Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Factors affecting post-logging debris flow initiation in steep forested gullies of the Southwestern Canadian… Brayshaw, Drew Devoe 1997

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

Item Metadata


831-ubc_1997-0508.pdf [ 28.78MB ]
JSON: 831-1.0087952.json
JSON-LD: 831-1.0087952-ld.json
RDF/XML (Pretty): 831-1.0087952-rdf.xml
RDF/JSON: 831-1.0087952-rdf.json
Turtle: 831-1.0087952-turtle.txt
N-Triples: 831-1.0087952-rdf-ntriples.txt
Original Record: 831-1.0087952-source.json
Full Text

Full Text

FACTORS AFFECTING POST-LOGGING DEBRIS FLOW INITIATION IN STEEP FORESTED GULLIES OF THE SOUTHWESTERN CANADIAN CORDILLERA, FRASER VALLEY REGION by Drew Devoe Brayshaw B.Sc , University of British Columbia, 1994 A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF THE REQUIREMENTS FOR THE D E G R E E OF MASTER OF SCIENCE in THE F A C U L T Y OF G R A D U A T E STUDIES (Department of Geography) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A September 1997 © Drew Devoe Brayshaw, 1997 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 The University of British Columbia Vancouver, Canada Date DE-6 (2/88) 11 ABSTRACT This thesis examines the factors which influence debris flow initiation in gullies. The analysis covers 53 gullies located in two neighboring, but separate, regions: the Chilliwack Valley region of the Cascade Mountains, and the Norrish Creek area of the southern Coast Mountains, respectively located to the south and north of the Fraser Valley in southwestern British Columbia. Lithology varies greatly between the two regions, but climate and logging histories are similar. All gullies studied had been logged during the period 1981 to 1990. Gullies were selected on the basis of their having experienced slope failures which potentially could have caused debris flows to occur. A wide range of geotechnical and morphometric terrain variables was measured in each gully system, at three different spatial scales: that of the entire gully system, that of the individual gully reach, and that of the individual slope failure. Gullies were also evaluated by the method of assessing Debris Flow Initiation Potential (DFLP) used in the Gully Assessment Procedures (GAP) section of the B.C. Forest Practices Code. Analysis of data showed that the most important variables influencing debris flow initiation in gullies were surficial material, gully wall gradient and length, and channel gradient, although not exactly in the manner which the DFLP assessment had predicted. Volume of the initial failure, amount of sediment stored in the gully channel, and angle of entry of the failure into the channel were other parameters which proved to be important. Regional variation in lithology seemed mainly to affect gully morphology, and not to influence debris flow initiation directly. Revisions were suggested for the DFLP assessment based on the observed behaviour of the most important and most easily and accurately measured variables. The current method uses the parameters of gully wall slope angle and surficial material type to qualitatively estimate the chance that a wall failure will occur, and the parameters of gully wall length and channel gradient to Ill qualitatively estimate the chance that a slope failure entering the channel will initiate a debris flow. The two estimates are then combined into an overall estimate of the qualitative probability of debris flow initiation, using a three-class system of high, medium and low hazard. The revised method of DFLP evaluation devised in this thesis is similarity qualitative. The revised estimate of the chance of slope failure is determined using gully wall length in addition to wall slope gradient and surficial material type, and the chance of debris flow initiation from failure is determined using channel gradient only. The resulting estimate of debris flow initiation potential uses only two qualitative hazard classes, high and low. iv Table of Contents Abstract ii Table of Contents iv List of Symbols vii List of Tables viii List of Figures ix List of Variable Names and Descriptions xii Acknowledgements xiv Chapter 1-Introduction. 1 1.1 Gullies as a landform component of forested hillslopes. 1 1.2 Debris flow as a geomorphic process characteristic of gullies. 5 1.3 The Forest Practices Code and Gully Assessment Procedures. 8 1.4 Thesis statement of purpose and definition of research goals. 9 Chapter 2- Theory of Debris Flow Initiation in Gullies. 14 2.1 Mechanics of slope failure in gullies. 14 2.2 Behaviour of debris flow in gully system. 16 2.3 Meterorological antecedents to debris flow initiation. 18 2.4 Effects of forest harvesting on slope failure and debris flow initiation. 21 2.5 DFLP evaluation of gully. 23 Chapter 3- Description of Study Areas: Chilliwack Valley and Norrish Creek. 25 3.1 Physical parameters and geographic description. 25 3.2 Bedrock and surficial geology, Quaternary history. 34 3.3 History of logging activity. 38 3.4 Precipitation record and extrapolation to study area. 40 3.5 History of slope instability in region. 41 3.6 Impact of mass movement events. 42 3.7 Observed regional variation in morphology of gullies and location of debris flow initiation. 43 3.8 Conclusion. 47 Chapter 4- Experimental Design and Measurement. 49 4.1 Experimental design and choice of measured parameters. 49 4.2 Methods of measurement: quantitative, qualitative. 53 I. Terrain variables. 53 II. Gully variables. 55 III. Failure variables. 61 4.3 Measurement error and limits to accuracy/precision. 65 4.4 Summary of fieldwork and data collection 68 V Chapter 5- Analysis of Field Data: Reach Data. 71 5.1 Methods of analysis: statistical and graphical. 71 5.2 Test of DFLP vs. observed results. 72 5.3 Test of GWFP vs. observed failures. 75 5.4 Test of GGP vs. observed mobilization of failures as debris flows. 82 5.5 Comments on reach data and DFLP. 87 Chapter 6-Analysis of Field Data: Failure Data. 89 6.1 Methods of analysis: statistical and graphical. 89 6.2 Test of GGP using at-a-point channel characteristics. 90 6.3 Test of failure parameters vs. observed results: significant qualitative parameters. 91 6.4 Test of failure parameters vs. observed results: significant quantitative parameters. 100 6.5 Regressions and results: V O L U M E and AOE. I l l 6.6 Between-parameter correlations of non-failure variables: significant correlations. 116 6.6.1 Date of logging and date of failure. 116 6.6.2 Drainage, seepage and lithology. 118 6.7 Comments on analysis of failure data. 121 Chapter 7- The Gully Assessment Procedures & Debris Flow Initiation Potential Assessment: Suggested Modifications. 122 7.1 Observed weaknesses of current DFIP methods. 122 7.2 GWSD as a parameter influencing gully wall failure potential, and NEWGWFP. 123 7.3 Parameters influencing debris flow initiation, and NEWGGP. 128 7.3.1 Method using T A N C G and ICSS (NEWGGP 1) 130 7.3.2 Method using T A N C G only (NEWGGP2) 133 7.3.3 Method using TANCG, AOE, and LOCN2 (NEWGGP3) 134 7.3.4 Conclusions with regard to NEWGGP 137 7.4 Revised method of determining DFIP, and a test of NEWDFIP. 137 Chapter 8- Regional variation in observed parameters and results. 144 8.1 Observed variation in morphology of gullies and location of debris flow initiation. 144 8.2 Statistically significant regional variation at the reach scale. 144 8.3 Statistically significant variations in failure data. 149 Chapter 9- The GAP and Factors affecting Debris Flow Initiation: Conclusions. 159 9.1 Conclusions. 159 9.2 Further recommendations to improve performance of GAP. 160 9.3 Limits of confidence in conclusions. 162 9.4 Suggestions for further research. 163 References 166 Appendices Appendix 1- The DFLP Assessment. Appendix 2- Ministry of Forests Terrain Data Cards. Appendix 3- Analysis of Field Data - Sediment Samples. Appendix 4- Correlation Matrix for Failure Variables and DFRESULT. LIST OF SYMBOLS vn c cohesion (Pa). The total cohesion in a soil mass resulting from soil properties only (true cohesion) c' apparent cohesion (Pa). Sum of true cohesion, c, and soil reinforcement, c p c p soil reinforcement due to roots (Pa). Acts like cohesion, but is not truly a cohesive force, rather the reinforcement of soil against movement by the interconnection of roots D depth; typically the smallest of the three dimensions of an object (m) D 5 0 median diameter of a sediment sample (mm). Half the sediment in the sample is smaller, half is larger H thickness of channel fill (m) F F-statistic in an A N O V A test. Higher F values indicate a greater degree of confidence in the result of the test F.S. factor of safety. Ratio of shear strength to shear stress L length; typically the longest of the three dimensions of an object (m) p p-value of an A N O V A test: probability that the observed distribution is random, and that separate classes in dependant variable are not distinct from each other r correlation coefficient S shear strength (Pa). Sum of forces resisting movement of soil on a slope u pore water pressure (Pa) W width; typically the intermediate dimension of the three dimensions of an object (m). z a Cartesian axis; elevation in the vertical plane above a datum (m). a angle of entry of slope failure into gully channel (degrees) a confidence level. Probability at which results of A N O V A tests with p-value < a will be accepted as being significant (3 slope angle (non-specific) (degrees) y unit weight (N/m3) X. the slope angle of the less steep of two gully walls (degrees) cp the slope angle of the steeper of two gully walls (degrees) (J)' angle of internal friction (degrees) T shear stress (Pa). Sum of forces driving movement on a soil slope Q angle at which fall line of gully wall intersects channel axis vn viii LIST OF T A B L E S Table 3.1 Sub-basins and areas, Chilliwack River valley 25 Table 3.2 Sub-basins and areas, Norrish Creek watershed 28 Table 3.3 Distribution of sites in each main study area by region 28 Table 4.1 Variables measured in gully systems 52 Table 5.1 Number and percent of total - gully reaches by DFIP rating 75 Table 5.2 Number of total reaches and reaches with failures by GWFP rating 76 Table 5.4 Distribution of reaches with debris flow by GGP class 84 Table 5.5 Influence of the threshold value (tan C G = 0.5) on debris flow initiation 85 Table 7.1 New GWFP classification system 126 Table 7.3.2 Results of A N O V A tests of NEWGGP2 vs. GGPOOR and GGP2FAN 134 Table 7.4 Possible new DFIP classes with expected fraction of each class experiencing debris flow initiation 138 Table 7.5 Number and percentage of reaches in each old and new DFLP class 139 Table 8.1 Results of A N O V A tests for variation in REGBASIN by variable 145 Table 8.2 Distribution of reaches and head reaches by region 148 Table 8.3 A N O V A results for regional variation in basin area variables 149 ix LIST OF FIGURES Figure 1.1 Gully with a steep headwall 2 Figure 1.2 Gully with an open-slope depression at its head 3 Figure 1.3 A typical gully system 4 Figure 3.1 Overview map of southwestern British Columbia showing locations of Norrish Creek and Chilliwack Valley 26 Figure 3.2.1 Map of Chilliwack Valley showing location of sub-basins studied 27 Figure 3.2.2 View eastwards along Chilliwack Valley showing physiography 29 Figure 3.2.3 View of the Cheam Range showing physiography 29 Figure 3.2.4 View of the Nesakwatch Cr. - Center Cr. divide showing physiography 30 Figure 3.3.1 Map of Norrish Creek and vicinity showing location of sub-basins studied 31 Figure 3.3.2 View upstream in Rose Creek showing physiography 33 Figure 3.3.3 View of south side of E. Norrish Creek showing physiography 34 Figure 3.4.1 Geologic map of Chilliwack Valley and vicinity 35 Figure 3.4.2 Geologic map of the Norrish Creek area 37 Figure 3.5 Gullies in the Tamihi Creek drainage showing chronic post-logging erosion 39 Figure 3.6.1 Cross section of a typical Chilliwack gully 44 Figure 3.6.2 Cross section of a typical Norrish Creek gully 44 Figure 3.6.3 A typical Chilliwack gully (FO-005-LB) 45 Figure 3.6.4 Another typical Chilliwack gully (BOR-002-RB) 45 Figure 3.6.5 A typical Norrish Creek gully (ENR-003-LB) 46 Figure 3.6.6 Two typical Norrish Creek gullies (NOR-006 and -007-LB) 46 Figure 4.1 Measuring the length of a gully reach with a hip chain 56 Figure 4.2 Cross-section of a gully channel showing method used to calculate ICSS 58 Figure 4.3 Gully reach (AP-002-RB-01) with no in-channel stored sediment 59 Figure 4.4 Gully reach (ROS-002-RB-01) with a low ICSS value 59 Figure 4.5 Gully reach (ESP-001-HW-01) with moderately high ICSS 60 Figure 4.6 Gully reach (TLM-001-LB-01) with extremely high ICSS 60 Figure 4.7 Measuring revegetation of a failure scar 62 Figure 4.8 Method of measurement of angle of entry of failure into gully system 64 Figure 5.1 Proportion of gullies producing debris flows out of reach and to fan 73 Figure 5.2 Proportion of gully reaches (rated by DFIP and GWFP class) with slope failures therein 77 Figure 5.3.1 Frequency scatterplot of S U R F F L vs. GWSA 79 Figure 5.3.2 Histogram of GWSA categorized by SURF_FL 79 Figure 5.4.1 A N O V A plot of means for significance test of GWSA vs. SURF_FL 81 Figure 5.4.2 Boxplot showing variation in mean GWSA by SURF_FL class 82 Figure 5.5.1a Result of A N O V A test of effectiveness of GGP for debris flows travelling out of the reach of initiation 83 Figure 5.5.1b Result of A N O V A test as for 5.5.1a, but for debris flow to fan 84 Figure 5.6 Qualitative diagram showing expected distribution of outcomes in a TANCG-GWSD parameter space 86 Figure 5.7 Scatterplot of T A N C G vs. GWSD, showing actual distribution of events 86 X Figure 6 1 Plot of T A N C G vs. GWSD for the failure data set 91 Figure 6 2a Photomosaic of the left wall of gully TLM-001-LB 92 Figure 6 2b The right wall of gully TLM-001 -LB 93 Figure 6 2c The fan of gully TLM-001-LB 93 Figure 6 3.1 Result of A N O V A test of influence of TYPE on DFRESULT 95 Figure 6 3.2a Result of A N O V A test of influence of L O C N on DFRESULT 95 Figure 6 3.2b Result of A N O V A test of influence of LOCN2 on DFRESULT 97 Figure 6 3.3 Result of A N O V A test of influence of SEEP on DFRESULT 97 Figure 6 3.4a Result of A N O V A test of inluence of LITHOL on DFRESULT 99 Figure 6 4.1a Scatterplot of T A N C G vs. V O L U M E 101 Figure 6 4.1b Histograms of L N V O L categorized by DFRESULT 102 Figure 6 4.2a Histograms of FAILSLP categorized by DFRESULT 103 Figure 6 4.3a Scatterplot of T A N C G vs. AOE categorized by DFRESULT and L O C N 105 Figure 6 4.3b Histogram of AOE categorized by DFRESULT 106 Figure 6 4.3c Result of A N O V A test of influence of LOCN2 on AOE 106 Figure 6 4.4a Plot of T A N C G vs. ICSS categorized by DFRESULT 108 Figure 6 4.4b Histogram of ICSS categorized by DFRESULT 108 Figure 6 4.4c Scar of failure FO-004-LB-01 F01 110 Figure 6 4.4d Depositional cone of failure FO-004-LB-01 FO1 110 Figure 6 4.4e Looking downchannel below FO-004-LB-01 F01 111 Figure 6 5.1a STATISTICA multiple regression summary for L N V O L 113 Figure 6 5.1b Plot of predicted vs. observed values for regression of L N V O L 113 Figure 6 5.2 Perspective and plan views of a gully wall showing relationship between angle of entry and Q for a failure 115 Figure 6 6.1 Result of A N O V A test of correlations between L O G D A T E and FALLDATE 116 Figure 6 6.2a Result of A N O V A test of the influence of SEEP on DRAIN 119 Figure 6 6.2b Result of A N O V A test of influence of DRAIN on SEEP 119 Figure 6 6.2c Result of A N O V A test of influence of LITHOL4 on D R A I N 120 Figure 7 1.1 Histograms of GWSD categorized by PASTFAIL 124 Figure 7 1.2 Plot of GWSD vs. GWSA categorized by PASTFAIL, with the suggested boundaries of the most effective NEWGWFP classes indicated 124 Figure 7 1.3 Plot of GWSD vs. GWSA categorized by PASTFAIL and SURF 125 Figure 7 1.4 Histograms of GWSD categorized by PASTFAIL and SURF 125 Figure 7 1.5a Result of A N O V A test showing correlation of SURF and GWSD 127 Figure 7 1.5b Result of A N O V A test of correlation of GWSD and S U R F F L 127 Figure 7 2.1a A N O V A plot of means for test of NEWGWFP vs. PASTFAIL 129 Figure 7 2.1b A N O V A plot of means for test of GWFP vs. PASTFAIL 129 Figure 7 3.1a Test of effectiveness of NEWGGP 1 in predicting debris flow out of reach 131 Figure 7 3.1b Test of effectiveness of NEWGGP 1 in predicting debris flow to fan 131 Figure 7 3.1c Scatterplot of L N V O L vs. ICSS categorized by DFRESULT 133 Figure 7 3.3a Histogram of a for sidewall failures only 135 Figure 7 3.3b Result of A N O V A test of effectiveness of NEWGGP3 in predicting debris flow initiation from failure 136 Figure 7 4.1a Result of A N O V A test of NEWDFLP vs. DFOOR? 140 Figure 7 4.2a Result of A N O V A test of DFIP vs. DFOOR? 140 Figure 7 4.1b Result of A N O V A test of NEWDFLP vs. DFTOFAN? 141 XI Figure 7.4.2b Result of A N O V A test of DFIP vs. DFTOFAN? 141 Figure 7.4.3 Bivariate histogram showing numbers of reaches in each old and new DFIP class combination 142 Figure 8.1.1 Histograms showing distribution of T A N C G in each studied region 146 Figure 8.1.2 Histograms showing distributon of SURF in each studied region 146 Figure 8.1.3 Histogram of GWSD categoriz4ed by REGBASIN 147 Figure 8.3.1a Histogram of B A S A R E A categorized by REGBASIN 150 Figure 8.3. lb Histogram of L N B A S categorized by REGBASIN 151 Figure 8.3.2 Result of A N O V A test of effect of SLPPOS on R E G B A S I N 152 Figure 8.3.3 Result of A N O V A test of regional variation in SLPCONFG 152 Figure 8.3.4 Result of A N O V A test of regional variation in C U R V E 153 Figure 8.3.5 Result of A N O V A test of regional variation in DRAIN 154 Figure 8.3.6 Result of A N O V A test of regional variation in LITHOL2 155 Figure 8.3.7 Histogram of LOCN2 categorized by REGBASIN 156 Xll LIST OF VARIABLE NAMES AND DESCRIPTIONS In this paper, the following convention is used. When a variable is discussed in a general sense, for instance gully wall slope length, its name is written out in full. When the reference is to the names of variables specifically measured in this study, the name of the variable used is given in capital letters in a specific variable code, i.e. GWSD. For example: "When examining the effect of gully wall slope length on reach sidewall stability, it was found that all of the reaches with GWSD > 40m had slope failures on their sidewalls." AOE angle of entry of failure into gully channel, a (degrees) ASPECT aspect of gully system (true north = 0°) AVSLOPE mean slope angle of open slopes outside gully (degrees) BASAREA drainage area of gully (ha) BASIN drainage basin in which gully is located CATAOE categorized angle of entry (>42 or <42) CATFS categorized failure slope (20-30°, 30-45°, or >45°) CATLNVOL categorized natural logarithm of volume of initial failure (<5, 5-6, or >6) CATLNVL2 as CATLNVOL, but <5 or >5 only CG channel gradient (degrees) CH DEPTH depth of gully water channel (m) CH WIDTH gully water channel width (m) COSAOE cosine of AOE CURVE horizontal slope curvature (straight, concave or convex) CURVE2 CURVE in two classes, concave vs. convex/straight D50 D 5 0 of sediment sample DFIP debris flow initiation potential, from GAP DFOOR? was there a debris flow out of the reach? (yes or no - reach data) DFRESULT did the failure initiate a debris flow? (yes or no - failure data) DFTOFAN? did the reach produce a debris flow which travelled to the fan of the gully? (yes or no - reach data) DRAIN soil drainage (imperfectly, well-drained, etc.) FAILAZIM aspect of the failure (true north = 0°) FAILDATE year in which slope failure occurred FAILNO number of the failure in the gully, numbered sequentially from top to bottom FAILSLP slope angle of the failure plane FPMTL material(s) exposed in failure plane/failure scar GGP gully geometry potential to initiate debris flow from failure, from GAP GGP2FAN whether or not failure initiated debris flow which travelled to fan (essentially DFTOFAN? for PASTFAIL ='yes' only) GGPFAIL PASTFAIL with all 'no' results removed (used to generate GGP2FAN & GGPOOR) GGPOOR like GGP2FAN, but for debris flow out of reach rather than to fan GULLY number of gully within basin GWFP gully wall failure potential, from GAP GWSA gully wall slope angle (degrees) GWSD gully wall slope length [distance] (meters), steeper of the two walls measured ICSS volume of in-channel stored sediment per meter of reach length (m3/m) ICSS2 categorized ICSS, <2 m3/m, >2 m3/m ICSS3 categorized ICSS, <4 m3/m, >4 m3/m LANDUSE landuse at point of failure (clearcut or cutlock boundary) LENGTH reach length (m) LITHOL bedrock lithology (exact) LITHOL2 bedrock lithology- igneous, sedimentary, or metamorphic LITHOL3 " - igneous or sedimentary/metamorphic LITHOL4 granodiorite/phyllite/andesite or mudstone/blueschist/sandstone/-metasedimentary LITHOL5 bedrock lithology- igneous/metamorphic or sedimentary LNBAS natural logarithm of gully drainage area in square meters LNVOL natural logarithm of initial failure volume in cubic meters LOC3 failure location- head/channel or side LOCN failure location with respect to gully (exact) LOCN2 failure location- head, channel or side LOGDATE year gully was logged NEWDFIP new debris flow initiation potential, as calculated in chapter 8 NEWGGP 1, 2, 3 new GGP's as calculated in chapter 8 NEWGGP the version of the NEWGGP's above which is chosen as the best of the three proposals (=NEWGGP2) NEWGWFP new gully wall failure potential as calculated in chapter 8 PASTFAIL does the reach have a failure in it or not? (yes/no) REACH reach number in the gully system (sequentially numbered from top to bottom) REGBASIN regional basin (Chilliwack Valley or Norrish Creek) SEEP seepage at failure scar SEEP2 categorized SEEP- at headscarp, or other SIDE which side of the basin is the gully on? (LB, RB or HW) SINAOE sine of AOE SLPCONFG morphology of slope (see Appendix 2 for list of possible types) SLPPOS where on the slope the failure is located SOIL the soil class of the soil at the point of failure STDVOL standardized volume, VOLUME/BASAREA SURF generalized surficial material of reach walls, rock (R), colluvium (C) or till (M) S U R F F L SURF subclassified by PASTFAIL, thus 'ran' means SURF = ' M ' and PASTFAIL = 'no' T2 TERRAIN categorized into types with till ('Till(+)') and without ('other') TANAVSLP tangent of AVSLP TANCG tangent of CG TANCG2 (also as CATTANCG in failure data) TANCG >0.5 or <0.5 TANFS tangent of FAILSLP TANGWSA tangent of GWSA TERRAIN terrain unit of failure polygon, material only ('till', 'till/colluv.', 'colluv.', 'rock', or 'till/glaciolac') TYPE failure type (see Appendix 2 for list of possible types) VOLUME volume of initial failure, from dimensions of scar (m3) X S A R E A cross-sectional area of gully water channel (CH DEPTH times CH_WIDTH) %FINES percentage of sediment sample composed of silt and clay, by weight %REVEG proportion of failure scar that is revegatated FAILIFY number of failures in reach, if greater than zero. xiv ACKNOWLEDGEMENTS This thesis required more time and a greater effort to complete than any other project I have attempted to date. Without the contributions of others, I would not have succeeded. My supervisor, Dr. Michael Bovis, was always available for consultation when needed, but I would especially like to thank him for allowing me to work as independently as possible. He also deserves credit for bringing this project to my attention. Drs. Michael Church, Olav Slaymaker and David McClung provided inspiration, constructive criticism and suggestions as well. Funding for this project was provided through a Forestry Renewal B.C. grant to the Ministry of Forests. I would like to thank Tom Millard for developing the concept behind this project and including me in it. In addition to his coordination of the project and financial support, I would like to thank him for letting me use his comprehensively annotated air photo collection, leff Ladd at the Chilliwack MoF district office provided further information with regard to the study area which was invaluable. Chad Rudiak stepped in at a key moment to provide essential information on the Norrish Creek area which contributed greatly to the project. The field work for this project took place under a wide range of weather conditions, and involved many long and arduous days of manual labour. I would like to thank my field assistant, Cam Campbell, for all his help, and especially for his off-roading experience which got us into many inaccessible spots. I would also like to thank Shane Cook and Chad Rudiak for their contributions, and the Varsity Outdoor Club in general for introducing me to the Chilliwack Valley in the first place. Fred Beckey's climbing guide to the area was an invaluable source of up-to-date acces information. Number crunching and writing associated with this thesis used up inordinately large amounts of computer time. I would like to thank the other grad students of the U B C Geography Department for graciously allowing me to monopolize the sediment lab computer for extended periods of time. Finally, I would like to thank my family, both for their encouragement and their financial support. By living at home I have been able to graduate debt-free. I couldn't have done it without them! 1 Chapter 1- Introduction 1.1 Gullies as a landscape component of forested hillslopes in coastal British Columbia. Gullies are steep headwater streams. They are a common geomorphic feature in the steep forested terrain of the coastal belt of British Columbia. They are defined as relatively steep, linear incisions in a hillslope, with roughly V-shaped cross-sectional form. Hydrologically, gullies are first- or second-order streams (Horton, 1945) with a well-defined stream channel and seasonal to episodic runoff regime. The upper portion of a gully system normally falls within the definition of a zero-order basin (Dunne, 1980; Dietrich et al., 1987) since it often lacks a well-defined channel. Gullies are relatively small drainage features which occupy only a small portion of the total landscape, but are disproportionally important as sediment sources and sediment delivery areas. The B.C. Terrain Classification System (Howes and Kenk,T988) recognizes this, and includes gullies as a distinct category. The past twenty years have seen a recognition of the geomorphic significance of gullies in the forested coastal region, and an increase in scientific research concerning gullies (Swanston and Swanson, 1976; Alley and Thomson, 1978; Dietrich and Dunne, 1978; Eisbacher and Clague, 1981; Swanson et al. 1982; Wilford and Schwab, 1983; Rood 1984, 1990; Krag et al., 1986; Howes, 1987; Chatwin et al. 1991; Rollerson, 1992; Millard, 1993; Oden, 1994; Sterling, 1997). Gullies have a distinctive morphology (Millard, 1993; GAP, 1995). The typical gully system has three distinct zones: the headwall or source zone, the transport zone (the main gully and its channel), and the depositional fan. The headwall includes "source areas that have very steep headwalls and sidewalls, are very susceptible to extensive erosion and landsliding, and often are the zone where debris flows start" (GAP, 1995) (Figure 1.1). Some gullies have heads that do not meet these criteria. These gullies, rather than having steep headwalls, have a zone where a slight open-slope depression, often associated with increased groundwater seepage, deepens 2 S I D B W A L U •_\ope yfto \ ey \V ^>Wp C M S ojoove. pV. a t \MV\\CV\ CV\O.V\MA W ^ V A S s. CHAKIMEL. Figure 1.1: A gully with a steep headwall, shown in simplified form. The headwall ends at the point at which the gully channel begins. The headwall is steeper than the gradient of the channel below it. At minimum, the height of the headwall is equal to the height of the sidewalls, but it may be much greater (as shown in the diagram). The headwall may be sharply delimited from the sidewalls, or merge into them. 3 Figure 1.2: An open-slope depression type of gully head, shown in simplified form. Above the point at which the gully channel begins, the sidewalls decrease in height and slope angle until the open-slope depression becomes indistinguishable from the slopes outside the gully. downslope until it becomes a gully (Figure 1.2). This latter type of gully head resembles the zero-order basin as described by Dunne (1980), or the similar slope hollow of Dietrich et al. (1987). The transport zone is that portion of the gully that best defines the gully. The transport zone has "steep gully sidewalls, confined channels with perennial or intermittent streams, unstable or partly unstable banks, and temporarily stored woody debris and sediment. The transport zone is often a confined, V-notch ravine " (GAP, 1995). Below the transport zone of the gully, a fan may or may not be found. The fan is where sediment and debris transported from the headwall and transport zone are deposited. If the gully enters a stream channel competent enough to transport material leaving the gully, then a fan will generally be absent. The deposition zone may be a gently sloping Figure 1.3: A typical gully system. Logging date is 1990; the slope materials consist of 2m of deeply weathered soil over basal till; bedrock ( phyllite) is exposed in the gully beds in reaches above the road. Below the road, the surficial materials are thicker and more colluvial in nature. The gully at photo right, FO-001-RB, experienced a slope failure in 1995 at the cutblock boundary, in the region of the gully head (no obvious headwall); a debris flow was initiated and travelled to the fan (at bottom of the photo). The tributary gully at photo left, FO-002-RB, experienced a smaller slope failure in 1990. The failure began on a sidewall halfway between the road and the cutblock boundary. A debris flow resulted, but flow material was deposited on the road, and the flow did not reach the lower gully. true fan or be a steeper cone form. The fan may have single or multiple channels ranging from deeply incised to unconfined (GAP, 1995). A typical gully system is shown in Figure 1.3 (however, most gullies in the study had only a single channel; tributary gullies were rare). The Gully Assessment Procedures handbook of the Forest Practices Code of British Columbia (1995) defines a gully in terms of channel dimensions. It is defined as having overall channel gradient greater than 25 percent, and at least one reach (not including the fan) of minimum length 100 m must have sidewall slopes steeper than 50 percent, a channel gradient of at least 20 percent, and a sidewall height exceeding 3 m as measured along the fall line. All gullies in this study meet these criteria. 1.2 Debris flow as a geomorphic process characteristic of gullies. Gullies are usually the most unstable areas of the forested coastal belt because of their steepness and wetness (Rollerson 1992, Bovis et al. 1997) . Due to their depth of incision into a hillslope, they are natural foci for the concentration of both surface runoff and groundwater seepage. Their sidewall and headwall slopes are typically 10° to 15° steeper than the slope on which the gully is incised. This combination of steep slopes and high water pressures results in a much higher incidence of slope failures on gully walls than on the planar hillslopes outside the gully system. The dominant processes of slope failure on gully walls are ravelling, debris slides and debris slumps. In coastal British Columbia, the period of highest precipitation is during late autumn and winter, and it is at this time that soil moisture and runoff are at maximum levels as well. Most slope failures in gullies occur during this period, and failures reaching the gully channel typically encounter enough water to generate a debris flow (Bovis and Dagg, 1988). Debris flow is the dominant geomorphic process occurring in gullies (Swanston and Swanson, 1976; Millard, 1993). The process of debris flow in gullies has received increased 6 scientific attention during the last twenty years, as attention on gullies has increased. With an increase in scientific attention has come a confusion of terminology. Debris flow is defined as the rapid movement downslope of a mass of saturated, poorly sorted, mineral and organic debris which behaves as a single-phase slurry (Jordan, 1994 after Varnes, 1958, 1978 and Costa, 1988). Debris flow is a process "intermediate between waterflooding and landsliding, with mechanical characteristics different from either of these processes" (Costa, 1984). In an attempt to distinguish debris flows occurring in gullies from those which occur outside the gully system, the terms "channelized debris flow" and "debris torrent" have been used in the past to indicate specifically those debris flows occurring in gullies (VanDine, 1985). The term "debris torrent" has also been used to indicate debris flows which have relatively coarse clastic sediment and a high proportion of woody debris, while the term "debris flow" is retained for fine-grained debris flows with lower woody debris content (Swanston and Swanson,1976; Slaymaker et al., 1988; Sterling, 1997). However, mechanisms of debris flow and debris torrent are largely identical, and the two terms represent two extremes of a continuous range of sediment size and content. Current practice is to refer to all such movements as debris flow, sub-classified, where necessary, into fine-grained and coarse-grained types. Hence, all such events in gullies will be called debris flows hereafter, rather than debris torrents. A related phenomenon, debris flood, is similar to debris flow, but with more water and less sediment. In debris flood, the water and sediment phases are not irreversibly intermixed, and separate upon deposition (Costa, 1988). Debris flood is, apparently, a rare phenomenon in coastal British Columbia (Rollerson, 1984; Rood, 1990). Jordan (1994) presents a classification system which recognizes the range of environments and initiating mechanisms which produce debris flows. His system recognizes eight separate sub^  types of debris flow. The two sub-types of interest with regard to this study are: "g) rainstorm-initiated debris flows in steep creek channels in humid, forested environments; 7 h) debris flows on forested slopes or in small gullies, initiated by shallow debris slides which are often caused by forestry activities." (Jordan, 1994) All of the debris flows represented in this study can be classified as either type g or type h. Many of the debris flows discussed in this study are "initiated by shallow debris slides... caused by forestry small gullies" (type h); those same "small gullies" are the uppermost reaches of "steep creek channels in humid forested environments" and the "shallow debris slides" are by and large "rainstorm-initiated", either directly or indirectly (type g). Given that the classification system of Jordan (1994) is implicitly exclusionary, in that the sub-types do not seem intended to overlap one another, it seems most reasonable to classify all the failures discussed herein as type h. Type g failures seem to be intended to represent a larger scale of debris flow, with little or no influence from human impacts (such as logging). Such debris flows were recently explicitly identified as being of Jordan's (1994) type g in a study by Sterling (1997), and debris flows in her study are manifestly of a different type from the majority of those studied herein. In addition, the focus of the GAP is on debris flows produced by forestry activities, and only Jordan's type h is explicitly so initiated. Debris flow represents a significant hazard in several respects. First, the delivery of large quantities of sediment to a higher- order stream may adversely impact salmonid populations through destruction of habitat and spawning beds, and may reduce water quality downstream (Tripp and Poulin 1986a, 1986b, 1992). Secondly, in mountainous terrain it is common for the fan portion of a gully to be the only area suitable for development. This has had the effect of concentrating human activity in precisely the area most likely to be struck by a debris flow. The magnitude of events is such that serious property damage, injury and/or fatality are the likely outcomes when debris flows intersect human activities (Hungr et al., 1987; Thurber Engineering, 1988). Thirdly, debris flow in areas of forest harvesting can destroy roads and bridges, remove 8 thin forest soils, and may often destroy all vegetation in the path of the flow. This increases the costs of logging and can remove areas from timber production for a period of decades or longer, creating a decline in long-term sustainability of timber harvesting. 1.3 The Forest Practices Code and Gully Assessment Procedures. The government of British Columbia introduced the Forest Practices Code in 1995 in an attempt to ensure that forest practices would follow environmentally sustainable procedures. The Code consists of four components: the Act, the regulations, the standards, and the guidebooks. The Act is the legislative umbrella which authorizes the Code's other components. The regulations lay out the forest practices that apply province-wide. The standards are established by the chief forester, when required, to expand on a regulation. Forest Practices Code guidebooks have been developed to support the regulations, but are not part of the legislation. They provide recommendations for procedures, practices and results that are consistent with the legislated requirements of the Code. Guidebooks are designed to help foresters and geomorphologists exercise their judgement in developing site-specific management strategies and prescriptions to accomodate local resource management objectives. One such guidebook is the Gully Assessment Procedures Guidebook (GAP, 1995). The GAP was designed to provide a framework for evaluation of debris flow and debris flood hazard in the gullies of coastal B.C. The Code requires that all areas planned for harvest be assessed for stability problems, and the GAP was developed to fill this need. The material in the GAP was taken from an earlier publication of the Watershed Restoration Program (Hogan et al., 1994) to which many B.C. terrain scientists contributed concepts and ideas. The GAP is intended to be used by forest technicians to provide a qualitative evaluation of the likelihood of debris flow or debris flood, and such use is considered to be only a preliminary step. If the GAP criteria indicate a moderate or high likelihood of a debris flow or flood, a professional geoscientist or engineer is 9 required to assess the gully if logging is to occur. Management strategies in the GAP become more conservative as the degree of risk associated with forest harvesting increases. The Debris Flow Initiation Potential (DFIP) section of the GAP was developed primarily from research carried out in the Queen Charlotte Islands under the auspices of the Fish/Forestry Interaction Program. The applicability of this research to the rest of coastal British Columbia is as yet undetermined, and, as explained below, is one of the objectives of this thesis. The Queen Charlottes are known to have extreme susceptibility to mass movement owing to their wet climate and poorly consolidated hillslope materials; thus it is suspected that the DFLP might be too conservative when applied in other areas of the coastal belt. A further problem with the DFLP is that there is little to allow a trained geoscientist to differentiate between gullies that pose a substantial risk from those that do not if a high to moderate DFLP rating is obtained. It is understood that this judgement is to be made based on the geoscientist's experience and knowledge of similar gullies, but there is a lack of published material for the professional to refer to should their experience not prove sufficient. The net result is that two geoscientists of equal experience might evaluate the same gully somewhat differently. 1.4 Thesis Statement of Purpose and Definition of Research Goals The objectives of this study are to study the factors affecting debris flow initiation in coastal gullies and to develop a greater understanding of the relative importance of the various geomorphic and geotechnical parameters in determining whether or not a gully is likely to experience a debris flow. Specifically, the objectives of this study are: 1) to test the effectiveness of the Debris Flow Initiation Potential section of the Gully Assessment Procedures in the region of the North Cascades and southern Coast Mountains of British Columbia; 10 2) to test other terrain parameters not presently included in the DFLP to assess the relative importance of various geomorphic and geotechnical variables in determining debris flow initiation potential, and to evaluate the feasibility of including such parameters that prove significant in a revised version of the DFLP; 3) to examine the degree to which debris flow initiation in gullies varies regionally within the study area, and to provide results so that the regional variability of factors comprising the DFLP may be evaluated for the entire coastal belt of British Columbia1. This study focusses exclusively on debris flows in gullies which have been subjected to timber harvesting. Unlogged gullies were therefore not included in the study. In the study area, undisturbed (old-growth) forest cover now comprises only a small fraction of the total area. Not all of such undisturbed forest is gullied terrain, and the old-growth areas are generally confined to relatively steep, inaccessible areas which are probably not representative of the bulk of the terrain in the forested coastal belt. The study sites were divided between two field areas: the portion of the northern Cascade Mountains drained by Chilliwack River, and the portion of the southern Coast Mountains drained by Norrish Creek. Regionally, climate is similar in both basins. Lithology and surficial geology are variable in each basin, and are characteristic of the ranges in which they occur. The similarity of climate between the two areas allows the effects of material type and terrain type to be examined in this study. The general approach is one of field investigation of gullies which have experienced at least one slope failure since harvest. Gully parameters were studied at three scales: the gully system as a whole, individual reaches within the gully, and at the point of initiation of each failure. At each scale, a wide range of geomorphic and geotechnical parameters was measured or 1 This study is one of three studies currently being conducted in the coastal belt to assess such variability. 11 evaluated and recorded. It was also determined whether or not the failure in question had initiated a debris flow. Statistical analyses of the observed parameters then followed with the aim of evaluating which of the measured parameters were most effective in determining whether or not a slope failure would occur in a reach, and whether or not a failure would produce a debris flow. The geomorphic and geotechnical parameters included all parameters in the DFLP evaluation, and others used in previous studies of terrain attributes (Howes, 1987; Rollerson, 1984, 1992). Measurement of parameters was consistent with the presently accepted field procedures in the Forest Practices Code, and an attempt was made to choose a measurement style which would allow quick and efficient evaluation of chosen variables. The general hypothesis underlying the study is that debris flow initiation in gullies is affected by geomorphic and geotechnical variables, and that by understanding the role those variables play in affecting debris flow initiation, an assessment of terrain variables for a particular gully will allow a prediction to be made of the likelihood of such a gully producing a debris flow after it has been logged, assuming that similar climatic conditions apply. In this study, the focus was on gullies having a morphology permitting timber harvesting, typically with a total channel length (excluding fan) of hundreds of meters and a depth of tens of meters. Volumes of initial failures were on the order of tens to a few thousands of cubic meters. Failures consisted primarily of surficial (colluvial or glacial) materials, and generally did not include a substantial bedrock component. A larger scale of debris flows also exists, in which gully lengths are on the order of one kilometer or more, depths are of a few hundreds of meters, and slope failures more frequently involve bedrock and volumes are of tens of thousands of cubic meters and greater (up to millions of cubic meters). Debris flows at this scale have been studied elsewhere by Jordan (1994), Jakob (1996) and Sterling (1997). The DFLP section of the GAP is 12 not intended to deal with gullies on this scale: typically the source and transport areas of such gullies are difficult, if not impossible, to log, and timber harvesting occurs only on the fan. None of the gullies included in this study were subject to snow avalanches. Although gullies in which snow avalanches occur were found within the study area, they were excluded for two reasons. First, snow avalanches transport clastic and woody debris, and this effect could have introduced a confounding factor into the study. Secondly, gullies which experience snow avalanches typically have a distinctive vegetation pattern dominated by slide alder; this is a noncommercial species and hence snow avalanche gullies typically are not logged. The assessment of DFLP is intended to be carried out prior to forest harvesting. To assess the utility of the DFIP, gullies which had been logged between six and fifteen years prior to the study were chosen. This time interval ensured that sufficient loss of strength would have occurred from post-logging root decay. This raises a question of the applicability of the study. Can results from gullies studied post-harvest be applied to a process intended to be carried out before harvesting takes place? Clearly, it must be assumed that most of the terrain variables measured would not be changed significantly as a result of harvesting. The other method of evaluating the DFLP would be to study gullies just before logging, then return in six to fifteen years to record debris flow events. Although this would be a valuable complement to the present study, it is not feasible for two main reasons. First, the need to evaluate the DFLP is immediate: the time scale required for the pre-harvest study is too long to be acceptable to forest managers. Secondly, the present study has climatic forcing (extreme precipitation events) during the period of maximum strength loss due to root decay; whether this would occur during a pre-harvest study is unknown. Prior to entering the field areas, aerial photographs and terrain maps were consulted to determine locations of gullies with potential failures. In the field, gullies were located, assessed for DFLP, and the relevant terrain parameters were measured. Upon completion of fieldwork, 13 field notes were collected, ordered, and checked for accuracy against photos and maps. The raw field data were organized into two overall data sets, one dealing with parameters variable at the reach scale, the other dealing with parameters variable at the scale of the individual failure. Analysis of the field data occurred in four stages. In the first stage, the effectiveness of the DFIP assessment was evaluated by comparing the DFIPs computed for the reaches as a group to the observed occurrence of debris flow within those reaches. Once this was done, the effectiveness of the components of the DFIP were assessed by comparing them to the observed results, allowing an assessment of the relative importance of the individual components of the DFLP to the performance of the whole. In the second stage, the various terrain parameters were considered at the scale of the individual failure. The relative importance of each parameter in determining whether or not a given failure would result in a debris flow was statistically evaluated at this stage. In the third stage of the analysis, the knowledge of the relative importance of the various terrain parameters in influencing debris flow initiation was used to create new methods of assessing DFIP. These new DFIPs were then tested at the reach scale for effectiveness In the same manner that the effectiveness of the original DFIP was tested. The most effective of the new DFIPs were then considered from the standpoint of practicality of application, and the method which combined the greatest degree of effectiveness with the greatest ease of application was recommended as the suggested revision of the DFIP. In the fourth and final stage of analysis, the two study areas, the Chilliwack River and Norrish Creek drainages, were compared in order to determine what effects regional variability in lithology and surficial material distribution exerted on debris flow initiation. Once data analysis had been completed, the results were summarized in a concluding statement. The limitations of the conclusions reached were stated, and suggestions were made on how further research could address these limitations, and improve the design of similar studies. 14 Chapter 2- Theory of Debris Flow Initiation in Gullies 2.1 Mechanics of slope failure in gullies. In coastal British Columbia, the majority of debris flows in gullies begin as small slope failures on either the gully headwall or sidewalls (Rollerson, 1984, 1992; Rood, 1990). Only a small proportion (less than 2%, according to Rollerson (1984)) initiate in the gully channel as a result of high water discharge. This is in contrast to conditions in other parts of the world: for example, Takahashi (1981) found that the majority of debris flows in Japan were triggered in-channel (although more recent Japanese research has found that slope failures play a greater role than was recognized in the 1981 study (Slaymaker, pers. comm.)). Thus, to understand the causes of debris flow initiation in B.C. gullies, it is necessary to begin by considering the conditions which produce sidewall and headwall failures. The equation most commonly used to model slope failures is the infinite slope model (Sidle and Swanston, 1982). This is typically arranged into the form of a calculation of factor of safety (F.S.), where the ratio of resisting forces (shear strength), S, to driving forces (shear stress), x, is calculated for a slope. F.S. greater than 1 indicates the slope is stable, F.S. < 1 indicates instability (theoretically the slope should fail when F.S. declines to unity). F.S. = S = c' + (yz cos2p - u)tan x yz sinJ3 cos(3 A simple consideration of the most important factors which can cause a decline in factor of safety, making a slope more unstable, results in the following list of effects: - c', apparent cohesion, includes soil reinforcement due to roots, c p ' (O'Loughlin, 1972). Logging causes roots to decay, lowering c p ', hence lowering S and F.S. As new growth takes place, c p ' increases again. Another effect of cohesion to be considered is that it is a material property of clastic surficial materials, typically increasing with increasing clay 15 content. So, in theory, surficial materials with higher cohesion will be more stable at a given angle than those which are cohesionless. Another process which can alter cohesion is weathering. Thus, a weathered layer in soil may be more likely to fail than the unweathered material beneath it. - P, slope angle. If the effects of cohesion and pore water pressure are not considered, then the equation simplifies to F.S. = tan <j)7tan p. So an increase in slope angle will decrease F.S.; therefore steeper slopes are more likely to fail. - u , pore water pressure. Increasing pore water pressure reduces shear strength and hence the factor of safety. The most common trigger for gully wall failures is climatic forcing (Caine, 1980; Church and Miles, 1987; Bovis and Millard, 1992). This can take place in several ways. Some examples are: 1) through a direct increase in soil moisture and pore water pressure in a homogeneous soil mass; 2) through a concentration of runoff in the gully channel, causing erosion of banks and consequent oversteepening of sidewalls; and 3) through local concentration of subsurface moisture along a boundary between regions of differing hydrologic properties (such as a boundary between unweathered and weathered glacial till), resulting in increasing pore water pressure along the boundary, leading to failure. Climatic forcing is discussed in more detail in section 2.3. Gully wall failure can also be caused by windthrow of trees on the gully wall, human or animal disturbance, and (much more rarely) by seismic activity. Snow avalanches in gully systems can also trigger gully wall failures, either immediately after the avalanche (if the avalanche destabilizes the gully's walls), or much later, when avalanche deposits melt, depositing unconsolidated colluvium into the gully bed at the same time that there is high runoff. Note that wall failures in a snow avalanche gully are unlikely to trigger debris flows if the gully into which the failure occurs is full of relatively new, dry snow; however, if the snow is old and wet, slushy debris flows may 16 result if the snow becomes incorporated into the debris flow, for example through pressure melting caused by impulsive loading during impact (Bovis and Dagg, 1992). Jordan (1994) observed that old avalanche deposits tend to be relatively impervious to debris flow entrainment, and that small failures entering a gully filled with old avalanche snow tend to run out on top of the snow without entraining it or producing larger debris flows. 2.2 Behaviour of debris flow in gully system. Once a gully wall has failed, the material will typically enter the gully channel. Due to the steepness and profile of gully walls, it is highly unlikely that a wall failure will come to rest before it can enter the channel. Once the failure enters the channel it will either come to rest as a mass of sediment, or begin moving down the channel as a debris flow. In the latter case it may, if conditions permit, stall as a debris jam before leaving the transport zone, or it may continue as a debris flow all the way to the fan. Rood (1990) observed that only about one-third of sidewall and headwall failures produce debris flows. If the initial failure is saturated, it may already be travelling as a debris flow when it reaches the bed of the gully. In the case of unsaturated failures (debris slides), there must be sufficient flowing water in the gully to saturate the failed mass for it to become a debris flow. Since most wall failures occur during periods of high precipitation and/or snowmelt, this condition is often satisfied. In cases where it is not, and a debris flow does not result, the failure mass will remain in the gully bed. If it dams the gully channel, water may build up behind it, and the failure of the dam may also be a trigger for a debris flow. If this does not occur, the failure mass will become part of the sediment stored in the gully channel, where it may be reworked fluvially, or possibly entrained by a future debris flow or flood. The angle at which the failure enters the channel has been shown to be an important factor in determining whether or not the failure becomes a debris flow (Benda and Dunne, 1987; Bovis 17 and Dagg, 1992). Failures entering the channel at a low subtended angle (a) are more likely to become debris flows than those which enter at a higher a. Failures which enter the channel at angles approaching the perpendicular (a = 90°) are the most likely to create debris dams in the channel. Channel steepness plays an important role in determining whether debris flows will result from failures into the gully, and in determining the speed at which the resulting debris flow travels. Various rheologic models are used to simulate debris flow dynamics and behaviour; all incorporate channel slope, though not all weight it equally (Jordan, 1994). If the velocity of the debris flow drops below a certain threshold, the debris flow will stall; the flow will stop moving and water will slowly drain from the mass (usually over a period of several days). Since debris flow velocity is dependent on channel slope, areas of gentler slope will be preferred locations for debris flows to stop: these are typically the fan zone, but can include benches or steps in the gully profile, either natural or those created by roads crossing the gully. The effects of inertia mean that the larger a debris flow is, and the faster it is moving, the less likely it is for a given step in the profile to halt its progress; thus, large debris flows are the most likely to travel to the fan, or even past the fan and down the higher-ordered stream below. When a debris flow travels down a channel, it flows in a series of surges, in which flow behaviour is predominantly laminar (Costa, 1984; Jordan, 1994). Larger clasts typically are found on the top and sides of the deposit, possibly even in the form of a rigid "plug". As the debris flow travels, it leaves behind it levees, and when it stops, the deposits display steep snouts and inverse grading. As a debris flow travels down-channel, it entrains all readily erodible material from the gully channel, both loose colluvial material on the gully bed, and large sediment wedges stored behind woody debris jams (tree trunks, root wads and similar items, known as coarse woody debris (CWD) or large organic debris (LOD)). This entrainment process means that a debris flow 18 can grow in volume by an order of magnitude or more from the initial failure volume by the time it reaches the fan (Swanston and Swanson, 1976; Bovis et. al 1997). After the passage of a debris flow, gullies are typically scoured of all stored sediment, and the channel shows either rock, or less erodible surficial material such as basal till. Some debris flows have been observed to entrain material so rapidly that they stall themselves: the "plug" becomes the front of a channel-wide jam of CWD and boulders. Thus, the amount of in-channel stored sediment (ICSS), including both colluvial debris and CWD, is an important factor regulating the behaviour of debris flows in gullies. When debris flows travel around bends in a channel, a phenomenon known as superelevation occurs, in which the flow banks around the curve, with the outside edge of the flow rising higher than the inner, since it is travelling at a greater velocity. Since most debris flows are observed after they occur, at which time other factors may have affected the deposits of the flow, superelevation is often measured from differential heights of levees, or from mud streaks or vegetation trim lines on channel walls. Superelevation may be used to back-calculate peak velocity (Pierson, 1985; Hungr et. al, 1984; Jordan, 1994) of the debris flow. Peak velocity and channel cross-sectional area may be used to calculate peak discharge, and this may be related empirically to the volume of the debris flow. All of these calculations have a large degree of error (Jordan, 1994). In this study, channels were, by and large, straight, and superelevation was not commonly observed. 2.3 Meteorological antecedents to debris flow initiation. Most debris flows are triggered by precipitation, generally either directly from rainfall or from a combination of rainfall and snowmelt. However, it is difficult to predict debris flows using only meteorological data (Church and Miles, 1987; Bovis and Millard, 1992). One reason for this is the coarseness of the sampling network: weather stations with rain gauges are typically located 19 in major valleys, where human activity is greatest and transportation corridors are found, and not in the mountainous areas where debris flows occur. The orographic effects of mountain topography mean that rainfall is typically greater in the mountains than in the adjacent lowland valleys. In addition, mountains act to selectively block and magnify storm cells such that it can be raining heavily in one drainage while on the other side of the ridge the sky is clear and the sun is shining. Thus, the rainfall records of the nearest gauge may have little resemblance to actual patterns of precipitation in the gully or gullies of interest. Furthermore, in two gullies with similar terrain characteristics, the intense spatial variability of mountain precipitation may mean that one will produce a debris flow while the other will not, even if the two are adjacent. Although it can be difficult to predict debris flows from precipitation records, the occurrence of heavy precipitation is nonetheless frequently associated with initiation of debris flows. A threshold value for precipitation intensity above which debris flow is likely to occur has been calculated to lie in the range of 80-100 mm in 24 hours (Caine, 1980), but due to the orographic intensification of precipitation and the coarseness of the sampling network, such a storm may only be recorded as delivering 40-60 mm over 24 hours; measured rainfall intensities in coastal British Columbia often exceed this value by a considerable amount. In the Queen Charlotte Islands, minimum 24-hour precipitation values required for debris flow initiation have been calculated to be as low as 20 mm for wet antecedent conditions and 30 mm for dry antecedent conditions (Hogan and Schwab, 1991). In addition to the effects of direct precipitation, the effects of rain-on-snow, or of sudden warming which causes thawing, can magnify the effects of storms which do not meet this threshold value (Church and Miles, 1987). Thus, while it may be difficult to predict whether or not an individual gully will produce a debris flow or not during a storm of given intensity, one is often able to predict with reasonable confidence that a storm of sufficient magnitude will have produced a debris flow somewhere. 20 Millard (1993) summarized this by stating that "[although 'apparently unremarkable' water inputs may produce debris flows, very remarkable events are more likely to produce debris flows". In some gullies, the supply of readily entrained debris is nearly infinite, and any precipitation event of sufficient intensity will trigger a debris flow. This is the transport limited case. In other gullies, debris supply is the limiting factor in determining whether debris flow will or will not occur. This is the supply limited case. These two behaviours have been described in detail by Jakob (1996) for large debris flow gullies of the southern Coast Mountains. In the small debris flow gullies described herein, the situation is slightly different. Since almost all debris flows in this study include a slope failure as a necessary trigger, neither purely supply limited or purely transport limited models of the behaviour of the gully seem completely applicable. Large rainstorms which do not produce a slope failure typically do not produce debris flows, even if there is a large quantity of debris already present in the channel (except for the very small fraction, approximately 2%, of gullies which do experience in-channel initiation). One could, if necessary, consider this as a case of transport limit with an extremely high threshold for initiation which is usually not exceeded, or as a supply-limited case with a requirement for slope failure into the channel as the only 'supply' to count. This departure from the supply or transport limit model is possibly a reflection of the important role played by CWD jams in stabilizing debris in the channel in gullies of this scale, where the size of the gully is of the same order of magnitude as the size of the largest CWD. In the larger supply- and transport- limited gullies, the CWD is so much smaller than the gully itself that it does not play as significant a role in stabilizing in-channel debris. The necessary meteorological requirements for debris flow initiation, then, are not so much a straightforward intensity of precipitation or amount of runoff as they are for an event or series of events which will raise soil moisture enough to cause slope failures into the gully in 21 question. Such events also typically produce enough runoff that any failure entering the channel can be mobilized as a debris flow should terrain conditions be favourable. Thus, slope failures need not directly follow a high precipitation event. Thurber Engineering (1988) described a slope failure in lower Chilliwack Valley, occurring approximately two weeks after a record rainstorm, which produced a debris flow. Site investigation showed that the storm had saturated the ground, but it was a later freeze which produced ground frost and ground ice, filling tension cracks and impeding drainage, which was the final triggering event for the slope to fail. 2.4 Effects of forest harvesting on slope failure and debris flow initiation. Many studies have shown a higher rate of slope failure on hillslopes which have been logged compared to those which remain unlogged. Howes (1987), in a study overlapping a portion of the study area of this study, found that clearcut areas had up to 6.6 times as great a rate of failure as did natural areas; Young (1992) found that the magnitude of the increase was of a similar size. Rollerson and Sondheim (1985) and Rollerson (1992) found significantly higher rates of failure in logged areas, as did Rood (1984, 1990). Several mechanisms exist by which logging can increase the rate of slope failure. One already mentioned is root decay. After trees are cut, their roots decay and there is a gradual loss of cohesive strength, which reaches a maximum between 6 and 15 years after logging, after which the effects of regrowth cause root strength to recover (Sidle et al., 1985; Rollerson, 1992). This decline can be quite important: O'Loughlin (1972) reported that root strength comprised 71% of the total shear strength on a hillslope of saturated till inclined at a 35° angle . During the period of diminished strength, slope failures are more frequent than under natural conditions, since root decay puts the slope in a critical state whereby an increase in pore pressures, which normally would have little or no effect on slope stability, can cause the slope in its weakened state to fail. Aside from the delayed effect of root decay on soil shear strength, the removal of trees can also 22 weaken slopes immediately: the slope loses the weight of the trees, so the normal forces on the slope decrease (Brown and Sheu, 1975). Methods of timber extraction can also increase the rate of slope failure. Yarding and hauling of felled trees disturbs the ground surface, exposing and disturbing the soil, facilitating both surface erosion and water infiltration (Sauder et al., 1987). Harvesting methods which result in uprooting of trees are particularly effective in increasing slope failure rates due to their increased disturbance of the surface of the ground. Increased amounts of windthrow along clearcut boundaries can also increase soil disturbance. Root decay and mechanical effects of timber harvesting can also have the secondary effect of changing soil moisture levels, in addition to their primary effects on shear strength mentioned above. Root decay and ground disturbance caused by logging affect the subsurface drainage network, creating new drainage pores and blocking old ones (Ziemer, 1992). The result that drainage is generally impeded and so slopes stay wet for longer periods, and are thus more prone to pore-pressure destabilization. Surface drainage can also be affected by harvesting, diverting water to areas it normally does not flow, causing unusual erosion or saturation of the ground. These effects are also time dependent: regrowth of vegetation and decay of old roots are two processes which will continue to influence soil moisture conditions long after the date of logging. In areas of undisturbed forest, windthrow is a common disturbance of the soil. Margins of cutblocks can experience large increases in rate of windthrow (Stathers et al., 1994). Craters left by windthrow of trees are often trigger points for slope failures, especially if they are located at breaks of slope such as at the top of gully side- or headwalls. In this study, several debris flows were observed to have begun from craters left by windthrow of trees along clearcut margins. In addition to increasing the rate of slope failure, logging can also increase the chance that debris flow will result from failures into a gully, or the size of debris flow produced by a gully. 23 Increased erosion can contribute unconsolidated sediment to the gully bed where it can be entrained by a debris flow initated from higher up the gully system. Yarding of logs through a gully system can disturb consolidated gully bed sediments, making them more entrainable. Falling of trees and bucking of logs can increase quantities of woody debris in the gully system, increasing the size of CWD jams and the amount of entrainable material in the gully, hence increasing potential maximum volume of a debris flow (Millard, 1993; Oden, 1994). 2.5 DFIP evaluation of gully. The GAP is designed to evaluate most factors in a gully which could be affected by logging. The section of the Procedures most relevant to this study is the assessment of Debris Flow Initiation Potential (DFIP). Water transport potential is of secondary importance. The method by which DFIP is assessed is described in detail in Appendix 1. A brief summary follows. DFIP is assessed by considering four factors. They are the steepness, length, and surficial material type of the gully walls, and the channel gradient. An ordinal scale of High, Moderate and Low is used for these factors to develop an overall assessment of debris flow initiation potential within a gully. In addition to these four factors, the history of the gully with respect to past debris flows is considered and overall DFLP is taken as whichever of the two is greater, so that if assessed DFIP were Low, but evidence existed of past debris flows initiating in the reach in question, then assigned DFIP would be High. The high, moderate, and low classes are subjective ratings only, based on a consensus of informed opinion (and current research standards of the time) among the geomorphologists who created the GAP. The rankings do not carry an associated numeric value, so that DFLP class High, for instance, is not associated with a specific percentage chance of debris flow initiation from the gully reach in question. In addition to determining 24 In addition to determining the potential for debris flow initiation in the reaches studied, the GAP attempts to determine the potential for debris flows from upslope (above the highest reach assessed- typically the top of the proposed clearcut) if the area of interest does not include the entire gully. The intent of the DFIP assessment is not to provide an evaluation of the precise probability that a debris flow will occur in the reach or reaches in question. Rather, it is intended to provide a simple, qualitative assessment of the chance that debris flows will initiate following harvesting. This assessment may be then used to suggest management strategies to minimize the impacts of forest harvesting on the gully system. Reaches asssessed as having High or Moderate DFLP require study in more detail by a trained geoscientist, who will suggest site-specific management options for these less stable sites. Thus, the DFIP acts as a screening device, discriminating between low-risk sites and those which require further study at a greater level of detail. In summary, debris flows in gullies of the coastal belt of British Columbia typically begin from slope failures on the gully walls. Initiation of debris flows is influenced by a variety of geomorphic and geotechnic variables, with a typical proximate cause being a heavy rainstorm or rain-on-snow event. Slope failures and debris flows are more likely to occur in gullies which have been logged. The DFLP assessment section of the GAP attempts to estimate the qualitative likelihood that a gully will experience a debris flow after logging, using a simple but effective technique based on parameters which previous research has shown to be important in influencing debris flow initiation. The DFIP assessment is intended to serve as a first step in the process of developing a management plan for gullies slated for forest harvesting, distinguishing between low-risk gullies and those which require further study by a professional geoscientist. 25 Chapter 3- Description of Study Areas: Chilliwack Valley and Norrish Creek. 3.1 Physical parameters and geographic description. The study was conducted during the summer of 1996 (June and July) in the drainages of Chilliwack River and Norrish Creek (Figures 3.1, 3.2.1 and 3.3.1). In Chilliwack Valley, the basins of Tamihi Creek (including the tributary "T.L.M." Creek), Borden Creek, Slesse Creek, Center Creek, Airplane Creek, "Spoon" Creek, Foley Creek and "E.S.P." Creek were studied. In Norrish Creek watershed, a small portion of the main Norrish Creek, East and West Norrish Creeks, "Cyr" Creek, "Hanson" Creek, and Rose Creek were studied, as was a small portion of uppermost Margaret Creek, which is located just outside the Norrish Creek watershed divide in the drainage of Statlu Creek. Basins in Chilliwack Valley (Fig. 3.2.1) occupied a total of 223 km2, ranging from 4 km2 for the smallest to 52 km2 for the largest, and counting only those parts of the drainage which were included in the study (areas south of the International Boundary were not included): Tab le 3 .1 -Sub-Bas ins and Areas , Chi l l iwack River Va l ley Bas in A r e a (sq. km) Notes Tamih i 49 includes T . L . M . sub-bas in Borden 17 S l e s s e " E . S . P . " " S p o o n " Center 52 4 7 36 42 16 sub-bas in of Ch ipmunk Cr. Fo ley (above Airp lane) A i rp lane Total 223 Basins in Norrish Creek watershed (Fig 3.3.1) occupied a total of 58 km2, ranging from 2 km2 for the smallest to 18 km2 for the largest: 26 28 Tab le 3 .2 -Sub-Bas ins and Areas , Norrish Creek watershed Bas in A r e a (sq. km) Notes W e s t Norr ish East Norr ish Margaret R o s e 15 12 2 18 6 drains to Statlu Cr. "Cy r " includes 2 gull ies draining to main Norr ish val ley but reached from Cyr Cr. road "i Hanson I I 5 Total 58 The total study area in both regions was 281 km 2 . 53 gullies were studied, and 62 separate slope failures (of greater than 25 m 2 surface area) were identified and studied in those gullies. A total of 92 individual channel reaches was identified and assessed for DFLP. The regional breakdown of these features is shown below: Tab le 3.3 - Distribution of Si tes in Each Main Study A r e a by Reg ion Reg ion Gul l ies Fai lures R e a c h e s Chi l l iwack 26 31 53 Norr ish 27 31 39 Total 53 62 92 Chilliwack River Valley is located in the Skagit Range of the Cascade Mountains roughly 10 km southeast of Chilliwack. All sub-basins examined are situated between Chilliwack Lake and the Slesse Park subdivision. Physiography is rugged, characterized by local relief up to 2000m, with low valley bottoms, high mountain ridges, and prominent, sharp peaks. (Fig. 3.2.2 and 3.2.3) Upper portions of stream valleys are of U-shaped cross-section, with a transition to V -shaped cross-section in the lower reaches. The main valley of Chilliwack River runs roughly east-west, and most tributary valleys run predominantly north or south into Chilliwack River. Alpine glaciation is not extensive, being confined to a few pocket glaciers, although several basins with Figure 3.2.2 Figure 3.2.3 Figure 3.2.2: View eastwards from "E.S.P." basin along axis of Chilliwack Valley towards the prominent horn of Williams Peak. Summit reaches 6900'. elevation, and valley floor is about at about 400' : total relief visible is almost 1700 m. A contact between granodiorite (unit Egd on Fig. 3.4.1) and the non-plutonic rocks it intrudes is visible on the ridgeline just west of Williams Peak (which is in the intrusive rock). Figure 3.2.3: The west faces of The Still and Welch Peak from Airplane Creek. Valley floor is at about 3000'; the summit of Welch Peak is at 8000'. Visible relief is about 1500 m. The cirque between the two peaks is not especially well developed owing to its southwest aspect, and there is no glacier in it at present, although it holds permanent snow. Although The Still appears as a prominent horn in this photo, it does not have a similar form when viewed from other angles. Rock is metavolcanic phyllites and schists. 30 headwaters in the USA have larger glaciers therein. Cirques, tarns, aretes and horns are common features in the mountains. Cirques are predominantly oriented north and east. Valley walls are steep, and many show slabs and walls of exposed bedrock (Fig. 3.2.4). Winter snowfall accumulations are on the order of 5 meters at 1500 meters elevation. Yearly precipitation totals are on the order of 2500 mm in the valley bottoms, with the majority of precipitation falling in the period October-April. Vegetation is typical of the Coastal Western Hemlock biogeoclimatic zone, with western hemlock, western red cedar and Douglas fir as the dominant tree species in the valleys, in association with devil's club, salmonberry and red alder. At higher elevations (above about 1,000m), this vegetation gives way to the Mountain Hemlock zone, featuring mountain hemlock, yellow cedar, Sitka alder and blueberry (Cannings and Cannings, 1996). Treeline is at Figure 3.2.4: The Illusion Peaks group looking west from Center Creek. The prominent east-facing cirque has a tiny, unmapped pocket glacier therein (not visible in the photo). Rock is Oligocene granodiorite of the Chilliwack Batholith. Total relief visible in the photo is about 1200 m; the rear wall of the cirque is about 500 m high. 31 Figure 3.3.1: Map of Norrish Creek and vicinity at 1: 125 000 scale showing location of studied basins; from 92 G/SE. [Blue grid spacing is 2km] 1) Margaret Creek 2) E. Norrish Creek 3) "Hanson" Creek 4) "Cyr" Creek 5) Rose Creek 6) W. Norrish Creek 32 approximately 1750m elevation ASL, and above treeline there are areas of alpine meadow along ridges. Logging activity in Chilliwack Valley began before 1900, extending gradually up the main valley, then into tributary valleys; current logging is concentrated on the side slopes of the tributary valleys. Norrish Creek valley is located in the Pacific Ranges of the southern Coast Mountains northeast of Mission (Fig 3.3.1). All basins studied are between the mouth of Dickson Lake and the confluence of Rose and Norrish Creeks. The area is rugged, with maximum relief expression on the order of 1000m, about half that of Chilliwack basin. Summits in the area are not prominent and ridges are rounded. The area is typical of the low mountains that flank the Coast Mountains along their southern and southwestern edges (Fig 3.3.2 and 3.3.3). The main valleys are U -shaped in cross-section, although Norrish Creek valley becomes V-shaped below its confluence with Rose Creek. Norrish and Rose Creeks flow predominantly north-south, and tributary streams flow mainly east-west to meet them. There are a few small cirques, not especially prominent, with a strong northeast orientation. Norrish Creek drainage is not presently glaciated. Winter snow accumulations are of approximately 5 meters in depth. Yearly precipitation totals are on the order of 2500 mm, with most precipitation falling in the October-April period. One large lake exists, Dickson Lake, naturally dammed by a rockslide (Evans 1985) and now used as a reservoir. There are a few small tarns as well. Vegetation is typical of the Coastal Western Hemlock and Mountain Hemlock biogeoclimatic zones, with dominant species the same as for Chilliwack basin (Cannings and Cannings, 1996). Treeline is at approximately 2000m elevation, above the height of all but the highest summits, and there are not extensive meadows, although a few patches of sub-alpine shrubs (not true meadows) occur locally. Although the entire drainage is a watershed reserve of Figure 3.3.2: Looking up the valley of Rose Creek towards its head from about one-third of the way up. Visible relief is about 400 m. Visible failures, from left to right, are ROS-002 (at far left), -004 F O l (narrow linear failure), -004 F02 (large open-slope failure entering gully below road), -006, failure from road (not considered), -007, and -008. Upper slopes were logged in 1981; lower slopes were logged c. 1965-1970. The most recent logging, in upper right-hand corner of photo, dates from 1994. 34 Figure 3.3.3: The south side of the East Norrish Creek drainage seen from the north side. Visible relief is about 600m. Visible from left to right are a large, ungullied open slope failure; gully ENR-004-LB, with two gully head failures and one natural failure entering from above the clearcut; a gully with no failures in it (crossing lower road between -004 and -005); and gully ENR-005-LB, with a large open slope failure from above the gully entering it at its head. Note the unusual split in the gully above the lower road; the debris flow follows the right-hand branch, which is slightly deeper. The slope was logged in 1984: the inactive open slope scar above the upper road, left of and above the active open slope failure, dates from that time as well. The talus slopes at the upper right are supplied by the 50 m. high cliffs of weathered, frost-shattered diorite above them, which are in shadow in the photo and hard to see. the Fraser Valley Regional District, logging has nevertheless been quite extensive (chapter 3.3), and public access is permitted. 3.2 Bedrock and surficial geology, Quaternary history. Chilliwack Valley is primarily underlain by rocks of the Chilliwack Terrane (Fig 3.4.1), consisting of the Chilliwack Group and Cultus Formation. Chilliwack Group rocks are c.300 m.y. old ancient marine floor sediments and volcanic rocks, metamorphosed to varying degrees. Cultus Mgd: Mt. Barr batholith- granodiorite Ogd: Chiliwack batholith-granodiorite eTgd: early Tertiary granodiorite Es: sandstone, conglomerate, argillite KTc: Custer Gneiss Ms: schist, amphibolite, greenstone grade sandstone, pelite & broken fin. - meta-morphosed in Cretaceous. MSL: Slollicum Schist- meta-volcanics, phyllite and metaconglomerate Jk: Kent Fm.- conglomerate, sandstone,argillite Extension of SHUKSAN TJc: Cultus Fm.- argillite, sandstone, minor carbonate- some metamorphism DPc: Chilliwack Group- undifferentiated pelite, metasansdtone & metaconglomerate,mafic & felsic metavolcanics, phyllite -all metamorphic but grade varies (Pc: Permian & Pennsylanian carbonates) PMc: Cogbum Schist- metachert, pelite, amphibolite, marble; PMu, ultramafics PPy: Yellow Aster complex- ultramafic rocks and gabbro. Figure 3.4.1. Geologic map of the Chilliwack Valley and vicinity at 1:250 000 scale. From Monger, 1989. 36 Formation is more recent marine sediments (sandstone, mudstone and limestone). These sediments underly, via thrust-fault contact, the Shuksan Suite, c.175 m.y. old greenschists and phyllites. The eastern portion of the area (Center Creek basin, in this study) is underlain by the late Tertiary (30 m.y. old) Chilliwack Batholith, a much younger plutonic body consisting of mainly granodiorite. Overall, there is quite a bit of geological variation in the area, with much surface heterogeneity except for the area within the Chilliwack Batholith (Monger, 1970, 1989). By contrast, the Norrish Creek area (Figure 3.4.2) is homogeneously underlain by bedrock which is part of the Coast Plutonic Complex (Roddick, 1965), primarily Cretaceous granodiorite, quartz diorite, and diorite, and displays little geological variation except for occasional mafic dikes. Thus, there is a strong geological contrast between the two regions of the study area, with mostly non-plutonic rocks in the Chilliwack drainage contrasted with plutonic bedrock in the Norrish area. Quaternary history and surficial geology is likewise contrasted between the two areas, with differences which are perhaps not as great as the difference in bedrock geology. Norrish Creek area was completely overridden by the Cordilleran ice sheet during Fraser Glaciation (Davis & Mathews,1944; Clague, 1989; Ryder et al., 1991), and it was this ice which produced the rounded ridges and minor summits and generally smoothed the topography of the area. These same glaciers pushed up against the North Cascade Range, but only overtopped it briefly and locally, and erosive activity was not great (Waitt and Thorson, 1983; Waitt, 1977); maximum ice penetration occurred up the major valleys, leaving moraines at its highest extent; these moraines now impound Chilliwack Lake (Saunders, 1987). Alpine glaciers also descended from their cirques during the onset of Fraser Glaciation, but retreated before the advance of the Cordilleran ice sheet (Mackin, 1944; Crandell, 1963; Waitt, 1977, 1975). Evidence of the Cascade alpine glaciation remains in the form of well-developed cirque and horn topography (Holland, 1964); 37 W A S H I N G T O N Figure 3.4.2: Bedrock geological map of the Norrish Creek area at 1: 253 440 scale (1 inch to 4 miles). From GSC Map 1151A (Roddick, 1965). Pink and brown colours represent plutonic rocks of various compositions [h, H = hornblende dominant (h)/only (H) mafic mineral, b, B= biotite (same convention), 2=granodiorite, 3=quartz diorite, 4=gabbro]. Green colours represent sedimentary and metavolcanic rocks of various types, increasing in age with decreasing number (l=oldest). Yellow colours represent Quaternary sedimentary cover. Unit 4, near Nicomen Mtn, although it outcrops in Norrish Creek drainage, did not occur in the locations which were studied. 38 where rock was mechanically weak or well-jointed, it was extensively quarried, producing the lower elevations and more subdued topography of the summits composed of weaker rocks; where rock was too massive to be easily quarried by glaciers, the ice merely abraded and polished, producing the smooth valley walls and slabs of the Slesse-Nesakwatch and Nesakwatch-Center divides (Fig. 3.2.4). This contrasting glacial history accounts for the contrasting surficial geology of the area. In the Norrish Creek region, where glaciation was most extensive, the majority of hillslopes are till-mantled. In Chilliwack basin, glaciation was less intense and till makes up less of the surficial cover; correspondingly, the Chilliwack area has a greater amount of colluvium, primarily derived from post-glacial weathering and mass wasting. The Chilliwack Valley also has a greater proportion of surface area with bedrock exposed. In both the Chilliwack and Norrish valleys, main valley floors and portions of subsidiary valley bottoms have recent fluvial sediments and older glaciofluvial and glacio-lacustrine materials, either buried under current fluvial sediments or exposed in terraces. These latter materials, although often gullied, and an important source of slope failures due to their low stability, were not included in this study for two reasons. First, they occupy only a small portion of the total study area; secondly, they are typically found only in valley bottoms, rather than on the hillslopes which are the focus of contemporary timber harvesting. 3.3 History of logging activity. Logging in the Chilliwack and Norrish valleys has a long history, and consequently little old-growth is left in either drainage. The main Chilliwack valley was logged during the first period of intensive logging, between 1900 and 1939. After the Second World War, most major valley bottoms had been logged,and interest turned to the bottoms of tributary valleys. In the 1950's and 1960's, tributary valley bottoms, such as Nesakwatch Creek, Slesse Creek, and Foley 39 Figure 3.5: Gullies on the left bank of an unnamed tributary of Tamihi Creek, opposite site TAM-004-LB. The slope was logged in the mid-1960's and slope failures appear to have been a chronic ocurrence since that time. Similar slopes higher up the tributary creek which were not logged do not display a similar degree of gullying. Surficial material is thick basal till, and bedrock is not exposed. Creek were logged. The main valley of Norrish Creek was also logged at this time. As the tributary valleys were depleted of timber, interest turned to the valley sides, and logging entered its third phase (1970's to present ). In this latest phase, timber harvesting has proceeded higher and higher up tributary creeks, and up main valley sideslopes: the trend is towards harvesting of steeper slopes which were previously ignored due to inaccessibility. It is in this third phase that interest in slope stability has grown, as the effects of slope failures on forest roads and slopes, and of increased sedimentation of valley floor streams, have become noticed. In particular, areas of Center Creek, Rose Creek (Figure 3.3.2), and Tamihi Creek (Fig 3.5) were logged with little 40 regard for slope stability in either harvesting methods or road construction, and the result is chronic slope instability which chokes the main valley streams with coarse sediment and has permanently removed large tracts of forest land from production. The situation has been further complicated in Norrish Creek by the construction of power lines and water mains, neither of which were completed with any regard for erosion or slope stability, but only to maintain the long-term stability of the structures. These non-logging developments have had an impact on Norrish Creek basin with magnitude comparable to, if not greater than, that of the impact of logging. 3.4 Precipitation record and extrapolation to study area. A number of precipitation gauges exist within or near the study area, although none are located at ideally representative locations with respect to most landslide initiation zones. In Chilliwack Valley, the most representative location is at the Chilliwack River hatchery, located just upstream of the confluence of Slesse Creek and Chilliwack River (Fig 3.2.1). At Norrish Creek, there is no station located within the basin; the closest stations are at Mission and Aggasiz (Fig 3.1). Mission is closer, although Aggasiz was considered by Howes (1987) to be more representative of Norrish Creek. The study area is characterized by strong local variation in precipitation intensity due to complex air movement patterns caused by the mountainous topography (Thurber Engineering, 1988). For instance, a storm in January 1984 exceeded a 100 year intensity at Aggasiz, but was of only moderate intensity at many other locations. A storm in July 1983 caused severe flooding and several debris flows in an area centered around Wahleach (Jones) Lake (Evans and Lister, 1984), but produced only moderate runoff elsewhere. While the above comments apply to summer convective storms, winter occluded fronts affect the region for weeks at a time. During these storms, precipitation is predictably high throughout the study area, and variability is less 41 from place to place. Overall, the regional climate is wet and relatively mild. Total annual rainfall is around 2500 mm, being slightly less in the Chilliwack Valley, especially downvalley, and slightly higher in the Norrish Creek area. One of the largest storms within the period of interest, that of Nov. 8-10 1990, had a total precipitation over 72 hours of 200 +/-10 mm at Chilliwack River Hatchery, Chilliwack townsite, Mission and Agassiz. This is almost one tenth of mean annual precipitation in only three days, and over half of that fell on one day (the 9th). This storm exceeded the 100 year return interval, and triggered many debris flows. A pair of similarly large storms occurred in the area in November and December of 1995. Most of the recent slope failures observed in the study are attributed to these three events. 3.5 History of slope instability in region. Both the Chilliwack and Norrish valleys are known for their unstable slopes which have produced spectacular failures in the past. Some of the factors which contribute to this instability are the steep topography, the strength of the surficial and bedrock geology (in particular the large quantity of unstable materials), the history of logging, and the high precipitation, as detailed above. Past slope failures in the Chilliwack area and immediate vicinity have occurred at all scales, ranging from the huge rock avalanche from Cheam Peak c. 11 000 B.P. (Naumann, 1990) down to minor slumps and ravels which involve only a few cubic meters of material. Previous hillslope geomorphic studies in the area have tended to focus on either very large failures (located mostly adjacent to but outside the current study area), both of rock (Naumann and Savigny, 1992; Clague and Shilts, 1993) and debris (Evans and Lister, 1984; Slaymaker et al., 1987; Jakob, 1996), or on smaller mass movements from the glaciofluvial and glaciolacustrine valley floor sediments (Thurber Eng., 1988). The latter have produced the greatest recent historical impact on the community of Slesse Park and on the fishery in Chilliwack River. In 42 addition, several studies have performed slope stability analyses for the region. Jordan (1987) mapped slope stability in lower Airplane Creek in a report for Herman Sawmills; Millard (1994) completed a Terrain Attribute Study for the Chilliwack valley. These studies highlighted gullies as the most unstable features in the forested landscape, and debris flows from those gullies as a primary hazard. Large storms in 1983, 1990 and 1995 are known to have produced many debris flows in the study region. In Norrish Creek, a long history of mass movement events also exists, from the prehistoric rockslide which dammed Dickson Lake (Evans, 1985) up to the present day. Slope failures in the period 1950-1981 were studied by Howes (1987) in an early Terrain Attribute Study. He showed that gullied slopes were one of the most likely terrain types to produce failures, that the rate of failure was higher on clearcut than on unlogged slopes, and that clearcut slopes showed an increase in rate of failure during periods of wetter climate, while natural slopes did not. 3.6 Impacts of mass movement events. Both Chilliwack River and Norrish Creek are important fisheries streams; in fact, the sport fishery on Chilliwack River is one of the most heavily fished in the province, and the most valuable (Whyte and Schubert, 1990). There is a fish hatchery at the confluence of Chilliwack River and Slesse Creek, and one at the mouth of Norrish Creek. Debris flows can damage a fishery in several ways (Tripp and Poulin 1986a, 1986b, 1992): increased sedimentation buries spawning gravels and kills juvenile salmonids, and has the overall effect of making streams shallower and wider (Hogan, 1986), which increases water temperature and exposure of juvenile salmonids to predators. Both of these effects cause decreases in fry survival rate. Sedimentation in Norrish Creek has proceeded to such an extent that much of the channel consists only of exposed boulders, and the water intake for the hatchery is perpetually clogged or buried (C. Rudiak, pers. comm.). 43 In addition to impacts on fisheries, debris flows in Chilliwack and Norrish valleys also pose danger to humans. Chilliwack Valley is one of the most popular sites for the Lower Mainland weekend recreationist to "get away from it all," and every weekend, hundreds of people visit the valley, rain or shine. Official campgrounds are full every weekend, and overflow campers have been observed encamped on debris flow fans during periods of heavy rain. Norrish Creek, though not as well known or popular as Chilliwack Valley, is nonetheless often visited by hunters and off-road vehicle enthusiasts. Active logging continues in both valleys, and loggers are also often exposed to potential debris flow hazard, both during working hours and also at times when in camp. Although much of the old-growth forest has been harvested in the two study areas, forestry is likely to continue unabated in the future. The staggered history of logging means that the first areas cut are now reaching maturity with second-growth timber; indeed, second-growth will soon become the most important source of timber in the region, an anomaly when compared to the rest of the province. Debris flows destroy vegetation in their path, and the scars tend to remain unvegetated, or covered with successional alder, for long periods of time (Howes, 1987). Thus, debris flows in the study area have had a detrimental impact on the economics of forestry. Debris flows have also destroyed logging roads and bridges, increasing the cost of maintaining such structures in cases where they were not decommissioned after logging. 3.7 Observed regional variation in morphology of gullies and style of debris flow initiation. During fieldwork, the subjective observation was made that the gullies studied in the Chilliwack Valley area differed in morphology from those in the Norrish Creek area, and that debris flow initiation in the two areas occurred in differing manners. Gullies in the Chilliwack region were underlain by mainly non-plutonic rocks. In this setting, the typical gully tended to be the surface expression of a bedrock irregularity, such as the intersection of two joint sets. Over these irregularities were draped the surficial materials of the region (Figures 3.6.1, 3.6.3-3.6.4). Such gullies tended to be long and deep, with a distinct headwall at the upper limit of the gully differentiating it from the slopes outside the gully. They most resembled the gully type shown in Figure 1.1 of Chapter 1. Slope failures in the gully tended to be sidewall failures of the surficial materials draped thereon. Fig. 3.6.1: Cross section of a typical Chilliwack gully showing surface expression of bedrock features and surficial material covering. Gully location corresponds to topographic low in bedrock. Fig. 3.6.2: Cross section of typical Norrish Creek gully showing bedrock and surficial material cover. Gully is developed in surficial materials with little, if any, influence from bedrock on form or location. Figure 3.6.3 Figure 3.6.4 Figure 3.6.3: A typical Chilliwack gully, in Foley Creek (FO-005-LB). Gully channel has exposed bedrock, while sidewalls are of glacial till. Bedrock controls channel gradient (note waterfall in upper reach). Sidewall failure on left gully wall dates from 1995: failed material was diverted into channel at low angle of entry (a = 15°) and deposited on bedrock steps in channel; no debris flow reached the road. D for the reach was about 45°. ICSS was low (c. 0.1 m3/m). Figure 3.6.4: Another typical Chilliwack gully, in Borden Creek (reach BOR-002-RB-02). Surficial material is thick glacial till, dissected by parallel gullies. There are two small debris slides dating from 1995 on the left gully wall, and one partially revegetated one on the right dating from 1990 (areas are c. 8m long by 4m wide, but depth is only c. 0.5m- volumes are -25-30 m3). None of the failures triggered a debris flow. Measured ICSS in the gully was approx. 4.5 m3/m. No bedrock was exposed in walls or gully floor. Figure 3.6.5 Figure 3.6.6 Figure 3.6.5: A typical Norrish Creek gully (ENR-003-LB). A till-mantled slope with a failure at the gully head (in this case, with a distinct headscarp). The failure resulted in a debris flow which crossed the lower road and ran out on lower-angled slopes below without reaching East Norrish Creek. The distance from the headscarp to the lower road is c. 125 m. Note the two tracks of alder to the left of the failure: the right-hand one occurs below a small failure scar halfway between the roads, while the left-hand one begins from old windthrow in an unlogged tree patch immediately adjacent to the upper road. Figure 3.6.6: Two typical Norrish Creek gullies, in Hanson Creek (NOR-006 and NOR-007-LB). Gully NOR-006-LB is on the right and joins NOR-007-LB at the lower road crossing. There is a small sidewall failure in NOR-006 which is only partly visible in the photo c. 15m above the upper road which did not produce a debris flow. A large failure in the open-slope depression at the head of gully NOR-007 initiated a debris flow which incorporated material from both the roads it crossed, entered Hanson Creek and continued downstream to Norrish Creek. The surficial material is a weathered, colluviated till: the slope is dissected by parallel gullies (NOR-008-LB is just out of the photo to the left). The distance from the failure scar to Hanson Creek is c. 250m. Note the change in date of logging immediately left of gully 007. 47 By contrast, the gullies of the Norrish Creek area were developed in areas underlain by relatively massive granodiorite, and tended not to be the expression of a bedrock feature (Figure 3.6.2., 3.6.5-3.6.6). These gullies are conjectured to have developed in places where fluvial processes (such as rilling) or a previous slope failure had allowed erosion to occur at a more rapid rate than on the surrounding slopes. Such gullies appeared to cut down through the surficial materials they had developed on until they reached bedrock, at which point they would begin to widen; their heads in some cases actively retrogressing. The gully head was not a distinct headwall as often as it was merely an open slope depression, or hollow, transitory between the open slope above and the gully below. Thus, this gully type is most like the gully shown in Figure 1.2 of Chapter 1. Failures in this open-slope depression area at the head of the gully were more commonly the source of debris flow in the gully system than were sidewall failures in the gully. In Norrish Creek, open slope failures entering the gully either at its head, or at some point on the sidewalls, were an important secondary source of material triggering debris flows, whereas this was relatively rare in the Chilliwack Valley area. When field data were analyzed, an attempt was made to quantify this subjectively observed regional variation in style and pattern of debris flow initiation and the degree to which it existed. Once its presence had been objectively verified, the causes of the regional variation in debris flow initiation were sought. A fuller description of this process, and the conclusions reached thereby, are given in Chapter 8. 3.8 Conclusion. Norrish Creek and Chilliwack Valley are drainages with high relief, steep valley walls, and high precipitation. In both areas, bedrock and surficial lithologies are composed of materials which are partly or wholly of low stability. Consequently, both areas have a history of slope instability at all scales. Slope instabilities have been exacerbated by timber harvesting which was mostly conducted without regard for slope stability. The most commonly occurring recent slope failures in both areas have been debris flows. Both study areas are representative of the geographic regions in which they occur, and both are representative of the forested coastal perhumid belt of southwestern British Columbia. Dates of logging and of extreme precipitation events have overlapped in such a fashion that heavy precipitation occurred during the period of maximum post- logging slope strength loss. Thus, not only are the study areas suitable for testing the GAP on the basis of regional representativeness, but they also meet criteria of suitability with regards to history of timber harvesting and climatic forcing. 49 Chapter 4- Experimental Design and Measurement. 4.1 Experimental design and choice of measured parameters. As stated in Chapter 1.4, the objectives of the study were threefold. The first objective was to evaluate the effectiveness of the DFIP section of the GAP in the area of the North Cascades and southern Coast Mountains. The second objective was to further study the factors affecting debris flow initiation in logged gullies, and to evaluate these factors for inclusion into a revised DFLP assessment, if such inclusion would be feasible. The third objective was to test the data for regional variability, by comparing the results from the Cascades to those from the Coast Mountains, and to forward the results of the study to MoF to be included in a coast-wide study of regional variation in DFLP evaluation effectiveness. At the outset of the study, consultation with geoscientists specializing in terrain stability assessment suggested that the Chilliwack Valley, Silverhope/Skagit Valley, and Norrish Creek watersheds might be appropriate sites for the study, as they have known debris flow gullies within them, logging during suitable times, and a history of extreme precipitation events during the times of maximum post-logging slope weakness. To achieve the greatest confidence in experimental results, an ideal study design would have been to select gullies before they were logged and perform gully assessments in them at that time, then wait fifteen years and re-evaluate them to see how actual events had occurred. A comparison of observed with predicted events would evalute the success or failure of the DFLP assessment. Time pressure and monetary constraints made this approach, while attractive, impractical. It has the additional weakness of being dependent on unknown meteorological conditions: if no extreme storms occurred during the fifteen year lag time, the study area might not produce sufficient debris flows, even in gullies assessed as highly hazardous. 50 A decision was made to select gullies which had been logged six to fifteen years previously (between 1981 and 1990). This interval was based on the recommendations of Sidle et al. (1985), as modified for the B.C. coast by Rollerson (1992), and was intended to ensure that the period of post-logging slope weakness had overlapped with the two known extreme forcing events (1990 and 1995). Gullies would be assessed in their present state with the assumption that terrain parameters measured after logging would be substantially unchanged from their pre-logging condition. As these parameters were measured over each gully reach as a whole, the possible effects a failure might have had on the reach parameters would be minimized. The desired sample size was set at a minimum of fifty gullies, each having at least one failure. The decision was made to begin the study in the Chilliwack Valley, for two reasons. Firstly, the Chilliwack area was known to have a heterogeneous bedrock geology, making it a good place to test for the effects of lithological differences. Secondly, as a result of the Terrain Attribute Study completed by Millard (1994), a complete set of 1992 and 1993 air photos was readily available for the Chilliwack area, with some gullies highlighted. This made pre-field planning of site selection easier. Should Chilliwack Valley yield less than fifty gullies, the study area would be expanded as necessary to one of the other basins. Within each gully, the goal was to measure as many different geomorphic and geotechnical parameters as possible, subject to two constraints. Firstly, each parameter must be measurable within an appropriate time frame; spending a disproportionate amount of time on one parameter of undetermined significance might compromise the ability to sample several dozen gullies. Secondly, the chosen parameters should, where possible, be ones which were familiar to the practice of geoscience in British Columbia, so that the results would be easily comparable to those of similar studies. 51 The initial set of parameters was suggested by T. Millard (pers. comm.), and was derived from standard Ministry of Forests gully, terrain and failure assessment cards. These parameters included those known to be significant in similar studies (Howes, 1987; Rollerson, 1984, 1992). The measured parameters were chosen to provide a comprehensive description of the gully (Table 4.1). The GAP is designed to be applied at the scale of the individual gully reach, but the focus of the study was split equally between the reach scale and the failure scale. To evaluate the effectiveness of the DFLP assessment, reach-scale data would be sufficient, but data at the scale of individual failures were necessary for the second part of the study, focussing on factors affecting debris flow initiation. It is at the latter scale, where two failures in the same reach exhibit different properties, that the greatest potential exists to study why some failures cause debris flows while others do not. Parameters were of three types: those homogenous at the scale of the entire gully system (terrain variables); those homogenous at the scale of an individual gully reach (reach variables); and those measured at each failure site within the gully (failure variables). As can be seen, the majority of parameters included were those which are found on the relevant MoF cards (Appendix 2). These cards were designed in an attempt to maximize and standardize the amount, of information gathered by terrain scientists working with a wide variety of previous experience and backgrounds; their use is becoming common in B.C. geoscientific practice. Generally speaking, parameters which are measured at a larger scale are applicable at smaller scales, unless the parameter is specifically remeasured at the smaller scale; thus slope curvature, measured at the whole gully scale as a terrain variable, is expected to apply to each reach within a particular gully. Table 4.1: Variables Measured In Gully Systems 52 For each sully system (terrain variables): Location: watershed, gully code #, bank, number, UTM coordinates Schematic plan of gully system showing location of failures and reach boundaries. Relevant sections of MoF Terrain Card: Terrain Unit (symbols) Slope Position Slope Configuration Slope Curvature Forested Disturbance Indicators Drainage Class Soil type, texture, and average depth Surficial Material(s): type, texture, average depth and struxcture (where existent) Depth to Bedrock (if visible) Bedrock Lithology Bedrock Structure Slope Aspect (of gully system) Elevation Mean, Max., Min. Slope Angle Logging Date For each reach within sully system (reach variables): Relevant sections of MoF Gully Assessment Procedure Field Data Sheet: Gully code # & location Reach number Length of reach Upslope Debris Flow Potential (presence or absence of evidence of previous flows) Channel Width Channel Depth Channel Cross-Sectional Area (W*D) Channel Gradient (degrees and percent slope) Channel WPI (from gradient, XS) Debris jams: presence or absence of woody debris, size of WD in jams, size of stored sediment in wedges Gully Wall Slope Angle(degrees and percent) Side/Headwall Surficial Material Gully Wall Slope Distance GWFP, GGP, DFIP (calculated from table) Amount of in-channel stored sediment (ICSS), estimated from cross-section (assumed triangular, 1/2WH used) arid averaged over reach length Number of failures in reach (minimum 25 square meters surface dimension) Whether or not a failure in reach initiated a debris flow, fate of flow if initated. For each failure in a sully system (failure variables): Relevant sections of a MoF Landslide Profile Data Card: Gully code, failure number For each segment of failure: scour dimensions, fill dimensions, path slope, path azimuth, % revegetated Estimated date of failure Angle of entry of failure into gully system Type of failure Elevation Aspect of failure Location of failure with respect to gully system Land use Present erosion Observed seepage Drainage class Slope gradient: @ origin, of failure plane, of gully at point of entry of failure Pos'n of failure on slope Slope configuration Slope curvature Terrain Unit (symbol) Material exposed in failure plane Height of headscarp (if significant) Dimensions of failure path Routing and fate of failure (i.e. stalled in channel, DF out of reach, or to fan?) Any other written comments, sketches, diagrams necessary to expand or clarify the above information (eg. tree ring count to verify age of regrowth) Sediment description (estimated composition) and collection of ~8 kg. sand and finer sediment where possible, for lab analysis. Representative facsimilies of the Terrain Card, GAP Field Data Sheet, and Landslide Profile Data Card are reproduced in Appendix 2. The variables were given specific codes (abbreviations or acronyms) as names in the analysis stage. A list of the variable codes used , together with the units in which each is measured (if applicable), is included in the Foreword. 53 4.2 Methods of measurement: quantitative, qualitative. As can be seen from the gully assessment cards (Appendix 2), some of the information is general and qualitative in nature, and is recorded by the user simply examining prevailing conditions, deciding which categories best apply, and circling those categories on the data card. In the following description of methods of measurement, such parameters are listed with the following symbol: (q), to indicate that they are qualitative measurements only. Other parameters are discussed below in more detail in order to explain how they were measured and the rationale behind such measurement. I. Terrain variables: Terrain Unit, Slope Position, Slope Configuration, Slope Curvature, Forested Disturbance Indicators, Drainage Class, all (q). Location, a multi-component identity number was assigned to each gully indicating name, number, bank, reach and location. For instance, TAM-001-LB-Ol indicates that the gully is located in the drainage of Tamihi Creek (TAM); it is the first gully measured within the drainage (001); the gully is on the left bank of Tamihi Creek (LB); and the first reach is indicated (01). Reaches are numbered sequentially from highest to lowest in a gully. The second part of the gully identification is a UTM coordinate recorded from a GPS unit to 100 meter accuracy (the coordinate is generally a point on the road closest to the gully, eg. where the gully crosses the road). Soil class and texture: an already present soil exposure was found (in a failure, gully wall, or road cut) and soil class and texture were qualitatively estimated. Only three soil types were identified in the field (podzols, brunisols and folisols). Such identification was based on my previous soil identification experience, and involved a simple observation of soil horizons present in exposures, rather than the use of quantitative methods (such as colour charts). Soil texture was 54 intended to be descriptive, eg. sandy loam, and was based on hand texturing of a sample of the soil in question. Soil average depth was measured to the boundary between the soil and the unweathered material beneath (the B/C horizon), or to bedrock in cases where such a boundary was not present (for instance, the folisols). In cases where a boundary was not exposed, the depth was reported as "deeper than (maximum exposure)." Surficial material type(s), textures and depths were recorded as for soil. An attempt was made to record all surficial materials present in the order of their stratigraphic sequence, if more than one material type was present (usually not the case). Surficial materials were generally either till or colluvium, with the distinction being made based primarily on degree of consolidation and colour. The till most commonly observed was basal till, hence bouldery clasts in a fine matrix, and well consolidated; the majority of observed colluvium was derived from ablation tills and consisted of similar materials, usually poorly consolidated, forming either sheared layers over basal till or thicker deposits at bases of slopes and in colluvial slope hollows. Colluvium was also generally lighter in color than till. A few cases of colluvium of a different nature were observed: these were much coarser than the till-derived colluvium and consisted of angular rock clasts in a coarse sandy matrix, with the clast lithology indicating derivation from adjacent bedrock. Surficial material structure was recorded descriptively, such as "layered", "sheared", "massive", etc. Depth to bedrock was the sum of soil depth and the depths of the various pedogenic and subjacent stratigraphic materials. If bedrock was not exposed, depth to bedrock was recorded as "deeper than (maximum exposure)" and some attempt was made to estimate what that depth might be. Bedrock lithology and structure were described according to previous experience with geologic mapping and rock climbing. If bedrock was not exposed, nearby exposures and geologic maps were consulted. Structure terms were recorded descriptively, but precisely, eg. "highly sheared", "massive", "joint plane 150/20". (The latter numbers refer to strike and dip.) 55 Aspect of a gully system was determined with a compass, standing at the top of the highest reach and facing down the fall line of the open slope outside the gully (not of the uppermost gully reach, if the two differed). Numbers were corrected for magnetic declination (recorded numbers are in true coordinates) and checked against aspect recorded from the GPS unit. Elevation was recorded off the map, and was recorded for the point at which the UTM coordinates were measured. Average, maximum and minimum slope angles were recorded for the slope as a whole. Maximum slope was often gully wall slope, and minimum slope was often the channel gradient. Slopes were measured with a clinometer, and recorded to the nearest degree. Date of logging was initially one of the more difficult assessments to make. Initially, ring counts were made on new growth (a tree was cut and rings counted). This indicated that estimates of regrowth age made by my field assistant (a Forestry major) were correct to within one year (plus or minus). This method was subsequently used, and additionally checked by interviewing loggers working in the area, as well as the staff of the Chilliwack Forest District Office, and Forests Forever information signs located in the clearcuts themselves. These additional confirmations of date allowed dates to be accurately determined for most sites. Sites which were possibly logged before 1981, or after 1990, (that is, sites where estimated age was 1981 or 1990, for which no corroboration of age existed) were excluded from the study. II. Gully variables: Length of reach was measured with a hip-chain (Fig. 4.1). Reach boundaries were taken to be where one of the gully parameters changed significantly (for instance, sidewalls deepened or channel gradient changed). At least one reach of 100 meters length had to meet all GAP criteria for a gully in order for the gully to be included in the study; however, all reaches in the gully were not required to meet the GAP criteria. 56 Figure 4.1: Measuring the length of a gully reach with a hip chain (AP-004-RB-02: fallen tree visible just below waterfall at top of reach marks point of entry of a very large open-slope failure [AP-004-RB-02 F01] into the gully). Note the steep rock gully sidewall at photo right. Upslope debris flow potential was evaluated through assessing whether any debris flows had entered the reach from above the highest measured reach. This was generally not the case, although at some sites it was not clear. Upslope terrain stability classes were estimated in the field, as the terrain map for the area was not available during fieldwork. 57 Channel width and depth were measured from the dimensions of the observed water-washed channel in the gully bed, not on actual observed water flow. Mean peak flow was estimated as the flow necessary to fill the observed water channel to the water-washed limits. Channel cross-sectional area was simply width times depth, as per the GAP Water Power Index assessment (Appendix 2). Channel gradient was measured with a clinometer and a Brunton compass in % slope and degrees. Channel water power index (WPI) was calculated from gradient and cross-section according to GAP Table A (Field Data Card, Appendix 2). Observed debris jams were described qualitatively according to the categories on the cards: (q). Gully wall slope angle (GWSA) and length ('distance', thus GWSD) were measured with clinometer and either measuring tape or hip-chain, whichever worked better. Measurement was made along the fall-line of the gully wall, from the channel bed to the highest point of the sidewall (namely the break of slope with the slope outside the gully), and thus was not necessarily measured orthogonal to the gully channel. The steeper of the two gully walls was chosen if one was steeper than the other, except for a few special cases (in Figure 4.1 above, the sidewall at photo left was measured). Gully wall surficial material was described in this section by a choice of the most appropriate category from GAP Table C (Appendix 1). The category of "failure scars" as seen on the GAP card was not used, since it is intended to be applied under pre-logging conditions. GWFP, GGP and DFIP were calculated from observed parameters according to the appropriate GAP table (Appendix 1). In-channel stored sediment (ICSS) was another parameter that posed difficulty initially. This subject was studied in detail by Millard (1993) and Oden (1994). The most accurate way to determine ICSS is through excavation of a cross-section of the gully, followed by detailed surveying of the reach and integration over the length of the survey. This approach was too time-consuming to be used in this study. Instead, the following simplifying assumptions were used. 58 The shape of the gully channel (the incised channel in bedrock or consolidated sediments) was assumed to be of triangular cross-section, defined by the angles of the gully sidewalls, with a width equal to the width of the exposed gully channel (Fig. 4.2): in-channel debris Figure 4.2 Cross-section of a gully channel showing method used to calculate ICSS W=gully width, H=depth of fill, and X, (p are the slope angles of the gully sidewalls. Knowing W and the two slope angles, the lengths of the other two sides can be found using trigonometry, and with the three sides and angles, H can be estimated. Then cross sectional area of the gully fill is simply ICSS=1/2WH. This is calculated three times per reach and ICSS for the reach is taken to be the average of the three values, expressed as a volume per unit length of the reach, i.e. m3/m. For cases where the gully bed was exposed and ICSS was small (ICSS< -1.0 m3/m), its volume was estimated based on the observed average,dimensions of material in the gully. For example, in a case where the channel bed was exposed bedrock, with a few large stones, fluvially worked deposits of finer clastic sediment, and perhaps a piece or two of logging slash thereon, the volumes of the individual components were estimated and summed to estimate ICSS. In a reach cross section with two boulders 30 cm on a side, a lm by 50 cm by 20 cm bed of smaller stones, and one section of tree-trunk lm long by 40cm in diameter, this gives total volumes in m3 of 0.054 for the boulders (cubic shape assumed, so 2 x (0.3)3), 0.1 for the small stones, and 0.16 for the tree-trunk (assumed to be a rectangular prism for simplicity and conservative estimate, thus volume = 1.0 x 0.4 x 0.4), which sums to an ICSS measurement over a 1 meter unit length of gully channel of 0.314 m3/m, which would be simplified to 0.3 m3/m. Figure 4.3 Figure 4.4 Figure 4.3: gully reach AP-002-RB-03, with ICSS of 0 m3/m. This reach was scoured to bedrock by a debris flow in 1990. Note the young alders growing on the levees flanking the channel, which provide a corroborating estimate of the date of the failure. Figure 4.4: gully reach ROS-002-RB-01. A small gully which produced a minor debris flow, one of the few in this study to initiate in-channel. ICSS was measured at approximately 0.8 m3/m. Figure 4.5 Figure 4.6 Figure 4.5: Gully reach ESP-001-HW 1- 01. A familiar scale object is not immediately apparent (the approximate dimensions of the log in the foreground are length 4 m, diameter 20 cm). ICSS in the gully (above failure F01) is approximately 3.5 m3/m. Figure 4.6: Gully reach TLM-001-LB-01. The gully is extremely large (GWSD = 65 m), with approximately 30% of the sidewalls' surface unvegetated and actively ravelling. There are many large CWD jams and wedges, with cobbles and boulders behind them. ICSS is very high (approximately 15 m3/m). 1 The "HW" code in a gully reach identity code refers to the gully's position in the headwaters of its drainage, where neither left bank (LB) nor right bank (RB) designations are appropriate 61 (Figures 4.3 through 4.6 provide examples). A minimum of three such measurements were calculated per reach and the values averaged to produce the final estimate of sediment volume. ICSS includes fluvially reworked sediment, single large clasts and ravelled material, and all woody debris, including small slash and CWD. However, the measurement of ICSS does not take into account the potential erodibility of the gully bed material itself, in cases in which the gully is not incised to bedrock. In each reach, the number of failures exceeding 25m2 area were noted, as this is the smallest failure size identifiable on conventional air photos (Howes, 1987; Rollerson, 1992). Each failure was assigned a number (eg. TAM-003-RB-01 F02 would be the second failure in gully TAM-003-RB, and would be located in reach #1 of that gully) sequentially from highest to lowest observed failure in a gully system. Numbers increased sequentially down-gully, and did not begin anew at each reach boundary. Whether or not a failure in the reach initiated a debris flow or not was also recorded under reach data. (The results of individual failures were recorded as failure data.) The fate of any initiated flow was also recorded ( for example, did the flow stall in the channel or reach the fan?) III. Failure variables The location of each failure within a reach was noted on the schematic gully plan drawing. Each failure was divided into segments and a schematic plan of the failure was drawn. For each segment, dimensions of scour and fill (length, width and depth), path slope, path azimuth and % revegetation were measured (Figure 4.7). The length, width and depth were measured with a measuring tape. Path slope and azimuth were measured with either a Brunton compass or a clinometer and a compass. The amount of revegetation was estimated to the nearest 5%, with the aid of a handheld percent estimation sheet originally used to depict varying percents 62 Figure 4.7: Measuring revegetation of a failure scar in gully FO-002-RB-01. Revegetation was estimated at 25%. This is the reach just below the cutblock boundary in the gully on photo left in Fig. 1.3. for estimation during microscopic examination, now commonly used for a variety of similar purposes. All vegetation was counted, from grasses up to alders and similarly 'large' trees. The dimensions of each of the failure segments were found by multiplying length times width times depth (conservative since it gives maximum dimensions), and the total volume of the 63 failure mass was calculated from total scour minus total fill. This was recorded as Initial Failure Volume and used as an estimate of the mass of the failure at the time it reached the gully channel. The date offailure was estimated by the following method. Airphotos taken in 1992 and 1993 allowed all failures to be designated as occurring either before or after the photo date. Prior to 1990, there was an interval without large storms dating back to 1984. This showed as a clear distinction between failures from the 1990 event, which were readily obvious on the airphotos, and earlier failures, which were already substantially revegetated with alder. These earlier failures, which generally were not re-activated in the 1990 or 1995 storms, were ignored since their age (12 years old at the time of study) placed them outside the interval of vulnerability for most of the dates of logging; furthermore, they were very difficult to access due to the thick growth of twelve-year old alders. Thus most failures visible on airphotos were typically assigned dates of 1990 for failure date. Sites which were not visible on the air photos were evaluated for age on-site. Of those failures, 1995 failures had failed immediately preceding the field season, and were readily distinguished by their fresh appearance. Other failures were assigned to 1993 or 1994 based on their degree of revegetation. This is not necessarily the best technique, but it was used on only a small number of failures. In cases where a landslide scar had yielded more than one failure, only the most recent event was taken into consideration to avoid data censoring2. Angle of entry of failure into gully system (a) was measured as the acute angle subtended between the gully channel and the failure path. Failures that enter almost in the same direction as the gully channel have low a, failures that enter almost perpendicular have high a, as shown . below (Figure 4.8): 2 Data censoring , as defined by Jakob (1996), refers to the effect of multiple slope failures. If a subsequent failure is larger than the initial one, evidence for the initial failure (in the form of deposits, vegetation trim lines, etc.) is destroyed or concealed by the later failure. In the current study, the effect of data censoring was most important with regards to estimation of size of failure scars. 64 gully a failure path Figure 4.8: measurement of angle of entry of failure into gully system. (See also Figure 6.5.2) Type of failure was (q), and chosen from the possible codes on the failure card (Appendix 2). It was observed that there is a correlation in this study between failure type and failure size: small failures are slides or flows (codes ds or df), while large failures tend to be classed as avalanches or flow-avalanches (codes da, dfa). Elevation and aspect of failure were respectively taken from a TRJM contour map, and measured with a Brunton compass. Location of failure with respect to gully system, land use, present erosion, observed seepage, drainage class, position of failure on slope, slope configuration and slope curvature at failure point were all (q). Their possible codes are shown on the cards in Appendix 2. Slope gradients at the point of origin of the failure, of the failure plane, and at the point of entry of the failure into gully were all measured with clinometer or Brunton compass. Terrain unit is the B.C. Terrain Classification (Howes and Kenk, 1988) symbol for the terrain unit containing the failure scar, and is a shorthand representation of the morphology and stratigraphy of the surficial materials. A direct description of this material was then given in the box "Material exposed in failure plane", which is (q) with a range of choices. Multiple choices were often circled (eg. wt/ut/ur would indicate that weathered and unweathered till and unweathered rock were all present in the failure plane). 65 Dimensions of the failure path refers to length, width (and depth if necessary) of the transport zone or path of the failure (the region between the scar and the gully in which scour equals fill or deposition, i.e. the failure had neither gained nor lost significant mass), if such a zone were present. This was usually the case if the initial failure had begun travelling as a debris flow or avalanche while still on the wall slope, before the gully bed was encountered. Routing and fate offailure: did the failure stall in the channel (without leaving its own reach) or did it become a debris flow? If so, did the flow stall in a lower reach, or did it travel to its fan or a higher order stream? If the flow stalled in a lower reach, why? (For instance, did it encounter a logging road?) Other comments were also made, as necessary, and were generally qualitative. The height of run-up of the initial failure on the opposite gully wall was originally intended to be measured, but was only observed in 2 out of 62 cases; similarily, the intent was to measure superelevation of levees around bends, but most gully reaches were straight and superelevation was not observed. For each failure, an 8 kg. sediment sample was taken from the wall of the failure scar, if it was feasible to do so. The intent was to sample the sand size and smaller fraction, so in cases of bouldery colluvium where there was very little matrix, for instance, no sediment sample was collected. In locations where access was difficult (for instance, Rose Creek and Center Creek), sampling was either reduced to one composite sample for several adjacent failures in homogeneous material, or discarded altogether if the situation warranted. Carrying 72 kg. of sediment out of Rose Creek, for instance, would have been, if not impossible, at least very difficult. 4.3 Measurement error and limits to accuracy and precision. The preceding listing of measured variables and methods of measurement describes many different variables and many different styles of measurement. Fortunately, there are fewer 66 possible types of measurement error associated with the measurements. These measurement errors are dependent to a large degree on the measurement instruments and their methods of operation. Firstly, measurements of slope and direction. Slope angles and directions were measured with a pair of related instruments, the compass and clinometer (the Brunton compass combines the functions of these instruments). Slope angles were measured to the nearest degree. Comparison of measurements between myself and my field assistant indicated that such measurements carry an error of plus or minus one or two degrees (a slope measured at 37 degrees by one might be measured at 35 by the other). Directions were measured with a compass corrected for 1996 magnetic declination to the nearest degree, and were measured to the nearest five degrees. Thus, such measurements are accurate to plus or minus two and a half degrees. Secondly, measurements of distance. Distance was measured variously with hip chains, measuring tapes, and in a few instances (a day when the hip chain broke) with calibrated pacing. Hip chains and measuring tapes have an estimated error of plus or minus ten centimeters (measurements were to nearest tenth of a meter); calibrated pacing has a similar error per pace, but positive and negative errors over many paces tend to cancel, so overall error is probably half a meter over reasonable distances. However, another source of error is introduced by the siting of the start and end points of the measurements, so total error in distance measurements is probably plus or minus half a meter, considering all factors, for distances, and perhaps a fifth of a meter for exposure depths, where definite horizons or transitions make boundaries easier to define. Volume measurements of failures and in-channel stored sediment contain larger errors. Volume was calculated through multiplication of LxWxD; ICSS was calculated from ViWH. Thus, failure volume shape is implicitly considered to be rectangular, and shape of channel fill to be triangular. In fact, failure volume was more accurately a three dimensional half-ellipsoid with 67 L, W and D measured along the three axes. Similarly, the shape of the channel fill was more likely to be a half-ellipse than a triangle in cross-section. This means that the measured volumes are maxima. Actual volume could be less, though by exactly how much is unclear, perhaps by as much as 30% of measured volume. Measurement of ICSS carries another source of error in that no correction was made for the porosity of channel fills. Porosity is typically highest for gullies with large amounts of fresh woody debris (logging slash) therein, possibly as high as 50% (Millard, 1993). As residence time of woody debris in the gully increases, fluvial reworking of ICSS and transport of fines tends to fill the gaps between large CWD and fluvially immobile boulders and cobbles with finer sediment, and to group CWD into jams. Thus old, stable CWD jams with sediment wedges behind them have typically low porosities. The ICSS measurements must be considered with this caveat in mind. Measurements of dates present another problem. Dates of logging were initially determined to within plus or minus one year, based on estimates from tree regrowth, if no more accurate indicator of date was available (for instance, a Forests Forever sign). Later consultation allowed constraint to an exact year for many, but not all, cases. Exact month could not be determined. Thus, in theory, two sites dated 1983 and 1984 could have as few as one (Dec. '83 to Jan. '84) or as many as twenty-three (Jan. '83 to Dec. '84) months separating them. In practice, logging takes place between spring and fall, so the maximum and minimum separation is probably less; however, one "year" of separation could be anywhere from six to eighteen months. Dates of failure were determined by the process described in Chapter 4.2. This process had objective constraints only with regards to air photo dates and discovery of fresh (c. 1995) failure material, so there is more latitude for error in some of the dates. As most failures occur during late fall and winter, the dates are reasonably correlated within the year, however, a 68 conjectured failure date of 1995 could still refer to any time betwen late October 1995 and March 1996. Dates of 1990 and 1995 are most likely to correlate with the extreme storm events of Nov. 1990 and Nov. and Dec. 1995, of course. Subjective, qualitative (q) data has different types of error associated with it. As it is recorded according to the experience and discrimination of the observer, there should not be a lot of variation within classes. My definition of "seepage observed at failure scarp" should not change much between sites (although there is the possibility that such seepage may normally be present but not during the day the site was examined). The question of error here revolves around the underlying basis on which my judgements are made, and whether they are consistent with the judgements others would make. The use of standardized forms reduces, but does not remove, this personal bias. Both my field assistant and I attended a Gully Assessment Course in June 1996, which should have helped to standardize our responses with those of the larger geoscience and forest science communities. Geographic coordinates were recorded from a Global Positioning System unit which gave UTM coordinates to within ten meters. Coordinates were recorded to the nearest hundred meters only, and thus can be accepted with a great deal of confidence at this level. Elevations were estimated from TRIM maps at 1:20 000, and from two portable altimeters. It is worth noting that none of these three measurements ever agreed with each other, nor with the low-confidence elevation measurement of the GPS unit, with less than fifty meters' error. The estimate from the TRIM map, on which horizontal position was fairly well established, is probably the most reliable of the three measurements. Elevation was largely ignored during the data analysis, since it was probably the parameter which was measured with the least reliability, namely low accuracy, low precision. 69 4.4 Summary of Fieldwork and Data Collection. Fieldwork in Chilliwack Valley began in early June 1996 and continued through mid-July. The greatest difficulty encountered was finding debris flow gullies to study. Although a complete set of air photos was available for the study area, it was found to be difficult to distinguish gullies with recent debris flow activity from older, vegetated failure scars and alder-filled gullies. Additionally, debris flows which occured in 1995 were not visible on the 1992-1993 air photos. The process which proved most effective was to highlight areas of probable debris flow occurrence on the map and then visit those sites by road. Access to side drainages in the Chilliwack area is not a simple matter. Roads are not maintained up many of the side drainages except during periods of active logging, and washouts compete with alders to render inactive roads undriveable. The Chilliwack Valley is not logged by a single company or corporation, but by a number of small companies, who have conflicting policies on gating roads. Even the local Forest District Office does not have keys to many of the gates on these roads. Those roads which are driveable typically require four-wheel drive and high clearance. Some roads have vehicle barriers in the form of large ditches or earthworks low down, but are perfectly driveable beyond. Mountain bikes were found to be helpful here. Maps and air photos of the drainages show roads which no longer exist, and new roads appear on a regular basis. The net result of this difficult access was to make finding gullies more difficult than actually surveying them. Several days were wasted exploring road complexes which appeared to lead into areas of high gully density and logging of the required age, but which in fact led into completely different drainages, or into thick alder stands, then vanished completely. However, when gullies were located, it was easy to survey them. Progress made was roughly counterclockwise about the Chilliwack Valley, beginning in Foley Creek and ending five weeks later at Center Creek. 70 After completing the survey of Chilliwack Valley, only twenty-seven suitable gullies having recent debris flows had been located. The study area had to be expanded in order to attain the desired target of fifty gullies. The first areas to be considered were the Silverhope Creek-Skagit River and Wahleach (Jones) Lake drainages. Both areas were rejected as unsuitable after an initial survey. The access situation in Silverhope-Skagit valley makes Chilliwack access seem reasonable by comparison. Furthermore, the Skagit area is much drier than the Chilliwack River drainage (parts of the region have ponderosa pine growth), and this raises concerns about climatic inhomogeneity. Once the Wahleach and Silver-Skagit drainages had been rejected as unsuitable, the Norrish Creek area was chosen. Initial reconaissance of the drainage was promising. A discussion with Chad Rudiak, then a contract employee of Terrastar Resources, was highly rewarding. Terrastar was just completing a Watershed Restoration Program rehabilitation assessment contract on the watershed, and the knowledge Mr. Rudiak had gained of the study area allowed him to provide dates of logging for the basin and a map showing the locations of many gullies which had produced debris flows. Thanks to this information, and a two week spell of good weather, fieldwork in Norrish Creek was completed much more rapidly than was the fieldwork in the Chilliwack area. Twenty-six gullies were mapped in two weeks to complete field data collection. 71 Chapter 5- Analysis of Field Data: Reach Data. 5.1 Methods of Analysis: Statistical and Graphical. The first step in analyzing the gully attribute data was to look at the reach data. Accordingly, the contents of the field notebooks were entered into several spreadsheets and subjected to analysis. Initially, QuattroPro was used to simply display various relationships to examine relations among individual variables. During this stage, obvious relationships were noted, as were general trends. Interpretation was deliberately minimized in this phase. The next stage was to examine some of the relationships more closely, with an eye to examining results with an expected outcome in mind, and to determine if what had been observed matched those expectations or not. This was first done simply, with QuattroPro, and then more stringently using Statistica. The effectiveness of the current DFLP, GWFP and GGP were evaluated during this period. The next stage of analysis involved a re-examination of the data stratified by result. Looking specifically at gully wall failures and debris flow initiation from failures, various parameters were examined to see which were significant determinants of debris flow occurrence. The method of ANOVA (analysis of variance) was used here to test the significance of the observed effects. Then a conceptual basis was sought to explain why certain parameters were significant while others were not; this involved both references to the theory of slope failure, and to the work of previous researchers studying terrain stability and debris flow initiation. The final stage of analysis of the reach data involved a reassessment of the parameters which had been found to be significant in terms of their practical utility in a revised GAP. Various combinations of parameters were put together to form proposed new GWFP and GGP evaluation methods, and the success of each combination was considered to find the set of parameters most 72 effective at predicting gully wall failure and gully channel initiation of debris flow at the reach scale (this last stage of analysis is dealt with in Chapter 7). 5.2 Test of DFLP vs. observed results. The first question considered was how well the assessed DFLP predicted the actual occurrence of debris flows. This question was answered by plotting the percentage of reaches producing debris flows by assigned DFLP class, as shown in Figure 5.1. Fig. 5.1 shows two different values: the percent of gully reaches of each DFLP class that produced a debris flow which travelled out of the reach (first bar, abbreviated 'o/o rch'); and the percent of reaches which produced a debris flow that travelled to the fan or the bottom of the gully (for gullies feeding directly into a higher order stream, without a fan) (second bar, abbreviated 'to fan'). The two values are slightly different, with "to fan" numbers being slightly smaller than the number of out-of-reach debris flows. The difference between the 'out-of-reach' and 'to fan' bars is due to the depositional effects of gully inflection points and of roads crossing the gully. One interpretation of these results gives the 'to fan' results a higher degree of importance, since the fan is the area of highest potential impact damage; the other school of thought notes that any debris flow which flows out of the reach could potentially reach the fan, hence the 'out-of-reach' data is a more accurate representation of debris flow occurrence. Gullies which produced debris flows that travelled only a short distance before stalling, and did not leave the reach, are considered not to have produced a debris flow, since some characteristic of the reach caused the flow to stop again before it had travelled a significant distance. Considering Figure 5.1, it is immediately observed that reaches assessed with DFLP of H or M behave in a similar fashion, and are more active than L reaches. H and M reaches average about a 50% chance of producing debris flows out of reach, and a 40% of producing a debris flow which will reach the fan; L reaches have a 28% chance for both outcomes. This shows that DFLP 73 Proportions by DFIP class M L Reach DFIP class o/o rch l l i to fan Figure 5.1: Proportion of gully reaches producing debris flows out of reach and to fan 74 works fairly well in that the H and M classes are more active than is the L class, as expected; however, class M has higher proportions of flows travelling out of reach and to fan than does class H, which is not as expected. This result should be interpreted with caution, as there are 19 reaches in M class and 48 in H; there is much more confidence in the H data than in the M data. The difference between the performance of the H and M classes represents about two gullies' difference in terms of raw data; thus, the difference is not extremely significant, and in practical terms, the H and M categories are behaving similarly. A point to consider is that all gullies considered in the study had at least one failure in them. This may have had the effect of biasing the GWFP towards higher values (see below). Since GWFP and GGP jointly determine DFLP, the net effect may have been to bias DFLP towards higher risk gullies; hence the data for proportions with debris flows may be biased upwards. This effect would be most noticeable in class L, less so in M, and least of all in class H. Thus, the fact that 28% of class L reaches produced a debris flow does not necessarily indicate that 28% of a large sample of L-rated reaches would produce debris flows; the actual number would probably be lower. The existence of this effect was only realized after fieldwork was complete. To fully understand its magnitude, it would be necessary to know the DFLP ratings for gullies without failures in the study area. Since return to the field area was not feasible, air photos, surficial geology maps and TRIM maps were examined. A random sample of gullies without failures was chosen and the four parameters used to calculate DFIP (GWSA, GWSD, TANCG and SURF) were estimated from the available material; for instance TANCG was found from TRIM elevations of upper and lower ends of the reach in question and the length of the reach (gives horizontal and vertical distances, vert/horiz = tan CG.) This assessment indicated that the distribution of H, M, and L ratings in reaches without failures was broadly similar to that observed 75 in this study. 25 reaches were chosen for this follow-up study, and results are shown in table 5.1 below: Table 5.1: Number and percent of total - gully reaches by DFIP rating # H #M #L % H % M %L R e a c h e s in study 48 19 25 0.52 0.21 0.27 Fol low-up map/photo 10 6 9 0.4 0.24 0.36 survey Although this is not a conclusive answer, since the potential degree of error in estimates of DFLP is large (made from four other estimated parameters each with large degree of error in estimate), it does suggest that the amount of possible bias may not be extremely great, since the estimated DFIP class distribution of reaches without failures is fairly similar to the observed DFIP class distribution of reaches with failures. 5.3 Test of GWFP vs. observed failures. When considering the distribution of debris flows in gullies by DFIP class, one must also consider that DFIP is the result of cross-tabulation of two separate parameters: GWFP and GGP. Thus, the error in DFLP is a combination of the errors found in those two parameters. So, a clearer understanding of the effectiveness of DFLP can be found by understanding the effectiveness of GWFP and GGP taken individually. GWFP evaluates the potential for failure of the gully walls in a reach, using two parameters: gully wall slope angle (GWSA) and gully wall surficial material (SURF). The rationale behind this approach is based on results from Terrain Attribute Studies such as those conducted by Howes (1987), which found that rate of landsliding increased with slope angle, and that there were different critical thresholds for different materials. In the GAP, rock is assumed to be most stable, and is given a low GWFP no matter what angle it holds. This is probably a good approach, since the stability of a rock mass will depend not only on the angle of its surface, but also on the presence or absence, and orientation, of any planes of weakness (joints, faults etc.) 76 within the rock mass and their intersection with the surface. Such structural details would be difficult to meaningfully include in an assessment such as the GAP. Colluvium and till are treated similarly, but till is assessed at a lower level of stability than colluvium, for two reasons: the observed frequency of slump-type, cohesive failures in till, which can occur at relatively low angles, and the observed presence of steep slopes in certain types of colluvial materials, indicating stability at higher slope angles. Examining the distribution of failures in reaches by DFIP shows that DFLP works fairly well in predicting gully wall failure (H: 73% of reach walls have failures; M: 63%; L: 48%). This indicates that GWFP may be working effectively. To further understand the effectiveness of GWFP, compare the distribution of failures in reaches by GWFP class to that of the failure distribution by DFLP class (Figure 5.2). As can be seen, the percentages are similar in GWFP and DFLP classes H and L (70% vs. 73% for H, 50% vs. 48% for L) but GWFP class M has a much smaller percentage of failures than does the DFLP M class (33% vs. 63%). This shows the effect of GGP on DFLP: many gullies with high GWFP have a GGP rating of M or L, and are downrated in the DFLP for that reason. Table 5.2 shows raw numbers and percents of total numbers for GWFP Another way to examine the effectiveness of GWFP is to determine the percentage of the data set occupied by each class, and the percentage of failures which occur in each class. Generally speaking, GWFP class H is above average in terms of numbers of failures (last column), while classes M and L are below average. Table 5.2- Number of total reaches & rch. w/ fai lures by G W F P C l a s s # reaches % total rch. # fai lures % tot. fl. % of rch w/ fail H 73 79.4 51 86.5 70 M 9 9.8 3 5.1 33 L 10 10.9 5 8.5 50 Total 92 100 59 100 64 77 Proportion of reaches with failures by DFIP and GWFP rating 0.8 Rating • % DFIP w/fail • % G W F P w/fail Figure 5.2: Proportion of gully reaches rated by DFIP and G W F P class with slope failures therein 78 Table 5.2 also shows that the H class contributes more to the total amount of failures than it does to the total amount of reaches (it is overrepresented), while M and L are both underrepresented. This is another indication that GWFP is successfully flagging the hazardous gullies by assigning them to class H. One effect which is not as it should be is the '% of reaches with failures' column. One in three M reaches have failures, but the figure is one-half for reaches rated L. Furthermore, if the 8 gullies which have R sidewalls are removed from the analysis (because they are all rated L automatically), then there are only 2 reaches left rated L, and both of those reaches have slope failures in them. This is either an example of a weakness in the classification system, or of the previously mentioned bias in sampling caused by only selecting gullies with failures in them. However, a sample size of two is insufficient to permit any conclusions to be made. To fully understand the effectiveness of GWFP in predicting gully wall failure, the relative contributions of its constituent elements, SURF and GWSA, must be examined. GWSA is quantitative and SURF is qualitative, so the two variables cannot be easily plotted against one another to examine the distributions of reaches which fail compared to those which do not; the categorical nature of SURF means that an attempt to do so produces a graph which is a set of overlapping bars. One way in which this difficulty can be avoided is by combining the variable SURF with the variable PASTFAIL to create a hybrid variable, SURFFL. In this fashion, when SURFFL is plotted against GWSA, the differing distributions of gully wall slope angles for reaches with and without failures may be clearly seen (Figures 5.3.1 and 5.3.2). In the DFLP assessment process (Appendix 2), GWFP is calculated using sidewall slope measured as percent slope (tan GWSA); however, in Figures 5.3.1 and 5.3.2, and in other figures which follow, GWSA is plotted, rather than TANGWSA. This was done because the use of GWSA was found to allow better discrimination between individual points at lower gully wall slope angles Figure 5.3.1: Frequency Scatterplot of SURF_FL vs. GWSA RY RN CY l or u CN MY MN 20 Q»O oO o o o Q o ° o oQo > O OO O ° O ° © s o o o o 30 40 50 60 GWSA (degrees) 70 80 1 case ° 2 cases o 3 cases o 4 cases O 5 cases Figure 5.3.1: Frequency scatterplot of S U R F F L vs. GWSA showing how the distribution of gully wall slope angles varies by surficial material type and failure result. F i g u r e 5 . 3 . 2 : H i s t o g r a m o f G W S A c a t e g o r i z e d b y S U R F _ F L 16 20 25 30 35 40 45 50 55 60 65 70 75 80 20 25 30 35 40 45 50 55 60 65 70 75 80 20 25 30 35 40 45 50 55 60 65 70 75 80 SURF_FL: SURF_FL: SURF_FL: CY RN RY G W S A ( d e g r e e s ) Figure 5.3.2: Histogram of GWSA categorized by SURF_FL showing another view of how the distribution of gully wall slope angles varies by surficial material type and failure result. 80 (especially between 20° and 35°, where many of the data points in this study lie) than was achieved with TANGWSA. Examining Figures 5.3.1 and 5.3.2, it can be seen that GWFP is functioning approximately as it is supposed to, in that the H class contains the majority of the failures. The distributions of till and colluvium reaches with failures (MY and CY) are strongly skewed towards the H class; for reaches with walls composed of the same materials, but without failures, the skew is not as great, and the median classification for MN and CN reaches is on the boundary between the M and H classes. In terms of the performance of the classes, more than 50% of the reaches in the H class have failures, about 50% of the reaches in the M class have failures, and less than 50% of the L class reaches have failures; however, as mentioned previously, the L class is dominated by R reaches, and the overall sample size is small, so the true performance of the L class cannot be accurately determined. Statistically, testing GWFP vs. PASTFAIL with the method of ANOVA gives a result which is just non-significant at a=0.05 (F (2,89)= 2.90, p<0.0605). This insignificance is probably the result of the small sample sizes of the M and L groups. When the M and L classes are combined into a composite class (ML), then ANOVA indicates that the difference between mean PASTFAIL for the H and ML groups is significant at oc=0.05 [F (1,90)= 5.23, p<0.025; mean PASTFAIL for the H class is 0.7 and for ML class is 0.42]. This confirms the speculation about the effects of small sample sizes given above. Testing the effect of GWSA vs. the composite variable SURFFL (PASTFAIL stratified by SURF) with ANOVA (Figure 5.4.1) produces a result which is significant at a=0.05. . Examination of this figure shows that the surficial materials have significantly different mean angular thresholds for failure, and that gully wall angle significantly affects whether a gully wall 81 Figure 5.4.1: ANOVA plot of means, F-statistic and p-value for test of significance of GWSA vs. SURF_FL. The difference in mean gully wall slope angles between gully walls with (Y) and without (N) failures is most significant for till (M) gully reaches. will fail only for till walls. The difference in mean GWSA for CY/CN and RY/RN gullies is less than one degree. Examining the distribution of GWSA by SURF-FL in a boxplot (Figure 5.4.2) confirms that the effects of wall slope angle are most significant for till and of lesser significance for colluvium. The effects of gully wall slope angle on failure initiation in rock-walled reaches cannot be accurately determined due to the small sample size. This indicates that GWFP best assesses the risk of wall failure for till, and is not as effective for assessing the risk of failure for sidewalls of rock or colluvium. Due to the large proportion of till reaches in the total number of reaches in the study, the lesser effectiveness of the GWFP assessment with regards to evaluating colluvial or rock walls does not show up in the ANOVA result for GWFP when it is tested against PASTFAIL for all the reaches in the study. 60 56 52 « 48 <u <u D ) < m 5 40 O 36 32 28 Figure 5.4.2: Distribution of GWSA by SURF_FL (Mean/ MN MY CN CY SURF FL RN RY 82 ±Std. Dev. • ±Std. Err. Q Mean Figure 5.4.2: Boxplot showing variation in mean GWSA by SURFFL. Difference in means is most significant for MN and MY reaches (standard error boxes do not overlap). 5.4 Test of GGP vs. observed mobilization of failures as debris flows. GGP assesses whether or not a failure reaching the gully bed will become a debris flow or not. The two parameters which are used to evaluate GGP are the length of the gully sidewalls (GWSD) and the channel gradient (in percent slope, i.e. TAN CG). The reason for the inclusion of channel gradient is relatively obvious, and was discussed in Chapter 2; the inclusion of GWSD is less obvious. The rationale behind including GWSD is that longer gully walls mean a failure has a longer travel distance before it impacts the gully channel; hence a failure initiating higher up a gully wall will likely be larger and have greater kinetic energy and inertia when reaching the channel. Due to their higher inertia, these larger failures are more likely to continue to travel down-channel (as a debris flow) than are small failures; due to their greater kinetic energy, they are also more likely to trigger a debris flow through impulsive loading of the in-channel sediment (Bovis &Dagg 1992). 83 When using the reach data set to examine the effectiveness of GGP, it is essential to exclude those reaches which did not contain failures: if there is no failure, then nothing reaches the channel, and GGP cannot be meaningfully evaluated. Out of the 92 reaches, 59 had failures; thus, the number of cases in the evaluation of GGP is 59. It was for this reason that the initial stipulation that all gullies must contain at least one failure was made. Since it was shown in section 5.3 that GWFP works fairly well, much of the error in DFLP must be the contribution of GGP. The effectiveness of GGP in predicting debris flow initiation from gully wall failure is shown in Figures 5.5.1a and lb. Figure 5.5.1a shows the results of an ANOVA significance test using GGP to predict debris flow out of reach; Figure 5.5.1b shows the results of a similar test examining debris flow to fan. The two figures (5.5.1a and b) show that GGP does a statistically poor job of predicting debris flow initiation. The H , M , and L classes all perform similarily, with the chance of debris flow initiation being 50% or greater for all three classes. Additionally, the highest percentage of Figure 5.5.1a: Plot of Means GGP Main Effect F(2,56)=1.21; p<3053 0.85 I 1 1 1 1 i» 0.45 ' 1 i i-H M L GGP Figure 5.5.1a: Result of ANOVA test of effectiveness of GGP in predicting initiation of debris flow travelling at least out of the reach in which it initiates; only reaches containing failures are considered. 84 Figure 5.5.1 b: Plot of Means . GGP Main Effect F(2,56)=.20; p<8208 s 0.66 r , , c o J 3 ro 'fe 0.48 [ 1 1 L > H M L GGP Figure 5.5.1b: Result of ANOVA test as for Figure 5.5. la, but debris flow to fan is considered instead of debris flow out of reach. debris flows initiate in the class M reaches. It makes little difference whether flow out of the reach, or to the fan, is considered; neither result is statistically significant at a = 0.05. Table 5.4 gives the raw numbers: Table 5.4- Distribution of reaches with debris flow by GGP: G G P No. No. w /DF No. w/ % total %W7 DF %w/ D F % of total % total rch O O R DF to fan reaches O O R to fan D F O O R D F T O F A N H 38 26 21 64.4 68.4 55.3 65 63.6 M 11 9 7 18.6 81.8 63.6 22.5 21.2 L 10 5 5 17.0 50 50 12.5 15.2 Total 59 40 33 100 67.8 55.9 100 100 Overall, the numbers in Table 5.4 indicate that GGP does a fair to poor job of predicting debris flow initiation. Two-thirds of observed failures triggered debris flows, of which more than half reached the fan. The majority of debris flows occurred in H rated reaches, but only because this was the most numerous class. Class M outperformed classes H and L in most other respects; classes H and L behaved similarly to each other. The percentage distributions of reaches with 85 debris flows out of reach and to the fan are approximately the same as the distribution of total reaches: for instance, the H class makes up 64% of the total number of reaches, 65% of the total number of reaches with debris flows o.o.r., and 63% of the total number of reaches with debris flows to fan. This indicates that none of the classes are significantly over- or underperforming (the M class slightly overperforms, and the L class slightly underperforms), which suggests that they are all at about equal risk of debris flow initiation. This is another indication that GGP is not an effective discriminant of risk of debris flow initiation from failure. One concludes that GGP is not performing as it should. As both GWSD and CG are quantitative parameters, it is easy to visually compare their relative effectiveness by plotting them on X and Y axes of a graph. The expected distribution of the points in this parameter space should then plot up as depicted in Figure 5.6. The actual plot of GWSD vs. TANCG is shown in Figure 5.7 with debris flows OOR distinguished from those which travelled to the fan. When Figures 5.6 and 5.7 are compared, two effects become immediately clear: first, the distribution of points on the graph is not what was expected; secondly, a strong control by tan CG is immediately evident with tan CG =0.5 (CG = 26.6°) being an obvious threshold value. For reaches with tan CG > 0.5, most failures trigger debris flow; for tan CG < 0.5, most do not. The percentage triggering by debris flow type (o.o.r. or to fan) is shown in Table 5.5. Table 5.5: Influence of the threshold value tan CG=0.5 on D F initiation Group # reaches # D F O O R # DF to fan % O O R % to fan T a n C G < 0 . 5 17 6 4 35.3 23.5 T a n CG>0 .5 42 34 29 81 69 Statistically, the effect of Tan CG on debris flow initiation is very significant. ANOVA gives p<0.001 for DF OOR, p<0.006 for DF to fan when those two results are tested against TANCG. The effect of the Tan CG = 0.5 break is even stronger: ANOVA gives p<0.0004 for DF 86 high Tan CG l o w expected L zone : most failures do not trigger debris flow expected M zone in which -50% fail will trigger df expected H zone in which most failures will trigger a debris flow Jlow GWSD high F i g u r e 5.6: Diagram of TANCG-GWSD parameter space showing how the DFIP assessment expects GGP to perform. The diagram is qualitative in nature, and the zone boundaries are intended to be guidelines only, and not to indicate the actual sizes of the respective zones. (Figure 5.6 is essentially a simplified representation of Table D in Appendix 1 with the axes transposed for clarity's sake.) Compare this figure to the observed results (Figure 5.7.) F i g u r e 5 . 7 : S c a t t e r p l o t o f T A N C G v s . G W S D s h o w i n g r e s u l t o f f a i l u r e 0 . 8 0 . 7 0 . 6 0 . 5 O O z < H 0 . 4 0 . 3 0 . 2 0 . 1 m • • o • • i • i j | mm t—i • ! • 1 | j • Oi • om\ o mmM am\ m • c • • i f o ~T ! O ! j i n ! j j o i .H 1 D V • j !° c j Q ! i Q : • • o I : O I \ \ ! O 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 0 G W S D ( m ) o N o d e b r i s f l o w G D . f . o u t o f r e a c h , b u t n o t t o f a n • D e b r i s f l o w t o f a n F i g u r e 5.7: Scatterplot of TANCG vs. GWSD for reaches with failures, showing observed distribution of reach parameters and results of failure. Compare this plot to the expected distribution shown in Fig. 5.6. The TANCG= 0.5 line is highlighted so as to draw attention to its effect. 87 OOR, p<0.001 for DF to fan when those results are tested against the variable CATTANCG (which has only two categories, TANCCK0.5 and TANCG>0.5). Unfortunately, there is no effect of GWSD comparable to that of Tan CG. Examination of Figure 5.7 seems to indicate that there is a trend for low GWSD to be associated with a higher chance of debris flow initiation, but there is also apparently high debris flow initiation in gully reaches with GWSD >40m. Unfortunately, the bulk of the measurements are for GWSD's of 30m and less, making it difficult to statistically determine the effect of such very high GWSD due to the low sample size. Statistically, GWSD is not a significant discriminant for debris flow initiation when tested with ANOVA: p<0.8231 for DF OOR (F statistic is only 0.05), which indicates that the hypothesis that GWSD is different for failures which do and do not initiate debris flows should be rejected. Why does GWSD not influence debris flow initiation from failure? The underlying rationale of the inclusion of GWSD in GGP is that large GWSD will give a failure more momentum when it reaches the channel, making it more likely to become a debris flow than to stall in the channel, and increasing the potential for impulsive loading. However, it is important to remember that not all failures will start from the top of a slope: some will start from the top, some from part way down, and some from low down. The total length of the slope will not affect a failure's momentum as much as will the height from which it starts. If fall height is a random variable, then GWSD is only a potential maximum, and will not have a direct bearing on the actual height at which a failure begins. That appears to be what has been observed here. 5.5 Comments on reach data and DFIP. Having examined the reach data, and the effectiveness of DFLP and its components on determining the chance of debris flow initiation at the reach scale, the following comments can be made to summarize this part of the study: 88 i) The DFLP assessment works moderately well in predicting which reaches will produce debris flows. Reaches with a DFIP rating of H or M behave in a similar manner, and are significantly more likely to produce debris flows than are L-rated reaches. M and L reaches may be skewed towards higher activity levels than actually occur due to the choice of only gullies which had at least one failure within them. ii) The GWFP assessment works fairly well at predicting where wall failure will occur. Both GWSA and SURF are significant in determining whether or not failure will occur. SURF is significant in determining the average angle at which failure will occur, and thus is correctly used to differentiate between materials for assessing GWFP. GWSA is of most significance in predicting whether or not failure will occur on till sidewalls, and seems to be of lesser significance for colluvial and rock gully walls. Although GWFP is successful in indicating the gully walls most likely to fail, it also classifies as high risk gully walls which are not as likely to fail. Due to the small sample sizes for M and L classes, their true performance cannot be determined, but it appears that they behave similarly to each other, and in a distinctly different fashion from the H class. iii) GGP is ineffective in predicting whether or not a failure entering a channel will become a debris flow. Of its two constituent factors, C G is strongly significant, especially so with regard to the threshold value of Tan C G = 0.5. GWSD is not significant in determining whether or not failures will become debris flows, and its inclusion in the GGP assessment is the factor which renders the GGP ineffective. 89 Chapter 6- Analysis of Field Data: Failure Data. 6.1 Methods of Analysis: Statistical and Graphical. The methods used to analyze the failure data were similar to those used to analyze the reach data. In the first stage of analysis, simple statistics were computed and graphs were plotted to examine the basic distributions of the various parameters. In the next stage, those relationships which had proven to hold at the reach scale were re-evaluated at the failure scale to compare the effects of the scale change. The third stage involved statistically testing the observed relationships to see if they conformed to expected outcomes. The next stage of the analysis was a general statistical evaluation of the various failure parameters to see which were most significant in influencing debris flow initiation. These parameters were then divided by two criteria: whether they were qualitative or quantitative, and whether they could be measured before a failure, or only after failure had taken place. Once this division had been undertaken, attempts were made to develop predictive models for the important post-failure parameters using pre-failure parameters, both through use of regression equations for quantitative parameters and through A N O V A and M A N O V A for qualitative parameters which could not be quantified to allow regression. In this phase of the analysis, the interrelation of the variables at the failure scale was also examined. The final phase of the analysis involved a synthesis of the newly verified significance of various parameters with the question of prediction of debris flow initiation (described in Chapter 7). An attempt was made to find the smallest number of variables with the maximum power to predict debris flow initiation. This set of variables was then subjected to conceptual analysis in order to explain its function through the theory of terrain stability and previous research into gullies. 90 6.2 Test of GGP using at-a-point channel characteristics. To examine the differences between the reach data and failure data sets, GGP was re-evaluated for the failure data. As was done for reach data, channel gradient (in % slope) was plotted against gully wall slope length. Note that throughout this portion of the analysis, any resultant debris flow which travelled out of the reach was treated as a postive debris flow result, thus avoiding the effects of lower reaches on the fate of the debris flow; the reasoning here is that any debris flow travelling out of the reach could potentially arrive at the fan. Thus, in this section, all positive (DFRESULT = 'yes') debris flow results are equivalent to the "Debris Flow out of reach" (DF OOR) result of the reach data. The plot of TANCG vs. GWSD for the failure data, using channel characteristics at the point of failure entry, is shown in Figure 6.1 (compare this to Figure 5.7). The general relations are the same in Figure 6.1 as they are in Figure 5.7. However, there are a few more points on Figure 6.1 (62 failures vs. 59 reaches w/ failures) and there may be a slightly stronger correlation of low GWSD with high chance of failure. One prominent difference is the data point at (GWSD 65 m, TANCG=0.505), which is indicated as having produced a debris flow on Figure 5.7 but as not having caused a debris flow on Figure 6.1. This point represents Failure 01 in the gully reach TLM-OOl-LFJ-01. This gully was highly eroded, with at least one third of its sidewalls having failed (Figures 6.2a through 6.2c). A debris flow was produced in the reach, but it was likely not from the best preserved failure, which was the one measured (this failure still had a debris cone below it in the gully bed, evidence that it had not produced a debris flow). Thus, the reach produced a debris flow, but the failure measured in the reach did not. Most of the other failures visible in Figures 6.2a and 2b appear to have failed in 1990 or earlier, and to have reactivated in 1995, making their initial failure volumes and similar parameters difficult to distinguish. Most of the other points of difference between Figures 5.7 and 6.1 are similarly explained. 91 F i g u r e 6 . 1 : P l o t o f T A N C G v s . G W S D f o r f a i l u r e d a t a s h o w i n g r e s u l t o f f a i l u r e 0 . 8 O O 0 . 7 0 . 6 0 . 5 0 . 4 0 . 3 0 . 2 0 . 1 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 G W S D ( m ) b • o /-yr-! /ft P • ! C D j t_>:l_J Q_r • • ; • o _ __! n a • i • T alaa o ; • o • : o < n D O • n i m • C • J o o i • q 1 o b i ° j o 1....Q j i o 0 ! • i o 1 • T A M - 0 0 1 - L B - 0 1 F 0 1 o D F R E S U L T : n o 9 0 1 0 0 ° D F R E S U L T : y e s Figure 6.1: Plot of TANCG vs. GWSD for the failure data set, using channel characteristics at the point of entry of the failure into the channel. Compare this figure to figure 5.7 (the most prominently different point is highlighted and discussed in the text). DFRESULT is equivalent to the 'debris flow out of reach' (DFOOR) result in the reach data. Statistically, for the failure data, ANOVA indicates TANCG is significant with regards to DFRESULT (F(l,60)=5.85, p<0.0186), while GWSD is not significant (F(l,60)=1.09, p<0.30). This indicates that one should accept the hypothesis that channel gradient is different for the failures which do and do not produce debris flows, and reject the hypothesis that gully wall slope distance differs between the two groups. This result agrees with the conclusion reached in Chapter 5. 6.3 Test of failure parameters vs. observed results: significant qualitative parameters. In order to determine which parameters werre or were not significant, a correlation matrix was produced for the failure data parameters using STATISTIC A. Due to its size (61 variables) Figure 6.2a: Photomosaic of the left wall of gully reaches TLM-001-LB-01 and -02 showing extensive sidewall erosion. Failure 02 is visible at the lower right-hand end of the photo (boundaries highlighted). For scale puposes, sidewall height is 62m. Surficial materials consist of relatively thick colluvium (weathered, colluviated till) over basal till. The till-colluvium boundary is visible as a faint inflection point on the gully wall c. 15m above the channel. 93 Figures 6.2b and 6.2c: Figure 6.2b shows the right wall of gully reach TLM-001-LB-01. The upper part of failure 01 is visible in the center of the slope, and is marked. A portion of the fan of the gully is visible at center left. The Canada-US border is the line of trees (a clearcut boundary) visible beyond the fan. Figure 6.2c provides a better view of the fan as seen from open slopes outside the gully. Recent (1995) debris flow deposits are visible, and apparently clammed TLM creek for a brief period of time. this matrix is not easily presentable; the most important column, showing the correlations of parameters with the variable DFRESULT, is presented in Appendix 4. Parameters indicated as significant in the correlation matrix were then tested more stringently to determine their effect on debris flow initiation. Several of the parameters which the correlation matrix indicated were not significant influences on DFRESULT were also tested more stringently to verify that they were insignificant. With regards to DFRESULT (whether or not the failure resulted in a debris flow), significant qualitative parameters were TYPE, LOCN, SEEP2, and LITHOL (LITHOL4 only). CURVE2 was just nonsignificant at the confidence level a = 0.05. The relationship of TYPE to DFRESULT is shown in Figure 6.3.1. In this and succeeding figures, DFRESULT is coded with 'yes' (debris flow resulted from failure) =1.0 and 'no' (no debris flow from failure) = 0.0. Thus the plot of mean DFRESULT by TYPE shows the observed fraction of failures of each TYPE class which resulted in debris flows. The variable TYPE is the observed type of failure entering the channel, as reconstructed from examination of the site after the fact, 'ds' refers to a debris slide, 'rxs' to a rockslide, 'df' to a debris flow (observed failure track has levees, etc.), 'dfa' to a large debris flow or avalanche, and 'da' to a large debris avalanche. The influence of TYPE on DFRESULT may be explained in the following manner: failure types df, dfa, da/rxs and da were all failures which originated on open slopes outside the gully system. These failures were generally larger and faster-moving than the typical gully wall failure; hence were most likely to trigger debris flows. Conversely, the ds class includes all the minor failures, large ravels and similar events, and thus it is the class least likely to cause debris flows. This explanation (TYPE is related to failure volume and location) explains the observed variation well. Statistically, ANOVA indicates TYPE correlates strongly with volume (as 95 F i g u r e 6 . 3 . 1 : P l o t o f M e a n s T Y P E M a i n E f f e c t F ( 5 , 5 6 ) = 2 . 9 3 ; p < 0 2 0 4 1 . 1 r , , '. : , 0 . 4 [ • 0 . 3 1 1 1 ' ' ' ' 1 d s d s / r x s d f d f a d a / r x s d a T Y P E Figure 6.3.1: Result of ANOVA test of influence of TYPE on DFRESULT, showing mean DFRESULT by TYPE, F statistic with degrees of freedom, and p-value. F i g u r e 6 . 3 . 2 a : P l o t o f M e a n s L O C N M a i n E f f e c t F ( 6 , 5 5 ) = 3 . 2 6 ; p < . 0 0 8 2 1 . 1 0 . 3 1 1 : ' 1 '• 1 1 — g h / o s d g s g h g c o s - g / g h o s - g o s d / g s N o . o f c a s e s : g h / o s d ( 1 0 ) , g s ( 3 4 ) , g h ( 8 ) , g c ( 2 ) , o s - g / g h ( 3 ) , o s - g ( 3 ) , o s d / g s ( 2 ) L O C N Figure 6.3.2a: Result of ANOVA test of influence of LOCN on DFRESULT, with plot of mean DFRESULT by LOCN, F statistic, p-value, and number of cases (observations) for each LOCN class. 96 LNVOL, natural log. of initial failure volume in cubic meters: F(5,56)=7.85, p<0.00001) and location (as LOCN, location of failure initiation in gully system: F(5,56)=5.47, p<0.0004), confirming this assumption. LOCN is also directly significant in influencing DFRESULT, as shown in Figure 6.3.2a. In LOCN, the code'gh' refers to gully headwall failures, 'gs' to gully sidewall, 'osd' to open slope depressions, 'gc' to gully channels, and 'os-g' to open-slope failures entering the gully. Thus a code of 'os-g/gs' indicates a failure which began on an open slope outside the gully system and entered the gully via a sidewall. The eight variables of LOCN were consolidated into three in LOCN2: 'head', 'side' or in-'channel'. LOCN2 is of greater significance than is LOCN (Figure 6.3.2b). Note that only two failures initiate in-channel. The variable SEEP records where seepage appeared on the failure scar: on the headscarp (hs), on the failure plane (fp), on the ground surface (outside the scar) (surf), or no seepage was present (ab). If more than one type of seepage is present, then only the one representing the wettest conditions was recorded. For instance, if there was seepage on the ground surface, the failure plane and the headscarp, SEEP would be recorded as 'surf. The hierarchy of worsening seepage conditions is surf-hs-fp-ab. The initial test of SEEP vs. DFRESULT was not significant at the a = 0.05 interval (Fig. 6.3.3), but indicated that the 'hs' group behaved distinctly from the other three results. The variable SEEP2 reclassifies SEEP into 'hs' and 'non-hs' groups. Unlike SEEP, SEEP2 is significant at a = 0.05 (F(l,60)= 6.1, p<0.017; mean DFRESULT for 'hs' is 0.31 and for 'non-hs' is 0.68). Exactly why failures with seepage at the headscarp should become debris flows much less frequently than other failures is uncertain: it may have to do with the most commonly observed cause of headscarp seepage, the intersection of a macropore by the headscarp. 97 Figure 6.3.2b: Plot of Means LOCN2 Main Effect F(2,59)=7.63; p<0011 1.1 1 0.9 _ l 0.8 0.7 \ : RESL 1 \ U_ Q 0) r> 0.6 \ ro ro > 0.5 1 \ : 0.4 \ 0.3 i ; ; head channel side LOCN2 Figure 6.3.2b: Result of ANOVA test of influence of LOCN2 on DFRESULT, showing plot of means, F statistic and p-value. Figure 6.3.3: Plot of Means SEEP Main Effect F(3,58)=2.01;p<1226 0.75 r 1 1 0.25 1 ' '< J ' 1 ab fp hs surf SEEP Figure 6.3.3: Result of ANOVA test of influence of SEEP on DFRESULT, showing plot of means, F statistic and p-value. The SEEP='hs' class seems to behave in a different fashion than do the other three classes. 98 The parameter LITHOL is not significant at a = 0.05 with regards to DFRESULT (Figure 6.3.4a). The seven lithologic codes in LITHOL are 'md' (mudstone); 'b', blueschist; 's' , sandstone; 'm', undifferentiated metasedimentary rocks; 'gd', granodiorite; 'a', andesite; 'phy', phyllite. The most obvious reclassification scheme, using general type of rock (sed./meta./igneous), LITHOL2, increased F but still was not significant at the a = 0.05 level (F(2,59)=1.68, p<0.19). Combining metamorphic and sedimentary rocks into one group did not significantly improve p-value (LITHOL3: F(l,60)=2.67, p<0.11), nor did combining metamorphic and igneous rocks (LITHOL5: F( 1,60)= 1.91, p<0.17). However, a system comparing the grouping of phyllite, andesite and granodiorite to a group comprising the other rock types (LITHOL4) was significant at a = 0.05: F(l,60)=4.59, p<0.036; mean DFRESULT for the group of granodiorite, andesite and phyllite was 0.68, vs. mean DFRESULT of 0.38 for the group of mudstone, blueschist, sandstone and miscellaneous metasedimentary rocks What distinguishes the grouping of granodiorite, andesite and phyllite from that of mudstone, sandstone, blueschist and metasedimentary rocks? Certainly, it seems an unlikely division: phyllite and blueschist are almost the same type of rock. However, examination of the photographic record of the area showed that rocks of the granodiorite/ andesite/ phyllite group were relatively massive and unbroken, while rocks of the other group were either highly jointed, faulted and broken, or relatively porous in the case of sandstone and mudstone. Furthermore, comparing the sample site locations to the geologic map (Figure 3.4.1) indicated that many, but not all, of the 'blueschist' locations were within the outcrop of the Cultus Formation, while many of the 'phyllite' locations were within the outcrops of the Chilliwack Group and Slollicum Schist, formations which are more highly metamorphosed and broken than are the rocks of the Cultus Formation. Thus, LITHOL4 may actually be a surrogate division of the rockmass by hydraulic conductivity, with the gd/a/phy group having relatively low hydraulic conductivities, and that of 99 Fig 6.3.4a: Plot of Means LITHOL Main Effect F(6,55)=1.26; p<2923 granodio sandston phyllite metased. blueschi andesite mudstone No. cases: gdio(34),sst(5),phyl(11),metased(4),blue(6),and(1),mud(1) LITHOL F i g u r e 6.3.4a: Result of ANOVA test of influence of LITHOL on DFRESULT, showing plot of mean DFRESULT by LITHOL class, F statistic and p-value. the other group being higher. If this was the case, then the rocks with lower conductivity, being more impermeable, would create higher pore pressures in the sediments above them, and higher runoff in gullies would result; so failures in those lithologies would be wetter, and encounter more water in the channel, making them more likely to become debris flows than failures in the more permeable lithologies. However, this explanation must be regarded as provisional until further research specifically examines the effects of hydraulic conductivity of the rock mass with regard to debris flow initiation; indeed, it is only conjecture (at this point) that links LITHOL4 with hydraulic conductivity in the first place. In addition, the results are most probably site-specific, and not applicable to all granodiorites, phyllites, etc., only to the ones in the study area. The interrelations of the variables SEEP, DRAIN and LITHOL are discussed in greater in section 6.6.2 of this chapter. 100 The variable C U R V E is non-significant (F(2,59)=1.76, p<0.18), despite the fact that previous research has shown it to be a significant influence on slope failure. The secondary variable CURVE2 separates slopes which are concave (cc) from those which are convex or straight (cv/st). The result of the A N O V A test of CURVE2 is just non-significant at the a = 0.05 level (F(l,60)=3.36, p<0.072; mean DFRESULT for concave slopes is 0.48, while for the grouped convex and straight slopes it is 0.71). This indicates that CURVE2 may play some role in influencing debris flow initiation, but the evidence is not conclusive. It is also hard to see why slopes that are concave (in horizontal cross-section) would be less likely to produce a debris flow from a failure than would slopes that are convex or straight. Conceptually, the opposite result should occur: concave slopes should be at the highest risk of debris flow initiation, since a concave slope curvature focuses water into the gully, and convex slopes should be at lowest risk, with straight slopes being intermediate. The latter result was observed in a previous Terrain Attribute Study (Rollerson, 1992): slope curvature was found to be significant at the a = 0.01 level, and concave slopes had the greatest observed frequency of slope failure. 6.4 Test of failure parameters vs. observed results: significant quantitative parameters. The most significant quantitative parameters were found to be channel gradient (as TANCG), natural logarithm of initial failure volume (LNVOL), angle of failure slope (FAILSLP), angle of entry of failure, a (AOE), and in-channel stored sediment (ICSS). The effect of channel gradient (as TANCG) on debris flow initiation was shown for the failure data in section 6.1. The effect of initial failure volume is as important: examine the plot of V O L U M E vs. T A N C G (Figure 6.4.1a). The observed distribution of volumes is non-normal: most of the volumes are 2000 cubic meters or smaller, but there are two which are over 15000 m 3. By taking the natural logarithm of the volume (LNVOL), a normal distribution is achieved. 101 [LNVOL was used in preference over LOG10VOL because it gives a wider scale over the range of volumes studied, hence allowing greater resolution.] The effect of increasing failure volume on whether or not failures initiate debris flow is quite visible. The distributions of LNVOL by DFRESULT are noticeably different for the 'no' and 'yes' groups (Figure 6.4. lb). Statistically, ANOVA gives this result a very high significance (F( 1,60)= 13.41, p<0.0005). Examining the distributions of VOLUME and LNVOL shows that the most important threshold value is around 100-150 cubic meters (VOLUME= 100 m3 or LNVOL = 5). Statistically, both breaks are significant, with the LNVOL = 5 break being slightly more significant (F(l,60)=14.68,p<0.0003 for LNVOL=5 and F(l,60)=14.12, p<0.0004 for VOLUME = 100 m3 ). Note that LNVOL = 5 is equivalent to a volume of 148 m3. [Given the possible error in the volume measurements, the figure of 100 m3 is probably better to use as a threshold value, as it is understood to be of less precision than the figure of 148 m3.] Mean DFRESULT for the <100 m3 failures is 0.29, while for the > 100 m3 failures it is 0.72. F i g u r e 6 . 4 . 1 a : P l o t o f T A N C G v s . V O L U M E ( 3 O 0 . 8 0 . 7 0 . 6 0 . 5 0 . 4 0 . 3 0 . 2 0 . 1 • • o ; o • ; o -6 o a ••; • • o • o o • • ; •-D-D • :•• • • • 1 0 o •> • • • ; o o • Dm co • D • .a I O....D o; o o ...a o o 1 0 0 1 0 0 0 V O L U M E ( c u b i c m e t e r s ) 1 0 0 0 0 o D F R E S U L T : n o • D F R E S U L T : y e s Figure 6.4.1a: Plot of channel gradient in percent slope (TANCG) vs. volume of slope failure (VOLUME) with result of failure (debris flow initiation) indicated. 102 Figure 6.4.1b: Histogram of LNVOL categorized by DFRESULT 1 2 1 0 8 iS 6 o **— o o 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 3 4 5 6 7 8 9 1 0 1 1 D F R E S U L T : D F R E S U L T : n o y e s LNVOL Figure 6.4.1b: Histogram of LNVOL categorized by DFRESULT, showing the differing distributions of volumes for failures which did and did not initiate a debris flow. That the volume of the initial slope failure should affect whether or not a debris flow initiates has already been noted above in conjunction with the significance of TYPE. It makes sense that larger failures should be more likely to initiate debris flows than should smaller failures, since the larger failures are more able to entrain debris in the channel by reason of their size. They are also travelling at higher velocities, and have greater inertia than smaller failures, and therefore are less likely to stall when they enter the channel. Additionally, large failures also have a greater impact force when striking the channel bed, and are therefore more likely to initiate a debris flow by impulsive loading of the in-channel sediments (Bovis and Dagg, 1992). The variable FALLSLP measures the slope gradient of the failure plane. Mean FALLSLP varies significantly for the two possible outcomes of DFRESULT (F(l,60) = 5.3, p<0.025; mean 103 Figure 6.4.2a: Histogram of FAILSLP categorized by DFRESULT —r 1 r— 15 20 25 30 35 40 45 50 55 60 15 20 25 30 35 40 45 50 55 60 DFRESULT: DFRESULT: no yes FAILSLP Figure 6.4.2a: Histograms of FAILSLP categorized by DFRESULT. The histograms are different, but not exceptionally so: the primary differences occur mainly in the numbers of extreme values, and the modes of the two distributions are similar. FAILSLP for DFRESULT ='no' is 40.5° and for DFRESULT='yes' is 36.8°. However, the magnitude of the difference is not large (less than four degrees). Strangely, the average failure slope is steeper for failures which did not initiate debris flows than for those failures which did initiate debris flows. Examining the distributions of FAILSLP (Figure 6.4.2a) for the two groups shows that above 45°, there are only 5 data points, four of which did not result in a debris flow, one of which did. The other major contributing factor to the skew in distributions between the two groups is the 20-30° group, where there are 8 'yes' results and only one 'no'. These two zones at the endpoints of the distribution are what produce the variation between the 'no' and 'yes' groups. Analysis of variance with FAILSLP categorized into these three groups (20-30°, 30-45°, and >45°), gives F(2,59)=2.42 and p<0.0981- not significant at the a = 0.05 level, likely due to the small sample sizes of the two extreme groups. Neither isolating the 20-30° or the >45° 104 renders the distinction significant at a = 0.05, although the >45° threshold is almost significant (F(l,60)= 3.52, p< 0.061). Assuming for the moment that the results are not an artifact of the small sample sizes at the extreme high and low slopes, why would relatively steep failures be less likely, and relatively gentle failures be more likely, to result in debris flows? The answer may have to do with the characteristic failures on those slopes. Failures with steep failure planes seem more likely to be those which result from gravitational effects only (for instance, failures caused by toppling); those with gentler failure planes seem more likely to result from the effects of increased pore pressures (for instance, failure along a weathering horizon). Thus, the relatively less steeply sloped failures may be more likely to be saturated, and the steep failures less likely. Therefore, when the failure mass enters the channel, it may be more likely to be already travelling as a debris flow if it is from a gently sloped failure, and less likely to be so travelling if it is from a steep failure; hence more and less likely, respectively, to continue down channel as a debris flow. This explanation is advanced as a hypothesis only. Given the relative uselessness of FALLSLP as a measure of predicting debris flow initiation, and its low statistical significance when compared to other variables such as channel gradient or failure volume, FALLSLP will not be discussed further herein. The variable AOE, angle of entry of failure into the channel (measured as a), has already been discussed in Chapter 4. AOE was found to be significant in influencing debris flow initiation by Bovis and Dagg (1992), Benda and Cundy (1990), and Benda and Dunne (1987). Failures which entered the channel at low a were found to be more likely to initiate debris flows, while those at high a were less likely to do so (although in the latter case, debris flows caused by formation and collapse of a dam from the failure deposits are more likely to occur). Failures at low a are more often associated with headwall failures, while those with higher a's are most likely to be sidewall failures. Plotting TANCG vs. AOE (Figure 6.4.3a) shows that the two share a weakly linear inverse relationship: it also shows a major threshold at approximately AOE = 42° Since AOE was only measured to the nearest 5°, this threshold is really between 40° and 45°. Examining the distribution of AOE categorized by DFRESULT (Figure 6.4.3b) also shows this trend, and shows how much the distributions of AOE differ for DFRESULT 'yes' and 'no'. Note that the two figures (6.4.3a and 3b) appear to differ slightly because 6.4.3a does not show frequency counts for the points (some points overlap). The difference in the two distributions is highly significant (F(l,60)=24.3, p<0.000007; mean AOE for DFRESULT='no' is 50° and for DFRESULT ='yes' is 22°. The "42°" threshold value is also highly significant (F(l,60)=24.4, p<0.000006; for AOE < 40°, the mean DFRESULT is 0.83, while for AOE >45°, mean DFRESULT is 0.3). Especially of note on these figures is the very high proportion of 'yes' F i g u r e 6 . 4 . 3 a : P l o t o f T A N C G v s . A O E b y D F R E S U L T a n d L O C N 2 o D F R E S U L T : n o L O C N 2 : h e a d • D F R E S U L T : n o L O C N 2 : c h a n n e l o D F R E S U L T : n o L O C N 2 : s i d e A D F R E S U L T : y e s L O C N 2 : h e a d • D F R E S U L T : y e s L O C N 2 : c h a n n e l • D F R E S U L T : y e s " " - 1 0 1 0 3 0 5 0 7 0 9 0 1 1 0 L O C N 2 : s i d e A O E ( d e g r e e s ) Figure 6.4.3a: Scatterplot o f TANCG v s . AOE, categorized b y DFRESULT and LOCN2. Threshold value of AOE ~ 42° is highlighted. u.o 0 . 7 A A A j I O o o 0 . 6 A m ft 0 . 5 CD O • A o o A 0 . 4 0 . 3 0 . 2 o o o n 1 106 Figure 6.4.3b: Histogram of AOE categorized by DFRESULT. The two distributions are obviously quite different. F i g u r e 6 . 4 . 3 c : P l o t o f M e a n s L O C N 2 M a i n E f f e c t F ( 2 , 5 9 ) = 4 8 . 3 9 ; p < 0 0 0 0 0 0 1 6 0 | : 1 - 1 0 h e a d c h a n n e l s i d e L O C N 2 Figure 6.4.3c: Result of ANOVA test of influence of LOCN2 on AOE, with mean AOE by location class, F statistic and p-value. 107 results for angles of 0-10 degrees. These are almost all headwall and in-channel failures. Statistically, AOE correlates very strongly with location of failure. Figure 6.4.3c shows this strong correlation (note very high F statistic and very low p-value (p<lxl0~7)). Even when only sidewall failures are considered, the "42°" threshold is still significant (F(l,36) = 9.28, p<0.0043; mean DFRESULT for AOE < 40° is 0.75 and for AOE >45° is 0.27). The last significant quantitative parameter of interest is ICSS, in-channel stored sediment. The methodology behind the measurement of ICSS was given in Chapter 4. With regards to failures, ICSS for the reach was measured above the point at which the failure entered the channel. Thus, ICSS measurements recorded as failure data provide an estimate of the channel conditions and volume of sediment which the failure would have encountered, not the channel conditions after scour by debris flow or after the failure has contributed its sediment to the channel. A plot of TANCG vs. ICSS (Figure 6.4.4a) shows a threshold value for debris flow initiation of between 2 and 4 m3/m, below which debris flow initiation is much more likely to occur. The mean ICSS values for the two DFRESULT groups are significantly different (F(l,60)= 11.1, p<0.0015) and fall on either side of the threshold value (mean ICSS for DFRESULT ='no' is 4.5 m3/m, and for DFRESULT ='yes' is 1.6 m3/m . A histogram of ICSS categorized by DFRESULT (Figure 6.4.4b) shows how differently the two variables are distributed, and also indicates that a critical threshold lies in the 2 to 4 m3/m range. Testing the significance of the thresholds finds that both are significant, though the value of 4 m3/m is more significant (for the threshold value of 2, F(l,60) = 11.2 and p<0.0014; for the value of 4, F(l,60) = 12.6 and p<0.0007). Why does the apparent likelihood that a debris flow will initiate decrease with increasing ICSS? Several factors must be considered. It would appear that the more debris there is in the 108 F i g u r e 6 . 4 . 4 a : P l o t o f T A N C G v s . I C S S b y D F R E S U L T O o 0 . 8 0 . 7 0 . 6 0 . 5 0 . 4 0 . 3 0 . 2 0 . 1 • • D CD ! • D(D I - O -• • • ; -o i '• • m • ; o mon • o o m •• • OD o • 0 • D O 2 6 1 0 1 4 1 I C S S ( c u b i c m e t e r s p e r m e t e r o f c h a n n e l l e n g t h ) o D F R E S U L T : n o 8 • D F R E S U L T : y e s Figure 6.4.4a: Plot of TANCG vs. ICSS categorized by DFRESULT. An apparent threshold value for ICSS with regard to DFRESULT lies somewhere in the range of 2 to 4 m3/m. Figure 6.4.4b: Histogram of ICSS categorized by DFRESULT. The histogram for DFRESULT ='no' has a proportionally thicker tail for the high ICSS values than does the 'yes' histogram. 109 channel, the more debris there is to be entrained by a potential debris flow. If the flow entrains too much debris, the plug at the front of the flow may become large enough to form a stable jam and the debris flow may stop itself. If the volume of debris in-channel is high, and the failure volume is small, then the flow of water in the gully may be through the pores in the ICSS and thus the failure may not encounter this water, instead building up a cone on top of the already existing debris (such a situation was observed in the case of FOL-004-LB-F01, Figures 6.4.4c-4e). Conversely, the measure of ICSS is likely itself a surrogate for other measures of channel stability: in a gully reach that is at risk of slope failures, but which has low potential for those failures to initiate a debris flow, the failed material may build up in the gully bed with no ability to be transported out of the reach, gradually increasing ICSS over time. Thus higher ICSS values may simply indicate channels which are more stable for other reasons, and may not be an active factor in debris flow initiation. If GWFP and GGP were known to work accurately as predictors of gully wall stability and potential for debris flow initiation, then the highest ICSS reaches wouldbe expected to be those with high GWFP and low GGP, and lowest ICSS would be in reaches where the reverse is true. Of course, this mainly applies to mineral ICSS, for other factors such as logging history influence the amount of woody debris ICSS. It is tempting to conclude that the amount of sediment stored in the channel is an effective indicator of the stability of the reach with regard to debris flow initiation. However, to do so would be wrong. The recorded occurrence in this study of debris flows initiating in reaches with high ICSS, and reaching the gully fan, indicates that high ICSS by itself cannot successfully indicate gully stability. In fact, the ability of in-channel sediment to influence debris flow initiation is likely dependent on a number of other factors, such as the volume and velocity of the failure entering the gully, the porosity and degree of saturation of the sediment, its size range and material 110 Figure 6.4.4c Figure 6.4.4d Figure 6.4.4c: Scar of failure FO-004-LB-01 F01. Failure date was 1995, failure volume was measured as approximately 140 m 3 (scar measures 12m. long by 8m. wide by 2m. deep). Surficial material is colluviated till, fairly well consolidated at this site (but becoming less so as one progresses down the gully). Figure 6 .4.4d: Depositional cone of failure FO-004-LB-01 F01 in gully channel. Sediment and CWD from the failure have built up a cone on top of already present ICSS. Stream flow in the gully is approximately 0.75m below visible surface of ICSS layer. ICSS is predominantly (-60%) woody slash and CWD, with relatively high porosity. I l l Figure 6.4.4e: Looking down gully channel below the cone shown in Fig. 6.4.4e. The mean ICSS value calculated for this reach is on the order of 8.0 m3/m. The edge of the depositional cone is visible to the right of the person. Note that the sediment in the channel is beginning to become overgrown with vegetation, indicating its relative immobility. Despite the high ICSS value for the gully, the large failure FO-004-LB-03 F02 (-2000 m3), entering the gully two reaches lower down, was able to initiate a debris flow in a channel which was not as steep and had higher ICSS. composition, and other factors. This problem is discussed in more detail in section 7.3.1 of Chapter 7. 6.5 Regressions and results: VOLUME and AOE. Of the five quantitative parameters initially identified in section 6.4 as being significant, one (FALLSLP) proved to be largely unimportant when studied more closely. Of the remaining four parameters, two (channel gradient, as TAN CG, and ICSS) can be measured in a gully reach before occurs: the other two (failure volume as VOLUME or LNVOL, and AOE) may only be measured after failure has occurred. To be effective in predicting whether or not a failure will initiate a debris flow, a parameter must be measurable before failure occurs. Thus, for the post-112 failure parameters, VOLUME / LNVOL and AOE, the question to be answered is whether or not these parameters may be estimated before failure through an empirical equation based on only pre-failure parameters. To develop empirical equations to estimate the post-failure parameters before failure occurs, the method of multiple regression was used. Using the forward stepwise regression option in STATISTIC A, all parameters which could potentially be related to the two parameters to be estimated were selected. STATISTICA then calculated a regression equation in which the statistically significant parameters from those chosen were addded sequentially to the regression equation, from most to least significant, with the equation recalculated and an r-squared statistic redetermined at each step. Thus, the point at which the addition of new parameters produced only minimal improvements in accuracy could be determined, allowing selection of the regression equation which provided the greatest predictive capacity with the least number of variables. The first regression to be attempted was that for volume. Since VOLUME is not normally distributed, several other measures of volume (LOGVOL, LNVOL, SQRTVOL (square root of volume)) were also used in attempts to find the regression with the highest predictive power (as measured by r-squared value) for the resultant equation. In general, the results were poor. VOLUME and its derived parameters are not easily predicted from pre-failure variables. The best result was obtained for LNVOL (regression equation shown in Figure 6.5.1a and graph of predicted vs. observed results in Figure 6.5.1b), but r2 value was only 0.32. Failure volumes are not easily predicted for several reasons. Most importantly, in each gully reach, a range of failures can occur, each of a different volume. It might be easier to predict average, or potential maximum, failure volume from terrain parameters, but to estimate maximum volumes of past failures, gully fan deposits would have to be studied, a relationship between debris flow volume and failure volume established, and frequency-magnitude 113 curves constructed. A similar approach was successful for Jakob (1996) but was beyond the limits of this study to perform. STAT. Regression Summary for Dependent Variable: LNVOL MULTIPLE R= .56657073 R2= .32100239 Adjusted R2= .27335344 REGRESS. F(4,57)=6.7368 p<00016 Std.Error of "estimate: 1.3472 St. Err. St. Err. N=62 BETA of BETA B ofB t(57) p-level Intercpt .264 1.75 .151 .880 LNBAS .403 .138 .593 .203 2.93 .005 TANFS -.453 .143 -3.57 1.13 -3.16 .002 TANGWSA .255 .135 2.21 1.17 1.88 .064 GWSD .226 .137 .020 .012 1.64 .105 Figure 6.5.1a: STATISTICA multiple regression summary for LNVOL. The other post-failure parameter of interest, AOE, was anticipated to be easier to predict from pre-failure terrain variables than was failure volume. Angle of entry of sidewall failures should be largely a matter of geometry: the angles of the sidewalls and the angle of channel gradient F i g u r e 6 . 5 . 1 b : P r e d i c t e d v s . O b s e r v e d V a l u e s , R e g r e s s i o n o f L N V O L D e p e n d e n t v a r i a b l e : L N V O L ; r - s q u a r e d = 0 . 3 2 i ; o o ° jo o o o 0 O : O „ : „ --O O :0 O- -<^5 o \° ° .P-j^er^^---" o°l„..^'A><r^'"""n Q "** o 0 Q-- Oo o - - o ° i * o o : o o O 0 - o o i ° o "^ c< R e g r e s s i o n 1.1 • • 1 2 . 5 3 . 5 4 . 5 5 . 5 6 . 5 7 . 5 P r e d i c t e d V a l u e s Figure 6.5.1b: Plot of predicted vs. observed values of LNVOL for the regression equation given in Figure 6.5.1a. R 2 is only 0.32 - not very good. 114 together determine at what angle the fall line of the sidewalls intersects the channel. For instance, in a hypothetical gully with a flat channel (CG = 0°), the sidewall angle is unimportant: all failures will enter the gully at right angles to the gully channel. As channel gradient steepens, the angle the fall line (vertical) on the sidewalls makes with the channel will decrease as long as sidewall gradient remains constant. Sidewall angle also plays a role in influencing a: for instance, if sidewall angle is 0°, then there is effectively no gully, and slope failures are open slope failures, (perpendicular to the channel). A further complication when attempting to calculate a from geometric principles is that in real gullies, the channel is not a one-dimensional line, but an irregular planar surface; consequently, the sidewalls may not be exactly perpendicular to the gully, or parallel to each other. In a gully reach, the angle of entry of a failure can be estimated through measurement of the angle, O, at which the fall line of the sideslope intersects the channel. Unfortunately, this was not done in this study: I did not fully understand the importance of this parameter until after the field portion of the study was complete. In theory, for a gully with planar sideslopes, a for a failure should equal O for the reach. In actuality, gully wall slopes are not perfect planes, and the vegetation growing on them can divert a sidewall failure, reducing its a below what Q predicts it should be. Remnant levees from older debris flows in the gully can have a similar effect, as can large pieces of CWD or other objects found in the gully. Figure 6.5.2 illustrates the relationship between a and O. In attempting to derive a regression equation for AOE, my first step was to attempt to find a formula for prediction of a from CG and GWSA. It was at this stage that I discovered the importance of Q: if Q. is not known, then any equation for a simplifies to an identity rather than an expression of a in terms of CG and GWSA. Thus, the only option left was to attempt 115 Figure 6.5.2: Perspective and plan views of a gully wall and channel showing relationship between a and Q. for a failure. A similar situation is visible in Figure 3.6.3 of Chapter 3. empirical regression techniques: a wide variety of parameters were entered into a stepwise regression program and the program determined which were significant. Complex equations, using terms such as (cos2 GWSA + cos2 CG-2)/(2(cos GWSA cos CG - 1) gave maximum r2 values of-0.65; simple equations (using a single regression with tan CG) produced only slightly lower r2 values of -0.5. In both cases, the parameters used were ones which are already included in the DFLP assessment in other ways: channel gradient and gully wall slope angle. To include a calculation of AOE based on a regression equation using these parameters would be to include an 116 extra level of redundancy, without significant improvement in accuracy; if desired, the same effect could be accomplished by changing the weighting of the parameters in the calculation of DFLP. In summary, of the two after-failure variables which have been shown to be significant, one is difficult to predict accurately from pre-failure measurements (volume) and the other one (AOE) may be calculated from variables which are already incorporated into the GAP, so it does not require the inclusion of any new variables. 6.6 Between-parameter correlations of non-failure variables: significant correlations. Aside from the parameters which correlate directly with debris flow initiation, the failure data set also contains parameters which are intercorrelated significantly with each other, which, when examined, can provide additional insight into the factors influencing debris flow initiation. 6.6.1. Date of Logging and Date of Failure It is expected that there should be a relationship between date of logging and date of failure. Sidle et al. (1985) have shown that the period of maximum slope weakness due to root F i g u r e 6 . 6 . 1 : P l o t o f M e a n s F A I L D A T E M a i n E f f e c t F ( 4 , 5 7 ) = 4 . 3 5 ; p < 0 0 3 8 1 9 9 0 1 9 8 9 6- <X \ "J\ \ \ 1 9 8 8 / \ \ \ U I 1— 1 9 8 7 1 9 8 6 \ / . \ LOGDA" I / | \ I LOGDA" 7 \ i h i 1 9 8 5 ro 1 9 8 4 \ / \ 1 9 8 3 ; ; ; j ; 1 9 8 2 G _ 1 : 1 9 8 8 G _ 2 : 1 9 9 0 G _ 3 : 1 9 9 3 G _ 4 : 1 9 9 4 G _ 5 : 1 9 9 5 F A I L D A T E Figure 6.6.1: Result of ANOVA test of correlation between date of logging and date of failure. Mean LOGDATE for each FAILDATE class is shown as are F statistic and p-value. 117 decay is between six and fifteen years after logging has occurred. Thus, when examining the observations of logging date and failure date within the failure data set, such a relationship should exist. That is, the average age of logging should be later for the failures which occurred more recently, and so on. The relationship between FAILDATE and LOGDATE is shown in Figure 6.6.1. It is significant (F=4.35, p<0.004) at a = 0.05: that is, the difference in LOGDATE between the various failure dates is significant. But, looking more closely at the data, what does such significance indicate? The numbers of observations per group in FAILDATE are: 1988, 1; 1990, 13; 1993, 4; 1994, 3; 1995; 41. The great majority of the failures occurred in the 1990 and 1995 storms. Note that FAILDATE records only the last date of failure; earlier failures from the same scar are generally victims of data censoring, and hence not recorded. The single failure that occurred in 1988 was in a gully logged in 1983 (delay of five years): this is just outside the beginning of maximum weakness. The failure was not reactivated in the later storms. As this is but a single datum, it should not be subject to much interpretation. The average date of logging in the 1990 storms was 1984, a delay of six years. This is just within the period of maximum weakness. Note that logging in the studied gullies took place between 1981 and 1990. Thus, the maximum delay for 1990 storms could have only been ten years, and the minimum one: the observed mean delay of six years is likely also a reflection of this fact. In the 1993 and 1994 failures, represented by only a few data points (four and three respectively) delay is short (four to five years). These failures are in areas logged in the late 1980's, not earlier. These failures are not associated with major storm events. Rather, they seem to represent failures in areas especially susceptible to failure, such that only a minor decrease in slope stability is necessary for the slope to fail. The 1988 failure is probably also of this type. 118 The average logging date in the 1995 failures is 1985, representing a delay of ten years. This is directly in the middle of the period of maximum weakness (6-15 years delay); it is also in the middle of the period of logging studied (1981-1990), also a delay of six to fifteen years for 1995 failures. This overlap of the forcing with the period of weakness may explain why so many failures are observed during this event, relative to other failure dates. A confounding factor is the effect of data censoring. As 1995 events were the most recent failures observed, it seems likely that some of these failures represent reactivation of scars of earlier failures. Such reactivation removes much of the evidence of the earlier failure; thus, the observed date of failure is biased towards 1995, and the average delay observed for the 1995 failure group is also overestimated. 6.6.2. Drainage, seepage and lithology. The variables which measure soil moisture levels are strongly intercorrelated. SEEP varies significantly with DRAIN and vice versa (Figures 6.6.2a and 2b). Average seepage is highest in imperfectly drained soils, and lowest in rapidly drained soils. Likewise, average drainage is highest in soils without seepage ('ab'), and lowest in failures with seepage at the ground surface ('surf). This is as might be expected. In section 6.3, the significance of LITHOL4 with respect to debris flow initiation was discussed. It was hypothesized that the parameter LITHOL4 might be correlated with bedrock hydraulic conductivity, which was not measured. If that is so, then there should be a correlation between DRAIN and LITHOL4: the 'more permeable' group of rocks should have a higher average DRAIN and vice versa. LITHOL4 and DRAIN are significantly correlated (Figure 6.6.2c), but the relationship is the exact opposite of the one which was expected. The relationship between LITHOL3 (plutonic/non-plutonic) and DRAIN is also statistically significant at a = 0.05 (F(3,58)=3.54, p<0.02), confusing the issue further. The group of lithologies conjectured in section 6.3 to be of higher hydraulic conductivity 119 Figure 6.6.2a: Result of ANOVA test of the influence of SEEP on DRAIN, showing mean DRAIN for each SEEP class, F statistic and p-value. The DRAIN codes are T=l , 'm'=2, 'w' =3 and 'r' =4. Figure 6.6.2b: Result of ANOVA test of influence of DRAIN on SEEP, with mean SEEP for each DRAIN class, F statistic and p-value. The SEEP codes are 'ab'=0, 'fp'= 1, 'hs' =2 and 'surf=3 120 F i g u r e 6 . 6 . 2 c : P l o t o f M e a n s D R A I N M a i n E f f e c t F ( 3 , 5 8 ) = 6 . 9 8 ; p < 0 0 0 4 0 . 8 1 1 1 1 L i m w r D R A I N [ c o d e s f o r L I T H O L 4 : 1 = ' m d / b s / s s / m e f ; 2 = ' g d / a n d / p h y " ] Figure 6.6.2c: Results of ANOVA test of influence of LITHOL4 on DRAIN, showing mean LITHOL4 for each DRAIN class, F statistic and p-value, and LITHOL4 numeric coding. (the 'porous/broken' rocks, mudstone, sandstone, blueschist and metasedimentary groups) have a strongly significant lower average DRAIN value than do the 'low hydraulic conductivity' (granodiorite, andesite and phyllite) group. Why would the more conductive rocks have soils atop them which were less well drained? A possible answer might be that the studied portions of the slope were gullies. Groundwater and surface water flow direction is typically towards gullies from elsewhere on the slope. Thus, the observed higher soil moisture conditions on the rocks of higher hydraulic conductivity results from groundwater flow upwards from the rock to the overlying soil in the vicinity of the gully. However, this argument contradicts the hypothesis given in section 6.3 for the observed higher resultant debris flow rate from failures in the 'lower hydraulic conductivity' group. It is likely that neither hypothesis explains the observations fully. Note that DRAIN itself is statistically insignificant with regards to DFRESULT (F=0.64, p<0.58), and SEEP is not significant with regards to either LITHOL3 or LITHOL4. Another confounding factor in the relationship of seepage, drainage and lithology is the effects of differing surficial 121 materials on seepage and drainage; the parameter TERRAIN (representing surficial material) was just non-significant at the confidence level a = 0.05 when tested with ANOVA against SEEP and DRAIN, with p-values lying in the range 0.08-0.12. The conclusion which must be drawn from these observations is that the various LITHOL parameters, SEEP, and DRAIN should not be incorporated into revised evaluations of GGP, since their significant effects are contradictory and not fully understood. 6.7 Comments on analysis of failure data. Analysis of the reach data showed that the method used in the GAP to calculate GGP was flawed, and did not work very effectively. Using the failure data, parameters which significantly affect the chance that a debris flow will result from a slope failure into a gully system were identified and separated into qualitative and quantitative, pre-failure and post-failure groups. The next stage in the process is the synthesis of these factors to attempt to create a new measure of GGP which will perform significantly more effectively than the system currently in place. This is examined in Chapter 7. 122 Chapter 7- The Gully Assessment Procedures & Debris Flow Initiation Potential Assessment: Suggested Modifications. 7.1 Observed weaknesses of current DFD? methods. Chapters 5 and 6 have evaluated the GWFP and GGP as they are presented in the GAP. In summary, the current GWFP is fairly successful in identifying which gully walls are at risk of failure, but is overconservative, because it classifies many walls which appear to have only moderate failure potential into the high risk class. An improvement of GWFP would allow it to differentiate among the walls currently rated 'high', and reduce the ratings of those walls which did not fail. The current GGP is inadequate at predicting which failures will become debris flows, because one of the two parameters it uses is statistically insignificant, but is treated as if it were significant. A refined GGP would be an improvement if it could successfully predict in which gully reaches failures will generate debris flows. To devise new systems for evaluation of GWFP and GGP, a logical starting point is a conceptual analysis of the parameters which comprise the current GWFP and GGP. GWFP is currently evaluated from GWSA, indexed by surficial material. In that different materials have different stabilities at different angles, this approach seems reasonable. Current GGP is evaluated from the cross-referencing of GWSD and channel gradient (as TANCG). Channel gradient has been shown to be an effective predictor of whether slope failures entering the gully will or will not become debris flows; however, GWSD functions extremely poorly in predicting debris flow initiation. The rationale for inclusion of GWSD in GGP was that on longer slopes, a failure can achieve greater momentum before entering the channel, hence higher GWSD would be more likely to result in failures which initiate debris flows. Unfortunately, GWSD is only a potential maximum height from which failures may begin; in reality, many failures begin quite low on the walls of the gully, and the measured GWSD has little to do with whether or not these failures will 123 become debris flows. It has been shown that GWSD is ineffective in predicting debris flow initiation from failure. However, GWSD should not be excluded from the DFLP assessment, for it will be shown in the next section (7.2) that it is an effective parameter when evaluating the potential for gully walls to fail. Conceptually, it makes more sense to include GWSD, a gully wall parameter, with other gully wall parameters in an assessment of the stability of gully walls than it does to include it with a channel parameter in an assessment of the potential for failures to become debris flows. 7.2 GWSD as a parameter influencing gully wall failure potential, and NEWGWFP. Statistically, when GWSD is tested against variable PASTFAIL in the reach data, the test result indicates that GWSD is significant. The difference in mean GWSD between the reaches with failure and reaches without failures is strongly significant at the confidence level of a = 0.05. [F(l,90) =7.95, p<0.006; mean GWSD for PASTFAIL = 'no' is 15.5m and for 'yes' is 25m] . Examining the distributions of GWSD for the two categories of PASTFAIL (Figure 7.1.1) shows that, although the mode is similar for both, the 'yes' plot has a much longer tail. Failures are unlikely (less than 50% chance) only for low GWSD's (0-10m) although the chances of the two results are equal for GWSD of 30-40m. All reaches with GWSD>40m had failures. This effect is important, and is not included in the current GWFP at all. The current GWFP uses GWSA expressed in percent slope (tan GWSA). A plot of GWSD vs. TANGWSA is not very good for analyzing how GWSD and GWSA interact, because the vast majority of gully wall slope angles fall in the range between 60% and 100% slope. Using the actual measure of GWSA in degrees allows better resolution (Figure 7.1.2). If GWSD is an efficient predictor of slope failure, how may it be incorporated into the current GWFP rating system, which was shown in Chapter 5 to be reasonably effective in 124 Figure 7.1.1: Histograms of reach gully wall slope length categorized by failure occurrence. 110 90 70 E d 50 CO 30 10 -10 Figure 7.1.2: Scatterplot of GWSD vs. GWSA by PASTFAIL (with boundaries of most effective NEWGWFP classes) | | a | H: 76 ! • % PASTFAIL= 'yes' b \D a 35 m D i I I " i n fl a 6 X • a • n < CL O Q D M: 56% yes ° 9 g°9 Onp • a c n p n n D n p o .. 0 r t..9..n9 ! ° i UB Y °8 8 L: 38% yes J u ° o o ] 5 20 30 / 40 50 60 37.5° GWSA (degrees) 70 o PASTFAIL: no 80 ° PASTFAIL: yes Figure 7.1.2: Plot of GWSD vs. GWSA categorized by PASTFAIL result, with the boundaries of the most effective NEWGWFP classes indicated. 125 F i g u r e 7 . 1 . 3 : S c a t t e r p l o t o f G W S D v s . G W S A b y S U R F a n d F A I L 2 0 • • • • • • • o # o 2 , ' • - • 8 * • • ••• »• • • • « ^ •£> » A o A A A • 3 5 5 0 G W S A ( d e g r e e s ) 6 5 8 0 P A S T F A I L : n o S U R F : M P A S T F A I L : n o S U R F : C P A S T F A I L : n o S U R F : R P A S T F A I L : y e s S U R F : M P A S T F A I L : y e s S U R F : C P A S T F A I L : y e s S U R F : R Figure 7.1.3: Plot of GWSD vs. GWSA as for Fig. 7 . 1 . 2 , but material type is indicated in addition to PASTFAIL result. F i g u r e 7 . 1 . 4 : H i s t o g r a m o f G W S D b y S U R F a n d P A S T F A I L 21 r 18 -15 -10 G W S D ( m ) Figure 7.1.4: Histograms of GWSD categorized by PASTFAIL and SURF, showing that distribution of GWSD differs not only for reaches with and without failures, but also for difiering surficial materials. 126 predicting which reaches were at high risk of slope failure? Examining the GWSA vs. GWSD plot of Figure 7.1.2 with differentiation by surficial material (Figure 7.1.3) shows that the three surficial materials behave quite differently. The histograms of GWSD stratified by PASTFAIL and SURF (Figure 7.1.4) reinforce this observation: the effect of GWSD on PASTFAIL is most pronounced for colluvial slopes (SURF = C), is of moderate intensity for till slopes (SURF = M), and is only of minor significance for rock slopes, although the sample of rock-sided gully reaches is quite small (only 8 reaches out of 92). Note the strong correlation that exists between SURF and GWSD (Figure 7.1.5a): the average reach in a C gully has walls almost twice as long as those in an M or R gully reach. This is likely the case because colluvium tends to form thicker deposits (at the base of slopes) than does till, hence the possible maximum depth of the gully is greater, increasing the mean depth. Till can occur in both thin and thick deposits; rock gullies' sidewall length is controlled by other factors (eg. joint spacing). Testing GWSD vs. SURFFL (Figure 7.1.5b) confirms that gully wall length is most significant as an influence with regards to colluvial sidewalls, and has a lesser degree of importance for till and rock sidewalls. What does this imply for a revised GWFP? The rating system (assigning of H, M and L classes) should be stratified by SURF as it currently is, and within each group, hazard should be assessed based on both GWSD and GWSA. It is difficult to define boundaries for a new GWFP which will provide the same percentage chance of failure for all three reach types. Synthesizing the various factors discussed above, a new GWFP classification system is presented below (Table 7.1): Table 7.1: New G W F P Classi f icat ion System Surficial Material Gully Wall Failure Potential Rock (R): L Till (M) or Col luv ium (C): G W S A < 3 7 . 5 G W S A > 3 7 . 5 G W S D > 3 5 m H H 10m<GWSD<35m M H G W S D < 1 0 m L H 127 F i g u r e 7 . 5 . 1 a : P l o t o f M e a n s S U R F M a i n E f f e c t F ( 2 , 8 9 ) = 7 . 5 7 ; p < 0 0 0 9 M C R S U R F Figure 7.1.5a: Result of ANOVA test showing correlation between SURF and GWSD, with plot of mean GWSD for each SURF class, F statistic and p-value. F i g u r e 7 . 5 . 1 b : P l o t o f M e a n s S U R F _ F L M a i n E f f e c t F ( 5 , 8 6 ) = 6 . 0 2 ; p < . 0 0 0 1 45, : ; ; i 4 0 h 1 0 1 1 1 1 1 • -M N M Y C N C Y R N R Y S U R F F L Figure 7.1.5b: Result of ANOVA test of correlation between GWSD and SURF_FL, showing mean GWSD for each SURFFL class, F statistic and p-value. Statistically, the performance of NEWGWFP is compared to that of GWFP in Figure 7.2. la and lb. Note how NEWGWFP is significant at the a = 0.01 level, whereas GWFP is not significant at a = 0.05. The F-statistic for NEWGWFP is almost double that of GWFP (4.79 vs. 2.90). This indicates that NEWGWFP performs significantly better at assessing the hazard of a gully wall failure than does GWFP. Additionally, in NEWGWFP, the probability of wall failure increases steadily from class L to H , whereas in GWFP it is higher in L than it is in M. Furthermore, there are more reaches classed as NEWGWFP M or L than there are GWFP M or L reaches. Thus, NEWGWFP represents an improvement over GWFP in several respects: it is more statistically significant a division of gully reaches, it misclassifies fewer reaches than does GWFP, and its classes yield an increase in probability of failure with increasing assessed hazard, as they should. On the downside, it uses one more variable than does GWFP; however, this variable is already part of the DFLP assessment, so the overall complexity is not increased by replacing GWFP with NEWGWFP. 7.3 Parameters influencing debris flow initiation, and NEWGGP. In Chapter 6, several parameters were found to be significant with regard to debris flow initiation from slope failure. Of these parameters, only two, channel gradient (TANCG) and amount of in-channel stored sediment (ICSS), are both quantitative, and measurable pre-failure. Thus, it seems that a revised GGP should incorporate these two parameters. However, it will be shown that there are reasons why ICSS should not be included in a revised GGP. What other variables could possibly be included in a new GGP? Of the other significant quantitative variables (other than channel gradient), volume is difficult to predict accurately pre- failure, and AOE is primarily a function of channel gradient. Of the qualitative variables, TYPE and SEEP are difficult to predict pre-failure; LITHOL4 is significant, but has an unexplained mechanism, and in F i g u r e 7 . 2 . 1 a : P l o t o f M e a n s N E W G W F P M a i n E f f e c t F ( 2 , 8 9 ) = 4 . 7 9 ; p < . 0 1 0 6 0 . 8 5 r • ! 0 . 3 5 1 1 1 L M H N E W G W F P F i g u r e 7 . 2 . 1 b : P l o t o f M e a n s G W F P M a i n E f f e c t F ( 2 , 8 9 ) = 2 . 9 0 ; p < . 0 6 0 5 0 . 7 5 r — — ! ! G W F P Figures 7.2.1a and lb: Comparison of ANOVA test results of the performance of GWFP and NEWGWFP in determining which reaches are at risk of wall failure. Figure 7.2. la shows the performance NEWGWFP; Fig. 7 . 2 . 1 b shows the performance of GWFP. 130 any case is not comprehensive with respect to rock types. However, LOCN2 (head, side or channel), is very highly correlated with AOE, and is not dependent on channel gradient. Another parameter, O, was not measured; however, Q and a are conceptually related, and a was shown to be significant in influencing debris flow initiation, a is a post-failure parameter, but Q. can be measured pre-failure. Thus, Q seems to be another parameter which could be included in a revised GGP. It will be shown that inclusion of Q does not produce a significant improvement in the accuracy of a revised GGP, since it is mainly a function of channel gradient and sidewall angle. 7.3. J Revised GGP model using TANCG and ICSS (NEWGGP 1). A plot of TANCG vs. ICSS, categorized by DFRESULT, was shown in Figure 6.4.4a. Thresholds of 0.5 for TANCG and 2.0 m3/m for ICSS are used to delimit the boundaries of the proposed NEWGGP 1 classes. NEWGGP 1 class H: TANCG > 0.5 and ICSS < 2.0 m3/m. class M: TANCG > 0.5 and ICSS > 2.0 m3/m, or TANCG < 0.5 and ICSS < 2.0 m3/m class L : TANCG < 0.5 and ICSS > 2.0 m3/m. The statistical effectiveness of this GGP rating scheme, NEWGGP1, is shown in Figure 7.3.1a and lb. NEWGGP 1 is tested against the variables GGPOOR and GGP2FAN in the reaches dataset (whether or not the reach in question produced a debris flow which travelled out of reach or to fan, for reaches with failures only). NEWGGP1 is significant at the a = 0.0005 level for GGPOOR and at a = 0.01 for GGP2FAN, while GGP was not significant at a = 0.05 for either variable (ref. Figures 5.5.1a and lb). This shows that the classes of NEWGGP1 are statistically distinct from each other with regards to debris flow initiation from failure. The tests above beg the question of whether or not it is valid to include ICSS in the process of estimation of a new GGP. ICSS is a parameter which may be easily changed during logging, through addition of slash to, or cleaning of slash from, a gully. To include ICSS in a 131 F i g u r e 7 . 3 . 1 a : P l o t o f M e a n s N E W G G P 1 M a i n E f f e c t F ( 2 , 5 6 ) = 9 . 0 3 ; p < 0 0 0 4 H M L N E W G G P 1 F i g u r e 7 . 3 . 1 b : P l o t o f M e a n s N E W G G P 1 M a i n E f f e c t F ( 2 , 5 6 ) = 4 . 8 6 ; p < 0 1 1 4 0 . 8 | 1 s r 0 . 1 I i i L H M L N E W G G P 1 Figures 7.3.1a and lb: Results of ANOVA tests of the effectiveness of NEWGGP1 in predicting likelihood of debris flow initiation for debris flows travelling, respectively, out of the reach they initiate in (Fig. 7.3.1a) and to the gully's fan (Fig. 7.3.1b) 132 revised GGP in the manner described above for NEWGGP1 would be to suggest that the risk of debris flow in a reach can be reduced through the addition of CWD and colluvial sediment to that reach. In fact, such an addition would increase the potential size of a resultant debris flow, and hence its hazard. ICSS as it was measured in this study was likely a surrogate measure for other stability parameters. Part of the significance of ICSS with regard to debris flow initiation can be explained as a result of such parameter replacement: the gullies with higher ICSS were those which were most stable (least likely to transport debris) for other reasons; hence they had both high ICSS values and low chances of debris flow initiation. Thus, there is probably no cause and effect relationship between high ICSS and low chance of debris flow initiation, in which case increasing ICSS (for example, by dumping logging slash into a gully) would have no effect on whether or not a debris flow would initiate. Another liability of using ICSS in a revised GGP is that its effectiveness is dependent on the size of the initial slope failure. This effect is shown in a graph of LNVOL vs. ICSS (Figure 7.3. Ic). The two lines on the graph show the upper boundary of the DFRESULT = 'no' group and the lower boundary of the DFRESULT = 'yes' group. Thus, the plot is divided into three regions: in the uppermost, debris flow almost certainly initiates from a failure; in the middle, debris flow initiation is possible, though not certain; in the lowest region, debris flow initiation is extremely unlikely, and may be impossible. Note the paucity of data above ICSS values of 6.0 m3/m, which makes an extension of the boundaries into this region of low confidence. (Lines were drawn by eye, and are estimates only; curved lines could have also been drawn.) As it was shown in Chapter 6 that volume of initial failures is very difficult to predict from pre-failure measurement of terrain parameters, and that a range of failure volumes can occur in a given reach, Figure 7.3.1c also indicates another weakness of using ICSS to predict debris flow initiation, 133 Figure 7.3.1c: Scatterplot of LNVOL vs. ICSS by DFRESULT with upper and lower bounds of 'yes' and 'no' estimated o dfresult='no' • dfresult='yes' ICSS (cubic meters per meter) Figure 7.3.1c: Scatterplot of LNVOL vs. ICSS categorized by DFRESULT, showing inferred boundary conditions for debris flow initiation (boundaries were drawn by eye, and are not intended to be exact- curved lines could also have been drawn) namely that a large enough failure may be able to trigger a debris flow no matter what the ICSS value is for the reach. 7.3.2 Revised GGP model using TANCG only (NEWGGP2). The NEWGGP model using TANCG only (NEWGGP2) is simply stated as follows: TANCG > 0.5: NEWGGP2 = H; TANCG < 0.5: NEWGGP2 = L. As this model is the most conservative of the three revised GGPs proposed, it can be used as a standard against which the other two proposals can be tested. If they do not produce a substantial improvement in performance over the model using only TANCG, then they should 134 receive no further consideration. In order for this to be done, the success of NEWGGP2 itself must first be evaluated. Testing NEWGGP2 against the variables GGPOOR and GGP2FAN for the reach data set reveals that it is strongly significant (Table 7.3.2). Compare this with the effectiveness of GGP in predicting the same thing (Figures 5.5.1a and lb). Table 7.3.2: Resul ts of A N O V A tests of N E W G G P 2 vs. G G P O O R and G G P 2 F A N M e a n var iable for N E W G G P 2 c lasses : Var iab le F-stat deg. free. p-value L H G G P O O R 13.88 1,57 <0.0004 0.35 0.81 G G P 2 F A N 11.88 1,57 <0.0011 0.23 0.69 F-statistic values are 10 times higher for NEWGGP2 than they are for GGP; p- values are in the p<0.001 range for NEWGGP2 vs. p<0.5 for GGP. All of this leads one to conclude that NEWGGP2 is highly effective at discriminating between reaches with high hazard of debris flow initiation from those in which such hazard is low, whereas GGP was not nearly as effective. Also, NEWGGP2 is of greater significance (higher F statistic) than is NEWGGP1, although NEWGGP1 had a slightly higher fraction initiating debris flows for its H class, and a lower fraction for its L class. 7.3.3 Revised GGP model using TANCG and Location/Predicted Angle of Entry (NEWGGP3) Almost all head and channel failures become debris flows (ref. Figure 6.3.2b), and have an AOE of 10° or less (ref. Figure 6.4.3c). By contrast, sidewall failures have a wide range of possible AOE's (Figure 7.3.3a). Although the planimetric angle of the gully wall/channel intersection (O) was not measured in this study, it seems highly likely that this variable is strongly correlated with AOE for sidewall failures. Thus, one possible revision of GGP could be made by using the measured AOE values as surrogate measurements of Q, and developing a new GGP 135 classification based on TANCG, LOCN2 and Q. The model of NEWGGP which uses LOCN2 and Q (as AOE) in addition to TAN CG may be stated as follows (NEWGGP3): LOCN2 = 'head', NEWGGP3 = H; LOCN2 = 'side', use TAN CG and L> TAN CG>0.5 andQ<45°: NEWGGP3 = H; TANCG < 0.5 and Q < 45°, or TANCG > 0.5 andQ>45°: NEWGGP3 = M; TAN CG<0.5 andQ>45°: NEWGGP3 = L. F i g u r e 7 . 3 . 3 a : H i s t o g r a m o f A O E f o r L O C N 2 - s i d e ' o n l y o Wmm fm \W Wow I^lllllIlP HP WW WW mm HH ma \ 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 0 L 0 C N 2 : s i d e A n g l e O f E n t r y ( d e g r e e s ) Figure 7.3.3a: Histogram of a for sidewall failures only, showing distribution of observed values, with the majority falling in the 30°-60° range. There are three disadvantages to using this system. First, the relationship between Q. and a (AOE) has not been shown to exist in field data, only through geometrical consideration, and hence must be regarded as conjecture. Secondly, it was shown in Chapter 6.5 that AOE is at least partly a function of channel gradient, and it seems probable that the relationship is stronger for D. and TANCG than it is for a and TANCG; therefore, the NEWGGP3 classification system contains a redundancy. Thirdly, the failure location variable is not directly applicable to the reach 136 scale. A reach at the head of the gully may have a portion equivalent to the LOCN2 location 'head' (includes headwalls and open slope depressions at the head of gullies), as well as a portion corresponding to the LOCN2 location 'side' (gully sidewalls and open slope depressions entering into the gully from the side). Thus, assigning a LOCN2 code to an entire reach can be difficult. For this reason, NEWGGP3 was not tested against the reach data for significance, but rather against the similar failure data set (variable DFRESULT). The results of this are shown in Figure F i g u r e 7 . 3 . 3 b : P l o t o f M e a n s N E W G G P 3 M a i n E f f e c t F ( 2 , 5 9 ) = 1 1 . 0 3 ; p < 0 0 0 1 N E W G G P 3 Figure 7.3.3b: Result of ANOVA test of effectiveness of NEWGGP3 in predicting debris flow initiation from failure. As NEWGGP3 was calculated for the failure data set (see text), the test is against DFRESULT, which is most similar to GGPOOR of the failure data set (ref. Figures 7.3.1a and 2a) 7.3.3b. NEWGGP3 has a p-value of p<0.0001: better than for NEWGGP2 vs. GGPOOR, although the F-statistic is lower (due to the slightly different sample size and number of classes (H,M,L for NEWGGP3 vs. H and L only for NEWGGP2)). NEWGGP3 has a slightly higher fraction initiating debris flows in class H (probably the same when the effect of the different sample sizes is considered) than does NEWGGP2, and a lower fraction in L (25% in NEWGGP3 vs. 30% in NEWGGP2). However, NEWGGP3 also has an M class which NEWGGP2 does not. 137 The M class is the most problematic class of a three-class system because it is ambiguous, being neither low enough risk to allow harvesting without precautions, nor high enough risk to be excluded from harvesting. Thus, the inclusion of an M class in the NEWGGP3 classification system can be seen as a potential disadvantage of that classification from the viewpoint of forest management. 7.3.4 Conclusions concerning proposed revisions of GGP. Of the three classification systems proposed in this section, the first (NEWGGP 1, using TAN CG and ICSS), although it produced a statistically significant division of results, was based in part on the use of a parameter, ICSS, whose use was conceptually unjustified. The other two proposed classifications (NEWGGP2 [using TAN CG only] and NEWGGP3 [using TAN CG, LOCN2 and O]) produced results with a relatively similar degree of statistical significance; NEWGGP3 produced slightly better resolution at the upper and lower ends of its classification system, but used an intermediate classification which NEWGGP2 avoided, and was also based in part on an untested, though reasonable, hypothesis. For these reasons, the system used in NEWGGP2 becomes the most acceptable choice for a revised method of estimating the hazard of debris flow initiation from slope failure in a gully reach. For simplicity's sake, NEWGGP2 will be referred to as NEWGGP only in the following discussion. 7.4 Revised method of determining DFIP, and a test of NEWDFIP. Having created revised methods for determining GWFP and GGP, it becomes necessary to combine them in order to create a new DFIP. The GAP combines GWFP and GGP ratings on the principle of "lesser hazard dominating": a reach with high GWFP and low GGP gets an overall DFIP rating of low, and vice versa. This system (equal weighting of GWFP and GGP) is reasonable since there are no numerical evaluations of hazard associated with a class. In the revised DFIP, NEWGWFP and NEWGGP have numerical values (percents which failed or 138 produced a debris flow) associated with each class; these can be used to determine the appropriate weightings for each possible combination. The classes of NEWGWFP and NEWGGP are shown below, cross-tabulated in matrix form with the resultant percent of debris flow initiation computed for each possible combination (Table 7.4). The percentage values with failures and debris flows are taken from Figures 7.2.1a and Table 7.3.2: Table 7.4: Poss ib le new DFIP c lasses with expected fraction of each c lass exper iencing DF initiation (= N E W G W F P % x N E W G G P %) N E W G W F P c lass and % w/ fai lure N E W G G P c lass and % w/d.f . H (76%) M (57%) L (39%) H ( 8 1 % ) 0.62 0.46 0.16 L ( 3 5 % ) 0.27 0.2 0.14 If one translates the subjective H, M, and L classes of the DFIP assessment into expected percentage chances of debris flow initiation, it seems reasonable that H should be >50% (preferably higher), M in the -40% range (below one-half but greater than one-third) and L < ~25%>. The numbers in Table 7.4 then suggest that the system used for assigning DFLP class from GWFP and GGP may also be used in the revised DFLP, producing the following proposed NEWDFLP classification system: NEWGWFP H and NEWGGP H: NEWDFLP H NEWGWFP M and NEWGGP H: NEWDFLP M NEWGWFP L or NEWGGP L : NEWDFLP L. (note that there is no NEWGGP M class) The ANOVA test of this proposed classification system (NEWDFLP) vs. DFOOR and DFTOFAN for the set of reach data is shown in Figures 7.4. la and lb. Compare these to Figures 7.4.2a and 2b, which show the performance of DFLP in the same tests. NEWDFLP has a higher degree of confidence than did DFIP: F-statistics are higher and p-values are lower. Once again, testing against DFOOR produces a higher degree of significance than does testing against 139 DFTOFAN. NEWDFLP is unexpectedly similar to DFLP in that class M and H behave in a similar fashion to each other, with approximately the same chance of debris flow initiation (it is about 10% higher for class M), while class L is much less likely to have debris flow initiate in a reach than the M or H classes. This similarity of the M and H classes was not predicted in Table 7.4. The difference between the two results comes from the fact that the '% w/DF' numbers in Table 7.4 come from a test against GGPOOR, while those in Fig. 7.4.1a are from a test against DFOOR, which includes reaches without failures, resulting in slightly different numbers. Thus, in Table 7.2, the NEWDFLP M class is considered to have a 57% chance of a slope failure, and an 81%> chance that that failure will produce a debris flow, for an estimate of overall chance of debris flow of 46%. However, in the actual reach data, 10 out of 16 of the NEWDFLP M reaches have a slope failure (63%), and 100% of those failures produce debris flows travelling out of the reach, so the actual performance of the NEWDFLP M class is 63% in Figure 7.4.1a. Table 7.5- Numbers and %'s of Reaches in Each DFIP Class C l a s s No. DFIP % DFIP No. N E W D F I P % N E W D F I P H 49 5 3 . 3 % 34 37 .0% M 18 19.6% 16 17.4% L 25 27 .2% 42 45 .7% The distribution of reaches in NEWDFIP is compared to that of DFLP in Table 7.3 and Figure 7.4.3. As seen in Figure 7.4.3, a bivariate histogram of DFLP and NEWDFIP, the NEWDFIP class L includes all the reaches which were classified as L under the DFLP classification, plus some which were M and some which were H under DFLP. The NEWDFIP H class is mostly DFLP H reaches, plus some DFIP M reaches; the NEWDFIP M class is mostly reaches which were DFLP H, plus a few which have stayed M during the reclassification. NEWDFIP has a higher percentage of L reaches than did DFIP, a lower percentage of H reaches, and the percent of reaches rated M is almost unchanged between the two systems; overall, about 70% of reaches in 140 F i g u r e 7 . 4 . 1 a : P l o t o f M e a n s N E W D F I P M a i n E f f e c t F ( 2 , 8 9 ) = 3 . 8 9 ; p < 0 2 4 1 O J 0 . 6 5 [ • 0 . 2 5 1 > '< L H M L N E W D F I P Figures 7.4.1a and 2a: Results of ANOVA tests of effectiveness of NEWDFIP and DFLP in predicting debris flow out of the reach of initiation. 141 F i g u r e 7 . 4 . 1 b : P l o t o f M e a n s N E W D F I P M a i n E f f e c t F ( 2 , 8 9 ) = 2 . 5 2 ; p < . 0 8 5 9 0 . 5 | , , : 1 0 . 2 1 '< '< = — L H M L N E W D F I P F i g u r e 7 . 4 . 2 b : P l o t o f M e a n s D F I P M a i n E f f e c t F ( 2 , 8 9 ) = . 4 5 ; p < . 6 3 8 1 0 . 4 l : , 0 26 I • • • H M L D F I P Figures 7.4.1b and 2b: Results of ANOVA tests of the effectiveness of NEWDFIP and DFLP in predicting the initiation of a debris flow travelling to the fan from the reach in question. 142 Figure 7.4.3: Bivariate Histogram of NEWDFIP and DFIP Figure 7.4.3: Bivariate histogram of DFIP and NEWDFIP showing numbers of reaches in each old and new class combination. DFLP are rated H or M, while only about 55% of reaches are so rated in NEWDFIP. The reaches rated H and M in NEWDFIP have a higher fraction which initiate debris flow than do their counterparts in DFLP; the reaches rated L in NEWDFIP have almost the same chance of producing a debris flow as do the reaches rated L in DFLP, despite there being over 50% more reaches in NEWDFIP class L than in DFLP class L. These points together indicate that the proposed NEWDFIP classification is more effective at discriminating between reaches at high risk of debris flow, and reaches where the risk is low, than is the DFLP classification. Given the similarity of the NEWDFIP H and M classes in terms of performance, it seems that the two classes should be combined and given an overall rating of H, as both classes have a 143 greater than 50% chance of debris flow initiation. This would also have the effect of shrinking the three class system to a two class system, allowing more effective management of gully systems for timber harvesting by removing the middle class and creating a high risk- avoid harvesting/ low risk- harvest acceptable split. 144 Chapter 8 - Regional variation in observed parameters and results. 8.1 Observed regional variation in gully morphology and location of debris flow initiation. As described in Chapter 3.7, the Chilliwack Valley and Norrish Creek regions appeared to have distinctive gully morphologies, and the locations in which gully wall failures initiated were distinctly different for the two differing morphologies. Gullies in Chilliwack Valley generally had a distinct headwall at their upper boundary (as shown in Figure 1.1); in many cases the gully appeared to be the surface expression of a bedrock feature, as shown in Figure 3.6.1. Slope failures in these gullies were typically sidewall failures. In contrast, gullies in Norrish Creek often lacked a distinct headwall; instead, the gully tended to lose definition at its upper boundary, becoming an open-slope depression that gradually merged into the surrounding slopes, as shown in Figure 1.2. Norrish Creek gullies generally did not seem to be the surface expressions of bedrock topography, but rather to be developed in surficial material alone, as shown in Figure 3.6.2. Slope failures in Norrish Creek gullies frequently occurred in the open-slope depression at the gully head. This chapter attempts to determine if the observed variation in location of slope failures between the two regions is a result of the differing gully morphologies, or if results from some other factor which also varies significantly between the two regions, such as location of logging activity with respect to the gully. 8.2 Statistically significant regional variation at the reach scale. During the last phase of analysis, once the data had been examined to determine the important factors affecting debris flow variation, the data were re-examined with regard to regional variation. Statistical confirmation was sought for the subjective observations described above. At the reach scale, the only parameters which vary significantly (at the confidence lavel cc= 0.05) between Norrish Creek and Chilliwack Valley are SURF (and SURFFL; PASTFAIL is not 145 significant), NEWGGP and NEWDFIP. The statitics for the ANOVA tests are shown in Table 8.1. T a b l e 8.1: Resu l ts of A N O V A tests for variation in R E G B A S I N by var iable Var iab le F-stat. deg. free. p-value M e a n R E G B A S I N for e a c h var. S U R F 4.48 (2,89) <0.014 M=1.5, C=1.3, R=1.0 N E W G G P 7.69 (1,90) <0.0067 L=1.21, H=1.51 N E W D F I P 3.16 (2,89) <0.047 L=1.28, M=1.56, H=1.53 Note: REGBASIN=1 for Chilliwack and 2 for Norrish, so that a mean REGBASIN result of 1.5 (for instance) means that the variable in question is evenly distributed between the two basins. These results are interesting: SURF contributes to NEWGWFP, but the variation in NEWGWFP between the two regions is not significant at the a=0.05 level. Likewise, channel gradient is the only contributor to NEWGGP, but mean CG itself does not vary significantly between the two areas. What varies is the distribution of CG around the TANCG = 0.5 threshold (Figure 8.1.1). Although the mean TANCGs of both regions are just over 0.5, and the difference in means is not statistically significant (F(l,90)= 1.05, p<0.31), a far greater propportion of the total number of reaches have a TANCG > 0.5 in the Norrish Creek region than do those in the Chiliwack Valley. The variation in NEWDFIP is due to the variation in NEWGGP, its dominant contributor. That the variation in NEWDFIP is not as significant as it is for NEWGGP is a result of the contribution of NEWGWFP to NEWDFIP. The variation in SURF is due to the absence of rock-walled reaches in the Norrish Creek area, and the relatively higher proportion of till with respect to colluvium there (Figure 8.1.2). The variation in SURF does not influence NEWGWFP significantly because the effects of SURF are greatest for rock-walled reaches only; till and colluvium are treated the same in NEWGWFP. The proportion of rock walled reaches is small enough that their uneven regional distribution does not influence NEWGWFP significantly. The variations in NEWGGP and NEWDFIP are not reflected in a similar variation in either GGPOOR nor DFOOR, and likely represent a statistical artifact (caused, as described above for channel 146 F i g u r e 8 . 1 . 1 : H i s t o g r a m o f T A N C G c a t e g o r i z e d b y R E G B A S I N 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 REGBASIN: chwk 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 REGBASIN: norrish T A N C G ( t a n g e n t o f c h a n n e l g r a d i e n t ) Figure 8.1.1: Histograms showing distribution of TANCG in each studied region. F i g u r e 8 . 1 . 2 : H i s t o g r a m o f S U R F c a t e g o r i z e d b y R E G B A S I N C REGBASIN: chwk C REGBASIN: norrish S U R F ( s u r f i c i a l m a t e r i a l o f g u l l y w a l l s ) [ M = t i l l , C = c o l l u v i u m , R = r o c k ] Figure 8.1.2: Histograms showing the distribution of gully wall surficial materials in each studied region. 147 gradient, by the application of a boundary based on a mean value to a group not evenly distributed about that mean) rather than a true measure of variation. Thus, the classification system varies significantly between the two regions in the study area, but the observed results, on which the classification system is based, do not so vary. This emphasizes that the relationships found to be significant for the whole data set may perform differently when regional subsets are considered, and implies that applicability may be partly site-specific, a disappointing conjecture. Of the other parameters, some, which do not show significant variation in their mean values, do show variation in their extremes; for instance, mean GWSD does not vary significantly between the two regions (F(l,90)= 1.02, p<0.32), but the majority (13 out of 17) of the reaches with GWSD > 30m are located in the Chilliwack Valley (Figure 8.1.3). The subjective observation that two separate gully types, each with a distinctive location of initiation of debris flow, exist in the study area is neither supported nor contradicted by these results. Figure 8.1.3: Histogram of GWSD categorized by REGBASIN 25 20 15 in n 0 1 1° m i 0 10 20 30 40 50 60 70 80 90 100 110 0 10 20 30 40 50 60 70 80 90 100 110 REGBASIN: REGBASIN: chwk norrish GWSD (meters) Figure 8.1.3: Distribution of GWSD in each studied region. Note that it was shown in Chapter 7 that longer sidewalls are associated with colluvium than with till or rock, and there is more colluvium present in Norrish Creek (compare this figure to Fig. 8.1.2 above) 148 During fieldwork, reaches which included gully head zones were not specifically recorded as such unless a failure occurred therein. Thus, unfortunately, it is difficult to distinguish between reaches which include gully heads and those which, while located high in gully systems, do not include gully heads when examining the reach data. Locations were recorded for failures, but these results are not directly transferable to the reach scale, except as minimum values, for two reasons. First, the location recorded was the location of the failure. Thus, a failure at the gully head would have been recorded as 'gh', but a failure on the sidewall would have been recorded as 'gs', even if the sidewall failure occurred in a gully head reach. Secondly, reaches without failures do not appear in the failure data, so any head reaches without failures would not be included should the data on failure locations be transferred to the reach scale. With these two points in mind, an examination of the failure location at the reach scale can still give us an estimate of the minimum number of gully head reaches occurring in each region. Twenty-two out of the 62 failures observed were gully head failures; each occurred in a separate reach, so there are a minimum of 22 head reaches out of 92 reaches in the study. The distribution of gully reaches and head reaches by region is shown in Table 8.2: T a b l e 8.2: Distribution of R e a c h e s and Head R e a c h e s By Reg ion Reg ion # reaches # head Min. % head rch fai lures Norr ish Cr. 39 17 43 .6% Chi l l iwack 53 5 9.4% Total 92 22 2 3 . 9 % Even considering that these are only minimum values for the number of gully head reaches in each region, the disparity between the two regions is great: almost certainly, there are more gully head failures in Norrish Creek than there are in Chilliwack Valley. If there are a greater number of gully head failures in Norrish Creek, the next question to be asked is why might this be the case. Specifically, with reference to section 8.1, is it the 149 differences in lithology and surficial materials between the two regions which have influenced the locations of slope failure and debris flow initiation in gullies therein, or some other factor? Unfortunately, there is still not enough information at this stage to provide a definite answer. Two unfortunate confounding factors are the difference in relief and logging between the two regions. Stated simply, the overall lengths of slopes in the Norrish Creek region are shorter than they are in the Chilliwack Valley region; in the Norrish Creek area, ridgetops are below treeline, and logging has occurred right up to ridge crests, while in the Chilliwack Valley area, logging has not yet reached treeline in many areas, and some gullies extend to above treeline in any event. The result of these two factors is that many of the gully systems in the Norrish Creek region have had their heads logged, while many in Chilliwack Valley region have not. Thus, it is not possible to determine if the differences in minimum numbers of gully heads with failures observed between the two regions is due to differences in lithology and surficial materials, as initially suspected, or due simply to differences in frequency of logging of gully heads between the two regions. 8.3 Statistically significant regional variations in failure data. At the scale of individual failures rather than that of reaches, over which parameters represent an at-a-site value rather than the mean conditions in a reach, the regional variation in the data becomes much more significant. Gully drainage area (as variables BASAREA and LNBAS) is the first parameter to show significant variation (Figure 8.3.1a and lb). The ANOVA results for BASAREA and LNBAS vs. REGBAS are given in Table 8.3: T a b l e 8.3: A N O V A results for regional variation in bas in a rea var iab les Var iab le F-stat. deg. free, p-value M e a n s by A r e a B A S A R E A 10.23 (1,60) O . 0 0 2 2 Chwk=2.4 ha, Norrish=0.8 ha L N B A S 8.09 (1,60) O . 0 0 6 1 Chwk=9.5, Norr ish=8.75 The mean drainage area of the Chilliwack area gullies is almost three times greater than that of the Norrish gullies (Figure 8.3. la) and the range of drainage areas is wider for the 150 F i g u r e 8 . 3 . 1 a : H i s t o g r a m o f B A S A R E A c a t e g o r i z e d b y R E G B A S I N _Q O «*— O o 25 20 15 10 0 1 2 3 4 5 6 7 8 9 10 11 0 1 2 3 4 5 6 7 8 9 10 11 REGBASIN: REGBASIN: chwk norrish B A S A R E A ( g u l l y d r a i n a g e a r e a i n h e c t a r e s ) F i g u r e 8 . 3 . 1 b : H i s t o g r a m o f L N B A S c a t e g o r i z e d b y R E G B A S I N 12 10 « 6 o 1.— o 0 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 6 7 8 9 10 11 12 6 7 8 9 10 11 12 REGBASIN: REGBASIN: chwk norrish L N B A S ( n a t u r a l l o g . o f g u l l y d r a i n a g e a r e a i n s q . m e t e r s ) Figures 8.3.1a and lb: Histograms of basin area for each studied region; Fig. 8.3. la shows the distribution of basin area, in hectares, and Fig. 8.3.1b shows the distribution of the natural logarhitm of basin area, in square meters. (Thus, values in Fig. 8.3.1b are equal to ln ([value in 8.3.1a]*10 000)).The two chosen scales of presentation, BASAREA and LNBAS, allow the best contrast in resolution. 151 Chilliwack area (Fig. 8.3.1b). This is a function of the much greater local relief in the Chilliwack area: the gullies are longer, and the drainage area is a function of gully length (area was calculated by summing reach length times twice GWSD for each reach, and checked against TRIM maps). This effect is also visible in the next parameter to show significant regional variability: SLPPOS, slope position of the failure (Figure 8.3.2). The Norrish failures are much more likely to occur near the head (the position just below slope apex) or on the uppermost slopes, while the Chilliwack failures occur mostly at mid-slope and below. This is partly an effect of the scale of local relief, as mentioned above, and partly as a result of logging history. In the Chilliwack Valley drainages, there are significant alpine areas along ridge crests, and most logging has taken place lower down on these slopes partly because there may be no trees on the upper slopes. The lengths of the relative positional zones also increases with relief: on a 700 m valley slope in the Chilliwack Valley, the lower, mid, and upper slopes may each occupy 200-250 vertical meters, where on a Norrish slope of 400m total relief, the vertical range of each positional zone may be only 100-150m. As logging has occurred over similar time scales in both areas, the net result is that in Norrish Creek logging has reached the ridgetops, while there is as yet a strip of a few hundred meters between the upper clearcut boundary and treeline on the longer slopes in the Chilliwack Valley. Slope configuration (SLPCONFG) also varies significantly between the two regions (Figure 8.3.3): dissected slopes ('di'), or slopes with single gullies ('sg') are found in both regions, but broken ('brk'), benchy ('be'), irregular ('irr') and faceted ('fac') slopes are only found in the Chilliwack valley. This is not surprising: the Chilliwack valley is located in an area of heterogeneous, dominantly non-plutonic rocks, so there is more likelihood that changes in lithology will have a surface expression in Chilliwack basin than there is that such changes will occur in the more lithologically homogeneous Norrish Creek basin. 152 F i g u r e 8 . 3 . 2 : P l o t o f M e a n s S L P P O S M a i n E f f e c t F ( 5 , 5 6 ) = 6 . 5 1 ; p < 0 0 0 1 2 - 2 | , : '. , ro > 0 . 8 1 1 ' ' i i i 1 h d u p u p / m i d m i d m i d / l o l o w S L P P O S Figure 8.3.2: Results of A N O V A test of degree of regional variation in SLPPOS, showing f statistic, p-value, and plot of mean R E G B A S I N for each SLPPOS class. In this and following plots, R E G B A S I N of 1 = Chilliwack and 2 = Norrish, so a R E G B A S I N value of 1.5 indicates that the class is equally distributed between Norrish Creek and Chilliwack Valley. F i g u r e 8 . 3 . 3 : P l o t o f M e a n s S L P C O N F G M a i n E f f e c t F ( 5 , 5 6 ) = 3 . 4 7 ; p < 0 0 8 4 1.8 [ ! ! • . . 1 0 . 9 d i b r k s g b e i r r f a c S L P C O N F G Figure 8.3.3: Result of A N O V A test of regional variation in SLPCONFG, showing F stat., p-value and plot of means. 153 Figure 8.3.4: Plot of Means CURVE Main Effect F(2,59)=4.34; p<0175 1.75 r '. : — , , L25 1 ' : ' i i straight concave convex CURVE Figure 8.3.4: Result of ANOVA test of regional variability of horizontal slope curvature (CURVE). Horizontal slope curvature (CURVE) varies significantly with REGBASIN, though the result is not as easily interpretable as it was for slope position of failure or gully basin area (Figure 8.3.4). Concave slopes are most common in the Chilliwack region, convex slopes are slightly more common in the Norrish area than in the Chilliwack Valley, and straight slopes are dominantly found in the Norrish area. This differing distribution may reflect the underlying lithological differences between the two areas (and their resulting influence on drainage patterns, in turn influencing the horizontal curvature of valley walls), or it may reflect the differing history of glaciation between the two areas, in the following manner: in Norrish Creek, which was glaciated by thicker ice for a longer period of time, valleys are dominantly U-shaped, whereas in Chilliwack Valley they are U-shaped in the upper reaches and become V-shaped lower down. It seems possible that straight valley slopes (those without curvature in the horizontal plane) are more frequently found in U-shaped valleys, where walls have been ice-shaped, than they are in V-154 Figure 8.3.5: Plot of Means DRAIN Main Effect F(3,58)=3.16; p<0314 (i=imperfectly, m=moderately, w=well, r=rapidly drained) DRAIN Figure 8.3.5: Result of ANOVA test of regional variability in soil drainage (DRAIN), shaped valleys, which have been eroded by stream flow (due to the greater sinuousity of water flow as compared to ice flow). Soil drainage (DRAIN) also (somewhat unexpectedly) varies significantly with region (Figure 8.3.5). Despite the greater frequency of colluvium in the Chilliwack basin, and the presence of the "broken" rock subgroup of LITHOL4 therein, imperfectly and moderately drained slopes are much more common in the Chilliwack region than they are in Norrish Creek, while fast-draining slopes are slightly more common in the Norrish Creek region than they are in Chilliwack Valley. As was noted in Chapter 6.6.2, however, the relationships between observed seepage, drainage, surficial materials and lithology are ambiguous and partly contradictory, and are probably beyond the scope of this study to fully investigate or explain. As is noted in Chapter 9, this topic is worthy of further research. Examining the variable LITHOL2 (which subdivides observed bedrock lithologies into igneous, sedimentary and metamorphic groups) shows the drastic lithologic difference between Figure 8.3.6: Plot of Means LITHOL2 Main Effect F(2,59)=99.56; p<.0000 0.8 sediment metamorp igneous LITHOL2 F i g u r e 8.3.6: Result of ANOVA test of regional variability of general bedrock litholgy (LITHOL2). [Note the F-statistic for this test. There is no doubt whatsoever that lithologies vary between the two areas.] Chilliwack and Norrish drainages (Figure 8.3.6). All the sedimentary and metamorphic rocks are found in the Chilliwack Valley, while except for the one gully developed on andesite (TAM-001-RB) and the three gullies on granodiorite in Center Creek, all igneous gullies are in Norrish Creek. The significant regional variation (F(2,59) = 5.9, p<0.0046) in failure location (as LOCN2; Figure 8.3.7) is predominantly caused by a high proportion of gully head failures in Norrish Creek, and a similarly high proportion of sidewall failures in Chilliwack Valley. This regional variation in failure location significantly affects regional mean A O E (F(l,60)=7.85, p<0.0068); mean a for Chilliwack valley, where most failures are sidewall failures, is 42°, while in Norrish Creek, where gully head failures outnumber sidewall failures, mean a is only 25°. As might be expected, logging date also significantly varies between the two regional basins when considered year by year (F(9,52)=2.28, p<0.0309). When average date of logging is considered for the two basins, however, there is little variation between the two means: mean 156 logging date is 1985 for both Chilliwack and Norrish Creeks. This is a result of the two regional basins being composed of a number of sub-basins, most of which were logged at different times. Figure 8.3.7: Histogram of LOCN2 categorized by REGBASIN 30 25 20 15 10 1 1 ! I M P head channel REGBASIN: chwk side head channel REGBASIN: norrish side LOCN2 F i g u r e 8.3.7: Histograms showing distribution of failure location in each studied region. Thus, in any one year, logging in a Chilliwack basin might take place, while it might not in any Norrish basins, thus producing a significant regional variation for that year. Over time, these variations balance out, so when the two basins are considered over the ten year period of interest, mean logging date does not vary significantly. Although mean failure volume (VOLUME) differs greatly between the two basins (the average failure volume in Chilliwack is almost three times as large as it is in Norrish Creek), the variation is not significant. The difference in mean VOLUME is due to the fact that both of the largest initial failures (AP-004-RB-F01 and TLM-002-LB-F01) were in the Chilliwack Valley region: as these failures are over ten times larger than the mean volume of the other failures, they skew the mean values when they are included. When a normally distributed measure of volume 157 (LNVOL) is tested, no significant variation in mean failure volume appears between the two basins. The fact that DFRESULT does not vary significantly with regional variation is at first puzzling, when other factors which significantly influence DFRESULT (such as LOCN2) have been shown to vary. However, since several parameters vary at the regional scale, and the effects of their variation do not vary in the same manner with regard to influence of DFRESULT (for instance, Chilliwack Valley has higher mean AOE, but also has wetter slopes [mean SEEP is lower]), it is probable that the effects of significant variations cancel each other out when DFRESULT is considered at the regional scale. With regard to the sediment samples collected, the variation in D 5 0 by region is significant (F(l,37) = 4.73, p<0.036), although the difference in means is not great (0.1 mm difference between mean Chilliwack and Norrish D50's, with Chilliwack area the coarser). There is no corresponding variation in % fines. Clast lithology, though not recorded in the sediment data collection process, was observed to vary with region, as might be expected when the two bedrock lithologies differ so greatly. With regard to the assertion advanced in section 8.1, namely that there was variation in the morphology of gullies and characteristic failure styles between the two regions, the results of the significance tests performed in this section seem to indicate that such variations truly exist, but are likely not related. Specifically, the significant variations in lithology and failure location observed for the failure data set back up the qualitative observation that such parameters differ between the two regions. However, as was pointed out in section 8.2, the differing proportions of logging activity on gully heads between the two regions means that one should not conclude that the observed difference in style and pattern of slope failure into gullies is due to lithological and surficial material variation; in fact, the logging of the gully heads in the Norrish Creek area is 158 probably the prime cause of the greater proportion of gully head failures observed there relative to the Chilliwack area. This also explains why the Norrish Creek gully heads, which are moderately steep open slope depressions, have more slope failures than do the Chilliwack gullies, which have steep headwalls. In summary, the gullies of the Chillwack Valley and Norrish Creek areas differed in morphology. Many of the Chilliwack Valley area gullies were surface expressions of a bedrock feature, as shown in Figure 3.6.1 of Chapter 3; in cases where the gully heads were visited, many were of the steep headwall type, as shown in Figure 1.1 of Chapter 1. In Norrish Creek, most of the gullies were developed in surficial materials only, with little bedrock expression, as shown in Figure 3.6.2 of Chapter 3, and most of the gully heads were of the open-slope depression type, as shown in Figure 1.2 of Chapter 1. Initially, it seemed that these regional variations in gully morphology might explain the variation observed in location of slope failures which triggered debris flows; however, the results of the closer examination performed in this chapter seem to indicate that it is most likely the variation in logging of the gully head region which has resulted in the greater number of gully head failures observed in Norrish Creek. The regional variations in distribution and type of lithology and surficial materials likely account for the regional variation in gully morphology, but have little to do with the regional variation in failure location. 159 Chapter 9- The GAP and Factors affecting Debris Flow Initiation: Conclusions. 9.1 Conclusions. This report has examined the factors affecting debris flow initiation in logged gullies of the Cascade Mountains and southern Coast Mountains of British Columbia. Initially, the method of assessing the potential for debris flow initiation (DFLP) used in the Gully Assessment Procedures (GAP) guidebook of the Forest Practices Code was tested to see how well it performed. The results of this evaluation indicated that the DFIP was functioning adequately, but that problems existed, most notably that the section designed to evaluate gully geometry potential [for slope failures to initiate debris flows] (GGP) was not working as it should, thus compromising the overall effectiveness of the DFIP assessment. The second stage of the analysis was to examine the data gathered at the scale of reaches and individual failures and to try to determine the most important factors influencing slope failure and debris flow initiation from slope failure at each scale. A number of parameters were found to be significant at each scale. Further investigation showed that some of these parameters were functions of others, and that some could only be measured after failure had occurred, and thus could not be adequately assessed from pre-failure conditions. Some of these parameters, though significant in influencing debris flow initiation, might themselves be surrogates for other parameters already measured. Once these parameters were eliminated, those that were left were then combined into new evaluations of gully wall failure potential and gully geometry potential. NEWGWFP was based on three parameters (GWSD, GWSA and SURF). NEWGGP was based on channel gradient (as TANCG) only. These new parameters were tested against their respective observed results and found to be of greater significance than the 'old' GWFP and GGP of the GAP. The NEWGWFP and NEWGGP were then combined to produce a new DFLP which was tested against the observed occurrence of debris flow initiation. It was found that despite the 160 improved performance of its new constituent parameters, NEWDFIP had similar performance for classes H and M. However, these two classes had significantly higher rates of debris flow initiation than did the L class, and the behaviour of the H/M vs. L classes was statistically of much greater significance than were the behaviour of the same classes in the original DFLP assessment. For these reasons it is recommended that the new GWFP and GGP be used as described in Chapter 8, but that the new DFIP be reduced to a two-class (High and Low) system only. 9.2 Further recommendations to improve performance of the GAP. In addition to the recommendations made above with regard to the new GWFP, GGP, and DFIP, the following suggestions are made of possible ways to improve the accuracy and performance of the GAP assessment of debris flow initiation potential: - In a reach in which a gully head is located, two GWFP assessments should be made, one for the head and one for the sidewalls, and the higher assessed GWFP should be used in assessing the DFIP. This step is implicit in the current GAP, but it should be made explicit. Regarding this comment, it seems appropriate to ask the question, do threshold values for slope failure differ between gully heads and sidewalls? Although this question cannot be definitively answered due to the difficulty, described in Chapter 9, of translating locations from the failure data set to the reach data set, at which scale GWFP is tested, there is enough subjective evidence1 to say that it seems that gully heads are slightly more likely to fail at a given angle than are gully sidewalls, and that gully heads are also more susceptible to failure if slope length is low than are gully sidewalls. Given these observations, and lacking any more concrete data, it is suggested that the minimum GWFP rating for gully heads be set at M. 1 Personal observation, and a test of the 'minimum number of gully head reaches' data from Chapter 8. 161 - In reaches with rock sidewalls (SURF = 'R') , an assessment should be made of the failure potential of any surficial materials located on top of these sidewalls, that is, outside of the gully proper, but still located such that failures in the material in question could enter the gully. This recommendation is made based on the observation that in the eight gully reaches with rock sidewalls in the study, the sidewall failures in the three reaches which had failures involved surficial material failures on 'open slopes' outside the inner gully system. - More generally, an assessment of the stability of the open slopes outside the gully proper, but still able to contribute material to the gully, should be made and included in, or appended to, the assessment of GWFP. Several cases were noted in this study of failures which began outside the gully on open slopes, travelled outside the gully channel on the open slope for a distance, then entered the gully, producing debris flows; these failures were larger than the typical gully failure, and apparently had a higher probability of initiating a debris flow than did failures which began within the gully system. - In this study, the final method chosen for evaluating GGP used only channel gradient. However, there was evidence that further improvement could be achieved by using a method to predict the angle of entry of failures into the channel. Geometric considerations showed that a measurement of the planimetric angle made by the fall line of the gully sidewall and the gully channel (Q) could likely be used to so predict angle of entry with a higher degree of confidence than was possible in this study. If a further study determines that Q can be so used, then the method of evaluating GGP which used Q (NEWGGP3 in Chapter 7) should be used in place of the proposed new GGP system proposed here. This would mean that the new DFIP system would have to be re-evaluated to assess the differing contributions of the newer GGP system, and might lead to a return of the three-class (H, M , L) DFIP, with an increase in the chance of debris flow initiation for the H class relative to the M class. 162 9.3 Limits of confidence in conclusions. The greatest problem with the study design was that gullies without failures were not included in the study. Thus, the overall frequency of gully reaches with moderate and low debris flow initiation potential may have been underestimated, and the initiation potential of such reaches overestimated as a result. An attempt was made to compensate for this lack by evaluating debris flow initiation potential of gullies not included by means of field notes, TRIM maps, and aerial photographs. The results of this attempt suggested that that there may not have been a great difference between the gullies which were selected for the study and those which were excluded. However, this does not mean that no bias exists in the study. The gullies included in this study had all been logged between six and fifteen years prior to the study. During that time, two large storms had occurred. Thus, climatic and temporal requirements for maximum vulnerability to slope failure were met in the study area. However, this also meant that the study took place after logging, whereas the GAP is primarily designed to take place before logging. Although none of the geometric and geological parameters included in the revised evaluations were ones which would change significantly from pre-logging conditions to post-logging conditions, nonetheless this (reasonable) assumption that post-logging conditions approximate pre-logging conditions was untested, creating another possible loss of confidence in the results. Finally, an attempt was made to exclude any influence that logging roads might have exerted on the study by not including gully reaches with obvious signs of road influence (for instance, drainage ditches diverting water into the gully). As well, all failures located below a road and within 25 m of that road were excluded. Unfortunately, given the intensity of logging and road building in the studied areas, it was not possible to exclude all failures located below logging roads from the study. Thus, a possibility exists that roads in the study area did make 163 some unnoticeable contribution to some of the failures observed. However, the overall influence of roads on the failures and gully reaches studied should be minimal. Despite these three caveats, the conclusions the study reaches are reasonable, and the results obtained seem to generally agree with both established theory and previous studies on debris flow in gullies. For these reasons, the overall confidence in the significance of the results of the study should be high. 9.4 Suggestions for further research. Several questions raised during the process of this study seem deserving of further research, in that they would produce results which would help to clarify some of the questions posed in this study, and might also produce further improvement in the method of assessment of debris flow initiation potential. The first subject worthy of greater attention is the issue of initial failure volume. This was shown to be of significance in influencing whether or not slope failures would become debris flows, yet was very difficult to estimate from pre-failure parameters. It would seem that the development of some sort of general frequency-magnitude relationship (perhaps subdivided by a gully type rating defined on the basis of similar terrain parameters), as per Jakob (1996), would be an obvious avenue to pursue. In this study, most failures had initial volumes of 2000 m3 and less, with the exception of two failures which had initial failure volumes of just over 15000 m3. (Both of the latter were open slope failures which entered gully systems near their heads.) There are two possible explanations for this volume gap. One is that, due to the scales of the gullies in question, the intermedate sizes of failures were simply unlikely to occur and thus were not observed. The other possibility is that the 'volume gap' observed represents some sort of scale break phenomenon, and that the larger failures are representative of a different set of processes than are the small failures, processes which are possibly unrelated to the effects of forest harvesting. 164 Further research could resolve this issue by sampling failures whose volumes were between 2000 and 15000 m3, if they exist. The ability to successfully predict initial failure volumes could lead, not only to an improvement in the accuracy of the GGP, but to benefits in other areas of forest management, for instance with regard to design of structures such as culverts, where debris flow volume must be estimated. The second issue worthy of special attention is the prediction of angle of entry of sidewall failures into the gully channel. This was also shown to influence probability of initiation of debris flow, but was difficult to estimate from pre-failure parameters. The best regression equations developed from the data set had r2 values of approximately 0.5. A hypothesis was put forward that for cases of planar slopes, the angle of entry, a, of the failure should be equal to Cl, the planimetric angle at which the fall line of the gully sidewall intersects the centre line of the channel, except in cases where the failure is diverted away from the fall line. This assumption would be easily tested by measuring Q for a large number of reaches with failures and then determining at what angle a the failures in those reaches entered the channel. An obvious complement to this study would be to measure the same variables for a number of gullies about to be logged, and then to return to the study area after a ten-year or fifteen-year lag, during which one or more large storms would have had to have occurred. This would then allow the hypothesis that this study operated under, namely that the pre-failure, pre-logging parameters are not significantly changed when measured post-logging and post-failure, to be tested. It is also possible that the interval of waiting could be avoided through the use of an already collected data set (or sets) from a previous study (or studies), performed before the study area was logged. Finally, the question of whether or not the observed regional variation in lithology and surficial material type influenced debris flow initiation and gully morphology was not answered 165 completely. In Chapter 9, it was concluded that regional lithologic variation (possibly modified by surficial material type) was the primary influence on regional variation in gully morphology, and that variation in logging location relative to the gully system was the primary influence on the observed regional variation in location of debris flow initiation. It seems likely that any influence of gully morphology on location of debris flow initiation was masked by the influence of logging. Further research could clarify this subject by studying gullies of differing morphology with similar logging pattern, thus isolating the effect of morphology on flow initiation location; this study would also allow confirmation of the effect of lithology on gully morphology. Another possibility for research would be to isolate the effect of surficial material variation on gully morphology by studying gullies developed on similar lithologies, but with a wider range of surficial materials than existed in this study, where most colluvium was closely related to the till from which it had developed. 166 References Alley, N.F. and Thomson, B. 1978. Aspects of environmental geology, parts of Graham Island, Queen Charlotte Islands. B.C. Ministry of Environment, Resource Analysis Branch, Bulletin No. 2. 65p. Anonymous, 1995. Gully Assessment Procedures guidebook; Forest Practices Code of British Columbia. B.C. Ministry of Forests, 40p. 1995. Mapping and assessing terrain stability guidebook; Forest Practices Code of British Columbia. B.C. Ministry of Forests, 34p. Benda, L. and Cundy, T.W. 1990. Predicting deposition of debris flows in mountain channels. Canadian Geotechnical Journal, 27: 409-417. Benda, L. and Dunne, T. 1987. Sediment routing by debris flow. In Beschta, R.L. et al. eds. Erosion and Sedimentation in the Pacific Rim. IAHS Publication no. 165: 213-223. Bovis, M.J. and Dagg, B.R. 1988. A model for debris accumulation and mobilization in steep mountain streams. Hydrological Sciences Journal, 33: 589-604. 1992. Debris flow triggering by impulsive loading; mechanical modelling and case studies. Canadian Geotechnical Journal, 29: 345-352. Bovis, M.J. and Millard, T.H. 1992. Precipitation thresholds for debris flow activity in the southern Coast Mountains of British Columbia. In Geohazards '92: Geotechnique and natural hazards, p.309. Bovis, M.J., Millard, T.H. and Oden, ME. 1997. Gully processes in coastal British Columbia: the role of woody debris. Unpublished report, 29 p. Brown, C.B. and Sheu, M.S. 1975. Effects of deforestation on slopes. Journal of the Geotechnical Engineering Division, American Society of Civil Engineers, volume 101 GT2: pp. 147-165 Caine, N. 1980. The rainfall intensity duration control of shallow landslides and debris flows. Geografiska Annaler, 62A: 23-27. Cannings, R.J. and Cannings, S.G. 1996. British Columbia: a natural history. Vancouver, Greystone Books, 310p. Chatwin, S.C., Howes, D.E., Schwab, J.W. and Swanston, D.N. 1991. A guide for management of landslide-prone terrain in the Pacific Northwest. B.C. Ministry of Forests, Land Management Handbook No. 34, 212p. Church, M. and Miles, M.J. 1987. Meteorological antecedents to debris flow in southwestern British Columbia. In Costa, J.E. and Wieczorek, GF. eds. Debris flows and avalanches. Geological Society of America, Reviews in Engineering Geology, 7, pp. 63-79. 167 Clague, J.J. 1989 Cordilleran ice sheet. In Fulton, R.J. ed. Quaternary geology of Canada and Greenland, Chapter 1. Geological Survey of Canada, Geology of Canada No. 1, pp. 40-42. Clague, J.J., and Shilts, W.W. 1993. Two landslide-dammed lakes in the Cascade Mountains, southwestern British Columbia. In Current Research, Part E, Geological Survey of Canada, Paper 93-1E, pp. 47-54. Costa, J.E. 1984. Physical geomorphology of debris flows. In Costa, J.E. and Fleisher, P.J. eds. Developments and applications of geomorphology. pp. 268-317. 1988. Rheologic geomorphic, and sedimentologic differentiation of water floods, hyperconcentrated flows, and debris flows. In Baker, V.R., Kochel, R.C. and Patton, P.C. eds. Flood geomorphology. pp.113-122. Crandell, D.R. 1963. Surficial geology and geomorphology of the Lake Tapps quadrangle. U.S. Geological Survey Professional Paper 388-A, 84p. Davis, N.F.G. and Mathews, W.H. 1944. Four phases of glaciation with illustrations from southwestern British Columbia. Journal of Geology, 52: 403-413 Dietrich, W.E. and Dunne, T. 1978. Sediment budget for a small catchment in mountainous terrain. Zeitscrift fur Geomorphologie, Supp. 29, pp. 191-206. Dietrich, W.E., Reneau, S.L., and Wilson, C.J. 1987.Overview: "zero-order basins" and problems of drainage density, sediment transport and hillslope morphology. In Beschta, R.L. et al. eds. Erosion and Sedimentation in the Pacific Rim. IAHS Publication no. 165:pp. 27-37. Dunne, T. 1980. Formation and control of channel networks. Progress in Physical Geography, 4: 211-239 Eisbacher, G.H. and Clague, J.J. 1981. Urban landslides in the vicinity of Vancouver, British Columbia, with special reference to the December 1979 rainstorm. Canadian Geotechnical Journal, 18: 206-216. Evans, S.G. 1986. Landslide damming in the Cordillera of Western Canada. In Landslide Dams: Processes, Risk and Mitigation, R.L. Schuster (ed.), Amer. Society of Civil Engineers, Geotechnical Special Publication, no. 3. pp. 111-130. Evans, S.G. and Lister, D.R. 1984. The geomorphic effects of the July 1983 rainstorms in the southern Cordillera and their impact on transportation facilities. In Current Research, Part B, Geological Survey of Canada, Paper 84-lb, pp. 223-235. Hogan, D.L. 1986. Channel morphology of unlogged, logged and debris torrented streams in the Queen Charlotte Islands. B.C. Ministry of Forests, Land Management Report No. 49, 94p. 168 Hogan, D.L. and Scwab, J.W. 1991. Meteorological conditions associated with hillslope failures on the Queen Charlotte Islands. B.C. Ministry of Forests, Land Management Report, No.73, 36p. Hogan, D.L., Rollerson, TP. and Chatwin, S.C. 1994. Gully Assessment Procedure for British Columbia forests (interim methods). B.C. Ministries of Environment and Forests, Watershed Restoration Technical Circular No. 5, 46p. Holland, S.S. 1964. Landforms of British Columbia, a physiographic outline. B.C. Ministry of Mines and Petroleum Resources, Bulletin No. 48. 138 p. Horton, R.E. 1945. Erosional development of streams and their drainage basins; hydrophysical approach to quantitative morphology. Geological Society of America Bulletin, 56:275-370. Howes, D.E. 1987. A terrain evaluation method for predicting terrain susceptible to post-logging landslide activity: a case study from the southern Coast Mountains. B.C. Ministry of Environment, Technical Report No. 28, 38p. Howes, D.E. and Kenk, E. 1988 (eds.) . Terrain classification system for British Columbia (revised edition). B.C. Ministry of Environment, Manual No. 10, 90p. Hungr, O. Morgan, G.C. and Kellerhals, R. 1984. Quantitative analysis of debris torent hazardsfor design of remedial measures. Canadian Geotechnical Journal, 21:663-677. Jakob, M. 1996. Morphometric and geotechnical controls of debris flow frequency and magnitude in southwestern British Columbia. Unpublished Ph.D. thesis, Department of Geography, UBC, 232 p. Jordan, R.P. 1987. Airplane Creek terrain and slope stability assessment. Unpublished report, 7p. Jordan, R.P. 1994. Debris flows in the southern Coast Mountains, British Columbia: dynamic behaviour and physical properties. Unpublished Ph.D. thesis, Department of Geography, UBC, 260 p. Krag, R.K., Sauder, E.A., and Wellburn, G.V. 1986. A forest engineering analysis of landslides in logged areas on the Queen Charlotte Islands, British Columbia. B.C. Ministry of Forests, Land Management Report No. 43, 138p. Mackin, J.H. 1941. Glacial geology of the Snoqualmie-Cedar area, Washington. Journal of Geology, 49, 5: 449-481. Millard, T.H. 1993. Sediment in forested and logged gullies, coastal British Columbia. Unpublished M.Sc. thesis, Department of Geography, UBC, 202 p. 1994. Terrain Attribute Study of the Chilliwack River drainage. Unpublished report. 1996. Personal communication. 169 Monger, J.W.H. 1970. Hope map-area, west half, British Columbia. Geological Survey of Canada, Paper 69-47. (map, 1:100 000 scale). Monger, J.W.H. 1989. Geology, Hope, British Columbia. Geological Survey of Canada, Map 41-1989, sheet 1. 1: 250 000 scale. Naumann, CM. 1990. The Cheam Slide: a study of the interrelationship of rock avalanches and seismicity. Unpublished M.A.Sc. thesis, Department of Geological Engineering, UBC, 203p. Naumann, CM. and Savigny, K.W. 1992. Large rock avalanches and seismicity in southwestern British Columbia. In Bell, D.H. (ed.), Landslides, Proc. 6th International Symposium, v.2, pp. 1187-1192. Oden, M.E. 1994. Debris recharge rates in torrented gullies on the Queen Charlotte Islands. Unpublished M.Sc. thesis, Department of Geography, UBC, lOOp. O'Loughlin, CL. 1972. A preliminary study of landslides in the Coast Mountains of southwestern British Columbia. In Slaymaker, O. and McPherson, H.J. (eds.) Mountain Geomorphology pp. 101-111. Pierson, T.C 1985. Initiation and flow behaviour of the 1980 Pine Creek and Muddy River lahars, Mount St. Helens, Washington. Geological Society of America Bulletin, 96: 1056-1069. Roddick, J.A. 1965. Vancouver North, Coquitlam, and Pitt Lake map areas, British Columbia. Geological Survey of Canada: Memoir 335, 276 p. and Map 1151A (1:253 440 scale) Rollerson, T.P. 1984. Terrain stability study - TFL 44. Land Use Planning Advisory Team. Woodlands Services, MacMillan Bloedel Ltd. 17p. 1992. Relationships between landslide attributes and landslide frequencies after logging: Skidegate Plateau, Queen Charlotte Islands, B.C. Ministry of Forests, Land Management Report No. 76, lip. 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, 55p. 1990. Site characteristics and landsliding in forested and clearcut terrain, Queen Charlotte Islands, British Columbia. B.C. Ministry of Forests, Land Management Report No. 64, 46p. Rudiak, C. 1996. Personal communication. 170 Ryder, J.M., Fulton, R.J., and Clague, J.J. 1991. The Cordilleran Ice Sheet and the glacial geomorphology of southern and central British Columbia. Geographie physique et Quaternaire, 45: 365-377. Sauder, E.A., Krag, R.K., and Wellburn, G.V. 1987. Logging and mass wasting in the Pacific Northwest with application to the Queen Charlotte Islands, B.C.: a literature review.B.C. Ministry of Forests, Land Management Report No. 53, 26p. Saunders, I.R., Clague, J.J. and Roberts, M.C. 1987. Deglaciation of Chilliwack River Valley, British Columbia. Canadian Journal of Earth Sciences, 23: 273-287. Sidle, R.C., Pearce, A.J., and O'Loughlin, CL. 1985. Hillslope stability and land use. American Geophysical Union, Water Resources Monograph 11, 140 p. Sidle, R.C. and Swanston, D.N. 1982. Analysis of a small debris slide in central Alaska. Canadian Geotechnical Journal, 19: 167-174. Slaymaker, H.O. 1988. The distinctive attributes of debris torrents. Hydrological Sciences Journal, 33: 567-573. Slaymaker, H.O., Hungr, O., Desloges, J., Lister, D.R, Miles, M.J. and VanDine, D.F. 1987. Debris torrent and debris flow hazards, Lower Fraser, Nicolum and Coquihalla valleys, British Columbia. 19th General Assembly, Intl. Union of Geodesy and Geophysics, Guidebook Excursion B2, 70p. Stathers, R.J., Rollerson, T.P. and Mitchell, S.J. 1994. Windthrow handbook for British Columbia forests. B.C. Ministry of Forests Research Program Working Paper 9401, 3 lp. StatSoft, Inc. 1993. STATISTICA Users' Manual. 559 p. 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, Department of Geography, UBC, 114 p. Swanson, F.J., Fredrickson, R.L., and McCorison, F.M. 1982. Mterial transfer in a western Oregon forested watershed. In Edmonds, R.L. (ed.) Analysis of coniferous forest ecosystems in the western United States, pp.233-266. Swanston, D.N. and Swanson, F.J. 1976. Timber harvesting, mass erosion and steepland forest geomorphology in the Pacific Northwest. In Coates, D.R. (ed.) Geomorphology and engineering, pp. 199-221. Takahashi, T. 1981. Debris flow. Annual Review of Fluid Mechanics, 13: 57-77. Thurber Engineering, 1988. Geotechnical assessment of slope stability in the Chilliwack Valley. Unpublished report to the Regional District of Fraser-Cheam, 30p. 171 Tripp, D.B. and Poulin, V.A. 1986a. The effects of mass wasting on juvenile fish habitat in streams in the Queen Charlotte Islands. B.C. Ministry of Forests, Land management Report No. 45, 48p. 1986b. The effects of logging and mass wasting on salmonid spawning habitat in streams in the Queen Charlotte Islands. B.C. Ministry of Forests, Land Management Report No. 50, 29p. 1992. The effects of logging and mass wasting on juvenile salmonid populations in streams on the Queen Charlotte Islands. B.C. Ministry of Forests, Land Management Report No. 80, 38p. VanDine, D.F. 1985. Debris flows and debris torrents in the southern Canadian Cordillera. Canadian Geotechnical Journal, 22, 44-68. Varnes, D.J. 1978. Slope movement types and processes. In Schuster, R.L. and Krizek, R.J. (eds.) Landslides, analysis and control. Special Report 176, Transportation Research Board, National Research Council, National Academy of Sciences, pp. 11-33. Waitt, R.B. 1975. Late Pleistocene alpine glaciers and the Cordilleran Ice Sheet at Washington Pass, North Cascades Range, Washington. Arctic and Alpine Research, 7: 25-32. 1977. Evolution of glaciated topography of upper Skagit drainage basin, Washington. Arctic and Alpine Research, 9: 183-192. Waitt, R.B. and Thorson, R.M. 1983. The Cordilleran Ice Sheet in Washington, Idaho and Montana. In Porter, S.C. (ed.) Late-Quaternary environments of the United States, Vol. 1, The late Pleistocene, pp.53-70. Whyte, I.W. and Schubert, N.D. 1988. An assessment of the fall Vedder-Chilliwack River sport fishery, 1988. Department of Fisheries and Oceans, Fisheries Branch. 24p. Wilford, D.J. and Schwab, J.W. 1981. Soil mass movements in the Rennel Sound area, Queen Charlotte Islands, British Columbia. Unpublished report, British Columbia Ministry of Forests. Young, S.E. 1992. Slope stability prediction techniques for forest management purposes- a case study from the Queen Charlotte Islands, British Columbia. Unpublished M.Sc. thesis, Department of Geography, UBC, 150p. Ziemer, R.R. 1992. Effect of logging on subsurface pipeflow and erosion: coastal California, USA. In Erosion, debris flows and envioronment in mountain regions (Proc. Chengdu Symposium, July, 1992). IAHS Publication 209: 187-197. 172 Appendix 1 - The DFD? Assessment. As described in Chapter 2.5, debris flow initiation potential is assessed using four parameters. These are: - GWSD, gully wall slope distance (length), in meters; - SURF, gully wall surficial material; - GWSA, gully wall slope angle, in percent slope; and - CG, channel gradient of gully channel, in percent slope. The four parameters are combined into two estimates of potential. They are: - GWFP, gully wall failure potential, the chance that a gully wall will experience slope failure; - GGP, gully geometry potential, the chance that a failure entering a gully will become a debris flow. Both GWFP and GGP are assessed on the ordinal scale of High, Moderate or Low. These ratings are subjective evaluations only, and do not carry associated numerical probabilities of failure or debris flow initiation. Once GWFP and GGP are assessed, they are combined to produce an overall asssessment of DFLP, which is also rated as being High, Moderate or Low. The exact system by which GWFP, GGP and DFLP are calculated is shown below, in tables taken from the GAP Handbook: 2) Upslope debris flow potential H M L Terra in stability c lass upslope V IV I to III Ev idence of landsl ides or debris Y e s Not clear No f lows in gully system Table C . Gully Wall Failure Potential (GWFP) Gul ly wall s lope angle Surficial material (%) R C M,F W , L N F >70 L H H H H 61-70 L M H H H 50-60 L L M H H <50 L L L M H % Enter result in shaded portion of Tab le E R = bedrock; C = colluvium; M = morainal (till); F = fluvial; W = marine; L = lacustrine; NF = natural failure scars present 173 Table D. Gully Geometry Potential for Debris Flow Initiation(GGP) Gul ly wall s lope distance Channe l g rad ien t .CG (%) (m) 0-30 31-40 41-50 51-60 61-70 >20 L H H H H >15-20 L M H H H >10-15 L L M M H >5-10 L L L M M 0-5 L L L L L m Enter result in shaded portion of Tab le E Table E. Debris flow initiation potential (DFIP) Gul ly wall fai lure potential Gul ly Geometry potential Low Moderate High High L M H Moderate L M M Low L L L 5) D e b r i s f l o w in i t ia t ion potent ia l H M L Debr is f low initiation potential H M L (DFIP, from table E) Pas t debr is f low initiation in this Y n/c N reach As can be seen, the DFLP assessment is relatively simple. The combination of GGP and GWFP into DFIP is based on a principle of equal weighting of subjective probabilities, so that if GGP is rated L and GWFP is H, then DFIP is assessed as L (the chance of a failure is high, but the chance of it initiating a debris flow is low, so overall debris flow initiation potential is thus low). The four factors used in the DFIP were chosen as a result of a synthesis of completed and in-process research conducted throughout the B.C. coastal belt (though most was conducted in the Queen Charlotte Islands).The combination of the various factors to produce the DFLP, and the weighting of each factor, was based on the following reasoning (Millard, pers. comm.): GWSA influences slope stability as shown in chapter 2.1. SURF has been shown to influence 174 critical angles for failures to initiate, with more competent materials generally supporting steeper slopes (for example, see Howes (1987)); till, which in general behaves in a similar fashion to colluvium, was assigned a lower stability ranking than was colluvium based on the observed presence of relatively large slump failures in till, whereas colluvium is more likely to fail by ravelling. That CG is directly related to whether or not a failure will become a debris flow is fairly obvious; the steeper the channel, the smaller the chance that a failure will come to rest upon entering the channel. GWSD was included in the computation of GGP based on the rationale that a longer gully wall length allows a failure to entrain debris (and hence grow) as it travels towards the channel. Larger failures have a greater mass, and hence momentum, than do small failures; these larger failures are thus more likely to initiate debris flows than are small failures. Appendix 2 - Ministry Of Forests Terrain Data Cards Province of Ministry British Columbia of Forests LANDSLIDE DATA CARD WATERSHED CODE POLYGON » FAILURE * COORDINATES FAILURE DATE TYPE: ds da df dl dfa su rxs rxa rxf *EL£VATION:[~ ] m j ASPECT: [ •LOCATION: os os->g gh gs gc esc osd j LAND USE: na rc rt rp cc cb| PRESENT EROSION: sh ri gu rtf j OLD FAILURE? yes no SLOPE GRADIENT: Origin) |° | fp| ] | Gutlyf" |° @ p.o.e. 'SLOPEPOSITION: macro apex up mid lo esc hd ANGLE OF ENTRY INTO GULLY: F Ir 'HILLSLOPE: Configuration: un be di fac irr sg brk| Curvature: Ccave cVex str •SOILDESCRIPTION: Close) | Texture (8) Q Depth | | m 'DRAINAGE CLASS: r w m i p vp | SEEPAGE: ab hf fp surf TERRAIN UNIT: FAILURE PLANE: wr ur wt ut c MATERIEL; wf uf wgf ugf ww uw fill 'STRATIGRAPHY AT HEADSCARP: HEADSCARP HEIGHT MATERIAL TEXT/LITHOL THICKNESS (m) COMPET. STRUCT, A.B.D. DIP. DIR. FAILURE PATH STREAM CHANNELS AFFECTED FAILURE VOLUME ZONE LENGTH (m) WIDTH DEPTH (m) GRADIENT REV EG ORDER GRADIENT LENGTH (m) ,N| 1 Transport CHANNEL Deposition 1 1% FAILURE ROUTING (fate of debris) COMMENTS ' ROAD CONSTRUCTION: Date (y/rrvd): / | Type: cat gs bri | Ditches: yes no •LOGGING: Date (y/nVd): / / I Type: gr hi si POSSIBLE TRIGGERING FACTORS: cd rr se dfe rtf flllsat RECORDED BY DATE ' Data lines referring to failure initiation poinL Others refer to the path in general. Figure H-1A: MoF Landslide Data Card 176 > CQ • LU Q il O o 3 O o o o L U L U C C STREAM CHANNEL VALLEY FLOOR WIDTH STREAM CHANNEL BANKFULL WIDTH §* t L U o g L U £ } GULLY FLOOR WIDTH / DEPTH BEDROCKI FLOW MATERIAL j DEPTH BEDROCKI FLOW MATERIAL j DEPTH SLOPE M0RPH. PATH AZIMUTH (o) PATH SLOPE (o) FILL (m) FILL (m) FILL (m) SCOUR (m) SCOUR (m) SCOUR (m) SEGMENT DISTANCE io 8 » J o H O CO — B £? £ c o S s 2 2 , = E E w c § ai C £ i i CO c O « - c ~ c -o O o c Z s cs CO U — » cr h-® a < > E Figure II-1B: MoF Landslide Profile Data Card Province of Ministry of British Columbia Fores is TERRAIN DATA C A R D PAGE OF WATERSHED AREA AIR PHOTO NO. SLOPE POSITION' Macro ap up mid lo esc hd HILLSLOPE CONFIGURATION: un be di fac in sg I CURVATURE: concave convex straight SLOPE CONDITION: mnl mccl mgwt I FORESTED TERRAIN: js pb It tf crp DRAINAGE CLASS: f w m I p vp SOIL 1 2 CLASSIFICATION TEXTURE TERRAIN COMPONENTS 1a lb 2a a 3a TEXTURE AVERAGE DEPTH(m) STRUCTURE BEDROCK: Item • 0 strike dip ASPECT SLOPE ANGLES: minimum GULLY: number ELEVATION depth gradient gully wall LANDSLIDES' » natural tt dearcut tt road cut # road fiH Length m | j Type lb eh cl | Stability: mrcl mtff tt Fill slope recovery: [ yes j | no j Logging date (year): RECORDED BY DATE Y FS 124 HRE 92/10 Figure II-2: MoF Terrain Data Card Terrain/Landslide Data Cards - Abbreviation Codes Landslide type ds - debris slide rxs - rockslide da - debris avalanche df - debris flow rxf-rockfall dt - debris torrent Location Land use os - open slope os -»g - os into gully esc - escarpment osd - open slope depression (incipient gully < lm deep) gh - guly headwall gs - gully sidewall gc - gully channel na - natural rf- road fill Present erosion sh - sheet p.o.e. - point of entry Slope position ap - apex (crest) lo - lower rp - road prism cc - clearcut ri - rill (<0.5 m) rtf - retrogressive failure up - upper esc - escarpment Hillslope config. un - uniform irr - irregular brk - slope break be - benchy fac - faceted rc - road cut cb - cutting boundary gu - gully (>0.5 m) mid - middle hd - headwater drainage sg - single gully di - dissected Curvature (along the contour) str - straight c cave - concave Seepage ab - absent fp - failure plane surf - adjacent land surface Drainage Class r - rapidly i - imperfectly w- well p - poorly c vex - convex hs - headscarp m - moderately well vp - very poorly Failure Plane wr - weathered rock ur - unweathered rock ut - unweathered till c - colluvium wgf - weathered glaciofluvial ww - weathered marine uw - unweathered marine fill - roadfill/sidecast uf - unweathered fluvial wt - weathered till wf - weathered fluvial ugf - unweathered glaciofluvial Road construction type: cat - cat; gs - grade shovel; bh - hydraulic excavator Logging type gr - grapple crane; hi - highlead; si - skyline; r/w - right of way Possible triggers cd - culvert drainage rtf - retrogressive failure rr - road runoff dfe - debris flow erosion se - stream erosion fillsat - fill slope saturation Slope Conditions mnf - minor natural failures (<500 m2) mcbf - minor cutting boundary failures js - juvenile soils It - leaning trees crp - soil creep ss - surface seepage mccf - minor clearcut failures mgwf - minor gully wall failures tc - tension cracks Figure II-3: Key to abbreviations used on Terrain & Landslide Data Cards Province of Ministry British Columbia of Forests G U L L Y A S S E S S M E N T P R O C E D U R E - FIELD DATA SHEET WATERSHED DATE Y u D I I GULLY NO. BLOCK NO. RECORDED BY REACH NO. DISTANCE (m) ROLL / PHOTO NO. Fan Assessment: do Sections 1, 2, 3, 4 (and 6, if post-logging). Transport Zone and Headwall: do Sections 1, 2, 3, 5 (and 6). Enter results in Management Strategies Tables 2 and 3 1) DOWNSTREAM IMPACT POTENTIAL Community watershed Dwellings, major installations, safety H Y Connectivity to fish streams or lakes or sensitive marine zones | Direct Terrain stability class upslope 2) UPSLOPE DEBRIS FLOW POTENTIAL H Indirect con. I to III N Evidence of landslides or debris flows in gully system N/C: not| clear TABLE A: WATER POWER INDEX (XS Area Index + CG Index) C W ( m ) X C D (m) = X S A r e a (m2) XSArea fm 2 ) n/a < 1 1 - < 2 2 - 5 > 5 XS Area Index 0 2 4 6 10 CG (%) n/a < 8 8 -20 > 20 WPI CG Index 0 2 4 6 3) WATER TRANSPORT POTENTIAL H M L Water Power Index (Table A) > 11 8-11 <8 Upstream catchment area (ha) > 9 <9 Water-transported woody debris accumulations LWD jams, or no WD SWD jams WD, no iams Largest sediment stored in wedges boulders cobble sand TABLE B: FAN DESTABILIZATION INDEX (FDI = CN Index + CI Index) Number of channels (CN) FDI 0 1 2 - 3 >3 CN Index 0 2 4 6 Channel Incision (CI, m) No channe 0 < 1 1 - 2 > 2 CI Index 0 8 6 4 2 4) FAN DESTABILIZATION POTENTIAL H M L Fan Destabilization Index (Table B) >8 5 - 8 < 5 Opt: Debris flow return period (years) < 11 11 -3q >30 FS 197B RVA 95/7 ure II-4A: Gully Assessment Procedure Field Data Sheet (front TABLE C: GULLY WALL FAILURE POTENTIAL (GWFP) GULLY WALL SLOPE ANGLE (%) SURFICIAL MATERIALS R c M, F W, L FS > 70 L H H H H 61 -70 L M H H H 50 -60 L L M H H < 50 L L L M H % Enter result in shaded portion of Table E R = Bedrock C = Colluvium M » Morainal F = Fluvial W = Marine L = Lacustrine FS = Failure scars present — any surficial material TABLE D: GULLY GEOMETRY POTENTIAL FOR DEBRIS FLOW INITIATION GULLY WALL SLOPE DISTANCE (m) CHANNEL GRADIENT (CG) 0-30 31 -40 41 -50 51 - 60 61 -70 >20 L H H H H > 15- 20 L M H H H > 10 - 15 L L M I M H > 5 - 1 0 L L L M M 0 - 5 L L L L L m Enter result in shaded portion of Table E TABLE E: DEBRIS FLOW INITIATION POTENTIAL (DFIP) GWFP (TABLE C) GULLY GEOMETRY POTENTIAL (TABLE D) Low Moderate High High L M H Moderate L M M Low L L L Enter result in Section 5 5) DEBRIS FLOW INITIATION POTENTIAL Debris Flow Initiation Potential (DFIP, from Table E) H M S L H M L Y n/c N Past debris flow initiation in this reach 6) POST-LOGGING CONDITIONS Years since logging < 1 2 - 5 6 - 10 > 10 Logging debris in channel Sparse Moderate Heavy V. Heavy Sediment stored behind logging debris Sparse Moderate Heavy V. Heavy Figure II-4B: Gully Assessment Procedure Field Data Sheet (bad 181 Appendix 3- Analysis of Field Data: Sediment Samples. I. Laboratory sediment analysis. Upon return from the field, the first task to be accomplished was analysis of the sediment samples. It was anticipated during initial design of the study that the sediment samples would not be an ideal source of data, as in the field they were only taken at sites at which collection was feasible, rather than at all sites, thus creating an incomplete data set. Hence, sediment data would probably be of lower reliability than the other data. A total of 30 separate sediment samples, representing 37 separate failures when composite samples are considered, were collected. Initially, sediment samples were oven-dried in a drying oven at 150 C for 24 hours. This technique was chosen over air-drying due to the high moisture contents of many of the samples; it was anticipated that oven-drying would be faster and more effective than air-drying. The approximately 8 kg. samples were then weighed to find total weight, following which they were split mechanically to a final weight of approximately 1kg., which was reweighed. The proportion of final to initial weight was recorded and the ~7 kg. not used was set aside as a backup, should it be required. The ~1 kg. split was placed in a sieve stack and mechanically shaken for 35 minutes. Sieve sizes were single-(j> intervals from 16mm to 63 u.m (thus covering the finest gravel sizes and the entire sand range). Following shaking, the contents of each sieve were weighed and recorded, as were the contents of the pan (containing the silt and clay-sized fraction). The results from this analysis were then entered into a spreadsheet. Percent weights for each sieve were computed. The gravel fraction (coarser than 2mm) was then removed from the analysis and percents were recalculated for the sand-size and smaller fraction. 'Cumulative percents finer than' were calculated for this subset and plotted, and the D50's were read off these curves. The D 5 0 and % Figure A3.1: Scatterplot of D50 vs. % fines y = 0.692-2.308*x+eps E c O o ; o o 0 r ° oi ° ! . . ; J ft oi 1 ! 1 1 i . . . . i . . . . ^ ^ - ^ ^ i ° 0.02 0.06 0.1 0.14 0.18 0.22 0.26 %FINES Figure A3.1: Plot of D 5 0 vs. % fines for sediment data. Line of regression equation is plotted: equation is (D50 = 0.692 - 2.308* %FINES), r-squared value is 0.47. fines values thus obtained for the <2mm fraction were entered into the data set for their respective failures. n. Comment on analytical results. Two separate results from the sediment analysis are worthy of comment: the distributions of D5o and % fines (silt and clay). A general inverse relationship is clear with samples with higher % fines having lower D 5 0 and vice versa (Figure A3.1), as might be expected, although there is a fair degree of scatter in the data. A linear regression on these data gives a linear equation with r-square of-0.5 (Figures A3.2, A3.3). Both D 5 0 and % fines show little variation from sample to sample, with the exception of a few outliers. This is particularily striking when one considers that there are two separate materials under consideration: till and colluvium. There is essentially no differentiation between these two materials when gradation is analyzed. The matrix fractions of both are coarse sands (D50 -0.5 mm) with only about 8- 10% silt and clay by weight (excepting a few outliers). This result must be interpreted with consideration of two points. First, extremely 183 Figure A3.2: Predicted vs. Observed Values for regression of D50 from % fines Dependent variable: D50 0.85 0.75 0.65 •= 0.55 T 3 I 0.45 tn £1 O 0.35 0.25 0.15 j i i o o i o _ ° , - ; •Cy--- --yS:- ofcr / f \ - f f ' o jo ) 1 o ] y\K : . . - : > C . - : ~ : : o \ _-K —.—. .— i—.——.—i—.—,—.—.—\—.—.—.—.—i—.——. i . ,—.—. 0.1 0.2 0.3 0.4 Predicted Values 0.5 0.6 0.7 Regression 95% confid. Figure A3.2: Plot of predicted vs. observed values for regression of D50 onto % fines, with 95% confidence boundaries indicated. STAT. MULTIPLE REGRESS. N=39 Intercpt %FINES Regression Summary for Dependent Variable: D50 R= .68903433 R2= .47476830 Adjusted R2= .46057285 F(l,37)=33.445 p<00000 Std. Error of estimate: .10125 St. Err. BETA of BETA -.689 .119 St. Err. B ofB .692 .037 -2.31 .399 t(37) p-level 18.5 .000000 -5.78 .000001 Figure A3.3: STATISTICA regression summary for regression of D50 onto % fines shown in Graph A3.2 above. coarse colluvium (consisting primarily of boulders to gravel sized clasts, with minimal matrix material) was not sampled, and so is not represented here. Thus, the colluvium sampled may not be representative of all colluvium in the study area. Secondly, it was recognized that the field distinction between till and colluvium was made primarily on the basis of consolidation. The two materials which were sediment sampled, till and colluvium, probably represent the same original material, a sandy till. Deposited as basal till, it remained as such. Deposited as ablation till (or another unconsolidated till form), it was rapidly reworked by gravity into the present colluvial 184 deposits without significantly changing its textural properties; for example, it was not significantly washed of fines. This accounts for the similar textural properties observed during sediment analysis. Since % fines was generally low, the effects of silt and clay fractions on slope stability were deemed to be small, and material was not further analyzed to distinguish between % silt and % clay (eg. through hydrometry). It is worth noting that the variation between adjacent sites (most notably between FOL-001-RB-01, F01 and F02) is in some cases greater than any other variation in the data set, even the regional variation. In other cases, however, (for instance SPN-001-FTW-Ol and SPN-002-HW-01) this is not the case. In light of this, the rationale by which composite samples were taken, namely that in adjacent sites in the same material variation would be minimal, becomes subject to criticism, as it is obviously not true in all cases. The regional variation in mean results shows that Norrish Creek samples are slightly coarser (higher D 5 0 , significant at p<0.05) than those of Chilliwack Valley; however, the regional variation in % fines (Norrish Creek has a slightly lower mean value) is not significant (p<0.25). These results are possibly due to the greater proportion of till in the Norrish drainage and the differing origins of the tills in the two locations. The Coast Mountains experienced a more intensive glaciation than did the Cascades, and Norrish Creek was covered by thicker ice than was Chilliwack Valley; the till in Norrish Creek is thus most probably derived from the granitic rocks of the Coast Mountains, whereas the till in Chilliwack Valley is locally derived and was composed of greater proportion of non-granitic rocks. Thus, the constituent lithologies of the till in Norrish Creek were likely slightly more mechanically and chemically resistant than those in Chilliwack Valley. This would result in a till which would be texturally coarser, with larger clasts, in Norrish Creek, which is what was observed. 185 In interpreting the results of the sediment analysis, it is important to remember that only 30 samples (representing 37 failures) were taken for 62 separate failure sites. Some failures which were not sampled were excluded due to material characteristics, and some were excluded due to inability to transport the samples properly (primarily gullies which were visited via long foot and bike approaches on days when many gullies were surveyed). Thus, some material does not show up in the sediment sampling at all, because it was excluded ( i.e. coarse colluvium), some gullies had no accessible surficial material (primarily rock walls, and failures which involved bedrock as a primary component) and those gullies which were sampled are biased towards those which were most accessible. So the sediment data may be seen to be both biased towards certain material types, and possibly skewed in its representation of those types. Any conclusions made about the effects of material properties on debris flow initiation herein must be considered with the two caveats made above. Despite this, the general lack of variation in what sediment data were collected can reasonably be interpreted within limits to indicate a fair degree of textural homogeneity over most of the study area, in. Textural data, (on next page following) 186 Drainage Gully Side Failure wt of weights (g) %fines d50 <2mm (mm) (g) 1mm .5mm .25mm .125mm .063mm Pan Airplane 4 RB 1 469.9 160.7 107 73.9 58.5 38 31.6 0.07 0.61 Airplane 1 RB 1 805.1 221.3 185.4 151.2 118.9 98.4 29.1 0.04 0.505 Borden 1 R B 1 757.3 178.9 149 108.4 113.8 121.9 86.4 0.11 0.35 Borden 2 R B comp. 438.6 132.8 106.3 70 65.1 44.2 19.3 0.04 0.55 Center all RB comp. 1145.2 213.8 269.9 264.2 188.6 141.6 65.5 0.06 0.39 Chipmunk 2 H W 1 551.3 164 126.7 99 71.6 49.9 40 0.07 0.52 Chipmunk 1 H W 1 523.8 168.5 112.7 88.3 73.2 38.2 40.9 0.08 0.55 E.Norrish 3 LB 1 536.9 123.9 105.5 90.8 83 72.1 61.1 0.11 0.36 E.Norrish 2 R B 1 721.4 206.6 161.8 123.5 95.5 56.7 78.1 0.11 0.505 E .S .P . 1 R B 1 650.3 161 105.1 91.2 126.4 100.7 62.9 0.10 0.3 E .S .P . 2 LB 1 590.8 193.8 129.7 91.9 72.1 55.9 47.1 0.08 0.58 Foley 1 R B 2 619.9 142 129.3 115.1 85.5 67.8 80.4 0.13 0.39 Foley 2 R B 1 815.3 150.8 146.6 124.7 114.8 119.4 158 0.19 0.275 Foley 1 RB 1 740.8 133.9 114.9 110 111.7 105 165.6 0.22 0.22 Foley 3 RB 1 1018.9 314.4 236.7 173.8 104 92.3 94.6 0.09 0.54 Foley 4 LB 1 575.4 145.7 113.3 88.7 81.4 67.6 76.9 0.13 0.4 Foley 4 LB 2 399 118.9 90.2 58 42.3 47.5 40.7 0.10 0.52 Foley 6 LB 1 968.8 314.7 217.9 177.8 126.1 78.8 51.5 0.05 0.57 Foley 5 LB 1 703.3 188.2 146.1 117.5 110.9 87.2 51.5 0.07 0.45 Margaret 1 H W 1 589.1 127.2 139.4 119.6 84.5 68.8 49.3 0.08 0.41 Norrish 3 LB 1 1006.7 252 193.7 162.2 153 123.1 121.3 0.12 0.38 Norrish 05\0 6\07 LB comp. 709.5 278.6 191.4 111.1 56.5 38.3 34.1 0.05 0.75 Slesse 1 LB 1 377.8 117.4 82.4 42.5 21.5 7.8 14.8 0.04 0.55 T .L .M. 1 LB 1 906.7 234.2 194.4 156.7 134 103.5 86.1 0.09 0.45 T .L .M. 2 LB 1 560 165.4 124 87.5 79 68.9 34.3 0.06 0.51 Tamihi 2 R B 1 529.2 164.5 126.6 95.2 58.9 27.6 56.4 0.11 0.58 Tamihi 4 LB 1 519.1 193.9 122.2 77.4 53.4 40.1 31.1 0.06 0.68 Tamihi 1 R B 1 1030.5 306.3 212.7 197 153.2 108.6 51.2 0.05 0.5 Tamihi 3 LB 1 997.7 222.5 179.6 186.1 183.5 88.5 135.4 0.14 0.34 W. Norrish 2 LB 1 763.5 192.8 172.6 144.7 98.9 73.7 80.6 0.11 0.455 187 Appendix 4- Correlation Matrix for Failure Variables and DFRESULT. STAT. BASIC STATS Correlations (guldata.sta) Marked correlations are significant at p < .05000 (Casewise deletion of missing data) Var. X & Var. Y r(X,Y) N BASIN DFRESULT .402748* .162206* 3.40832* .001173* 62* REGBASIN DFRESULT .098639 .009730 .76780 .445616 62 GULLY DFRESULT .122294 .014956 .95445 .343687 62 SIDE DFRESULT .265413* .070444* 2.13236* .037079* 62* FAILNO DFRESULT -.318729* .101588* -2.60471* .011576* 62* ASPECT DFRESULT -.146177 .021368 -1.14457 .256932 62 BASAREA DFRESULT .046161 .002131 .35795 .721640 62 LNBAS DFRESULT .055857 .003120 .43335 .666317 62 SLPPOS DFRESULT .085774 .007357 .66686 .507418 62 SLPCONFG DFRESULT .118513 .014045 .92452 .358922 62 CURVE DFRESULT .077749 .006045 .60407 .548076 62 CURVE2 DFRESULT .230159 .052973 1.83198 .071918 62 DRAIN DFRESULT .088451 .007824 .68784 .494207 62 188 SOIL DFRESULT .044002 .001936 .34117 .734166 62 TERRAIN DFRESULT .183803 .033783 1.44841 .152709 62 T2 DFRESULT .208012 .043269 1.64729 .104727 62 AVSLOPE DFRESULT .000717 .000001 .00555 .995589 62 TANAVSLP DFRESULT .001884 .000004 -.01459 .988405 62 LITHOL DFRESULT .087852 .007718 -.68314 .497150 62 LITHOL2 DFRESULT .231990 .053820 1.84739 .069622 62 LITHOL3 DFRESULT .206429 .042613 -1.63419 .107456 62 LITHOL4 DFRESULT .266632* .071093* 2.14290* .036186* 62* LITHOL5 DFRESULT .175787 .030901 1.38317 .171737 62 LOGDATE DFRESULT -.146376 .021426 •1.14617 .256276 62 GWSA DFRESULT .162156 .026295 •1.27290 .207962 62 GWSD DFRESULT .133355 .017784 •1.04227 .301467 62 CG DFRESULT .302982* .091798* 2.46264* .016680* 62* TANCG DFRESULT .298084" .088854* 2.41891* .018617* 62" TANCG2 DFRESULT .343490* .117985* 2.83303* .006270* 62* 189 TANGWSA DFRESULT -.147805 .021846 -1.15761 .251611 62 VOLUME DFRESULT .201758 .040706 1.59563 .115827 62 LNVOL DFRESULT .427340* .182619*, 3.66131* .000532* 62* CATLNVOL DFRESULT .465060* .216281* 4.06915* .000140* 62* CTLNVL2 DFRESULT .443306* .196521* 3.83082* .000308* 62* FAILSLP DFRESULT .285003* .081227* -2.30314* .024753* 62" TANFS DFRESULT .283193* .080198* -2.28724* .025724* 62" CATFS DFRESULT -.265851* .070677* -2.13614* .036757* 62* FAILAZIM DFRESULT -.140411 .019715 -1.09850 .276374 62 %REVEG DFRESULT .012506 .000156 .09688 .923143 62 AOE DFRESULT -.536577* .287915* -4.92540* .000007* 62* SINAOE DFRESULT -.534140* .285306* -4.89408* .000008* 62* COSAOE DFRESULT .511228* .261354* 4.60758* .000022* 62" CATAOE DFRESULT -.537999* .289443* -4.94376* .000006* 62* TYPE DFRESULT .421655* .177793* 3.60199* .000642* 62* LOCN DFRESULT .258334* .066737* 2.07136* .042633* 62* L0CN2 DFRESULT .434630* 190 .188903* -3.73817* .000416* 62* LOC3 DFRESULT -.407568* .166111* -3.45717* .001009* 62* LANDUSE DFRESULT .109077 .011898 .84998 .398717 62 FPMTL DFRESULT .099452 .009891 .77419 .441858 62 SEEP DFRESULT .195409 .038185 •1.54339 .127995 62 SEEP2 DFRESULT .303540* .092136* 2.46763* .016471* 62H FAILDATE DFRESULT .130573 .017049 -1.02015 .311752 62 ICSS DFRESULT .395034* .156052* -3.33083* .001485* 62* ICSS2 DFRESULT .396757* .157416* -3.34806* .001410* 62* ICSS3 DFRESULT -.415926* .172995* -3.54273* .000774* 62s1 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items