 355.4 386.3 409.6 431.7 5 156.8 245.9 291.8 331.3 368.9 405.0 434.5 473.1 497.9 523.3 10 175.9 279.0 330.9 376.7 417.0 454.7 486.9 530.6 556.4 584.0 25 200.0 320.7 380.3 434.1 477.6 517.6 553.1 603.3 630.2 660.7 50 217.9 351.7 417.0 476.7 522.7 564.2 602.2 657.2 685.0 717.5 100 235.7 382.5 453.4 519.0 567.4 610.4 650.9 710.7 739.4 774.0 126 Table D.2 continued Rainfall Total-Duration-Frequency at Selected AES Stations Cleveland (110EF56) 1969-1989 Duration (Days) 1 2 3 4 5 6 7 8 9 10 Rainfall Intensity (mm) Return Period (yrs) 2 79.5 117.3 140.9 163.5 185.0 202.8 220.0 233.4 245.4 260.0 5 101.4 147.7 179.2 205.9 236.3 254.3 276.3 294.1 309.2 327.5 10 115.9 167.8 204.5 233.9 270.3 288.3 313.6 334.3 351.5 372.3 25 134.3 193.2 236.5 269.4 313.2 331.4 360.7 385.1 404.8 428.8 50 147.9 212.1 260.2 295.7 345.0 363.3 395.6 422.7 444.4 470.8 100 161.4 230.8 283.8 321.8 376.6 395.1 430.3 460.1 483.7 512.4 127 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0090429/manifest

Comment

Related Items