UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The use of digital elevation models as a preliminary slope stability tool for the Clayoquot River Watershed Faucette, Dave 1995

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

Item Metadata


831-ubc_1996-0317.pdf [ 9.99MB ]
JSON: 831-1.0086991.json
JSON-LD: 831-1.0086991-ld.json
RDF/XML (Pretty): 831-1.0086991-rdf.xml
RDF/JSON: 831-1.0086991-rdf.json
Turtle: 831-1.0086991-turtle.txt
N-Triples: 831-1.0086991-rdf-ntriples.txt
Original Record: 831-1.0086991-source.json
Full Text

Full Text

THE USE OF DIGITAL ELEVATION MODELS AS A PRELIMINARY SLOPE STABILITY TOOL FOR THE CLAYOQUOT RIVER WATERSHED by Dave Faucette Bachelor of Arts The University of British Columbia, Vancouver, British Columbia, 1991 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES (The Department of Forest Resources Management) We ac^ep^i^me^s^^nforrnjng to the required standard THE UNIVERSITY OF BRITISH COLUMBIA DECEMBER 1995 ° Dave Faucette, 1995 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. The University of British Columbia j Vancouver, Canada I Department of v'?)AoJ Date DE-6 (2/88) Abstract The purpose of this thesis was to develop a preliminary slope stability tool for the Clayoquot River Watershed. Current techniques obtained from aerial photographs, topographic maps and field surveys are becoming increasing expensive to undertake. Testing new, cost effective, easily acquired methods for mapping hazardous areas needs to be undertaken. This project attempts to do this by using morphological variables acquired from Digital Elevation Models (DEMs) and Geographic Information Systems (GISs). In contrast to current hazard mapping methods used in British Columbia the procedure outlined in this project attempts to provide an indication of the fluxes taking place within the landscape. The Clayoquot River watershed situated along the west coast of Vancouver Island was the study site for this research project. Steep slopes, a harsh climate and the intense debate over logging throughout the entire Clayoquot Sound region makes this an ideal location to test a method as proposed in this thesis. Slope gradient, transient snow zone, across slope curvature and catchment area were the four variables that went into creating a slope stability image. The next stage was to combine this image with a rating based on surficial geology and then to test this product with the hazard maps produced by MacMillan Bloedel for the area. Unfortunately their project has been delayed, so testing the findings of this thesis is not possible at this time. With improved models for acquiring the variables and more accurate digital information along with the procedures outlined in this thesis, it is felt that preliminary slope stability assessments will become an invaluable tool to the forest managers throughout British Columbia. ii Table of Contents Page Abstract i i Table of Contents iii List of Tables v List of Figures vi Acknowledgments vii 1.0 Introduction 1 1.1 Thesis Statement 2 1.2 Objectives 2 1.3 The Physical Environment 4 2.0 Types of Mass Movement 13 2.1 Debris Avalanche-Debris Flow 14 2.2 Deep Seated Mass Movements 15 2.3 Dry Ravel and Rockfall 15 3.0 Soil Mechanics 16 3.1 Cohesion 20 3.2 Angle of Internal Friction (O) 21 3.3 Soil Moisture and Groundwater 21 3.4 Other Triggering Mechanisms 26 4.0 Slope Stability Assessment 27 5.0 Methods 36 5.1 Potentially Mapable Variables 38 5.1.1 Slope Gradient 41 5.1.2 Slope Shape 41 5.1.3 Aspect 43 5.1.4 Slope Position 44 5.1.5 Transient Snow Zone 45 5.1.6 Shallow, Linear Depressions or "Swales" 45 5.1.7 Surficial Material 47 5.2 Derivation of Morphological Variables 49 5.2.1 Slope Gradient 49 5.2.2 Across Slope and Downslope Curvature 49 5.2.3 Aspect 51 5.2.4 Transient Snow Zone 51 5.2.5 Catchment Area. 51 5.3 Creation of Decision Trees 52 6.0 Results and Discussion 61 6.1 The Issue of Subjectivity 68 6.2 How to Improve the Model 69 7.0 Conclusions 72 iii References 73 Appendices A Landslide Studies Conducted Throughout the Pacific Northwest 84 B FORM.FOR FORTRAN Program 105 C CATCH.FOR FORTRAN Program 108 iv List of Tables Page Table 4.1. Summary of hazard mapping methods 28 Table 4.2. Terrain stability hazard classification 32 Table 4.3. Surface erosion potential rating 33 Table 5.1. Checklist of slope attributes 40 v List of Figures Page Figure 1. Sequence of procedures for completing thesis 3 Figure 2. Clayoquot Sound in relation to British Columbia 5 Figure 3. The Clayoquot Sound area 6 Figure 4. The Clayoquot River watershed 8 Figure 5. Mean monthly temperature and precipitation for the Clayoquot region 10 Figure 6. Slide mechanism with translational failure surface 18 Figure 7. Failure envelope 20 Figure 8. Quasi one dimensional flow of water in soil 24 Figure 9. Rainfall intensities commonly associated with landslide activity 25 Figure 10. Hypothetical curves indicating root strength conditions over time 27 Figure 11 A. Profile curvature slope shapes 42 Figure 11B. Planform curvature slope shapes 42 Figure 12. Plot of source area-slope relationship for 9 watersheds along the Pacific Coast for channel initiation 47 Figure 13. Source area-slope relationship with a range of slopes 48 Figure 14. Flow chart of the FORM.FOR program 50 Figure 15. Flow chart of the CATCH.FOR program 53 Figure 16. Decision tree for Slope 0-20 degrees 56 Figure 17. Decision tree for Slope 21-29 degrees 57 Figure 18. Decision tree for Slope 30-35 degrees 58 Figure 19. Decision tree for Slope 36 degrees and above 59 Figure 20. Final image derived through decision trees and variables 60 Figure 21. Slope map of the Clayoquot River watershed 63 Figure 22. Location of Aerial Photographs 64 Figure 23. Aerial Photograph MB94010 65 Figure 24. Aerial Photograph MB94008 66 vi 1 1.0 Introduction Forecasting slope stability in south-coastal British Columbia is a complex task involving many variables. Natural processes in the form of intense precipitation (especially during the winter months), along with severe winds and occasional tectonic activity combine with steep slopes to make this rugged and variable landscape prone to mass movements. These natural processes are exacerbated by human activities such as forest harvesting and the construction of logging roads. Landslide inventories conducted throughout western North America suggest that roads have been the major management activity associated with shallow landslides related to logging (MOF Tim. Harv. Br. 1993; NCASI 1985). Slope failures, both naturally occurring and logging-accelerated, represent a loss of land base and have the potential to affect social, economic and resource values. The threatened values include water quality; fish habitat; wildlife habitat; forest productivity, health and investment; human life, private property, roads and railways; utilities; landscape values; cultural values; forest revenue; and recreation and range values (BC MOF Tim. Harv. Br. 1993). There is an urgent need for forest companies to be better able to forecast potential impacts of this nature. A review of the relevant literature offers a plethora of both quantitative and qualitative techniques for delineating hazardous areas. Much of the data collected for these techniques must be obtained from aerial photographs, topographic maps and field surveys. The disadvantage of these techniques has been the high cost of acquiring the necessary data. With the recent attention given to the comprehensive Forest Practices Code (1995), and a Forest Road and Logging Trail Engineering Practices Manual (1993) as well as a forest road deactivation program, now is an appropriate time to test potentially cost effective, easily acquired data for mapping hazardous areas. The highly emotional, ongoing debate between the logging industry and environmentalists 2 over the fate of the remaining old growth forests on the west coast, and the potential loss of slope stability associated with that harvest, seems an appropriate place to concentrate a slope stability hazard assessment. 1.1 Thesis Statement This thesis will attempt to analyze the causes of slope failure and develop a simple functional model that will facilitate analysis to forecast the location of slope failures in the Clayoquot River watershed. The model will be a prototype that could further assist forest companies in evaluating the risk of slope failure due to road construction and timber harvesting throughout the Clayoquot Sound area. 1.2 Objectives This thesis has two main objectives. First, it provides an analysis of the dynamics of slope stability. Second, the product developed provides a reconnaissance level, prototype terrain map that forest managers may use in assessing the hazards involved in planning the location of roads and cutblock layouts. To achieve these objectives the following must be completed: 1) Literature review of the necessary topics; 2) Selection of the necessary variables; 3) Creation of GIS images of the variables selected; 4) Development of decision trees; 5) Creation of final GIS map which incorporates the decision trees; and 6) Comparison of the results of the model with existing terrain stability maps developed in the study area. Figure 1 outlines the sequence of procedures needed in order to complete this thesis. LITERATURE REVIEW OF POTENTIAL MAPPABLE VARIABLES CREATION OF GIS IMAGES OF VARIABLES SELECTION OF RELEVANT VARIABLES CREATION OF DECISION TREES CREATION OF GIS IMAGE BASED ON DECISION TREES CREATION OF SURFICIAL GEOLOGY IMAGE COMPARE IMAGE WITH PRODUCED STABILITY MAP FOR WATERSHED COMBINE SURFICIAL GEOLOGY WITH DECISION TREE IMAGE Figure 1: Sequence of procedures for completing slope stability analysis 4 1.3 The Physical Environment Clayoquot Sound is located on the west coast of Vancouver Island (Fig. 2). It is part of a physiographic region which encompasses the entire Vancouver Island, the Queen Charlotte Islands, along with southeastern Alaska, referred to as Wrangellia (Jones et al. 1977). The 'Insular Belt' is the term used by Muller (1977a) to describe this geological terrane. The Island is the main component of the Insular Belt, the westernmost major tectonic subdivision of the Canadian Cordillera (Muller 1977b). The Wrangellia Terrane is a volcanic arc-oceanic plateau that accreted to North America during the Cretaceous period (Jones et al. 1982). These rocks are composed of a thick sequence of late Palaeozoic volcanics capped by a thin sequence of shallow-water marine shale, sandstone and limestone. The sedimentary rocks are succeeded by thick basaltic lavas that continue from late Permian to late Triassic time, after which shallow carbonates and elastics were deposited (McLaren et al. 1983). Along with this complex tectonic history has been the more recent period of glaciation and the subsequent glacio-isostatic recovery. The Fraser Glaciation (about 30-11 ka) is the last period of ice sheet glaciation in British Columbia (Ryder and Clague 1989). The map produced by Dyke and Prest (1987) indicates that almost all of Vancouver Island was covered by ice except for the majority of Brook's Peninsula and a few isolated areas along the west coast. The Fraser Glaciation, and the post-glacial period of weathering and erosion, have resulted in U-shaped valleys that penetrate the Vancouver Island Mountains. The valleys are steep, with rocky mid to upper slopes and gentler lower slopes usually covered by thick glacial till (Clayoquot Sound Scientific Panel 1994). The Vancouver Island Mountains, with peaks up to 2200 m in elevation, occupy all of the island except for the small Nanaimo and Nahwitti lowlands, which lie below 600 m (Clague and Bornhold 1980). Clayoquot Sound (Fig. 3), like most of the west 5 Figure 2: Clayoquot Sound in relation to British Columbia 7 coast of the island, has mainly a rugged, rocky shoreline both on the exposed Pacific coast and along the fjords and inlets. This high-energy environment results in beaches that are for the most part small and confined to the shelter of inlets and bays (Province of British Columbia 1993; Clague and Bornhold 1980). The entire Clayoquot Sound area is roughly 262,000 ha in size. The watershed selected for investigation is the Clayoquot River watershed (Fig. 4), approximately 7,733 ha in size and considered to be an undeveloped watershed. According to Wilkerson (1990) an undeveloped watershed is one in which no more than two per cent of the area has been disturbed by human activity. (The) Clayoquot River watershed, like many of the remaining pristine watersheds within Clayoquot Sound, is composed predominately of older forests (age class 8 and 9, 141 years and older) which represent between 98 and 100% of the forest area (Clayoquot Sound Scientific Panel 1994). The majority of this watershed falls within the General Integrated Management Areas category under the April 13, 1993 land use announcement (Province of British Columbia 1993). Under this designation, areas "will continue to support various types of economic activity, including timber use and management, fisheries, wildlife, tourism, recreation and mineral exploration and development" (Province of British Columbia 1993). Two areas within the watershed have been classified as New Protected Areas and will remain protected from resource development. Under the category of old-growth forest, 1,800 ha of unlogged land along the western shore of Clayoquot arm of Kennedy Lake, part of the lower Clayoquot River and the entire area surrounding Clayoquot Lake, will remain protected. The second protected area within the watershed is part of the Clayoquot Plateau. This high-elevation ecosystem contains 29 rare plant species and numerous limestone caves (Province of British Columbia 1993). 9 The temperate climate of Clayoquot Sound is a product of the region's proximity to the Pacific Ocean. The area is characterized by abundant rainfall and strong winter storms from the Pacific Ocean. Intense rains are often associated with winds as severe as 150 km/h generated by cyclonic winter storms moving from west to east (Province of British Columbia 1993; Rollerson and Sondheim 1985). Climate stations in the vicinity of Clayoquot Sound show that winter daily rainfalls of 10-15 mm are not uncommon (Clayoquot Sound Scientific Panel 1995). The maximum 24 hour rainfall is in the order of 200 mm. The mean annual precipitation is about 3,300 mm, with the winter months accounting for roughly 86% of the precipitation (Rollerson and Sondheim 1985). Figure 5 shows mean monthly temperature and precipitation for weather stations situated near Clayoquot Sound. (The) Clayoquot River watershed, as does the rest of the Sound, lies predominantly within two biogeoclimatic zones. Biogeoclimatic zones (and their more detailed levels of classification; i.e., subzones, variants and phases) are a form of classification based primarily on climate, soil and vegetative data (Pojar and Meidinger 1991). The Coastal Western Hemlock zone (CWH) is the predominant zone within Clayoquot Sound and occupies elevations from sea level to 900 m. This zone is generally regarded as the wettest and most productive forest zone in British Columbia (Jones and Annas 1986). Western hemlock (Tsuga heterophylla) is the most common tree species and is mainly found at mid to low elevations (Pojar et al. 1991). Western redcedar (Thuja plicata), Amabilis fir (Abies amabilis) also occur frequently throughout this zone. Minor species include Douglas-fir (Pseudotsuga menziesii), Sitka spruce (Picea sitchensis) and yellowcedar (Chamaecyparis nootkatensis) (Rollerson and Sondheim 1985). The CWH zone is further divided into subzones which can be separated along gradients of continentality (hypermaritime, maritime and submaritime subzones) and precipitation (very dry, dry, moist, wet, and very wet) (Pojar et al. 1991). The first 10 month ly p rec ip i ta t ion (mm) 500 dai ly m e a n temperature (°C) 20 — _ Figure 5: Mean monthly temperature and precipitation for the Clayoquot region (Clayoquot Sound Scientific Panel 1995). subzone within Clayoquot Sound is described as the Very Wet Hypermaritime subzone. This subzone is mainly situated along the outer coast of Clayoquot Sound in areas of moderate relief where the maritime influences dominate the climate. Sitka spruce can be found growing within this subzone along exposed shores where winds carry salt 11 spray inland. Two variants have been described for this subzone; however only the Southern Very Wet Hypermaritime variant occurs in Clayoquot Sound (Clayoquot Sound Scientific Panel 1995). The majority of (the) forests within Clayoquot Sound occur in the Very Wet Maritime subzone. Two variants are described for this subzone based on elevation. The Submontane variant occurs from sea level to about 600 m while the montane variant lies between 600 and 900 m. Within the submontane variant, two phases are said to exist: 1) old-growth forests dominated by western redcedar with a smaller component of western hemlock (CH phase), and 2) second-growth forests dominated by western hemlock and amabilis fir, that are even aged and have originated from windthrow of the old-growth forest (HA phase). Lewis (1982) hypothesized that HA forests were a serai stage of CH forests. Much research has taken place to attempt to explain the difference between CH and HA sites. The majority of this research has been the product of the Salal Cedar Hemlock Integrated Research Program (SCHIRP)(Prescott and Weetman 1994). The second variant of the Very Wet Maritime subzone is the Montane variant. This variant occurs at the transition between low and high elevation forests. In the Montane variant, the HA phase contains a comparatively high proportion of amabilis fir, while the CH phase is dominated by yellowcedar rather than by western redcedar (Clayoquot Sound Scientific Panel 1995). The Mountain Hemlock zone (MH) represents the second biogeoclimatic zone in the area. This zone represents subalpine elevations of 900-1800 m within this area. Mountain hemlock (Tsuga mertensiana), amabilis fir and yellowcedar are the most common tree species within this zone, with yellowcedar generally confined to habitats with abundant moisture (Jones and Annas 1986). The majority of the soils belong to the Ferro-Humic Podzol soil landscape (Jungen and Lewis 1986). The soils are moist to wet over most of the year and rarely 12 freeze to any significant depth. The weathering that takes place under these climatic conditions and the influence of the forest has resulted in the production of amorphous complexes of soluble organic matter with the mobile compounds of aluminum and iron. These materials form a discrete, dark coloured podzolic B horizon (Clayton et al. 1977; Jungen and Lewis 1986). While the organic layer developed within the Podzol soil landscape is roughly only 10-30 cm thick, it-is critical that this organic layer be maintained. The presence of the organic layer helps prevent soil erosion by protecting the mineral soil; it contains the vast majority of the nutrients; it has high water-absorbing and water retaining capabilities as well as improving soil porosity and permeability, it also supplies most available nutrients and supports life in the soil (Clayoquot Sound Scientific Panel 1995). Because the terrain is rugged and steep, the most common material in this landscape is colluvium and till, with smaller amounts of bedrock, glaciofluvial, glaciomarine, and fluvial deposits. The till is generally found in two strata: a pedogenically weathered ablation till, overlying an unweathered massive basal till. Where slides occur in till, they are often at the interface of the two strata or just above the interface. Less often the failure plane is within the basal till (Rollerson and Sondheim 1985). Till most often occurs on gentle to moderate slopes within Clayoquot Sound. These deposits typically flank the lower valley walls as a blanket or veneer. The main pedogenic distinction between till and colluvial soils is the presence of a tough, indurated pan (Bc horizon) in most morainal soils (Jungen and Lewis 1986). Typically, the colluvium shows strong soil development down to the bedrock surface and, on unstable slopes, it is this surface which frequently controls the failure plane (Rollerson and Sondheim 1985). These soils are shallow (often between 1 and 2 m), well to moderately well drained, loose to friable and do not contain any sign of cemented horizons (Jungen and Lewis 1986). Distinctive landforms have resulted from 13 the downslope movement of weathered glacial materials and bedrock such as debris flow fans and talus cones, as well as a thin covering of rubble on steeper rocky slopes. Glaciofluvial materials, consisting of sand and gravel deposited by glacial meltwater streams, comprise the raised deltas near the heads of most inlets. Glaciomarine sediments (chiefly silt and sand), which accumulated in shallow marine waters when sea levels where higher, now blanket gentle slopes close to sea level. They are prone to surface erosion and landslides where protective vegetation is removed (Clayoquot Sound Scientific Panel 1995). 2.0 Types of Mass Movement The dominant erosional process over much of the forested landscape in the Pacific Northwest is mass soil movement (Rice and Krammes 1970; Swanston and Swanson 1976; Sidle 1980). Mass (soil) movement is simply the downslope movement of large quantities of soil, water and debris in various combinations by the process collectively known as landsliding. The types of movements can be quite variable depending upon the type of material being moved, as well as the amount of water incorporated in the failure (the more water involved in the failure, the more rapid the movement). A distinction should be made between mass soil movements as described above and the more visible process of surface erosion. Surface erosion is a process in which soil particles are detached and transported downslope by water. This process is a significant factor in contributing sediment to streams from localized sites such as cut banks and road surfaces (Sidle 1980). Varnes (1978) distinguishes five basic types of mass movement. They include, falls, topples, slides, flows, and spreads, as well as complex movements which are combinations of the various types of movements. Sidle (1980) suggested that three types of soil mass movement predominate in the Pacific Northwest. They are 14 1) shallow, rapid failures such as debris avalanches, debris flows and debris torrents; 2) slow, deep-seated mass movements such as soil creep, slumps and earthflows; and 3) single particle erosion such as dry ravel and rockfall. 2.1 Debris Avalanche-Debris Flow There is confusion in the literature when it comes to defining or distinguishing the differences between these two types of movements. The USDA Forest Service even goes so far as to say that "from a land management standpoint, there is little purpose in differentiating among the types of shallow hillslope failures, since the mechanisms and the controlling and contributing factors are the same" (NCASI, 6). Debris avalanches and debris flows most often occur on steep slopes where shallow, cohesionless soils have developed over impermeable glacial tills or bedrock. Quite often, the triggering mechanisms are large inputs of water into the soil mantle, from either intense or prolonged storms or rapid snowmelt or combinations of these processes (rain-on-snow events). Other mechanisms prone to cause failure are surface loading from such activities as sidecasting and removal of mechanical support from slopes caused by excavation. Initiation of mass movements commonly occurs within shallow, linear depressions or local swales on hillslopes. These depressions serve to concentrate soil seepage during periods of storm precipitation or rapid snowmelt, as well as cause the build-up of temporary perched water tables. Positive pore water pressure builds up in these depressions reducing the frictAonal strength of the soil material. Unlike debris avalanches and debris flows, debris torrents are initiated in steep first- or second-order intermittent stream channels where deep gullies or V-notch drainages serve as collectors of debris avalanche and flow materials from adjacent hillslopes. These topographic features are usually very steep and possess highly 15 unstable side slopes which frequently deposit rockslides and small debris avalanches with both organic and inorganic material into these confined channels. The material deposited may become confined behind temporary log jams until enough precipitation causes the dams to burst, producing large-volume, high-velocity debris torrents. These events have the potential to scour tributaries down to the bedrock and accelerate sediment delivery to downstream, lower-gradient channels (California Geology 1983; Fannin and Rollerson 1993). Their impact can be even more devastating if they deposit material within important fish habitat. 2.2 Deep Seated Mass Movements Soil creep is the slow (millimeters to centimeters/year), "continuous" downslope movement of the entire soil mantle. Soil creep is not usually thought of as a mass movement because of its slow rate of movement but it has potential to contribute sizable amounts of sediment to streams. Creep may occur in cohesive or cohesionless materials. This type of movement is locally important in coastal British Columbia where glaciolacustrine sediments are present. The geology and recent glacial history of the province however does not favor the development of deep, cohesive materials in which this movement predominates. In landscapes where rapid movements, such as debris avalanches or flows are the dominant process, soil creep along with gully sidewall collapse and ravelling is often the major process that recharges failure sites. For the most part, however, soil creep is not a dominant erosional process. 2.3 Dry Ravel and Rockfall Dry ravel involves the downslope movement of individual soil grains, aggregates and coarse fragments by gravitational forces. This process commonly occurs on steep, denuded, or sparsely vegetated slopes where cohesionless or single 16 grained surface soils predominate. Dry ravel results from the loss of mterlocking frictional resistance among dry soil aggregates or grains. Rockfalls are characterized by free falling, leaping, bounding or rolling of newly detached material from a cliff or steep slope. Falls are rapid to extremely rapid mass movements. When a large volume of material is involved, this may trigger failure of the material on the lower slopes. Rockfalls are an important type of mass movement process along the Pacific coast where exposed rock cliffs or steep rock slopes are present. The process is facilitated by frost action or water pressure acting on well-jointed or fractured rock bodies. 3.0 Soil Mechanics There is a high degree of heterogeneity in soil properties (unit weights, thickness, moisture content, permeability and strength parameters) even across the smallest of areas, which consequently makes obtaining detailed knowledge of the principal factors leading to slope failure sometimes difficult. Despite this variability, applying the principles and techniques developed in engineering soil mechanics and rock mechanics have proved helpful in risk analysis of mass movement. In the simplest of terms, mass soil movements are said to occur when the stresses acting on any hillslope exceed the strength of the materials comprising that hillslope. The ratio between the actual strength available and the stresses acting at the most failure-susceptible shear plane on any hillslope, gives an index of relative stability termed the "Factor of Safety". FS = available strength or shear resistance of material magnitude of disturbing or mobilized shear stress 17 Theoretically, when the calculated ratio is less than one, the slope should fail. Failure will be the result of an increase in stress due to conditions external to the slope, which changes the denominator, or, by internal processes that weaken the material, which concurrently lowers the numerator. However, the natural variability in the factors influencing the stability of slopes leads to considerable uncertainty in the estimates of factors of safety. Figure 6 illustrates the slide mechanism with translational failure surfaces. In figure 6A it is assumed that the potential failure surface is parallel to the surface of the slope, and is at a depth that is small compared with the length of the slope. The slope can then be considered as being of infinite length, with end effects being ignored. The force transmitted between two bodies in static contact may be resolved into components normal N (perpendicular to the interface) and tangential T (parallel to the interface) to the contact surfaces (Fig. 6A). When the shear slipping movement takes place along this surface the ratio T/N will have reached a limiting value which is termed the coefficient of friction (n). If it is assumed that the average value of T/N will remain constant for a given material, the limiting value of T may be written: where ^ = is the angle of internal friction In actual fact however, it is impossible to keep track of the forces at each individual contact point. Therefore, it is necessary to use the concept of stress. The summation across a given plane of all the normal forces divided by the area of the plane is the normal stress (a) acting upon the plane. Normal stress as it pertains to Fig. 6B can be expressed by the formula: T l i m =Ntan^ [1] [2] where y = unit weight ysat = saturated unit weight x = shear stress at failure plane a = normal stress at failure plane a = slope of failure plane u = pore water pressure at failure plane c = cohesion at failure plane Ac = root strength or change in root strength z = soil thickness above failure plane mz = ground water height above failure plane Figure 6: Slide mechanism with translational failure surface (adapted from Craig 1992; Sauder et al 1987). Figure 6A: Overview Figure 6B: Detail at failure plane 19 Similarly, the summation over that same plane of the tangential components divided by the area of the plane is the shear stress (x). Shear stress as it pertains to Fig. 6B can be expressed by the formula: r = |(l - m)y + my sat }z s m a c o s a [3] The shear strength (S) along any plane is a function of the normal stress (a) on that plane, or: S~f(a) [4] In 1776 Coulomb defined the function / as a linear function of the normal stress as follows: S = c+-tan</> [5] where c = cohesion cr = normal stress on slip surface <t> = angle of internal friction The Coulomb criterion is shown as a straight line in Fig. 7 with the intercept on the x axis equal to c and the slope equal to tan <f>. These characteristics (c and tan <f>) are termed cohesion and the coefficient of friction respectively. Cohesion is also enhanced by the contribution of rooting strength (Ac). The sum of c + Ac is called total cohesion. The shear stress as defined by Eqs. 4 and 5 represents the maximum shear stress that may be sustained on any plane in a given material. Any combination of shear and normal stresses that plots as a point below the strength function (e.g. point #) represents a safe state of stress, whereas a point on the line (e.g. point *) represents stresses that will result in failure of the material. Stresses that plot above the line (e.g. point $) cannot persist in this material because failure will take place before such stresses can be attained. Hence, the Coulomb strength function defines the limiting stress and is often called the failure envelope (Wu 1966). 20 Vi Vi CD C/3 Normal Stress Figure 7: Failure envelope (Wu 1966) 3.1 Cohesion (c) A vertical cut through a block of moist silt or clay would remain standing unsupported for some time. The same attempt with a dry sand would, on removal of the cutting implement, result in the slumping of the sand until its slope equals the normal angle of repose for that material. In silts and clays, therefore, some other factor must contribute to shear strength. This factor is called cohesion and results from the bonding between two particles that tends to hold them together in a solid mass without the application of external forces (Smith 1990). These particle-dependent forces are termed true cohesion and are the result of numerous factors. These include such things as Van de Waals forces, electrostatic attraction between clay surfaces and edges, linking of particles through cationic bridges, cementation primarily by organic matter and iron and aluminum oxides, and 21 surface tension at air/water surfaces between particles in unsaturated soils (Beekman 1987). In some cases however, cohesionless soils are able to exhibit an "apparent" cohesion. This apparent cohesion may be due to the effects of organic colloids within the soils, tree root anchoring (Ac) and other agencies (Sidle et al. 1985). This apparent cohesion is an external force and is not a physical property of the soil itself as is the cohesion created by clays. 3.2 Angle of Internal Friction (<J>) A parameter important to the determination of shear strength is the angle of internal friction (<j>). This angle represents the frictional resistance of a material due to particle to particle contact and the degree of interlocking of individual grains or aggregates. It is an important factor contributing to the shear strength of a soil and is influenced by several characteristics. Particle shape is regarded as the most significant of these characteristics. The friction angle of angular particles tends to be larger than rounded particles because of their greater mterlocking capabilities (Hammond et al. 1992). Similarly, well-graded soils (soils which have a wide distribution of particle sizes) have a higher friction angle (())) then those of poorly graded soils (soils which contain only a narrow range of sizes) because there is more particle mterlrcking (Hammond et al. 1992). With increasing density and decreasing porosity, both intergranular contact and degree of interlocking increase, thus the friction angle (<f>) increases dramatically. 3.3 Soil Moisture and Groundwater The previous discussion implicitly emphasized the important role the shear plane angle plays in overall slope stability. Another prominent factor affecting the stability of any slope is soil moisture. In general, saturated soil mantles favor mass wasting. Regional analyses in many areas of the world have related both seasonal rainfall and 22 individual storms to shallow, rapid landslides (Caine 1980; Sidle et al. 1985). Shallow rapid failures, such as debris avalanches, are generally triggered by a single storm (either long duration or high intensity) or by extensive rain-on-snow events. The influence of pore water pressure (pressure produced by the head of water in a saturated soil and transferred to the base of the soil through the pore water) due to increasing groundwater levels is widely recognized as the triggering mechanism for most slope failures. Pore water pressure at the shear plane as it pertains to Fig. 5B can be expressed by the formula: u = /jnz cos2 a [6] where yw= unit weight of water Studies evaluating the influence of pore water pressure on slope stability have located piezometers at the base of the soil profile. The basal discontinuity often represents the failure plane of shallow landslides. In a fully saturated soil, when an external stress is applied to that soil, the load will be carried by: (i) the water in the soil voids which results in an increase in pore water pressure; or (ii) by the soil skeleton which results in grain to grain contact stresses; or (iii) it will be shared between the water and the soil skeleton. At some point soon after application of a stress load, the applied total stress must be balanced by two internal stress components or else failure occurs. Pore water is able to transmit normal stress, but not shear stress, and is therefore ineffective in providing shear resistance. For this reason, the pore pressure is sometimes referred to as neutral pressure. Since pore water is ineffective in providing shear resistance, the concept of effective stress was developed. Effective stress is the stress transmitted through the soil fabric via intergranular contacts. It is this stress component that is effective in controlling both volume change deformation and the shear strength of the soil, since both normal stress and shear stress are transmitted across grain to grain contacts. Effective stress may be defined quantitatively as the difference between the total stress and the pore water pressure (Whitlow 1990). Hence: 23 a' = a-u [7] where a' = effective stress a = total applied normal stress u = pore water pressure In this case the effective stress is not the actual grain to grain contact stress, but the average intergranular stress on a plane area within the soil mass. In 1955 Bishop suggested a similar formula for a partially saturated soil: cr' = a-[ua-z(ua-uw)] [8] where ua = pore water pressure in the air uw = pore water pressure in the water X = parameter related to the degree of saturation In addition to pore water pressure, seepage forces also act on the soil mass and influence the behavior of the soil, causing deformation or shear failure. Seepage is simply the movement of water through a soil mass. The movement of water through a soil mass is from zones of higher, to lower, pore water pressure. When considering problems of water flow, it is usual to express a pressure as a pressure head, or head, measured in meters of water. Bernoulli's equation states that total head (H) causing water flow is the result of three head components: where hz = position or elevation head j- = pressure head due to pore pressure u jz; = velocity head where the velocity of flow is v In problems of groundwater flow, the last component, velocity head is usually ignored because velocity and osmotic components are considered negligible. The first two terms represent the head causing the water to flow through a mass of soil. In 1856 Darcy showed that the velocity of a fluid through a porous medium is proportional to the hydraulic gradient. 24 vcc/'orv = ki [10] where v = water flux density k = the flow constant or coefficient of permeability A H i = the hydraulic gradient = /VL AH = difference in total head between two points in soil. When AH i s expressed per unit length over which the difference exists, it is called the hydraulic gradient. The quantity flowing is therefore given by: q = Av = Aki J-JJJ where q = quantity flowing in unit time A = area through which flow is taking place Figure 8 illustrates quasi one dimensional flow of water in soil. Figure 8: Quasi one dimensional flow of water in soil (Whitlow 1990) 25 The effect of upward flowing water in a soil is for seepage pressure on the particles to reduce the intergranular or effective stress. If a sufficiently high flow rate is achieved the seepage pressure can cancel out the effective stress causing a quick condition. This is essentially a condition in which the soil has no shear strength, since the intergranular stress has been reduced to zero. Chatwin (1991) suggests that the risk of slope failures caused through saturated soil conditions can be reduced substantially by avoiding logging activities during particularly intense single storms. Soil mass movements often occur once a threshold rainfall intensity has been exceeded. Figure 9 shows a histogram which suggests that there is a high probability of failure if the rainfall intensity (the total amount of rainfall over a specific time period) exceeds the values shown. These values are said to represent the minimum intensities needed to trigger a slide assuming that the soil is saturated and slopes are vegetated. 3 0 0 2 5 0 2 0 0 tn 1 1 5 0 1 1 0 0 5 0 0 12 2 4 4 8 7 2 9 6 T i m e (hr) Figure 9: Rainfall intensities (with wet antecedent conditions) commonly associated with landslide activities (Chatwin 1991) I R a i n f a l l i n t e n s i t i a s a s s o c i a t e d w i t h l a n d s l i d e s R a i n f a l l i n t e n s i t i e s b e l o w t h r e s h o l d l e v e l s 26 2.4 Other Triggering Mechanisms Local geology, rainfall and groundwater are not the only factors capable of producing mass movements. Seismic events may also trigger soil movements; however, the underlying cause of most slope movements is undoubtedly water pressure increases caused by prolonged or intense rainfall. After forest removal, the gradual decay of tree roots often predisposes soils on steep slopes to failure. Plant root systems are important stabilizing components of steep, vegetated slopes subject to mass failure. Research has revealed an increased frequency of shallow, translational landslides occurring about 3 to 10 years after clearcutting (Megahan et al. 1978). In areas severely burned following forest removal, minimum root strength may occur even sooner (Prellwitz 1989 cited in Hammond et al. 1992). This period of increased landslide occurrence corresponds to the minimum rooting strength of the site following initial decay of the roots from the original vegetation and prior to substantial regeneration of trees (Sidle 1992). Roots add strength to the soil by laterally tying the slope together across zones of weakness or instability and by vertically anchoring through the soil mantle into fractured bedrock, although this latter method of stabilizing slopes is rarely present in coastal British Columbia (Gray and Megahan 1981; Ziemer and Swanston 1977). Root reinforcement is only effective in stabilizing relatively thin soils (that is, about 1 m), such as those found on steep dissected slopes of coastal British Columbia and Alaska. Dense networks of medium and small sized roots reinforce the upper soil layer so that it acts as a membrane to provide lateral strength and increased slope stability (Sidle 1985). The combined effect of this membrane strength and possible vertical anchoring to the bedrock is evident by the location of landslide failure planes just beneath the zone of maximum root density on deforested slopes (Wu and Swanston 1980; Ziemer and Swanston 1977). O'Loughlin (1972) calculated that the root network accounted for 71 % of shear strength at saturation of the till soils on slopes of 35° within coastal 27 British Columbia. Reneau and Dietrich (1987) argued that lateral root strength controls the size of shallow colluvial landslides in hollows and that a critical depth and size of cohesionless colluvium was necessary to overcome lateral root strength and basal friction to induce instability. Figure 10 shows hypothetical curves indicating root strength deterioration following timber harvesting, rooting strength of regenerating site vegetation and net rooting strength. 0 2 4 6 8 10 12 14 16 18 20 Time since cutting (years) Figure 10: Hypothetical curves indicating root strength conditions over time (Sidle 1985) 4.0 Slope Stability Assessment Various methods have been used to assess potential slope instability. They range from simple inventories of existing mass movements to mathematical models (Niemann and Howes 1992). Gerath et al. 1993 provide an excellent summary of the various hazard mapping methods. Table 1 outlines the range of possible methods of 28 analysis and their strengths and weaknesses. Appendix A also reviews 17 slope stability studies that have been undertaken along the west coast of Table 4.1 Summary of hazard mapping methods (Gerath et al. 1993) Method of Analysis Map Scale* Strong Points Weak Points Landscape inventories S-D -objective -useful data base -qualitative -no prediction Landslide activity M-D -objective -useful data base -no prediction Landslide density S-L -objective -quantitative -no prediction Subjective geomorphic S-D -flexible -makes use of expert skills -useful data collection -qualitative -difficult to review -requires high skill Subjective rating S-L -flexible -makes use of expert skills -work can be delegated and checked -qualitative -danger of oversimplification Univariate susceptibility M-L -objective -shows effect of factors -qualitative -relies on data quality Univariate probability M-L -quantitative -flexible -danger of wrong factor selection -relies on data quality Multi-variate probability M-L -quantitative -precise -danger of wrong factor selection -high reliance on data quality -difficult Stability M-D -quantitative -can be reviewed -shows influence of some factors -danger of oversimplification -may conceal lack of knowledge * S - Small scale maps (> 1:100,000); M - Medium (1:25,000 to 1:50,000); L - Large (1:5,000 to 1:20,000); D - Detailed (1:5,000 to 1:500). 29 North America. Many of these studies fall within one of the mapping methods as outlined by Gerath et al.(Table 1). The main objective of the review of studies in Appendix A was to examine the range of variables or landslide parameters that were used in the assessment. Many of the 17 studies reviewed were simply landslide inventories. Landslide inventories offer limited information in that their applicability rests on the skills of the interpreters. In addition to this, the ability to extrapolate the results to areas that have not failed is difficult (Niemann and Howes 1992). In the landslide activity method, information from several time periods (usually serial aerial photographs) is used. A limitation of this type of analysis is that it does not recognize areas which are not active at present, but may be potentially unstable nonetheless (Gerath et al. 1993). The landslide density method of analysis sets out to calculate landslide densities for an area. This may be done in one of three ways: 1) average number of landslide sites per unit area in a map unit; 2) percentage unstable area in a map unit; and 3) contours of equal landslide density. The limitations of this method are the same as those encountered in landslide inventories (Gerath et al. 1993). Subjective (geomorphic) or qualitative techniques are commonly used by geologists in mapping landslide hazard sites (Duncan et al. 1987). The tools used for producing a map of this type include aerial photo analysis and ground truthing. The mapper then uses experience, field observations and best judgment before subjectively assigning a hazard rating to the various polygons. The experience of the mapper dictates the quality and usefulness of the map (Gerath et al. 1993). However, this technique lacks the site specificity often required by engineers (Duncan et al. 1987). Closely related to this is the method described as subjective rating analysis. In this method an algorithm is developed for the assignment of subjectively set hazard 30 classes (Gerath et al. 1993). The hazard classes are assigned a subjective weighting based on several relevant attributes. Duncan et al. (1987) conducted a study in the Coast and Cascade ranges of western Washington and western Oregon which blends the qualitative methodology of Swanston et al. (1979) with quantitative analytical methods. Each stability variable was assigned a relative weighting value. The final product was a multivariate equation that could "be used by logging engineers as an aid in road location design and construction in areas of questionable slope stability." (Duncan et al. 1987 p 152). In British Columbia two subjective rating methods are used operationally (Young 1992). They are Environmentally Sensitive Area (ESA) Mapping and Slope Hazard Mapping (SRS). Environmentally Sensitive Areas The objectives of the ESA classification are: A. To identify, for forest management, areas that are environmentally sensitive or have values for other resources, including: • Areas having actual or potential, fragile or unstable soils that may deteriorate unacceptably after harvesting (ES); B. To identify the importance of streams, or stream reaches, to fish and the sensitivity of streams to forest harvesting; C. To provide site-specific data on environmental sensitivity and on other resource values for consideration by forest planners and managers in the determination of the rate, location and timing of timber harvesting (Forest Inventory Manual 1991) Two ESA soil classes are recognized: high (ESj) and moderate (ES2). The high (ESj) class applies to areas having extremely fragile or unstable slopes where harvesting is likely to be restricted because it would lead to unacceptable site deterioration. Unacceptable site deterioration includes: i) severe lowering of site productivity; ii) extreme delay in the re-establishment of protective vegetation and 31 forest cover; iii) long term loss of the productive land base; and iv) severe lowering of the quality and downstream water and degradation of fisheries habitat. The moderate (ES2) class applies to areas having significantly fragile or unstable soils but less than those designated as ESi (Forest Inventory Manual 1991, Environmentally Sensitive Areas, p. 4). ESA sites are selected through aerial photo interpretation, air and/or ground truthing when possible and by specialists having knowledge of the local area. Slope Stability Mapping (SRM) This mapping procedure begins with terrain mapping. The Terrain Classification System (Howes 1988; Ryder and Howes 1984) is a scheme designed for the classification of terrain polygons homogenous with respect to materials (genetic origin, texture), surface expression (including slope angle, slope form and material thickness) and modifying geomorphic processes (Young 1992). Mapping and data collection are carried out by a combination of aerial photo interpretations and field surveys. These terrain inventory maps can then be used to prepare interpretive maps for slope stability and surface erosion potential and other management needs such as gravel sources (Schwab 1993). In the case of assessing slope instability, each polygon is subjectively assessed according to its failure potential following logging. Forest companies, the B.C. Ministry of Environment and the B.C. Ministry of Forests, have variously used two, three, four, and five class stability systems (Rollerson and Sondheim 1985). Today, the five class stability rating is the standard. Table 2, shows the major management implications expected within the classes. Similar tables are generated for surface erosion potential ratings (see Table 3). According to Schwab (1981) citing (Briere 1980), the stability hazard mapping system serves four important roles: 32 Table 4.2 Terrain Stability Hazard Classification (Schwab 1993) Class Stability Hazard Management Implications I Stable: -no stability related constraints; -no evidence of potential instability -regular road maintenance inspection and repairs. n Generally stable: -no stability related constraints; -minor inclusions of sensitive terrain -regular road maintenance inspection and repair; -road deactivation consideration in planning. m Moderately stable: -engineering review of roads constructed -presence of some sensitive sites but no through sensitive sites; natural slope failures; -regular road maintenance inspection and -low potential for slope failure; repair of ditches and culverts required; -minor stability and erosion problems should -seasonal deactivation during suspended road be expected particularly with road cut banks: use; -semi-permanent deactivation required when roads are no longer in regular use or not maintained. rv Marginally stable (ES2): -a field review by a qualified terrain -terrain units with similar geology to specialist or a geotechnical/geological Class V; generally steep slopes, a few engineer is required in order to assess in small natural slope failures are some detail the stability of the area prior to any time present. development. -a high potential for logging related -harvesting with lull suspension systems. slope failure; -road should avoid these sites; construction -extreme potential for road related is generally not permitted. failures; -roads where constructed, must be fully engineered; on site supervision required during construction. -regular road maintenance inspection and repair; carry out maintenance, before fall rains, after major storms and during spring runoff; -permanent road deactivation and re-sloping when roads are no longer in regular use or not maintained. V Unstable (ESI): -slope stability and erosion concerns take -the presence of active or recurrent slope precedence over timber development: in failures. general, these sites are considered protection forests and are not available for timber harvesting. -a detailed field review by a terrain specialist and/or geotechnical/geological engineer is required prior to any decision on timber harvesting. 33 Table 4.3 Surface Erosion Potential Rating (Schwab 1993) Class Rating Description Management Implications V L Very low potential flat or gently sloping terrain, organic soils, flood plain. disturbance of streams could initiate some bank and channel erosion. L Low potential gentle slopes, short slopes. expect minor erosion of fines from ditch lines and disturbed soils; exercise care not to channelize water on to more sensitive sites. M Moderate potential moderately steep slopes and long slopes; erodible soil textures, fine textured materials. plan preventive remedial actions for disturbed slopes; expect problems with water channelized down road ditches and across disturbed areas; water management is critical; plan for complete road deactivation; grass seed all disturbed sites. H High potential moderately steep slopes and highly erodible soil textures. major problems exist with water channelized on to or over these sites; problem avoidance may permit road development; immediate revegetation of all disturbed areas. V H Very high potential steep slopes erodible soil textures, active surface erosion or gully erosion. severe surface and gully erosion problems exist; erosion concerns take precedence over timber harvesting; protection of forest: no development Surface erosion potential ratings are a qualitative assessment of the potential for sediment generation during and after logging development. Areas of major concern are sensitive landforms, roads, recent landslides, and sites excessively disturbed by yarding or site preparation. Ratings are intended to "flag" potential problem terrain. 1) a basis for evaluating the spatial distribution of various stability classes; 2) a means for transferring experience and knowledge from a study area to similar but inaccessible or unstudied areas; 3) a framework for assessing management strategies and predicting outcomes; and 4) a means for communicating among managers, field personnel, researchers and the public. The most common landslide hazard assessment method in use today is that of modeling. The modeling exercises can range from correlating a simple, single variable 34 such as slope gradient or soil type with the occurrence of mass movements through multi-variate analysis (Niemann and Howes 1992). The univariate or single variable approach involves two fundamental premises: 1) A relative (susceptibility) correlation which suggests that land units which have similar critical attributes to areas which have failed previously are also likely to fail; and 2) An absolute (probability) correlation which suggests that future landslide frequency in areas which have similar characteristics can be predicted, based on known frequency of occurrence in failed units, over a given time period (Gerath et al 1993). The fundamental premise behind the univariate susceptibility method is that slope stability is dictated by a single variable or different variables which are examined separately. Duncan et al. (1987), examined each variable in their study and assigned a relative weighting value, which subsequently went into making up a stability index which combined all the variable values. Rollerson (1992) also tested variables using the univariate method in a study on the Queen Charlotte Islands. He analysed post-logging landslide frequencies as they related to individual landscape attributes. The results of this portion of the study were then directed toward the development of multifactor classifications for slope hazard mapping. One of the main findings of this report, as was the case in Rollerson and Sondheim (1985), was that different landslide attributes were significantly different for clearcut versus road failures. The need to develop separate predictive classifications was advocated. In the univariate probability analysis a direct statistical correlation is sought between the probability of landslide occurrence and a single variable. The variable is usually formed as a prescribed combination of several factors (Gerath et al. 1993). Howes (1987) presented a terrain evaluation method for predicting landslides following clearcut harvesting operations in the southern Coast Mountains of British Columbia. The fundamental premise of the study was that a detailed landslide inventory was to be 35 done in conjunction with a surficial geological mapping program. The data from the landslide inventory and the surficial geology map were integrated to identify hazardous areas, particularly after clearcutting has occurred. Fifteen terrain classes were defined though unique combinations of site factors. Various statistics (i.e. number of rapid movements from natural, clearcut and roadways, total areas and frequency rates) were then calculated for each of the terrain classes. Multivariate probability analysis, which uses multiple-regression techniques and discriminant analysis to correlate landslide frequency and a group of attributes, has become widely used. Discriminant analysis is a classification method that maximizes the separation between groups of data using functions of the original data values (Wadge et al. 1993). Rice and Lewis (1991) used linear discriminant analysis as a means of estimating the risk that logging or road construction would lead to a "critical" rate of erosion, (> 100 yd^ /acre) at a site. The study was conducted in northwestern California. Two equations were developed (one for roaded areas and one for logged areas). Furbish and Rice (1983) used linear discriminant analysis to produce an equation that could be used to provide an assessment of risk for undisturbed terrain. The results of their analysis reveal that post-logging failure is most likely to result near actively scoured terrain, just below major convex breaks of slope and within drainage depressions. The major drawback of multivariate approaches, as pointed out by Gerath et al. (1993), is that they remove any application of the mapper's experience and judgment when assessing the study area. The results are totally dependent on the quality of the data going into the model. The method of stability analysis, unlike the previous methods, relies on engineering techniques to directly assess the site. The analyses aim to establish an average factor of safety. This method is quite reliable and accurate for assessing the strength-stress relationships in a small geographic area. However, obtaining the 36 necessary data for the soil properties, geology, and ground water hydrology for even a small area is both costly and time consuming. These drawbacks make this method an impractical tool when broad hazard assessment is necessary (Duncan et al. 1987; Sidle et al. 1985; Swanston 1981). Swanston (1981) presents an approach which uses both the qualitative techniques mentioned earlier along with the quantitative technique mentioned above. This approach requires careful assessment of the landscape parameters obtained from aerial photos, maps and field observations, coupled with basic information on the engineering properties of specific soil materials. 5.0 Methods Slope stability studies have relied extensively on aerial photographs, topographic maps and field surveys to acquire the necessary data. These data collection methods however, are becoming increasingly expensive to undertake and can be quite time-consuming largely due to the field survey. The need is for a method of stability assessment that is both less expensive and more easily acquired. This thesis attempts to describe such a method and is intended to be used as a first approximation for assessing the landscape for the Clayoquot Sound region. The Forest Practices Code Mapping and Assessing Terrain Stability Guidebook (1995) mention that reconnaissance type assessments, such as that attempted in this thesis, are useful in identifying unstable or potentially unstable land areas from a broad perspective. They can be valuable tools for outlining areas needing further or more concentrated analysis through detailed mapping. At the heart of this assessment lies the use of Digital Elevation Models (DEMs) and Geographic Information Systems (GISs). A DEM is a computer-based ordered array of numbers representing the topography of all or part of a planetary surface and comprises a fundamental data layer in GISs (Mackey et al. 1994; Moore et al. 1991). The most commonly used data structure in DEMs is the regular square grid in which 37 elevations are available as a matrix of points equally spaced in two orthogonal directions (Tarboton et al. 1991). Other methods of data storage include triangular integrated network (TIN) or vectorized contours, stored in a digital line graph (Zhang and Montgomery 1994). A GIS is a computer-assisted system for the acquisition, storage, analysis and display of geographic data (Eastman 1992). According to Wadge et al. (1993), GIS has an important role to play in hazard assessment because of the following advantages over traditional methods: 1) spatial modeling and map creation can be done on the same computer; 2) a variety of models can be applied to a terrain and displayed to reflect different hazard scenarios in forms other than the traditional map; and 3) the implications of hazard in terms of risk and planning can be made understandable to planners. The use of DEMs and GISs has the potential to greatly reduce the time required to develop preliminary terrain evaluations (Niemann and Howes 1992). These authors also point out that when published maps of bedrock geology or surficial material characteristics are used in conjunction with data derived from a DEM, an approximation to the traditional multifactor map product is possible. This thesis describes such an approach. Morphological characteristics derived from a DEM are combined with surficial geology in a GIS context to classify (the) terrain within (the) Clayoquot River watershed. The method could assist the engineer or geoscientist in more detailed subsequent studies on those areas targetted as high risk. Creation of the DEM is the first stage in the acquisition of the variables needed for the hazard classification. Terrain Resources Inventory Management (T.R.I.M.) digital topographical data were obtained for the Clayoquot River watershed. This database was then transferred into Terrasoft 10.3c geographic information system software. Because the terrain of (the) Clayoquot River watershed is deeply incised with 38 many steep slopes, the Proximity of Inverse of Distance Squared (SIWDF-PROX) model was used to generate the most representative DEM possible. A 20 m X 20 m regular square grid was used. Zhang and Montgomery (1994) suggest that grid size significantly affects computed topographic parameters. As grid size decreases, landscape features are more accurately resolved; however the spacing of the original data used to construct a DEM effectively limits its resolution. Decreasing the grid size beyond the resolution of the original survey data does not increase the accuracy of the land surface representation of the DEM, and potentially introduces errors. Their analyses from watersheds in Oregon and California showed that a 10 m grid size presents a rational compromise between increasing resolution versus data handling requirements for modeling surface processes in many landscapes. Despite this argument, 20 m resolution was chosen because of the size of the database, the processing time involved, the computer memory required, and the accuracy of the T.R.I.M. data. This information was then imported into IDRISI. IDRISI is a grid-based geographic information and image processing system developed by the Graduate School of Geography at Clark University (Eastman 1992). It is a software system designed to facilitate professional-level geographic research on a low-cost, non-profit basis. Besides being low-cost and relatively easy to use, IDRISI allows users to develop independent modules in any computer language. This particular advantage was useful as programs in FORTRAN 77 programniing language were already written. These programs were then modified as required. A 586 and 486 computer platform was used in conjunction with IDRISI to produce the final output. 5.1 Potentially Mappable Variables The overriding observation from reviewing the studies listed in Appendix A was the plethora of mappable stability attributes that were involved. In some cases, the 39 attributes were selected in a subjective manner, and in others they were arrived at through quantitative or statistical analysis. Table 4 lists nine attributes which can be used as indices of slope instability. The list is far from complete, but is adequate for the purpose of demonstrating the process. Lewis and Rice (1990), in a study conducted on private timber lands in northwestern California, collected data for 172 variables within 415 timber harvest plans. However, only 44 of these 172 variables were intended for statistical analysis and of these 44, 12 were eventually discarded due to problems involved in field measurement or in analysis. Like any modeling exercise, practical mapping systems should attempt to make use of the smallest number of relevant variables that yield a usable forecast. Rice and Lewis (1991), in defense of the two, three-variable equations they derived through linear discriminant analysis, suggest that there has been a tendency to retain variables that may be useless or even detrimental to the quality of a forecast. As the main focus of this thesis is to forecast slope stability from digital elevation data, the variables of interest from Table 4 belong to the category labeled slope morphology. Digital Elevation Models can be used to derive a wealth of information about the morphology of a land surface (U.S. Geological Survey cited in Jenson and Domingue 1988). Reliable estimation of topographic parameters, which reflect terrain geometry, is necessary for geomorphological, hydrological, and ecological modeling because terrain controls the fluxes of mass in the landscape (Mitasova and Hofierka 1993). A complete topographic characterization of a point includes measures of both its surface geometry and its relative position within the landscape (Speight 1974 as cited in Martz and De Jong 1988). According to Martz and De Jong (1988) acquisition of a point's slope gradient, slope aspect, profile convexity (downslope curvature) and contour convexity (across slope curvature) on a surface describe that point. The relative position is generally described in the context of some process or flow assumed to be taking place within the landscape. Table 5.1 Checklist of Slope Attributes (Gerath et al. 1993) Bedrock lithology rock type/s, alteration physical properties strength, swelling potential etc. block size typ. joint spacing fabric attitudes dip, dip direction (relative to slope) weathering degree of weathering Soil origin genetic terms (e.g. fluvio-glacial) (unconsolidated texture typ. grain size, uniformity near-surface physical properties rel. density, strength, cementing etc. material) qualifying descriptors active/inactive modification of deposit form of deposit e.g. fan, blanket, terrace thickness typical or range Contrasting layer description e.g. loose veneer, or pedological type surficial layer thickness typical or range (including solum) substrate description e.g. dense t i l l , bedrock nature of contact e.g. weak horizon? Slope morphology angle typical or range uniformity or curvature e.g. straight, stepped, concave position relative to crest, toe or valley floor elevation typical or range length measured along fall line aspect fall line azimuth (typical or range) lateral curvature e.g. narrow gully, ridge, re-entrant Groundwater permeability typical or range seepage regime e.g. discharge area, concentration surface drainage relative facility or surface runoff limiting layer impervious layer depth of limiting layer typical or range depth to groundwater table typical or range precipitation intensity Vegetation forest type species, condition, spacing forest age avg. age or height harvest/fire history time since harvest, treatments other vegetation vegetation other than trees Identified processes slope instability type of instability (landslide class) erosion type of erosion other processes e.g. solifluction, permafrost frequency no. of events per unit area/time areal extent area affected by events thickness depth of instability (typ., range) Incipient processes instability signs e.g. cracks, bent trees, scarps frequency no. of signs/unit area Human activity type e.g. roads, f i l l , ditches quantity e.g. road length/unit area 41 The literature suggests that when it comes to selecting variables for mapping or modeling involving slope stability, one observation that can be made with a reasonable amount of certainty is that slope angle or gradient is nearly always the most important factor. 5.1.1 Slope Gradient Slope gradient, the first spatial derivative of elevation influences the rate of water and sediment flow by controlling the rate of energy expenditure or stream power available to drive the flow (Zevenbergen and Thorne 1987). In general, landslide hazard increases with increasing slope gradient (Megahan et al. 1978). On hillslopes with shallow, semi-cohesionless materials, slope gradient is a major indicator of natural landslide hazard (Howes and Swanston 1991; Swanston 1981). Most large landslides seem to occur on sites which fall within a relatively narrow range of slope steepness. O'Loughlin (1972) in his study in coastal British Columbia found that over 75% of the measured landslides occurred on slopes ranging from 31° to 39°. Similarly, Megahan et al. (1978), in a study in the northern Rocky Mountains of Idaho, found that most slides occurred on slopes of about 30°. In both studies, slopes which are too steep (> 40°) show little mass movement action simply because they are either too rocky or soil materials have little chance to accumulate and therefore move off the slopes as either dry creep or rock falls. 5.1.2 Slope Shape Another landscape feature which exerts a strong influence on overall slope stability is slope shape, the second derivative of elevation with respect to horizontal distance. The major influence of slope shape on the soil mass is the rapid recharge of water in certain parts of the soil mantle during rainstorms (Sidle et al. 1985). Slope shape is described as being concave, convex, or planar and can be measured in both 42 profile and planform. Profile curvature, (Figure 11 A) reflects the change in slope angle, and thus controls the change of velocity of mass flowing down along the slope curve and therefore influences aggradation and degradation. Planform curvature, (Figure 11B) the curvature of the land surface perpendicular to the gradient, reflects the change in aspect angle, and influences flow convergence and divergence (Mitasova and Hofierka 1993; Zevenbergen and Thorne 1987). CONCAVE CONVEX STRAIGHT Figure 11 A: Profile curvature slope shapes AIR A I R S O T L CONCAVE CONVEX A I R S O I L STRAIGHT Figure 11B: Planform curvature slope shapes Slopes that are concave serve to concentrate recharge water into small areas of the slope. These concentration points are more prone to develop perched water tables so that pore pressures rise more rapidly and to a greater extent than on other parts of the slope. The frequent association of shallow, rapid landslides with hillslope depressions and V-notch gullies is, therefore, not surprising (Sidle 1985; Sidle et al. 43 1985). Rollerson (1992) notes that curvature along the horizontal plane shows that concave (often in first-order headwater basins) have twice to three times as many failures as other landscape positions. Concave slopes have their steepest gradients in their upper portions (Swanston 1981). Convex slopes, on the other hand, serve to disperse subsurface water and therefore tend to be more stable than concave shaped slopes. Convex slopes have their steepest gradients in the mid-lower portions (Sidle 1985; Sidle et al. 1985; Swanston 1981). 5.1.3 Aspect Aspect is a site variable which has been found to be a useful indicator of site predisposition to failure (NCASI 1985). Aspect defines slope direction and therefore the direction of flow. Knowledge of how aspect varies throughout a catchment provides the information necessary to determine what upslope land area contributes to the flow at any point in the catchment (Zevenbergen and Thorne 1987). Aspect may be an indicator of site susceptibility to landsliding because it is commonly correlated with: a) climatic conditions; b) geologic and soil conditions; and c) type and amount of vegetation. Rollerson and Sondheim (1985), in their study of MacMillan Bloedel's TFL 44, included aspect as one of their six variables. The logic behind including aspect was that if significant local rainshadow effects exist, east-facing slopes in the study area might be less failure prone. Their results suggested that this indeed was true: east-facing polygons with no natural failures were rated as having a low sensitivity to post-clearcut slope failures (Rollerson and Sondheim 1985). O'Loughlin (1972) found that approximately 75 44 percent of the landslides investigated in coastal British Columbia possessed southerly aspects. This relationship however, may have resulted from Pleistocene ice sheet movement which resulted in north facing slopes being rocky and broken, a condition which discourages landslide formation, while south facing slopes are relatively uniform and underlain by an extensive unweathered till substrate. Furbish and Rice (1983) in a study in northwestern California found aspect was an excellent variable in predicting landslides for relatively small areas (54 mi^ ), but found that it was almost useless for a much larger area (1,027 mi2, spanning 2° of latitude). They concluded that, although aspect might be useful in a restricted climatological and geomorphological setting, its effect was masked by other, more important variables in a larger area with greater diversity. 5.1.4 Slope Position Rollerson (1984, 1992) found slope position to be a significant categorical variable. In his 1984 Terrain Stability Study of TFL-44 in both clearcut and road landslide failure frequency, stream escarpments, headwater drainage areas and lower-mid slopes were the slope positions most likely to fail. Similar conclusions were noted in his 1992 study of the Queen Charlotte Islands Duncan et al. (1987), in their slope stability index rating table, provide weighting values to represent the relative contributions of the individual causative variables to slope stability. Weighting values ranged from -10 for a ridge-top slope position to +8 for a slope angle in excess of 35°. They gave the upper third slope position a numerical value of 3, the middle third slope position a value of 2, and the lower third a 1. 45 5.1.5 Transient Snow Zone Beaudry (1984) suggests that the mountainous region of southwestern British Columbia could be divided into 3 zones of elevation based on the way in which precipitation reaches the ground. 1) A low elevation zone, where precipitation is almost exclusively in the form of rain. 2) A mid elevation zone, where snow will accumulate and melt throughout the winter responding to temperature fluctuations generally between -5°C and +5°C (the transient snow zone). 3) A high-elevation zone, where snow usually begins in early October, and continuous melt begins in late March. When precipitation in the form of rain falls on relatively shallow snowpacks (rain-on-snow events) within a mid-elevation zone, rapid' melting of the snowpack can result in higher rates of water input to a soil than would commonly result from rainfall alone. This can lead to higher pore-water pressure in soils on unstable slopes and higher peak streamflows than would occur from rainfall alone (Harr 1986). The transient snow zone within southwestern British Columbia ranges in elevation from 300 m to 900 m (A. Chapman, MoF, personal communication). The above mentioned process is said to become even more of a concern when logging takes place within this mid-elevation zone, since logging may alter snow accumulation and subsequent melt during rainfall. This could then increase hillslope and channel erosion by altering the rate of water input to soils and by increasing the size of peak streamflows (Christner and Harr 1982 as cited in Harr 1986). 5.1.6 Shallow, Linear Depressions or "Swales" . Recent work with DEMs has led to the development of algorithms for the a extraction of channel networks, flow vectors, delineation of watershed boundaries and even sub-basins (Donker 1992-93; Fairfield and Leymarie 1991; Helrnlinger et al. 46 1993; Jenson and Domingue 1988; Martz and de Jong 1988; Martz and Garbrecht 1992; McCormack et al. 1993). The extraction of this type of data can be potentially helpful in the assessment of hazards. One interesting by-product of the partitioning of watersheds into sub basins is the potential to forecast where swales may be located, or where they may be likely to form. Swales on hillslopes are common points of origin for debris avalanches and debris flows (Howes 1987; Swanston and Howes 1991). Without doing fieldwork or having large scale aerial photographs, acquiring knowledge of these locations is difficult. Montgomery and Dietrich (1988, 1992, 1994), in research on soil-mantled regions of western United States, have generated results which support a predictive relationship between source area and steepness of slope for humid landscapes where channel initiation occurs by landsliding, overland flow, and to a lesser extent seepage erosion and channel-head sloughing. Their results have shown that an inverse relation between the drainage area contributing to the channel head (source area) and the local valley slope defines a topographic threshold between channeled and unchanneled regions of a landscape (Montgomery and Dietrich 1992). This relationship indicates that the greater the local valley slope, the smaller the threshold area defining the initiation of a channel, as is generally observed. Instability of channel heads can generate destructive debris flows and torrents or gullying, and heavy sediment loading downstream (Dietrich and Dunne 1993). Data from nine study sites regarding source area slope relationships representing a broad range of geology, vegetation and climate are shown in Figure 12. These data show a steepening in the source area-slope relation at gradients greater than tanO =0.5 (26.5°) (where 9 is the local slope) which the authors associated with a change in process dominance from overland flow on gentler slopes to landsliding on steeper II)1 47 10 -I 1 I) slope (m/m) Figure 12: The source area-slope relationship for 9 watersheds along the Pacific Coast for channel initiation (Montgomery and Dietrich 1994b). slopes. Figure 13 illustrates the expected source area-slope relation in a landscape with a range of slopes in which both overland flow and landsliding control channel initiation (Montgomery and Dietrich 1994). 5.1.7 Surficial Materials Surficial materials (called engineer's soil) refers to unconsolidated deposits such as sand, gravel and glacial till, whereas soil refers to the uppermost 1-2 m of these materials that has been modified by physical and biological processes (Clayoquot Sound Scientific Panel 1994). Surficial material is the chief criterion used for subdividing the land surface into terrain units (Ryder and Howes 1984). Surface materials are an important component of a watershed because they support plant growth, regulate the flow of water and the supply of sediment to streams, and provide the substrate for roads. In addition, it is widely recognized (e.g. Ryder 48 73 I 0 l ) slope (m/m) Figure 13: Source area-slope relationship in a landscape with a range of slopes (Montgomery and Dietrich 1994b). and Howes 1984), that exanuning the characteristics and properties of terrain is an important step in planning for work on steep terrain. Knowledge of surficial materials in a physical sense can be used to assist land management practices in the avoidance of destabilization. Recent research in coastal British Columbia has found that different surficial materials have different critical angles at which they fail (Howes 1987). For example, 26° and 33° appear to be important slope breaks for open hillslope (hillslopes not dissected by gullies) till landslides, while 30° and 36° appear to be important on open hillslope colluvium landslides. Likewise, Rollerson (1992) in a study on the Queen Charlotte Islands observed some clear differences amongst terrain and surficial material categories. However, he went on to suggest that the differences in failure frequencies may have been due in part to the slope steepness on which they are located. For example, deep morainal materials that have relatively low failure frequencies are typically found on gentler slopes than colluvial materials that show higher failure frequencies. 49 5.2 Derivation of Morphological Variables 5.2.1 Slope In calculating the slope characteristics of the Clayoquot River watershed, the IDRISI module SURFACE was used. The SURFACE module determines slope by calculating the maximum slope around each pixel from the local slope in X and Y. Only the neighbours above, below, and to each side of the pixel are used in this procedure (Eastman 1992). This method is similar to that developed by Zevenbergen and Thorne (1987), which calculates slope and other variables from 9 elevations of a 3 X 3 submatrix using Lagrange polynomials. A FORTRAN program, FORM.FOR was written outlining the procedure used to derive slope and other necessary variables using Lagrange polynomials (see Figure 14 and Appendix B). Once a slope gradient image for the watershed was produced it was reclassified into useful categories. In a slope stability study in Skidegate Plateau on the Queen Charlotte Islands Rollerson (1992) used the following slope classification: 0-20°; 21-29 °; 30-35°; and 36°+. Similar slope classifications exist in other parts of the Pacific Northwest (Duncan et al. 1987; Swanston 1981). The slope-class breakdown presented above will suffice in this study. 5.2.2 Across Slope and Down Slope Curvature Across slope and down slope curvature were calculated with the FORTRAN program FORM.FOR (see Appendix B). The end products from the program were a range of small numbers. For convenience these results were multiplied by 100. These data then had to be reclassified into either concave, planar or convex categories. When confronted with this problem Dietrich et al. (1992), chose the smallest values estimated to be free of any artifacts of the model. In the present case, experimenting with R E A D IDRISI E L E V A T I O N M A T R I X C A L C U L A T E L A G R A N G E P O L Y N O M I A L C O E F F I C I E N T S C A L C U L A T E S P A T I A L D E R I V A T I V E S Figure 14: F l ow chart of the F O R M . F O R program 51 various cutoff points revealed that -2 and +2 provided the best representation of the landscape to resolve the fine-scale topography. Thus, every value <-2 is judged to be concave, every value > +2 is convex, and everything in between is classed as planar. 5.2.3 Aspect Aspect was derived using the SURFACE module of IDRISI. Its more complete technical derivation is shown in FORM.FOR in Appendix B. Aspect was then reclassified into 8 categories. 5.2.4 Transient Snow Zone The variables slope position and transient snow zone were combined into a single category, hereafter referred to as the transient snow zone, as it appeared that they might be representing the same characteristics. The transient snow zone literature, as mentioned previously, is based on elevation indices which can easily be calculated. As such, a transient snow-zone image was developed simply by reclassifying an IDRISI contour image into 3 zones: a low-elevation zone (0-300 m), a transient mid-elevation snow zone (300-900 m), and a high-elevation zone (>900 m). As mentioned in the literature (Rollerson 1984, 1992), stream escarpments and headwater drainage areas were the slope positions most likely to fail. Identifying stream escarpments through a DEM has not been done to date. Headwater drainage areas however have been identified and their importance emphasized. As such a separate image has been created in this study which represents catchment area. 5.2.5 Catchment Area In order to determine the local catchment areas in the watershed, a FORTRAN program, CATCH.FOR, was written (see Appendix C). The program analyses an elevation matrix to determine catchment area at every point represented by an element 52 in the matrix. The basic premise of the program is that a flowline advances (eight possible directions) downslope from every point represented by an element in the matrix, and extends as far as possible down the steepest path between adjacent points. Flowlines coalesce as they advance through the landscape, and counting the number of flowlines passing through each point gives its catchment area. Catchment area is expressed as the number of points upslope which can contribute flow to any particular point (Martz and De Jong 1988). Figure 15 outlines the logic followed in the CATCH.FOR program in flow chart form. The output from CATCH.FOR then had to be reclassified in order to specify a critical support area that defines the minimum drainage area required to initiate a channel. Figures 12 and 13 from Montgomery and Dietrich (1994) seem to indicate a critical support area of approximately 10,000 m^ (1 ha). Based of this information 1 ha will be the critical support area. 5.3 Creation of decision trees The original intent of this thesis was to review literature from previous studies (see Appendix A) and to develop an algorithm that would classify terrain along British Columbia's outer coast. This meant determining how the attributes within the different variables related to one another as well as how the different variables themselves related to each other. The literature reviewed provided data which would allow the attributes within the different variables to be classified at an interval level of measurement. However, deciding how the different variables relate to each other without additional data made creating an algorithm a risky procedure as well as a useless one. Therefore a method of assessment must be developed which is able to make use of what information is currently available. What this thesis attempts to do is to borrow ideas from the computer programming language Prolog (PROgramming in LOGic), which is based on 53 READ IDRISI E L E V A T I O N M A T R I X (TOP LEFT CORNER) NEW FLOWLINE A D V A N C E S DOWN STEEPEST SLOPE TO OTHER ELEMENTS Figure 15: Flow chart of the C A T C H . F O R program 54 heuristics. According to Townsend (1989), heuristics are rules that are useful in reaching a goal. A guaranteed solution is not always possible using heuristics nor do they always point out the most efficient solution as would an algorithm. However, they are useful when no algorithm exists. It is essentially the same idea as that of bounded rationality employed by humans in decision making. The principle behind bounded rationality is that decision making is aimed at satisfactory rather than optimal outcomes, simply because in most circumstances the individuals involved are unable to comprehend or identify the optimal choice (Lee 1993). In a situation such as this thesis, which is aimed at a preliminary level of analysis or where information may be unavailable, optimal outcomes are neither possible nor a reasonable place to start. In circumstances where data are hmiting, we must find ways to do the best we can with what is available to us. The fundamental data structure in Prolog is the tree (Saint-Dizier 1990). Every tree has a single root, which is shown above the rest of the tree. In the case of this thesis the root is the terrain map. From the root emanate a finite number of branches, which lead to nodes. The nodes can be subdivided again into branches and nodes. In a decision tree, the branches serve simply to establish dominance or sibling relationships between the nodes and the leaves (leaves are nodes at the bottom of the tree at the end of each branch). The approach then is to select the variables to be used, and to prioritize them or rank them, in terms of their importance. For present purposes aspect has been dropped from the variable list because the literature does not consistently cite it as a strong determining factor in slope stability analyses. Without knowing where landslides have occurred it is impossible to determine aspect's significance. Downslope curvature also has been dropped because, in the test area, it did not provide an image adequately representing the three types of downslope shape (concave, planar, and convex). It did represent convex ridgetops quite well but it did a poor job of representing concave 55 midslope locations which are critical to the analysis. With across slope curvature, the situation was opposite; concave topography was represented quite well while classifying convex topography was poor. The justification for leaving across-slope curvature in the model is because the goal of the assessment is to identify unstable terrain and, in the present context, this image does a better job of achieving this end. In prioritizing the variables that will make up the decision trees, the ranldng becomes a discretionary act, which may then incorporate subjective biases. Rejeski (1993) in discussing subjectivity and GIS models, suggests that we not attempt to eliminate subjectivity. He suggests modellers should recognize the need for open peer-review of ranking schemes as a form of "debiasing" the modeling approach. He further cautions that when ranking variables, scientists need to use extreme caution as weighting errors have a tendency to propagate in multi-layered map analysis. As mentioned previously, the literature constantly cites slope gradient as one of the most important variables in terrain instability. Therefore, slope gradient will assume the highest level in decision tree. For the reasons listed above, across-slope curvature will assume the lowest priority on the decision trees. Beneath slope, is the variable transient snow zone. Its acquisition was fairly routine and therefore represented reasonably reliable data. Catchment area assumes the position below transient snow zone. Decision trees for Slope 0-20°, Slope 21-29°, Slope 30-35° and Slope 36°+ are shown in figures 16-19 respectively. For decision trees within the Slope 0-20° tree, all terrain is arbitrarily considered hazard free in terms of sliding. Most of the literature from the Pacific Northwest would support such a decision. Each of the coloured end products (leaves) from the four different slope categories were then reclassified according to their hazard probability (Fig. 20). According to the Forest Road and Logging Trail Engineering Practices (1993) hazard refers to the likelihood of mass soil movements originating in upland areas. When hazards are multiplied by consequences (the probability that if a mass soil movement 56 59 event occurs, it will have an impact on an on-site/downslope/downstream resource, social and economic values) risk is determined. Green leaves (low hazard probability) were reclassified with a value of one; yellow leaves (moderate hazard probability) were reclassified with a value of two; and red leaves (high hazard probability) were assigned a value of three. Each individual coloured end product (leaf) was then overlaid on top of one another resulting in every pixel within the watershed being classified 6.0 Results and Discussion The final image derived through the use of the DEM and the decision trees does not resemble a traditional mulufactor hazard rating image because of the method used to create the image. Traditional mulufactor hazard maps are created using polygons which delineate contiguous homogeneous areas. In this thesis, a raster format was used in which each pixel has a unique value. In addition to this, the variables used in this thesis are not commonly used in mulufactor terrain analysis. More traditional variables might be such things as slope gradient (which is included here), soil type, rainfall intensity, geology, forest cover type, etc. However, most of the variables used here were generated through the FORTRAN programs, whose output were real numbers assigned to a pixel. Even after being reclassified, the output is still capable of a high level of variability since each pixel is capable of retaining a different value from its adjacent neighbours. x The results of the analysis (Fig. 20) would seem to indicate that the great majority of this watershed should not be logged (see Fig. 20). It should be remembered however, that this image is meant to be a preliminary reconnaissance level hazard assessment, whose main goal is to outline or delineate those areas needing further or more intensive investigation by a trained professional engineer or geoscientist. Throughout the Clayoquot Sound region as a whole, about 80% of the remaining older forests within the General Integrated Management Area are on slopes 61 steeper than 30°. Many of these slopes have a moderate to high likelihood of failure following clearcutting (M. Sondheim 1994 as cited in Clayoquot Sound Scientific Panel 1995). The landscape of the Clayoquot River watershed has very similar characteristics. For example, over 60% of the watershed falls within the last two slope categories (30-35° and 36°+), and of this 60%, 40% falls within the 36°+ slope category (Figure 21). Being as steep as it is and with the number of stream channels throughout the watershed, the number of catchments less than 1 ha is quite large. When exarnining the breakdown between categories within the transient snow zone image, almost 59% of the watershed falls within the 300-900 m elevation range, which is considered the high hazard zone. Only 12% of the entire watershed falls between 0-300 m. Obviously, the physical attributes of the watershed predispose it to a high level of slope instability. It should also be noted that 3 of the leaves account for 33% of the watershed being classified as high hazard. Within the slope category 36° + , within the transient snow zone and when the surface is planar, regardless of whether the catchment is less than 1 ha and when in the high elevation zone and the catchment is < 1 ha and the surface is planar, the possibility of failure is high. In order to further substantiate the methodology and findings of this thesis, additional tests were undertaken. The first test was conducted by obtaming two color aerial photographs of the Clayoquot Sound watershed from MacMillan Bloedel. Figure 22 gives an indication of the aerial extent covered by the two photographs. Both photos were taken in the summer of 1994. Aerial photograph MB94010 (Figure 23) covers the southwest portion of the watershed and takes in approximately 1,787 ha. Aerial photograph MB94008 (Figure 24) covers the northwest portion of the watershed and takes in approximately 1,210 ha. The occurrence of landslides, debris flows, etc, were then drawn on a mylar overlay in order to obtain some idea regarding the extent and frequency of events. Figure 22: Location of Aerial Photographs \3 66 Within aerial photograph MB94010, eleven clearly identifiable events were recorded covering approximately 60 ha or 3.3% of the watershed covered by the photograph. With regard to photograph MB94008, eight events were recorded covering approximately 62 ha or 5.1 % of the area covered by the photograph. The events delineated from the aerial photographs were fairly major events, which begs the question, how many smaller processes are being obstructed from view by the trees. I would assume that there are many more. Extrapolating just the clearly identifiable events from the two aerial photographs across the whole watershed would mean that approximately 4.0% of the watershed is in a state of post natural failure. The potential for further movement seems quite possible especially if logging and road construction were to take place. The events that have taken place were for the most part originating in steeply sloping, high elevation headwater areas. In order to further substantiate the model and the use of the decision trees as a means of forcasting slope instability, experts in the field of hazard assessment were sent copies of the decision trees with the last row remaining blank. A comparision of the results of this thesis with expert opinion may lend further support for the model developed. Three terrain specialists within British Columbia were faxed copies of Figures 17-19. One expert was not able to respond due to a heavy workload. The remaining two experts did not fill out the decision trees for various reasons. One common reasons that was cited was the fact that surficial geology was not included in the analysis. The importance of surfical geology has been noted in previous sections of the thesis and was to be a key variable in the assessment. However, as the Terrain Mapping project by MacMillan Bloedel for the Clayoquot Sound region has been temporarily put on hold (G. Horel, MacMillan Bloedel personal communication) obtaining the necessary data required to complete this thesis as outlined in the objectives in now not possible in the very near future. The intent was to achieve a 67 representation of both surficial geology for the watershed along with the final terrain map as generated by MacMillan Bloedel. As mentioned previously, different surficial materials have different critical angles at which they fail (Howes 1987). Further literature review on this topic would have allowed me to create another image depicting the findings which then could have been translated into a new series of decision trees which ultimately could have been used to create a new and more reasonable assessment of the terrain for the Clayoquot River watershed. It was also suggested by one of the experts that the slope categories were too wide and a smaller breakdown was suggested. It was also suggested that the categories listed were not the right ones and that the model as it stands now is oversimplifying the process. Overall, the methodology proposed in this thesis was so contrary to the way they conduct slope stability assessments that they felt they could not fill in the bottom row of the decision trees. 6.1 The Issue of Subjectivity Any scientist attempting to estimate the hazards or consequences and ultimately the risks involved in a terrain stability analysis is confronted with uncertainties. This thesis is no different. While the data available were not as good as one might hope for, no amount of data (about the past) can render the future certain. Where there is a lack of available data, subjective biases may enter into the heuristic analysis. In particular, for the decision tree tables it became necessary to weight the variables disproportionately according to the role I thought they play in identifying hazardous terrain. Since experimental evidence is impossible (unreasonable) to acquire, the sorts of analysis offered here, together with use to adapt the hypothesis over time represents the only feasible approach to guiding actions such as road building, which may influence 68 slope stability. In this context, one of the most important inputs into a terrain analysis is local knowledge. Those who espouse muiti-variate analysis or other statistical procedures in determining unstable terrain tend to overlook this important element. What makes the method used in this thesis appealing is the ease with which the trees are broken down inviting peer review of the ranking schemes, and to post hoc evaluation of real slides. The method of using decision trees allows uncertainties to be dealt with in an open manner as well as exposing the models limitations. Uncertainty is valuable information when dealt with in an honest and open manner. The model developed in this thesis meets the suggested requirements of Rollerson and Sondheim (1985). These authors suggest that when developing a hazard classification similar to what is proposed in this thesis, three points need to be addressed. First, the classification should be simple; second, a large number of classes should be avoided; and third, in order to be a useful planning tool, the classification must avoid placing too much of the study area into the category labeled as moderate risk. The first two requirements are certainly met in this thesis. In the case of the third requirement, as this thesis is a preliminary analysis a moderate hazard would not pose a problem as these areas would require further analysis by a professional engineer or geoscientist. It is essential in an analysis such as this to avoid committing a Type I error. A Type I error is an error of commission affirming a proposition that turns out to be false (Lee 1993). In the terrain mapping project, it is better to claim an area unsafe and leave it for further analysis then to let it be cut and have it fail at a later date. Soil integrity must be the number one priority. 6.2 How to Improve the Model One of the problems with the current model may stem from possible incompleteness of the mathematical functions used. For example, the standard algorithm for calculating catchment area uses only a limited number of flow directions 69 (in this case 8) from each grid cell, which has a tendency to create a false pattern of waterchannels flowing in parallel lines along preferred directions (Fairfield and Laymarie 1991). This is especially true in the case of gentle terrain. Fortunately, the majority of the Clayoquot River watershed is very steep with few areas of flat terrain. There were however, a few places on less steep terrain which exhibited this spurious pattern. Several new algorithms for determining channel networks and subsequent catchment areas are now being developed or have been developed which use a random eight node approach (Fairfield and Laymarie 1991) as well as 360 directions of flow based on a vector-grid algorithm (Mitasova and Hofierka 1993). Also problematic were the images across slope and down slope curvatures. In the case of down slope curvature, the image produced did not do a very good job of delineating concave terrain. With respect to across slope curvature, on the other hand, the image did not do a very good job of delmeating convex terrain. In some areas of known concavity (i.e. along major stream and river networks) it only provided a speckled pattern showing some pixels as being concave and some as planar. This can be the result of two possible factors. The first factor, as suggested by Mitasova and Hofierka (1995) is that the eight possible directions of flow from each cell may not be sufficient for preserving the othogonality of slope curves to contours. In order to combat this problem, these authors offered a method of curvature computation which used an interpolation function, completely regularized spline with tension, that is suitable for topographic analysis because of its accuracy and regular derivatives of all orders. These authors also suggest that for the study of flow divergence/convergence, it is more appropriate to measure curvature in the normal plane in the direction perpendicular to gradient. They call this measure tangential curvature because the direction perpendicular to the gradient is the direction tangent to contour of a given point. 70 The second and more likely factor contributing to the problems with the across slope and downslope curvature may be related to the accuracy and scale of the T.R.I.M. data. The first axiom in any (GIS) modeling exercise is that the accuracy of the model can never be greater than the original source of the data. According to a Ministry of Crown Lands information package regarding T.R.I.M. data, the digital map file accuracy has a horizontal resolution of 10 m and a vertical resolution of 5 m. This resolution is a product of the relative accuracy of ground controls and orientation to it (aerotriangulation), as well as instrument calibration and model set-up accuracies. At 1:20,000 scale, the map product is obviously not capable of deciphering the many small areas of concavity and convexity even if the T.R.I.M. data capture them. A larger scale map product is required in order to do a better job of identifying these area. What can be said based on information available, what has been generated, and from the feedback from the experts is that determining terrain stability from morphological variables derived through a DEM is not a very feasible procedure. It appears that the method developed is most likely overestimating the percentage of the watershed falling into the high hazard category. On the ground, this would translate into a more costly and time consuming procedure. The variables used however, are sound and intuitively logical. Slope indicates the rate of water and sediment flow by controlling the rate of energy expenditure available to drive flow; the transient snow zone represents areas where rain-on-snow events produce excess soil moisture necessary to initiate destructive mass soil movements; catchment area, based on recent literature has provided some long-needed quantifiable information regarding the source area-slope relationship; and across-slope curvature influences convergent and divergent topography. The importance of surficial geology in a slope stability assessment such as this cannot be underestimated. Its absence has left an important component of terrain assessment out. 71 At this point, use of morphological variables can provide an excellent starting point for conducting a hazard assessment. Unlike traditional hazard mapping variables, morphological variables provide an indication of the fluxes taking place within the landscape. With the recent provincial government decision accepting the recommendations of the Scientific Panel's report of April 1995, a much more detailed level of information will need to be collected regarding both the physical and ecological processes taking place within the watersheds. Delineating areas of potential instability would be greatly enhanced by generating images characterizing the morphological characteristics as outlined in this thesis. 7.0 Conclusions Comprehensible simplification was explicitly intended so as to give rise to understanding and gaining insight regarding the key phenomena involved in slope stability. The only disappointment of this thesis was that the model could not be subjected to the testing process. What tests were conducted emphasised the need for surficial geology in any stability assessment. Models can not be proved right; they can only be proved wrong (Holling et al. 1978). I however believe that this model provides provisional acceptance of belief as the knowledge gained through the literature review regarding slope stability and mass wasting is accurately reflected in the model. Despite not being able to invalidate the model, I still believe that I have a critical understanding of the weaknesses and strengths of the model as well as other models regarding slope stability. And most important, by knowing the extent and limits of the model's predictive capabilities I am in a position to comment on the design and evaluations of further development proposals for the Clayoquot River watershed in the most responsible manner possible (Holling et al. 1978). 72 References: Amaranthus, M.P., R.M. Rice, N.R. Barr and R.R. Ziemer. 1985. Logging and forest roads related to increased debris slides in southwestern Oregon. J. of For., 83(4):229-233. Ballard, T.M., and R.P. Wtilington. 1975. Slope instability in relation to timber harvesting in the Chilliwack Provincial forest. The Forestry Chronicle, 51(1):59-61. Beaudry, P.G. 1984. Effects of forest harvesting on snowmelt during rainfall in coastal British Columbia. M.Sc. thesis. University of British Columbia, Vancouver. 183p. Bedrossian, T.L. 1983. Watershed mapping in northern California. Calif. Geol., 140-147pp. Beekman, F. 1987. Soil strength and forest operations. Ph. D. thesis, Agricultural University, Wageningen, The Netherlands. 168p. Benda, L.E. and T.W. Cundy. 1990. Predicting deposition of debris flows in mountain channels. Can. Geotech. J., 27(4):409-417. Bourgeois, W.W. 1978. Timber harvesting activities on steep Vancouver Island terrain. In C.T. Youngberg, ed. Forest soils and land use. Proc., Fifth North Am. For. Soils Conf., Ft. Collins, Colo. 393-409pp. Bovis, MJ . and B.R. Dagg. 1992. Debris flow triggering by impulsive loading: mechanical modeling and case studies. Can. Geotech. J., 29(3):345-352. British Columbia Government. 1993. Clayoquot Sound: a balanced decision-a sustainable future. Victoria, B.C. British Columbia Ministry of Forests Timber Harvesting Branch. 1993. Forest road and logging trail engineering practices (Interim) and instructions for its implementation. Victoria, B.C. Burroughs, E.R., G.R. Chalfant and M. A. Townsend. 1976. Slope stability in road construction. U.S. Dep. of the Inter., Bur. of Land Manage., Portland, OR. 102p. Caine, N. 1980. Rainfall intensity-duration control of shallow landslides and debris flows. Geogrqfiska Annaler, 62(A):23-27. 73 Chatwin, S.C. 1991. Measures for control and management of unstable terrain. In S.C. Chatwin, D.E. Howes, J.W. Schwab and D.N. Swanston, eds. A guide for management of landslide prone terrain in the Pacific Northwest. B.C. Ministry of Forests. Land Management Handbook No. 18. 85-166pp. Clague, J.J. and B.D. Bornhold. 1980. Morphology and littoral processes of the Pacific coast of Canada. In S.B. McCann, ed. The coastline of Canada. Geol. Surv. Can. Pap. 80-10. 339-380pp. Clayton, J.S., W.A Ehrlich, D.B. Cann, J.H Day and LB. Marshall. 1977. Soils of Canada Volume 1 Soil Report. Canada Department of Agriculture Research Branch. 243p. Craig, R.F. 1992. Soil mechanics, 5th ed., Chapman & Hall, London. 427p. Donker, N.H. 1992-93. Automatic extraction of catchment hydrologic properties from digital elevation data. ITC Journal, :257-265. Dietrich, W.E. and T. Dunne. 1993. The channel head. In K. Bevan and M.J. Kirkby, eds., Channel network hydrology. John Wiley & Sons Ltd., Chichester. 175-219pp. Dietrich, W.E. C.J. Wilson, D.M. Montgomery and J. McKean. 1993. Analysis of erosion thresholds, channel networks, and landscape morphology using a digital terrain model. J. Geol, 101(2):259-278. Duncan, S.H. 1985. Road location in sensitive watersheds: An industry perspective. In D.N. Swanston, ed. Workshop on Slope Stability: Problems and Solutions in Forest Management. Gen. Tech. Rep. PNW-180. Portland, OR.: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 78-80pp. Duncan, S.H., J.W. Ward and R.J. Anderson. 1987. A method of assessing landslide potential as an aid in forest road placement. Northwest Sci., 61(3): 152-159. Dyke, A.S. and V.K. Prest.1987. Late Wisconsinan and Holocene retreat of the Laurentide Ice Sheet. Geol. Surv. Can., Map 1702A, scale 1:5 000 000. Eisbacher, G.H. 1982. Slope stability and land use in mountainous valleys. Geoscience Canada, 9(1): 14-27. Fairfield, J. and P. Leymarie. 1991. Drainage networks from grid digital elevation models. WaterResour. Res., 27(5):709-717. Fannin, R.J. and T.P. Rollerson. 1993. Debris flows: some physical characteristics and behaviour. Can. Geotech. J., 30(1):71-81. 74 Fredriksen, R.L. 1970. Erosion and sedimentation following road construction and timber harvest on unstable soils in three small western Oregon watersheds. Res. Pap. PNW-104. Portland, OR.: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 15p. Forest Inventory Manual. 1991. Forest classification/sampling and environmentally sensitive areas (Volume 2). Ministry of Forests, Victoria, B.C. 63p. Furbish, D.J., and R.M. Rice. 1983. Predicting landslides related to clearcut logging, northwestern California, U.S.A.. Mount. Res. Develop., 3(3):253-259. Gerath, R.F. 1995. Landslide hazards in BC's forests: the case for formal risk assessment. The BC Professional Engineer. 46(2): 10-13. Gerath, R.F., O. Hungr and D.F. VanDine. 1994. Landslide hazard mapping guidelines for British Columbia. Report to the earth sciences task force B.C. government resources inventory committee. Thurber Engineering Ltd., Vancouver, B.C. 55p. George, P., A. Carr and G. Lenz. 1994. Clayoquot: a heritage worth protecting. Western Canadian Wilderness Committee Educational Report. 13(5):4. Gimbarzevsky, P. 1983. Regional overview of mass wasting on the Queen Charlotte Islands. Fish/Forestry Interaction Program, Working Paper 3/83. Victoria, B.C. 34p. Gray, D.H., and W.F. Megahan. 1981. Forest vegetation removal and slope stability in the Idaho batholith. Res. Pap. INT-271. Ogden, UT.: U.S. For. Serv., Intermountain Forest and Range Experiment Station. 23p. Gresswell, S., D. Heller and D.N. Swanston. 1979. Mass movement response to forest management in the central Oregon coast ranges. Res. Bull. PNW-84. Portland, OR.: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 26p. Hammond, C , D. Hall, S. Miller and P. Swetik. 1992. Level I stability analysis (LISA) documentation for Version 2.0. Gen. Tech. Rep. INT-285. Ogden, UT.: U.S. For. Serv., Intermounatain Forest and Range Experiment Station. 190p. Harr, R.D. 1986. Effects of clearcutting on rain-on-snow runoff in western Oregon: A new look at old studies. WaterResour. Res., 22(7): 1095-1100. 75 Helmlinger, K.R., P. Kumar and E. Foufoula-Georgiou. 1993. On the use of digital elevation model data for Hortonian and fractal analyses of channel networks. WaterResour. Res., 29(8):2599-2613. Hogan, D.L. and D.J. Wilford. 1989. A sediment transfer hazard classification system: linking erosion to fish habitat. In Procs. of Watershed '89. A conference on the stewardship of soil, air and water resources. Juneau Alaska March 21-23, 1989. USDA Forest Service Alaska Region. 143-155pp. Holland, S.S. 1976. Laridform of British Columbia: a physiographic outline. British Columbia Department of Mines and Petroleum, Research Bulletin No. 48. Victoria, B.C. 138p. Holling, C.S., ed. 1978. Adaptive Environmental Assessment and Management. John Wiley & Sons. New York. 377p. Howes, D.E. 1987. A terrain evaluation method for predicting terrain susceptible to post-logging landslide activity. MOEP Technical Report 28, B.C. Ministry of Environment and Parks. 38p. Howes, D.E. and E. Kenk. 1988. Terrain classification system for British Columbia (revised edition). MOE Manual 10 B.C. Ministry of Environment and Ministry of Crown Lands, Victoria, B.C.. 90p. Howes, D.E. and D.N. Swanston. 1991. Slope stability evaluations using digital terrain models. In S.C. Chatwin, D.E. Howes, J.W. Schwab and D.N. Swanston, eds. A guide for management of landslide prone terrain in the Pacific Northwest. B.C. Ministry of Forests. Land Management Handbook No. 18. 19-84p. Hungr, O., G.C. Morgan, and R. Kellerhals. 1984. Quantitative analysis of debris torrents for design of remedial measures. Can. Geotech. J., 21(4):663-677. Jenson, S.K. and J.O. Domingue. 1988. Extracting topographic structure from digital elevation data for geographic information system data. Photogramm. Eng. Remote Sens., 54(11): 1593-1600. Jones, D.L., A. Cox, P. Coney and M. Beck. 1982. The growth of western North America. Sci. Amer., 247(5) :70-84pp. Jones, D.L., N.J. Silberling and J. Hillhouse. 1977. Wrangellia- a displaced terrane in northwestern North America. Can. J. Earth Sci., 14(ll):2565-2577. Jones, R.K. and R. Annas. 1986. Vegetation. In K.W.G. Valentine, P.N. Sprout, T.E. Baker and L.M. Lavkulich, eds. The soil landscapes of British Columbia. B.C. Ministry of Environment. 35-45pp. 76 Jungen, J.R. and T. Lewis. 1986. The coast mountains and islands. In K.W.G. Valentine, P.N. Sprout, T.E. Baker and L.M. Lavkulich, eds. The soil landscapes of British Columbia. B.C. Ministry of Environment. 101-120pp. Jungen, J.R. 1985. Soils of southern Vancouver Island. B.C. Ministry of Environment Technical Report 17. 198p. Korman, J., W.A. Kurz and D.R. Marmorek. 1992. Fish/Forestry interaction model workshop summary and model design. ESS A Environmental and Social Systems Analysts Ltd. Vancouver, B.C. 91p. Krag, R.K., E.A. Sauder and G.V. Wellburn. 1986. A forest engineering analysis of landslides in logged areas on the Queen Charlotte Islands, British Columbia. British Columbia Ministry of Forests and Lands. Land Management Report Number 43. 138p, Larse, R.W. 1971. Prevention and control of erosion and stream sedimentation from forest roads. In J. Morris, ed. Proceedings of a symposium- Forest land uses and stream environment. Oregon State University, Corvallis. 76-83pp. Lee, K.N. 1993. Compass and gyroscope: integrating science and politics for the environment. Island press. Washington, D.C. 243p. Lewis, J., and R.M. Rice. 1990. Estimating erosion risk on forest lands using improved methods of discriminant analysis. Water Resour. Res., 26(8): 1721-1733. Mackey, B.G., D.W. McKenney, C A . Widdifield, R.A. Sims, K. Lawrence, and N. Szcyrek. 1994. A new digital elevation model of Ontario. NODA/NFP Technical Report; TR-6. Sault Ste. Marie, ON. 26p. MacMillan Bloedel Limited. 1993. MacMillan Bloedel & Clayoquot Sound Briefing notes. Vancouver, B.C. Martz, L.W. and E. de Jong. 1988. CATCH: A FORTRAN program for measuring catchment area from digital elevation models. Comput. Geosci., 14(5):627-640. Martz, L.W. and J. Garbrecht. 1992. Numerical definition of drainage network and subcatchment areas from digital elevation models. Comput. Geosci., 18(6):747-761. McCashion, J.D., and R.M. Rice. 1983. Erosion on logging roads in northwestern California: How much is avoidable?/, of For., 81(l):23-26. 77 McCormack, J.E., M.N. Gahegan and S.A. Roberts. 1993. Feature-based derivation of drainage networks. Int. J. Geographical Information Systems., 7(3):263-279. McLaren, P., J. Harper and P. Hale. 1983. Coastal environments of southern Vancouver Island. Geological Association of Canada-Mineralogical Association of Canada-Canadian Geophysical Union, Joint Annual Meeting. Victoria, B.C. Field Trip 7: Guidebook. Megahan, W.F. 1988. Roads and forest site productivity. In J.D. Lousier and G.W. Still, eds., Degradation of forest land: "Forest soils at risk". Proc., 10th B.C. Soil Science Workshop, February, 1986. British Columbia Ministry of Forests. Land Management Report Number 56. 54-65pp. Megahan, W.F., and W.J. Kidd. 1972. Effect of logging roads on sediment production rates in the Idaho batholith. Res. Pap. INT-123. Ogden, UT.: U.S. For. Serv., Intermontain Forest and Range Experiment Station. 14p. Megahan, W.F., N.F. Day and T.M. Bliss. 1978. Landslide occurrence in the western and central northern rocky mountain physiographic province in Idaho. In C T . Youngberg, ed. Forest soils and land use. Proc., Fifth North Am. For. Soils Conf., Ft. Collins, Colo., 116-139pp. Mitasova, H., and J. Hofierka. 1993. Interpolation by regularized spline with tension:JJ. Application to terrain modeling and surface geometry analysis. Math. Geol. 25(6):657-669. Montgomery, D.R. (1994). Road surface drainage, channel initiation, and slope instability. Water Resour. Res., 30(6): 1925-1932. Montgomery, D.R. and W.E. Dietrich. 1988. Where do channels begin? Nature, 336(6196):232-234. Montgomery, D.R. and W.E. Dietrich. 1992. Channel initiation and the problem of landscape scale. Science, 255:826-830. Montgomery, D.R. and W.E. Dietrich. 1994a. A physically based model for the topographic control on shallow landsliding. Water Resour. Res., 30(4): 1153-1171. Montgomery, D.R. and W.E. Dietrich. 1994b. Landscape dissection and drainage area-slope thresholds. In M.J. Kirkby, ed. Process models and theoretical geomorphology. John Wiley & Sons, Chichester. 221-246pp. 78 Montgomery, D.R. and E. Foufoula-Georgiou. 1993. Channel network source representation using digital elevation models. WaterResour. Res., 29(12): 3925-3934. Moore, I.D., R.B. Grayson and A.R. Ladson 1991. Digital terrain modelling: a review of hydrological, geomorphological, and biological applications. Hydrol. Proc., 5:3-30. Moore, K. 1991. Coastal watersheds: an inventory of watersheds in the coastal temperate forests of British Columbia. F r^thlife Canada Foundation & Ecotrust/Conservation International. 54p. Muller, J.E. 1977a. Evolution of the Pacific margin, Vancouver Island and adjacent regions. Can. J. Earth Sci., 14(9):2062-2085. Muller, J.E. 1977b. Geology of Vancouver Island. Geological Association of Canada-Mineralogical Association of Canada, Joint Annual Meeting. Vancouver, B.C. Field Trip 7: Guidebook. National Council of the Paper Industry for Air and Stream Improvement (NCASI). 985. Catalog of Landslide Inventories for the Northwest. Technical Bulletin No. 456. New York. Niemann, K.O. and D.E. Howes. 1991. Slope stability evaluations using digital terrain models. B.C. Ministry of Forests. Land Management Report No. 74. 28p. O'l^ughlin, C L . 1972. The stability of steepland forest soils in the Coast Mountains, southwest British Columbia. Ph. D. thesis, University of British Columbia, Vancouver. 147p. Pojar, J., K. Klinka, and D.A. Demarchi. 1991. Coastal Western Hemlock Zone. In D. Meidinger and J. Pojar, eds. Ecosystems of British Columbia. B.C. Ministry of Forests, Special Report Series 6. 95-11 lpp. Pojar, J., K. Klinka, and D.A. Demarchi. 1991. Mountain Hemlock Zone. In D. Meidinger and J. Pojar, eds. Ecosystems of British Columbia. B.C. Ministry of Forests, Special Report Series 6. 113-124pp. Prescott, C E . and Weetman G.F. 1994. Salal Cedar Hemlock Integrated Research Program: A synthesis. Faculty of Forestry, University of British Columbia, Vancouver, B.C. 85p. Province of British Columbia. 1993. Clayoquot Sound land use decision: background report. 79 Province of British Columbia. 1995. Forest Practices Code of British Columbia: Mapping and assessing terrain stability guidebook. 34p. Rejeski, D. 1993. GIS and risk: A three culture problem. In M.F. Goodchild, B.O. Parks and L.T. Steyaert, eds. Environmental modeling with GIS. Oxford University Press, Oxford. 318-331pp. Rice, R.M. and J.S. Krammes. 1971. Mass wasting in watershed management. Proceedings of the Symposium on Interdisciplinary Aspects of Watershed Management, Bozeman, Montana. 231-259pp. Rice, R.M., and J. Lewis. 1991. Estimating erosion risk associated with logging and forest roads in northwestern California. Water Resour. Bull, 27(5):809-817. Rice, R.M., N.H. Pillsbury and K.W. Schmidt. 1985. A risk analysis approach for using discriminant functions to manage logging-related landslides on granitic terrain. For. Sci., 31(3):772-784. Rollerson, T. 1984. Terrain stability study- TFL 44. Land Use Planning Advisory Team. Woodland Services, MacMillan Bloedel Ltd., Nanaimo, B.C. Rollerson, T.P. 1992. Relationship between landscape attributes and landslide frequencies after logging: Skidegate Plateau, Queen Charlotte Islands. British Columbia Ministry of Forests, lip. Rollerson, T.P., and M. Sondheim. 1985. Predicting post-logging terrain stability: A statistical-geographical approach. Procs., Joint Symposium of the I.U.F.R.O. Mountain Logging Section and the Sixth Pacific Northwest Skyline Symposium Vancouver, B.C. 7p. Rood, K.M. 1984. An aerial photograph inventory of the frequency of yield of mass wasting on the Queen Charlotte Islands, British Columbia. B.C. Ministry of Forests, Land Management Report No. 34, 55p. Ryder, J.M. and J.J. Clague. 1989. British Columbia (Quaternary stratigraphy and history, Cordilleran Ice Sheet). In R.J. Fulton, ed. Quaternary geology of Canada and Greenland. Geol. Surv. Can., Geology of Canada no.l. 48-58pp. Ryder, J.M. and D.E. Howes. 1984. Terrain information, a user's guide to terrain maps in British Columbia. B.C. Ministry of Environment. 16p. Saint-Dizier, P. 1990. An introduction to programming in prolog. Springer-Verlag, NewYork. 184p. 80 Sauder, E. A., R.K. Krag and G. V. Wellburn. 1987. Logging and mass wasting in the Pacific Northwest with application to the Queen Charlotte Islands, B. C.: A literature review. FERIC Special Report Number SR-45. 25p. Schroeder, W.L. 1985. The engineering approach to landslide risk analysis. In D.N. Swanston, ed. Procs., Workshop on Slope Stability: Problems and Solutions in Forest Management. Gen. Tech. Rep. PNW-180. Portland, OR.: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 43-50pp. Schwab, J.W. 1981. Interpretive stability hazard classes: development and application for longterm planning in the Prince Rupert Forest Region. Presented at the Slope Stability Workshop. August 18-20, 1981. Portland, Oregon. 7p. Schwab, J.W. 1988. Mass wasting impacts to forest land: forest management implications, Queen Charlotte timber supply area. In J.D. Lousier and G.W. Still, eds., Degradation of forest land: "Forest soils at risk''. Proc., 10th B.C. Soil Science Workshop, February, 1986. British Columbia Ministry of Forests. Land Management Report Number 56. 104-115pp. Schwab, J.W. 1993. Interim terrain and slope stability mapping standards Prince Rupert Forest Region. B.C. Forest Service, Prince Rupert Forest Region. 26p. Sidle, R.C. 1980. Slope stability on forest land. Pacific NWExt. Publ. PNW-209. Oregon State, Corvallis. 23p. Sidle, R.C. 1985. Factors influencing the stability of slopes. In D.N. Swanston, ed. Procs., Workshop on Slope Stability: Problems and Solutions in Forest Management. Gen. Tech. Rep. PNW-180. Portland, OR.: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 17-25pp. Sidle, R.C. 1991. A conceptual model of changes in root cohesion in response to vegetation management. J. Environ. Qua!., 20:43-52. Sidle, R .C , A.J. Pearce and C L . O'Loughlin. 1985. Hillslope stability and land use. Water Resources Monograph Series 11. Washington, D.C.: American Geophysical Union. 140p. Smith, G.N. 1990. Elements of soil mechanics, 6th ed., BSP Professional Books, Oxford. 509p. Swanston, D.N. 1974. The forest ecosystem of southwest Alaska: 5. Soil mass movements. Gen. Tech. Rep. PNW-17. Portland, OR.: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 22p. 81 Swanston, D.N. 1974. Slope stability problems associated with timber harvesting in mountainous regions of the western United States. Gen. Tech. Rep. PNW-21. Portland, OR.: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 14p. Swanston, D.N. 1981. Watershed classification based on soil stability criteria. In D.M. Baumgartner ed. Interior West Watershed Management. Washington State University, Pullman Washington. 43-58pp. Swanston, D.N., and F.J. Swanson. 1976. Timber harvesting, mass erosion, and steepland forest geomorphology in the Pacific Northwest. In D.R. Coates ed. Geomorphology and Engineering. Dowden, Hutchinson & Ross, Inc. Stroudsburg, Pa. 199-22lpp. Tarboton, D.G., R.F. Bras and I. Rodriguez-Iturbe. 1991. On the extraction of channel networks from digital elevation data. Hydrol. Processes, 5(1): 81-100. The BC Professional Engineer. 1993. Forest roads: hydrological and geotechnical impacts. 44(9):3,14-15. The Scientific Panel for Sustainable Forest Practices in Clayoquot Sound. 1994. Progress report 2: Review of current forest practice standards in Clayoquot Sound. 82p. The Scientific Panel for Sustainable Forest Practices in Clayoquot Sound. 1995. Report 5: Sustainable ecosystem management in Clayoquot Sound: Planning and practices. 296p. Townsend, C. 1989. Introduction to turbo prolog, 2nd ed., Sybex, San Francisco. 325p. Varnes, D.J. 1978. Slope movement types and processes. In Schuster, R.L. and R.J. Krizek. eds. Landslides: analysis and control. Special report 176, National Academy of Sciences. National Research Council, Washington, D C , ll-33pp. Wadge, G., A.P. Wislocki and E.J. Pearson. 1993. Spatial analysis in GIS for natural hazard assessment. In M.F. Goodchild, B.O. Parks and L.T. Steyaert, eds. Environmental modeling with GIS. Oxford University Press, Oxford. 332-338pp. Ward, T.J. 1985. Computer-based landslide delineation and risk assessment procedures for management planning. In D.N. Swanston, ed. Procs., Workshop on Slope Stability: Problems and Solutions in Forest Management. Gen. Tech. Rep. PNW- 180. Portland, OR.: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 51-57pp. 82 Whitiow, R. 1990. Basic soil mechanics, 2nd ed., Longman Scientific & Technical, Burnt Mill. 528p. Wilkerson, J.F. 1990. Undeveloped watersheds on Vancouver Island larger than 1000 hectares. B.C. Ministry of Forests Recreation Branch Technical Report 1992:3. 9p. Wu, T.H. 1966. Soil mechanics. Allyn and Bacon,Inc, Boston. 43lp. Wu, T.H., and D.N. Swanston. 1980. Risk of landslides in shallow soils and its relation to clearcutting in southeastern Alaska. For. Sci., 26(3):495-510. Young, S.E. 1992. Slope stability prediction techniques. M.Sc theis, University of British Columbia, Vancouver, 15 lp. Zevenbergen, L.W. and CR. Thorne. 1987. Quantitative analyses of land surface topography. Earth Surf. Processes Landforms., 12(l):47-56. Zhang, W., and D.M. Montgomery. 1994. Digital elevation model grid size, landscape representation, and hydrologic simulations. WaterResour. Res., 30(4): 1019-1028. Ziemer, R.R. and D.N. Swanston. 1977. Root strength changes after logging in southeast Alaska. Res. Note PNW-306. Portland, OR.: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 25p. Ziemer, R.R., J. Lewis, R.M. Rice and T.E. Lisle. 1991. Modeling the cumulative watershed effects of forest management strategies. J. Environ. Qual., 20:36-42. 83 Appendix A: Slope stability studies conducted throughout the Pacific Northwest Debris aval an chi ng in thin soils derived from bedrock in Southeast Alaska Study Site: The study sites were chosen based on their predisposition to landsliding. Objectives: The objective of this study was to develop an understanding of the reasons for debris avalanche failures on steep, thin soiled slopes without glacial t i l l deposits in order to anticipate future problems when these sites are subject to harvesting. Area Description: Sites were chosen where glacial t i l l was absent, soils thin, slopes steep and landsliding frequent. Soils ranged from 65% to well over 100%. Soils were always less than 3 feet and as shallow as 1 foot deep. Bedrock ranged from graywacke, shales and slates, to granite and gneiss. Precipitation ranges from 1143 to 5511 mm annually on this region with 2540 mm being the approximate average for the area. Methods: Sixteen sites having no glacial t i l l were examined. At each location, configuration of the slide trace and bedrock type and attitude were determined, and the angle of slope at the point of slide initiation was measured. The soil depth was also measured and the probable cause of initial sliding was determined by reconstruction of events that apparently occurred when the slope failed. Results: In every case, failure occurred on extremely steep slopes, lying between 35° and 40° or more. Bedrock slopes are smoothed off and dip into the valley at angles that approach or exceed the angle of internal friction for these soil materials. The soils are coarse and permeable with a high proportion of angular rock fragments and overlain by a thin organic mat consisting of intertwining rocks and forest litter. The soils appear to drain very rapidly since overland flow was never observed, even during intense rainfall. Reference: Swanston, D.N. 1967. Debris avalanching in thin soils derived from bedrock. Res. Note PNW-64. Portland OR. :U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 84 The Queen Charlotte Islands, British Columbia Study Site: The study examines 102 landslides in 11 watersheds located throughout the Queen Charlotte Islands. The landslides were surveyed between April and September 1983. Objectives: The objectives of the study were to: 1) apply forest engineering expertise to identifying and documenting possible logging and road related factors that contributed to landslide initiation in logged areas of the Islands; 2) recommend practices that could reduce landslide frequencies on logged terrain; and 3) identify terrain features that might indicate potential failure sites prior to road construction. Area Description: The Queen Charlotte Islands include some of the most valuable forest lands in British Columbia. The Islands are divided into three physiographic regions; the lowlands, plateau and Queen Charlotte Ranges. The bedrock is composed primarily of volcanic, sedimentary and intrusive rocks, frequently exposed or mantled by a thin layer of glacial and colluvial material. The climate is characterized by cool summers, mild winters and strong winds. Precipitation ranges from 4,200 mm on the west coast to 1150 mm for sheltered locations and is influenced greatly by topography. Major tree species include western hemlock, Sitka spruce, western red cedar, yellow cedar and red alder. Methods: The goals of this study did not require a randomized statistically valid survey. Rather, the need was to provide data on a cross section of landslides within logged areas. Therefore, the sites selected for this study were selected non randomly. A subsample of the 27 watersheds was used to provide a reasonable cross section of the Island's geologic and climatic conditions, timber harvesting practices and ages of cut blocks. Following this, a subsample landslide population was chosen from each of the selected watersheds. The landslides chosen were to be representative of a range of sites, failure types, and land use practices. Variables Examined: Parameters of interest include ranges and modal values of initiation slopes, topographic position and location, surficial geology, bedrock, terrain and site moisture features. Results: The survey randomly collected 77 debris slides, 11 debris flows and 14 debris torrents. They made the following general descriptions of the failures: 1) Open slope debris slides developed on steep (70% +), convex or uniform slopes covered with colluvial veneers over bedrock. They often occurred in shallow, linear slope depressions or seepage zones and often initiated at or below convex breaks in slopes of 20% or more. Debris slides in gullies were commonly small sidewall failures or scars; 2) Debris flows occurred in wet depressional sites and deep morainal deposits on moderate to steep (45-60%), uniform or concave slopes. 3) Debris torrents occurred in steep-gradient V-notch gullies, but frequently were triggered by large debris slides or flows that initiated on open slopes adjacent to the gullies. The major factors contributing to these failures appeared to be overloading of steep slopes and inadequate control of road drainages, usually in combination. Reference: Krag, R.K., E.A. Sauder and G.V. Wellburn. 1986. A forest engineering analysis of landslides in logged areas on the Queen Charlotte Islands, British Columbia. B.C. Ministry of Forests and Lands. Land Management Report Number 43. 138p. 85 The Queen Charlotte Islands, British Columbia Study Site: The study site consists of the entire Queen Charlotte Islands. Objectives: The study presents a regional overview of slope failures derived from an analysis of small scale aerial photographs. The three major objectives for this inventory were to: 1) Document mass wasting processes on the Islands using 1:50,000 scale aerial photos; 2) Aid in the selection of representative study watersheds for the FFIP research component; and 3) develop a data base for slope stability research. Area Description: See Krag et al. Methods: Aerial photographs (1:50,000 and 1:60,000) from 1979 and (1:63,000) from 1976 were used in the verification of slope movement features. Additional photo missions of the Rennell Sound area and 20 additional watersheds were photographed in normal color at the scale of 1:12,000. Results: A total of 8,240 various slope failure forms were identified from the 1:50,000 scale aerial photographs. Three landslide intensity classes were identified. Moderate landscapes had 1-3 failures per km2 and occupied 7% of the islands. Severe landscapes had 4 to 7 failures per km2 and occupied 7 % of the islands. Extreme landscapes had more than 8 failures per km2 and occupied only 1 % of the total area. The majority of the slope movements identified in the study occurred on slopes steeper than 30%. Approximately 1/4 of the failures occurred on slopes from 30% to 60%, about 1/2 on 61-90% slopes and the remaining 1/4 on slopes steeper than 90%. Aspect was also hypothesized as playing a major role in the occurrence of slope movements. The southerly (S, SE and SW) exposed slopes account for approximately 1/2 of all landslides, followed by the northerly (N, NW and NE) aspect with about 1/3 of the failures. Two thirds of all slope failures occurred on slopes underlain by the soft (Masset formation) and hard (Karmatsen formation) volcanic rocks, which also dominate the bedrock geology of Islands. The remaining 1/3 of the failures are associated with the intrusive plutonic rocks and various carbonate and clastic sedimentary types. Reference: Gimborzevsky, P. 1983. Regional overview of mass wasting on the Queen Charlotte Islands. Fish/Forestry Interaction Program, Working Paper 3/83. 34p. 86 Rennell Sound, Queen Charlotte Islands, Brit ish Columbia Study Area: The Rennell Sound study area is located on the west coast of Graham Island in the Queen Charlotte Islands. It covers 160 km2. Objectives: The project was undertaken to achieve the following objectives: 1) inventory natural and accelerated (man caused) soil mass movements; 2) determine characteristic features of failure sites; and 3) provide recommendations for forest management. Area Description: The study area lies within the Skidgate Plateau. The bedrock geology of the study area is divided into two main halves. The Masset Formation covers the northern portion of the study area and is characterized by a thick accumulation of volcanic flows and pyroclastic rocks. The majority of the southern portion of the study area lies within the Yakoun Formation. This formation is primarily volcanic, dominated by pyroclastic rocks formed largely of porphyritic andesite. Colluvium is the dominant surficial material. The climate is characterized as humid mesothermal- mild winters, adequate precipitation in all months for plant growth, and no dry season. Extreme precipitation events usually occur with high winds. The study area lies within a wet subzone of the Coastal Western Hemlock Biogeoclimatic zone. Methods: Air photos were initially used to identify natural soil mass movements. The variables recorded are noted below. Field inspection of soil mass movements took place over a 20 month period from June 1976 to February 1978. Variables Examined: The following variables were recorded from aerial photographs: slope angle, elevation, aspect, slope position and slope length. Traverses into unidentified sites resulted in additional observations on mass movements not initially apparent on aerial photographs. Site factors measured under these circumstances included: aspect, elevation, slope angle, slope types, slope position, slope length, bedrock, landform and soil, rooting depth, drainage, vegetation and mass movement process. Results: The study examined 322 natural soil mass movements which represents a frequency of 20 events per square kilometer. Non significant site factors involved in identifying unstable sites in the study area included aspect, bedrock geology and elevation at the stating zone. The important site factors were thought to be slope angle, seismic activity, soil hydrology and the influence of trees. The majority of natural soil mass movements initiate on slopes between 31° and 45°. The mean slope being 36.4°. There was little surficial material on slopes over 45° and virtually non on slopes over 60°. The other three site factors were hypothesized as being important is soil mass movements for the area. Reference: Wilford, D.J., and J.W. Schwab. 1979. Soil mass movements in the Rennell Sound area, Queen Charlotte Islands, British Columbia. British Columbia Forest Service. 75p. 87 Rennell Sound, Queen Charlotte Islands, Brit ish Columbia Study Site: The Rennell Sound study area is located on the west coast of Graham Island, the major northern island of the Queen Charlotte Islands. The area lies along the western shores of Rennell Sound and covers 160 k m 2 of forest land. Objectives: The objectives of this study were to: 1) identify mass movement failures triggered by the October-November, 1978 storm; 2) quantify area disturbed and volume of material moved; and 3) determine the relative rate and frequency of mass movements on roads, clearcuts and forested terrain. Area Description: Most of the Rennell Sound study area is covered by a thick accumulation of volcanic flows and pyroclastic rocks. The topography of the study area has been extensively modified by glaciation and morainal deposits are located throughout. The morainal material has a clay loam to sandy loam texture, depending upon bedrock type and the degree of soil formation. Colluvium is the dominant surficial material. Stable terrain supports forest stands of western red cedar, yellow cedar, shore pine and western hemlock. The soils are organic or Humic Podsols developed over tills. Methods: The survey recorded all mass movements in the study area caused by the October-November 1978 storm. Only failures in excess of 25 nv' were recorded. The survey was conducted by traveling all access roads by vehicle or on foot, and by making foot traverses through some clearcuts. Variables Examined: Information collected at each failure site included: location (watershed, cutting permit, road), mass movement process, date of harvesting or road construction, time since harvesting or construction, landscape position, landform, bedrock formation, slope gradient (origin), size, volume and apparent triggering factor. The survey grouped slopes into three classes: 1) > 3 6 ° ; 2) 20-36°; and 3) <20° . Results: The October-November 1978 storm caused 264 mass movements within the study area. Forty six percent of all slope failures occurred on slopes greater than 36°, 13 % on slopes between 20° and 36° and 2% on slopes less than 20°. The additional 39% of the failures occurred in steep gullied terrain. There were also significant differences in the rates of mass wasting depending on the various bedrock formations. Mass movement since logging also showed the predictable 4 to 6 year high failure frequency. Roads contributed more mass wasting per unit area than do clearcuts, however, clearcut failures produced 5 times more failures. Reference: Schwab, J.W. 1983. Mass wasting: October-November 1978 storm, Rennell Sound, Queen Charlotte Islands, British Columbia. Province of British Columbia, Ministry of Forests Research Note 91. 88 Skidgate Plateau, Queen Charlotte Islands Study Site: The study area comprises a series of selected clearcut areas within the Skidgate Plateau in the Queen Charlotte Islands. Objectives: This paper addresses past logging landslide frequencies as they relate to individual landscape attributes and combinations of landscape attributes. The relationships described are directed toward the development of multifactor classifications for slope hazard mapping. Area of Description: The Skidegate Plateau is a long, narrow, northwest trending surface which separates the Queen Charlotte Ranges in the west from the Queen Charlotte Lowlands to the east. Elevation of the plateau surface ranges from a few metres in the east to 750 metres in the west. Bedrock formations are composed primarily of volcanics, sediments and metasediments. The plateau was overridden by ice during the Pleistocene, which has resulted in all but the steepest slopes being dominated by morainal materials. Precipitation occurs predominately during October and Apri l . Methods: The study was restricted to logged areas that were from 6 to 15 years old. The terrain within each selected area was mapped at a scale of 1:20,000 according to a terrain classification system in common use in British Columbia. Twenty eight randomly selected logged areas were studied, making up a total of 768 terrain polygons. The aggregate area mapped was approximately 3,380 ha. Variables Examined: The following variables were examined for each polygon: slope angle; presence or absence of natural landslides; presence or absence of minor natural failures; landscape position; slope morphology, dominant/subdominant terrain types; bedrock formation; horizontal slope curvature; dominant soil type; soil drainage; and aspect. Results: Analysis of Variance and the Kruskal-Wallis test were used to compare landslide frequencies with selected landscape attributes. The most common types of landslides within the study area are clearcut and roadfill landslides, at 84.5 and 14% respectively. With respect to roadfill landslides a select number of landscape attributes are statistically significant. In the case of the ANOVA test, slope angle, presence or absence of natural landslides, landscape position and slope morphology are the only significant variables. With respect to the Kruskal-Wallis test the significant variables were slope angle, presence or absence of natural landslides, presence or absence of minor natural failures, landscape position, slope morphology, dominant/subdominant terrain types and horizontal slope curvature. With respect to clearcut landslides a greater number of variables prove significant than do roadfill failures. With regard to the ANOVA test only slope aspect was not significant, while in the Kruskal-Wallis test, slope drainage and slope aspect proved to be insignificant. Reference: Rollerson, T.P. 1992. Relationships between landscape attributes and landslide frequencies after logging: Skidegate Plateau, Queen Charlotte Islands. British Columbia Ministry of Forests, Land Management Report 76. 89 Coastal Mountains of southwest Brit ish Columbia Study Site: The study area consists of 640 k m 2 of forested steeplands in southwestern British Columbia. Objective: The objectives of this Ph. D. thesis were to: 1) determine whether large scale harvesting in the Coast Mountains seriously influences the stability of steep slopes; and 2) to determine what natural and human caused factors are detrimental to the stability of steep forested slopes. Area Description: Three easily recognizable categories of steepland soils were distinguished on the basis of their drainage characteristics, parent material and profile development. Category 1 are soils developing on well drained sites over t i l l , colluvium or alluvium. Category 2 are soils developing on poorly drained sites over unweathered ti l l or bedrock. Category 3 are soils developing on bedrock. The predominant tree species are red cedar, western hemlock and Douglas fir. At higher elevations mountain hemlock, yellow cedar and amabilis fir gain in predominance. The study area possesses a cool, moist, mountain climate with significant amounts of winter snowfalls. Methods: Several hundred air photos of approximately 1:30,000 taken in 1939, 1952, 1966, 1967 and 1969 respectively and three series of 1:15,000 black and white air photos taken in 1957, 1963 and 1968 enabled the locations and approximate age of the larger landslides to be determined. Variables Examined: At each landslide site the mean width, mean depth and length of scar, the mean slope along each 15 m section of the scar, the altitude and aspect of the scar and the mean soil depth to impermeable bedrock or unweathered ti l l were measured. Results: Within the study site, 77 landslides of the debris avalanche or debris slide type were discovered and recorded. Forty slides were investigated in the field while the remaining 28 were identified on 1968 and 1969 aerial photos. Most of the landslides that were investigated in the field initiated in drainage depressions, small seepage hollows or on poorly drained parts of open slopes. Seventy five percent of the measured landslides occurred on slopes ranging from 31° to 39°. Sites with surface slopes greater than 40 ° are usually extremely rocky and not prone to large scale failures. Approximately 75% of the landslides possessed aspects with a southerly component. SSE was the most prevalent aspect, accounting for 20%. A Chi squared (%2) test indicates that aspect plays a part in the location of slope failure. However, there is a preferred orientation of slopes towards the south in both the Capilano and Seymour catchments. This is the result of the north facing slopes being rocky and broken. This condition discourages landslide formation. The south facing slopes are relatively uniform and underlain by an extensive unweathered ti l l substratum. Most landslides seemed to occur between the altitudes of 1,000 and 1,100m. This coincides with the fact that numerous roads associated failures occurred within this range of altitude. The uniformity of the bedrock across the study area would tend to rule against bedrock type as a cause of spatial variability in landslide numbers. However, the field study did show that bedrock differences may exercise some control over large slope failures related to road construction. Reference: O'Loughlin, C L . 1972. The stability of steepland forest soils in the Coast Mountains, southwest British Columbia. Ph. D. thesis, University of British Columbia, Vancouver. 147p. 90 Southern Coast Mountains, British Columbia Study Site: The study area is located 65 km east of Vancouver and includes the Norrish, Cascade and Deroche drainage basins. Objectives: The main objectives of this paper were to: 1) detail the terrain evaluation method for predicting landslide hazards; 2) present an example landslide inventory map and computer generated landslide clearcut rating map; 3) describe some of the results of the landslide inventory study; 4) define the types of terrain with the greatest landslide potential following clearcut logging; and 5) compare natural and clearcut landslide rates on different types of terrain as well as comparing changes in these rates over time (since 1940) with documented long tern changes in local precipitation records. Area Description: The study area is underlain by granitic rocks of the Coast Plutonic Rocks, except for the Deroche drainage basin, which is underlain by Jurassic meta-volcanic rocks. The landscape is the product of multiple glaciations, which have subsequently smoothed the ridges and peaks and oversteepened the valley side slopes. These higher elevation areas are mantled discontinuously by glacial t i l l or colluvium. Lower valley slopes are overlain by continuous and thicker t i l l deposits and pockets of glaciofluvial gravels. The area experiences cool, wet winters and dry, warm summers. Annual precipitation, based on records from Mission is approximately 1770 mm with 69 % of it falling between October-March. Methods: In order to develop a clearcut landslide rating scheme three programmes were followed. They were: 1) landslide programme in which landslides with a minimum area of 100 m2 were identified on 1:15000 and 1:12000 scale, black and white aerial photographs from 1940, 1962, 1968, 1979 and 1981. A landslide inventory map at a scale of 1:20000 was produced showing landslide locations and site characteristics; 2) terrain mapping programme mapped the surficial geology of the study area according to the Terrain Classification System (ELUC). Homogeneous areas are delimited on the map and referred to as terrain polygons. This was done with 1:20000 aerial photographs, field checked in the summer of 1984; and 3) landslide rating programme which has two objectives: a) to rate each terrain polygon on its likelihood of failing after clearcut logging; and b) to assess the likelihood of landslide debris impacting upon the stream system i f a rapid mass movement should occur within a terrain polygon. Variables Examined: Site factors used in the initial definition of the terrain classes were: type of material and drainage; depth and percent cover of surficial material; slope angle, which varies for different material types; slope configuration; presence of gully erosion; and aspect. Results: A total of 595 mass movements were recorded within the study area. The landslides occurred in gully sides and headwall, hillslopes not dissected by gullies (open hillslopes) and escarpments in unconsolidated sediments. Landslides were most common in areas undergoing gully erosion (44%), despite the total area of the study in this terrain classification (15%). Open hillslopes which accounted for 72% of the study area, accounted for only 48% of the landslides. Most failures occurred in either colluvium that is rapidly drained (42%) and t i l l which is moderately to rapidly drained (36%). The mean slope angle of natural and clearcut t i l l landslides generated on open slopes is 33° and 32.1°. The mean slope angle of natural and clearcut colluvial landslides is 38.2° and 36.5° respectively. Aspect appears to play a significant role in failure initiation. The preferred south aspect of landslide initiation is thought to be a function of the way in which local winds move up the valley in the study area, particularly during the winter months. The preferred north aspect is thought to be the result of different moisture conditions on these slopes versus other slopes. Reference: 91 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 of British Columbia. B.C. Ministry of Environment and Parks. MOEP Technical Report 28. 38p. 92 MacMil lan Bloedel Limited Holdings on Vancouver Island, Brit ish Columbia Study Site: The area inventoried for this study were MacMillan Bloedel Holdings located throughout Vancouver Island. Objective: The conditions, concerns and methods adopted by MacMillan Bloedel on Vancouver Island to minimize environmental impacts by timber harvesting on steep terrain are the main objectives. Area Description: Vancouver Island consists of extensive areas of steep terrain (greater than 30° slopes). This terrain is primarily covered by t i l l and/or colluvial deposits with bedrock outcrops interspersed. The mountains of the Island result in an uplifting of the air mass causing large amounts of precipitation to fall on the west coast areas. The climate on Vancouver Island consists of relatively high winter precipitation followed by low summer rainfall with the mountainous terrain playing a large role in the precipitation patterns. The steep forested areas are located primarily in the wet Western Hemlock and Mountain Hemlock zones. Methods: An aerial photo interpretation inventory, using 1:15,840 photos was conducted to identify management related (road or clearcut area) landslides. Results: A total of 272 failures (road and clearcut) were identified on MacMillan Bloedel holdings. Only 35% of the slides were road induced. The photo interpretation study indicated that 2.6 times more road induced slides per km of road occurred on 30-35° slopes compared to > 35° slopes. The inventory showed that areas with high intensity precipitation events (Northeast Coast, West Coast) had more frequent landslides than low intensity regions. There was an increasing frequency of failures for steep slopes compared to moderate slopes for clearcut area slides. Slides from roads in steep areas showed an opposite trend. The following tables were provided: Road induced slides/km of road on 30-35° and > 35° slopes Region 30-35° slopes > 35° slopes Southeast Coast .31 .32 Northeast Coast 1.15 .31 Central Island .22 .03 West Coast 1.03 .28 Vancouver Island (average) .54 .15 Number of clearcut induced slides/100 ha of logged area with 30-35° and > 35° slopes Region 30-35° slopes > 35° slopes Southeast Coast .79 4.45 Northeast Coast 3.34 4.00 Central Island .17 .54 West Coast 2.47 2.92 Vancouver Island (average) .67 .87 Reference: Bourgeois, W.W. 1978. Timber harvesting activities on steep Vancouver Island terrain. In C T . Youngberg, ed. Forest soils and land use. Proc, Fifth North Am. For. Soils Conf., Ft. Collins, Colo., 393-409pp. 93 MacMillan Bloedel Ltd.'s Tree Farm License 44, Vancouver Island, British Columbia Study Site: The study examines 178 polygons in eight selected clearcut areas located within the western portion of MacMillan Bloedel's Tree Farm License 44. These areas lie along the southwest coast of Vancouver Island and are located between 5 and 20 km inland from the coast. Objectives: The objective was to provide an alternative method to classical engineering approaches for developing slope stability maps in forested watersheds where detailed data on precipitation, slope hydrology and soil physical properties are limited, non existent or i f collected, are expected to be highly variable. Area Description: The study area lies within the Coastal Western Hemlock biogeoclimatic zone, wetter maritime subzone. The dominant tree species in the area are western hemlock, Pacific silver f ir and western red cedar with minor components of Douglas fir, Sitka spruce and yellow cedar. The majority of soils fall within the Ferro-Humic Podzolic Great Group. The temperate climate is characterized by abundant rainfall and strong winter storms from the Pacific Ocean. The bedrock lithology is highly mixed, consisting of extrusive and intrusive complexes and some metamorphics. Methods: The logged portion of each of the eight selected areas was mapped for surficial materials, land surface morphology and geomorphic process using the ELUC Terrain Classification System. The field mapping scale was 1:20,000. Each terrain unit mapped was visited in the field and a set of data describing the geometric and geomorphic attributes of the polygon recorded. For each terrain class several simple statistics can then be generated, based on the post-logging failures of the class members. Variables Examined: The variables entered into the data base for each polygon include: dominant surficial material; whether or not the hillslope in benchy or irregular; presence or absence of gullies; presence or absence of natural failures; average slope angle in degrees; and average aspect in degrees; polygon area; the number of presumed logging induced clearcut failures; and the number of road failures. Results: Landslides identified with clearcutting were found in 34 of 168 terrain polygons clearcut; these polygons contained a total of 77 failures. Landslides associated with roads occurred in 20 of the 133 polygons roaded; these 20 polygons contained a total of 42 failures. Three predictive classifications (classifications 1, 2 and 3) were developed for clearcut failures and two for road failures (classification A + B). The authors concluded that the best way to maximize the number of failures in a class while minimizing the number of polygons in that class is to develop separate predictive classifications for clearcut versus road failures. A classification developed specifically for clearcut landslide prediction allows the forest manager to minimize the land base that need be removed from harvesting while maximizing the total number of landslides prevented by such action. Similarly, a predictive classification developed specifically for road failures allows a manager to reduce the number of areas where it is necessary to either use specialized road building techniques or avoid road construction while mamtaining the number of failures prevented. Reference: Rollerson, T.P., and M. Sondheim. 1985. Predicting post-logging terrain stability: A statistical-geographical approach. Procs., Joint Symposium of the I.U.F.R.O. Mountain Logging Section and the Sixth Pacific Northwest Skyline Symposium. Vancouver, B.C. 94 Weyerhaeuser Company Lands in Western Washington and Oregon Study Site: Lands owned by Weyerhaeuser Company and intervening public lands west of the Cascade mountains between Snohomish River in Washington and the Millicoma River in Oregon were studied. Objectives: Specific landscape parameters associated with landslides from forest roads were examined in the coast and Cascade ranges of western Washington and western Oregon to develop a slope stability risk assessment technique that can be used by logging engineers as an aid in road location, design and construction in areas of questionable slope stability. Area Description: The study areas was divided into five physiographic Provinces: 1) foothills of the northern Washington Cascades; 2) southwestern Washington Cascades; 3) west central Oregon Cascades; 4) Washington coast range; and 5) Oregon coast range. Each physiographic area was further separated by the predominating geology and climate. Methods: Township-centered aerial photographs taken in 1972, 1974 and 1978 at a scale of 1:64,000 in combination with geologic maps were used to identify mass failures associated with forest roads for each province. To ensure uniform sampling throughout each province, soil survey reports were used to identify soil associations. Area in each association was determined and the number of individual study sites were chosen according to the weighted areas. This resulted in 134 50 ha sites. Each site was then delineated on 1:12,000 scale aerial photographs, all haul roads within the site boundary were divided into 6 1 m segments, existing landslides recorded and stability variables assigned to each segment by photo interpretation. The numerical values representing the individual stability variables for each segment by group were examined by stepwise discriminant function analysis. Variables Examined: The following variables were examined: slope angle; slope position, slope form and stand age; presence of perched ground water; soil depth and subsoil texture; bedrock type; elevation; and road placement. Relative weighting values were assigned to each of these stability variables. Results: The authors determined that to be useful the slope stability indexing system must be capable of an accuracy of at least 80 % with a 95 % level of confidence. The authors decided that the sum of the 9 individual variables could be considered to represent an index of stability for each site. This index was termed the "stability index". The application of this procedure on the 4,266 data points in a correct assignment of 88 % for the failed group and 66 % for the nonfail group and was significant at the 99 % level of confidence. The analyses were completed with development of the following equation: Failure Risk = 3.6236 + 1.0787 (index) + .034287 (index)^ The indexing technique, the authors warn is unique to the area and must therefore be considered as having limited geographic application. Reference: Duncan, S.H., J.W. Ward and R.J. Anderson. 1987. A method of assessing landslide potential as an aid in forest road placement. Northwest Sci. 61(3): 152-159. 95 Clearwater National Forest and Middle Fork of the Payette River, Idaho Study Site: Two sites within the western and central northern Rocky Mountain physiographic province in Idaho made up this survey. The first survey area was approximately 5,670 k m 2 within the Clearwater National Forest (CNF). The second survey area was approximately 800 k m 2 within the Boise National Forest occupying the Middle Fork of the Payette River drainage (MFPR). Objectives: The objectives of the study were to: 1) quantify the magnitude of landslide occurrence; 2) estimate damages caused by landslides; 3) characterize properties of landslide prone areas; and 4) evaluate mass erosion in relation to various land uses. Area Description: The sites are located in the Idaho Batholith. A variety of igneous rocks, collectively referred to as granitic, make up the batholith. Topography ranges from low to high relief. The weather patterns on the two study areas are typical of the Mediterranean climate that influences the Pacific Northwest. Both areas are almost entirely forested. The dominant tree species within the CNF are Grand fir and western red cedar. Within the MFPR, Douglas fir and ponderosa pine are the most common species. Methods: The study methods varied for the two areas. On the CNF, data was collected during each summer for a three year period to document only those landslides that occurred during the previous winter and spring. Data collection began in 1974 and continued through 1976. Data collection on the MFPR consisted of a detailed survey during the summer and fall of all landslides in the drainage regardless of age. The slides sampled were located by: 1) aerial reconnaissance from fixed wing or helicopter aircraft in inaccessible areas; 2) period aerial photography; 3) ground observations by forest workers familiar with the area; and 4) on site reconnaissance along roads. Variables Examined: A variety of data were collected to describe the slide site, including geology (bedrock type, weathering class, fracturing class and bedding); landform (slope gradient, aspect, slope location and size of watershed contributing area); soils (soil depth and texture); vegetation (crown cover, habitat type and rooting characteristics); and site disturbances (type of disturbance, age of disturbance, location in road prism and rock type). Results: A total of 1,418 landslides were inventoried during the study area- 629 on the CNF and 789 on the MFPR. Landslides were most common of slopes ranging from 51 % to 70%. Slides in this range accounted for 47 % of the total number. They found most landslides occurred on slope of about 30° (60%) while slides were rare on slopes greater than 41° (90%). Such is the case because the shear stress on extremely steep slopes is so great that soil materials do not have a chance to accumulate but rather move off the slopes as dry creep or rock falls. Slope gradients may also vary according to position on the slope and slope shape. Hence, it is likely that landslide occurrence varies with slope position. Landslide frequency increases from upper to middle and lower slope locations (18%, 31 % and 3 8 % respectively). The drainage area above a slide is an important factor regulating the soil water content at the slide site. The greater the drainage area, the greater the soil water. However, a stream channel is formed i f the drainage area becomes too large; the soil becomes saturated causing surface runoff. For the areas studied, 95 % of the landslides occur on drainages of 8 ha or less, and 90 % occur in drainages of 4 ha or less. The median size drainage area for landslides in this steep mountain area is only .6 ha. The single most important factor found contributing to landslides in the Northern Rocky Mountains was road construction, accounting for 58 % of the landslides. 96 Reference: Megahan, W.F., N.F. Day and T . M . Bliss. 1978. Landslide occurrence in the Western and Central Rocky Mountain physiographic province in Idaho. In C T . Youngberg, ed. Forest soils and land use. Proc., Fifth North Am. For. Soils Conf., Ft. Collins, Colo., 116-139pp. 97 Klamath Mountains of southwest Oregon Study Site: The Klamath mountains of southwest Oregon within the Siskiyou National Forest. The study covered 137,000 acres. Objectives: The objectives of this study were to estimate quantitatively the effects of forest management activities on the frequency and volume of mass movements and to collect information concerning conditions at landslide sites which might serve as a guide to appraising future risks of landslides. Area Description: The study areas consists predominantly of protertiary sediments and volcanics that have been folded, faulted and intruded by serpentinized masses of ultra basic and granitoid rocks at fault contacts between the bands. The area is generally between 2,000 and 5,000 feet in elevation and highly dissected with narrow canyons. Annual precipitation, mainly winter ram, ranges from 50 to over 150 inches. Methods: Aerial photographs taken in 1956 and 1976 at a scale of 1:15,840 were inspected. Site variables were recorded and the volume (to the nearest 100 yd 3 ) of mass movements entering drainage ways was estimated during the 20 year interval. Variables Examined: The following variables were examined: slope angle; position on slope; aspect; mean annual precipitation; and geomorphological erosion response unit (GERU). The GERUs were nine soil and geologic types grounded to reflect rock types or landforms expected to have similar responses to disturbance. Results: Almost 1.5 million y d 3 of debris slide erosion occurred during the 20 years spanned by the inventory. Slide frequency was about one slide every 4.3 years on each 1,000 acres and the erosion rate was about l /2yd 3 per acre per year. Roads, occupying only 2% of the area inventoried, were the sites of over half of the slides and 60% of the slide volume. Harvest areas, occupying 10% of the area inventoried, yielded 34% of the slides and produced 18% of the slide volume. The remaining 88% of the study area, which was in a natural condition, produced only 22% of the slide volume. Debris slide frequency were associated with slope. Terrain steeper than 70% made up 9% of the sample area but contained 76 % of the debris slide erosion. Over one third of the slides and nearly half of the slide volume occurred on the lower one third of slopes. However, slides occurring on the upper one third of the slope were, on average, larger than those on the lower two thirds of the slope. There was no definite relationship between aspect and erosion rate or slide frequency. Slide frequency showed a slight, though erratic, upward trend with increasing mean annual precipitation. The culluvial, hummocky and intrusive GERUs had the highest natural slide erosion rates, but only modest proportional increases when subjected to management. Reference: Amaranthus, M.P., R.M. Rice, N.R. Barr and R.R. Ziemer. 1985. Logging and forest roads related to increased debris slides in southwestern Oregon. / . of For. 83(4):229-233. 98 Mapleton Ranger District, Central Oregon Study Site: The Mapleton District, Siuslaw National Forest in the central Oregon ranges acted as the study site. Objectives: The primary objective of this study was to determine the influence timber management activities have on soil mass movements within the study area. Area Description: The Oregon Coast range is composed primarily of marine and estuarine sedimentary rocks with pockets of igneous rocks. Soils are derived from sandstone and siltstone and are poorly developed. Slope gradient varies considerably for different Soil Resource Inventory (SRI) land types. The major species of trees are Douglas fir, western hemlock, western red cedar, red alder, big leaf maple and wine maple. Methods: Following a major storm in November 1975, a field inventory of 245 mass movements, was conducted in the Mapleton Ranger District of the Siuslaw National Forest. Color aerial photographs at a scale of 1:15,850 were used to mark the location of individual mass movements. Only mass movements which entered a drainageway and whose volume exceeded 10 yd^ were counted. Approximately 70% of the District's land area was inventoried. Variables Measured: The following variables were measured for each site: Soil Resource Inventory (SRI); slope steepness; aspect; slope position; and time (years since cutting). Results: The majority of failures resulted from soils derived from residuum and colluvium over micaceous sandstone. These soils were characteristically thin, gravely, sandy loams to loams. Slope steepness was also an important factor in landslide generation with 95 % of all inventoried failures occurring on slopes greater than 70% (31.5°). In the case of natural failures the average gradient was 81 % and in the case of in unit failure 90% and above was the predominant slope. Aspect was the least important variable in the study. Natural landslides were nearly equally distributed on northerly and southerly slopes. However, approximately 94% of road related and in unit failure occurred on northern aspects. Slope position for in unit failures seemed to exert a considerable influence on landslide occurrence with 78% of the inventoried failures located on the mid-slope or at stream adjacent sites. Time since felling also exerted a significant control for in unit slope failure. Sixty three percent of failures occurred on clearcut units harvested within 3 years preceding the storm. Reference: Gresswell, S., D. Heller and D.N. Swanston. 1979. Mass movement response to forest management in the central Oregon coast ranges. Res. Bull. PNW-84. Portland, OR: U.S. For. Serv., Pacific Northwest Forest and Range Experiment Station. 26p. 99 Klamath Mountains, northwestern California and Oregon Study Site: Data were collected from 62 ha of clearcut patches on the English Peak Batholith, 7 km northwest of Sawyers Bar, California. Two test data sets also come from granitic batholiths of the Klamath Mountains. One is the Whits Rock Batholith, 12 km south of Tiller, Oregon while the other is the Ashland Batholith, 8 km south of Ashland, Oregon. Objectives: This paper reports a study to test how successfully the Rice and Pillsbury (1982) equation can be applied in other areas and explores how its performance is altered by differences between areas. It also describes a method for utilizing the results of a discriminant analysis, together with quantitative estimates of the volumes and risks at stake, to arrive an optimum management strategy. Area Description: Granitic plotons make up about 17 % of the Klamath Mountains. The slopes of all three study areas were steep, soil depths were shallow, generally less than 1 m, but were often underlain by 1 m or more of weathered parent material. The histories of the three batholiths are similar. Shortly after each study area was logged i t was struck by a storm exceeding Caine's (1980) landslide threshold for rainfall duration of 24 hours or more. Methods: In each of the study areas, only slides unrelated to roads , landings or stream channel undercutting were included in the analysis. Variables Examined: The Rice and Pillsbury (1982) equation was based on four variables: X = 13.24 - 4.63 log (SLOPE) - 3.04 log (DOM) - 1.03 log (HORSTM) + .69 log (DRAREA) where c is the discriminant score of a site SLOPE is the terrain slope of the slide or a 60 m slope segment centered on a stable site, expressed as a percent. D O M is the ground cover provided by the crowns on the dominant vegetation within a .405 ha circular plot, expressed as a percent. HORSTM is the horizontal distance in meters from the stream to the centroid of the slide or to the stable site DRAREA is the tributary surface drainage area above the slide scarp or stable sampling point in square meters. SLOPE indexes the magnitude of the force of gravity promoting failure. DOM is more complicated. Its coefficient indicates that sites supporting heavier timber are more prone to failure after logging. DRAREA seems to be indexing the same conditions as HORSTM because small drainage areas are indicative of instability. It may be however, that small drainage areas also index the possibility of subsurface delivery of water from the slope an the other side of the ridge. Results: The first statistical test of the data was a Chi-squared (c^) goodness of fit test addressing the question of whether the distributions of discriminant scores in the two test areas can be considered as coming from the same population of discriminant scores as those collected on the English Peak Batholith. Following the Chi square test a more crucial test was performed using BMD-P4F to compare the correct and incorrect classifications in each of the test areas in a three way contingency table. Only 25 % of the slide sites were correctly identified. Because DRAREA was the least significant variable in the discriminant function, they eliminated it and recomputed the discriminant function, still using English Peak Batholith data. c = 17.08 - 5.67 log (SLOPE) - 3.21 log (DOM) - 1.19 log (HORSTM) This equation predicted had an overall accuracy of 75 %. As a final step in the analysis, they recomputed this equation using all of the data c = 16.59 - 5.54 log (SLOPE) - 3.69 log (DOM) - .42 log (HORSTM) 100 This equation correctly identified 76 % of the stable sites and 91 % of the unstable sites, for an overall accuracy of 83 %. Reference: Rice, R.M., N.H. Pillsbury and K.W. Schmidt. 1985. A risk analysis approach for using discriminant functions to manage logging-related landslides on granitic terrain. For. Sci. 31(3):772-784. 101 Six Rivers National Forest, Northwestern California Study Site: Clearcut patches from the entire Six Rivers National Forest and the combined catchments of Hurdy-gurdy and Jones Creeks were used for this study. Objectives: This paper dicusses a method for predicting where landslides of the debris avalanche type are likely to occur after clearcutting of steep slopes in northwestern California. The techniques are based on discrimianat analysis of conditions associated with slide and non-slide sites. Area Description: The Six Rivers National Forest lies within the Douglas-fir region of the Klamath and Coast Range Mountains. The mean annual precipitation ranges from 100 to 315 cm. The Hurdy-gurdy and Jones Creeks catchments in the northern part of the SRNF are 140 km^ in area. Structurally complex, heterogeneous and highly fractured bedrock combine with steep topography and a Mediterranean climate to produce exceptionally rapid erosion rates. The Galice formation comprised of sandstones and argillaceous rocks and the Franciscan Group comprised primarily of sandstone and mudstone are the dominate geologic formations in the area. Methods: The sample included clearcut patches spanning 15 years of age, assuming that landslides resulting from clearcutting would occur within that time. The Hurdy-gurdy and Jones Creek catchments contained 21 clearcut patches having steepland slopes visible on 1:15,840 color air photographs. The patches included a range of geomorphic, meterologic and vegetative conditions and were relatively accessible for field verification. Site variables at 35 slide sites and 79 non-slide sites were measured. A discriminant function developed for this area was to be tested with data collected from the Six Rivers National Forest. The study used a series of discriminant analyses and multiple regression analyses to screen variables. Variable combinations were evaluated by R , adjusted R^ and percentages of correctly classified sites. An individual variable was considered valuable i f it significantly and consistently contributed to the correct classification of sites in a series of analyses and if, after split-sample testing, it maintained its relative value. Variables Examined: Nineteen variables describing environmental conditions at slide centriods and non-slide points were estimated. Eight additional variables were derived from transformations. Variables selected were those that could be easily measured on air photographs or maps and that required no field verification. The variables examines were: site (slide or non-slide); PTAREA- planametric photo area of slide; AGE- age of cutblock relative to date; ROCK- rock type; MOR- hillslope-contour morphology; SLOPE- slope of ground surface; LASP- local, first-order aspect; RASP- regional, second-order aspect; ORD- order of downslope stream; DOM- dominant canopy closure in a 1,000 m2 circle centered on site; DISDVD- horizontal distance to upslope divide; DISUCB- horizontal distance to upper cutblock boundary; DISUP- horizontal upslope distance to first convex slope break; DISST- horizontal downslope distance to stream; ELV- elevation above mean sea level; ELVBDVD- elevation below divide; ELVAST-\ elevation above stream; MAP- mean annual precipitation; MAE- mean annual evaporation; MAR- mean annual runoff; and TY24H- two year 24 hour storm depth. Variables created from transformations: MORPH- hillslope contour morphology; SINLASP- Sine(LASP); COSLASP- Cosine(LASP); SINRASP-Sine(RASP); COSRASP- Cosine(RASP); LASOUTH- §180- LASP e; RASOUTH- e 180- RASP e; SUMDIS- 1/DISUP + 1/ DISST. Results: The debris avalanche density was found to be about one slide per square kilometer of non-steepland and about 41.4 slides per km^ of steeplands. Therefore, 95% of the slides originated within steeplands, constituting less than 1/3 of the clearcut area in the SRNF. On the basis of the HJC data, the final discriminant equation was: Y = 3.439 + 4.350 log SUMDIS + .882 log LASOUTH + 1.593 MORPH 102 MORPH equaled .825 i f a site was located within a draw, or, -. 175 i f it was located on a convex or planar hillslope. This resulted in a correct classification of 76.3 % for the site and non slide sites. SRNF data resulted in a total correct classification of 73.8%. The analysis was complete with the development of the following equation based on SRNF data: Y = 3.245 + 3.017 log SUMDIS + 1.312 MORPH MORPH equaled .837 (draw) or -.163 (planar convex). This equation correctly classifies 81.0% of the sites. The discrimination of sites by the variable SUMDIS indicates that a high proportion of the sampled slides occurred near streams and immediately below major convex breaks in slope (usually the upper boundary of the steeplands. The high proportion of slides near streams likely reflects a general, continuous oversteepening of hillslopes by stream undercutting. High moisture levels likely account for the high proportion of landslides associated with draws, as indicated by the variable MORPH. Major slope breaks make a discontinuity where material is nearer to failure threshold conditions below the break than above. The regolith is thinner below the break because, due to steeper slopes erosion can more nearly keep pace with weathering. The high proportion of slides on north facing slopes in the HJC (as indicated by LASOUTH) may reflect locally pervasive orientations of rock foliation's conducive to failure or aspect related variations in the relative availability of moisture or both. Because moist northerly aspects, slide frequencies may reflect a greater dependency of such aspects upon denser tree root networks for their strength. Reference: Furbish, D.J., and R.M. Rice. 1983. Predicting landslides related to clearcut logging, northwestern California, U.S.A.. Mount. Res. Develop. 3(3):253-259. 103 Private Lands in Northern California Study Site: The study was conducted on private land in northwestern California. Objective: The main objective of this paper is to develop an equation(s) that may estimate the probability that a site wi l l yield more than 100 yd^ per acre of erosion i f the timber is harvested or i f a road is built there. Area Description: Not described Methods: The study is based on data collected from a sample of 1,104 Timber Harvesting Plans (THPs) completed in California between November 1978 and October 1979. Owners of the properties covered by 655 THPs granted permission to include their land in the study. Of these, 415 THPs made up the total of the sample population. Study plots were normally two acres in area. A total of 172 variables were recorded for each plot. Only 31 of these were used in the harvest area analysis and 25 in the road analysis. Variables Examined: Variables that were measured included descriptors of topography, geology and soil, hydrology, vegetative cover, weather and climate, and management impacts. The final discriminant function equations included only a small number of variables. The discriminant function for forest roads was: DS = -.0281 - .1142 * SLOPE + 75.16 * HCURVE + 1.0075 * HUE. The discriminant function for logged areas was: DS = 5.032 - .1633 * SLOPE + 67.88 * HCURVE - 1.215 * WEAKROCK SLOPE is the terrain slope in degrees. HCURVE is the horizontal curvature of the road centerline in the equation for roads and of the terrain in the equation of logged areas. Horizontal curvature was the reciprocal of the radius of a circle passing through the measurement site and two other points on the same contour about 60 ft on either side of the site. It was measured i f ft and coded negative for swales and positive for ridges, being zero on planar slopes. HUE is the Munsell hue of moist subsoil. WEAKROCK is coded + 1 i f a bedrock specimen crumbles or deforms under hammer blows and -1 i f it fractures. Slope indexes the partitioning of the force of gravity into two components: a normal component increasing friction and fostering stability and a tangential component promoting down-slope movement and failure. HCURVE may index the accumulation of unstable amounts of colluvium in swales and the convergence of subsurface water, which can lead to high pore water pressures that cause failures. HUE also probably indexes subsurface water. WEAKROCK indexes the strength of materials resisting failure. Results: The authors were satisfied with the equations derived. No road function containing more variables had a greater apparent prediction accuracy than that presented. The equation for the logged areas was exceeded in accuracy by four and five variable models, but they preferred the three variable function as a guard against outfitting the data. The estimated accuracy's of the equations, which were corrected for bias using bootstrapping, are 78 % for the forest road construction and 69 % for the logged area equation. Reference: Rice, R.M., and J. Lewis. 1991. Estimating erosion risk associated with logging and forest roads in northwestern California. Water Resour. Bull. 27(5):809-817. 104 Appendix B: FORM.FOR FORTRAN Program C FORM.FOR C C A FORTRAN-77 PROGRAM TO CALCULATE THE SPATIAL DERIVATIVES USING A THIRD C ORDER LaGRANGE POLYNOMIAL C C ORIGINAL PROGRAM WRTTTEN BY: C K.O. N IEMANN AND D.E. HOWES C C REFERENCE: C NIEMANN, K.O. AND D.E. HOWES. 1992. SLOPE STABILITY EVALUATIONS USING DIGITAL TERRAIN MODELS. B.C. MINISTRY OF FORESTS. VICTORIA, B.C.28p. C C MODIFIED BY: C DAVE FAUCETTE C DEPARTMENT OF FOREST RESOURCES MANAGEMENT C U N J ^ R S I T Y OF BRITISH COLUMBIA C PARAMETER (L=817,M=631) INTERGER*4 EL(L,M),NROW,NCOL,I,J REAL*4D,E,F,G,H,GS,SQG,SQH,TEMP,SLOPE(L,M),ASPECT(L,M),$DOWNSL(L,M), ACRSSL(L,M),TEMP1 CHARACTER*12 INPUT, OUTP1, OUTP2, OUTP3, OUTP4 C C INPUT CRITERIA C P R I N T V N A M E OF INPUT ELEVATION IMAGE:* READ(* 2) INPUT 2 FORMAT (A12) PRINT*, 'WHAT IS THE RESOLUTION OF THE IMAGE: ' READ(* *) GS 3 PRINT*, 'ENTER THE NUMBER OF COLUMNS:' READ(*,*) NCOL IF(NCOL.GT.M) THEN PRINT*,'SORRY, THE MAX. NUMBER IS: ' ,M GO TO 3 ENDIF 4 PRINT*,'ENTER THE NUMBER OF ROWS:' READ(*,*) NROW IF(NROW.GT.L) THEN PRINT*,'SORRY, THE MAX. NUMBER IS: ' ,L GO TO 4 ENDIF OUTPUT1 = 'SLOPE.IMG' OUTPUT2=' ASPECT.IMG' OUTPUT3 = 'DOWNSL.IMG' OUTPUT4=' ACRSSL.IMG' OPEN(3,FILE=INPUT) OPEN(5,FILE=OUTPl) OPEN(7 ,FILE=OUTP2) OPEN(9,FILE=OUTP3) OPEN(l 1 ,FILE=OUTP4) C 105 C READ IDRISI ELEVATION MATRIX I N COLUMN AND ROW ORDER C C D O 1 0 I = l , N R O W D O 2 0 J = l , N C O L READ(3,*)EL(I,J) 20 CONTINUE 10 CONTINUE C C START M A I N LOOP C C CALCULATE THE POLYNOMIAL COEFFICIENTS FOR 3 X 3 SUB MATRICES C DO 100 I=2 ,NROW- l DO 300 J=2 , NCOL-1 D=(( (EL( I , J-l)+EL(I,J+l))/2)-EL(I,J))/(GS*GS) E=(((EL(I-1 ,J )+EL( I+1 ,J))/2)-EL(L,J))/(GS*GS) F=(-1*EL(I-1,J-1)+EL(I-1,J+1)+ EL( I+ 1,J-1)-EL(I+ 1,J+ 1))/(4*(GS*GS)) G=(-1*EL<I,J-1)-fELil.J+1))/(2*GS) H=(EL(I-1,J)-EL(I+1, J))/(2*GS) C SQG=G*G SQH=H*H C C CLACULATE THE SPATIAL DERIVATIVES C SLOPE(I,J) = ((SQG) + (SQH))**0.5 IF((H.EQ.0.0.AMXG.EQ.0.0).OR.(G.EQ.0.0))THEN ASPECT(I,J)=361 ELSE TEMPI =ABS(H/G) .. .TEMP=(ATAN(TEMP1))*57.958 IF(TEMP.LT.0)THEN ASPECT(I,J)=361 ELSE IF(H.GT.0.0.AND.G.GT.0.0)ASPECTa,J)=270-(ABS(TEMP)) IF(H.GT.0.0.AND.G.LT.0.0)ASPECT(I,J)=90+(ABS(TEMP)) IF(H.LT.0.0.AND.G.LT.0.0)ASPECTa,J)=90-(ABS(TEMP)) JT(H.LT.0.0.A>TO.G.GT.0.0)ASPECT(I,r)=270+(ABS(TEMP)) ENDIF ENDIF IF(G.EQ.0.0.AND.H.EQ.0.0)THEN DOWNSL=0.00001 ACRSSL=0.00001 ELSE DOWNSL=(-2*((D*(SQG))+(E*(SQH))+(F*G*H)))/((SQG)+(SQH)) ACRSSL=(-2*((D*(SQH))+(E*SQG))-(F*G*H)))/((SQG)+(SQH)) ENDIF 300 CONTINUE 100 CONTINUE C C WRITE IDRISI IMAGE FILE DO 151=1 , NROW D 0 2 5 J = 1 , NCOL 106 WRTTE(5,111 l)SLOPE(I, J) 1111 FORMAT(F10.2) WRITE(7,1111)ASPECT(I,J) WRTTE(9,111 l)DOWNSL(I,J) WRTTE(11,1111)ACRSSL(I,J) 25 CONTINUE 15 CONTINUE C CLOSE(3) CLOSE(5) CLOSE(7) CLOSE(9) CLOSE(l l ) STOP END 107 Appendix C: CATCH.FOR FORTRAN Program C CATCH.FOR C C A FORTRAN-77 PROGRAM TO DETERMINE THE LOCAL AND GLOBAL CATCHMENT AT C EVERY ELEMENT OF A N ELEVATION MATRIX. C C IT TAKES IDRISI ELEVATION IMAGE I N ASCII FORMAT AND CALCULATES CATCHMENT C AREAS BELONGING TO ANY PIXEL I N THE IMAGE C C ORIGINAL PROGRAM WRITTEN BY: C DR. LAWRENCE W. MARTZ C DEPARTMENT OF GEOGRAPHY, UNIVERSITY OF SASKATCHEWAN C SASKATOON, CANADA S7N 0W0 C C REFERENCE: C MARTZ, L.W. AND JONG, E. 1988. CATCH: A FORTRAN PROGRAM FOR MEASURING CATCHMENT AREA FROM DIGITAL ELEVATION MODELS. COMPUTERS & GEOSCIENCES 5(14):PP. 627-640. C C MODIFIED BY: C DAVE FAUCETTE C DEPARTMENT OF FOREST RESOURCES MANAGEMENT C UNIVERSITY OF BRITISH COLUMBIA C C DIMENSION ARRAYS PARAMETER (L=817,M=631) INTEGER ELEV(L,M),IROW,ICOL,NROW,NCOL,IRl,ICl,DEPR(L,M) CHARACTER*15 INPUT,OUTPl,OUTP2 REAL CA(L,M) C C OPEN FILES AND GET INFORMATION C PRINT*, 'NAME OF INPUT ELEVATION IMAGE (WITH EXTENSION):* READ(*,22) INPUT 22 FORMAT(A15) 33 PRINT*,'ENTER THE NUMBER OF COLUMNS:' READ(*,*) NCOL IF (NCOL.GT.M) THEN PRINT*,'SORRY, THE MAX. NUMBER IS: ' ,M GO TO 33 ENDD? 44 PRINT*,'ENTER THE NUMBER OF ROWS:' READ(* *) NROW IF(NROW.GT.L) THEN PRINT*,'SORRY, THE MAX. NUMBER IS:*,L GO TO 44 ENDIF OUTP1 = 'CATCH.IMG' OUTP2='DEPR,IMG' OPEN(3 ,FILE=INPUT) OPEN(5,FILE=OUTP1) OPEN(7,FILE=OUTP2) C 108 C READ IDRISI ELEVATION MATRIX I N ROW AND COLUMN ORDER C DO 100 IROW= l.NROW READ(3,*)(ELEV(IROW,ICOL),ICOL=l,NCOL) 100 CONTINUE C C SCAN ELEVATION MATRIX.. . C DO 5 IROW=l ,NROW D 0 4 I C O L = l , N C O L C SET THE CURRENT POSITION OF THE FLOWLINE AT THIS ELEMENT I R l = I R O W I C l = I C O L C C CHECK IF CURRENT FLOWLINE POSITION IS EDGE ELEMENT 2 CALLEDGE(*10,IR1,IC1,NROW,NCOL) C C IF NOT EDGE ELEMENT, FIND NEIGHBOURS TO WHICH SLOPE IS GREATEST C CALL ADVANC(*3,IR1,IC1,ELEV,M,L,IR2,IC2) C C IF FLOWLINE ADVANCES, INCREMENT CATCHMENT AREA BY ONE AND CHANGE C CURRENT POSITION OF FLOWLINE C IF(IRl.NE.IROW.OR.ICl.NE.ICOL) CA(TR1,IC1)=CA(IR1,IC1)+1.0 IR1=IR2 IC1=IC2 GO TO 2 C C IF FLOWLINE CANNOT ADVANCE (I.E. I N A DEPRESSION), DECREASE CATCHMENT C AREA BY ONE AND BEGIN NEW FLOWLINE 3 CA(IR1,IC1)=CA(IR1,IC1)-1.0 DEPR(IR1,IC1)=1 GO TO 4 C C IF CURRENT FLOWLINE POSITION WAS A N EDGE ELEMENT, INCREMENT CATCHMENT C AREA BY ONE AND BEGIN NEW FLOWLINE 10 IF(IRl.NE.rROW.OR.ICl.NE.ICOL) CA(TR1,IC1)=CA(IR1,IC1) + 1.0 4 CONTINUE 5 CONTINUE C C WRITE LOCAL CATCHMENT AREA MATRIX TO OUTPUT FILE DO 13 IROW=l ,NROW DO 12 ICOL= l ,NCOL WRTTE(5,103) ABS(CA(IROW,ICOL)) WRTTE(7,104) DEPR(IROW,ICOL) 12 CONTINUE 13 CONTINUE 103 FORMAT(F10.4) 104 FORMAT(I2) CLOSE(3) CLOSE(5) CLOSE(7) STOP E N D C S U B R O U T I N E S * * * * * * * * * * * * * * * * * * * * * * * * * * C S U B R O U T I N E E D G E ( * , I R O W , I C O L , N R O W , N C O L ) C C H E C K S I F A N E L E M E N T IS A N E D G E E L E M E N T I N T E G E R J R O W . I C O L . I R . I C . N R O W . N C O L D 0 2 I R = - 1 , 1 D 0 1 I C = - 1 , 1 I F ( I R O W + I R . L T . 1 . 0 R . I R O W + r R . G T . N R O W . O R . I C O L + I C . L T . 1 . 0 R . I C O L $ . G T . N C O L ) R E T U R N 1 1 C O N T I N U E 2 C O N T I N U E R E T U R N E N D c S U B R O U T I N E A D V A N C ( * , I R 1 , I C 1 , E L E V , M , L , I R 2 , I C 2 ) C F I N D S G R E A T E S T S L O P E B E T W E E N E L E M E N T A N D ITS N E I G H B O U R S A N D C H E C K S I F S L O P E IS P O S I T I V E O R N E G A T I V E I N T E G E R m j C , m U C l , I R 2 , I C 2 , E L E V ( L , M ) R E A L S M A X . S L P D 0 2 I R = - 1 , 1 D O 1 I C = - 1 , 1 I F ( I R . E Q . O . A N D . I C . E Q . O ) G O T O 1 C A L L S L O P E ( I R l , I C l , r R , I C , E L E V , M , L , S L P ) I F ( T R . E Q . - 1 . A N D . I C . E Q . - 1 ) T H E N S M A X = S L P I R 2 = I R 1 - 1 I C 2 = I C 1 - 1 G O T O 1 E N D I F I F ( S L P . G T . S M A X ) T H E N S M A X = S L P I R 2 = I R 1 + I R A C 2 = I C 1 + I C ' E N D I F 1 C O N T I N U E 2 C O N T I N U E I F ) S M A X . L E . O . O ) R E T U R N 1 R E T U R N E N D C S U B R O U T I N E S L O P E ( I R O W , I C O L , I R , I C , E L E V , M , L , S L P ) C C C A L C U L A T E S R E L A T I V E S L O P E M A G N I T U D E B E T W E E N E L E M E N T S C I N T E G E R I R O W , I C O L , I R , I C , E L E V ( L , M ) R E A L S L P I F ( A B S ( I R ) . E Q . A B S ( I C ) ) T H E N S L P = ( E L E V ( T R O W , I C O L ) - E L E V ( T R O W + I R , I C O L + I C ) ) / ( 2 * * 0 . 5 ) E L S E S L P = ( E L E V ( i R O W , I C O L ) - E L E V ( I R O W + I R , I C O L + I C ) ) ENDIF RETURN END 


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