Landslide Susceptibility along the Sea to Sky Highway


	Discussion *Assumed values and Limitations of the Model Used* The model used is essentially a physically-based digital terrain model for mapping the relative slope stability potential across a landscape. By using digital elevation models, the simple equations which solve for slopes that are likely to be unstable can be modeled. This allows for an inexpensive and relatively easy method for modeling the stability of hillslopes. The model does require some background knowledge of the physical processes which act on the study area. This is where the model requires fairly large assumptions to run. Without the knowledge of acceptable values for the assumed parameters, the outcome of the model may be significantly altered. The values we used were taken from both the paper by Montgomery and Dietrich (1994) and from the advice of Dr. Brett Eaton, Instructor of Hillslope Geomorphology at the University of British Columbia. We came up with the following list of assumptions: Assuming the soil has no cohesion is acceptable for the case of saturated soils, but this does not take into account the role of cohesion from the roots of vegetation. To account for the cohesion provided by vegetation we increased the friction angle from the typical values of 30-40 degrees to 45 degrees. According to Montgomery and Dietrich (1994) this is acceptable in accounting for the role of cohesion from vegetation. The assumed value of the soil bulk density was taken from the advice of Dr. Brett Eaton, where typical values are anywhere from 1600 to 2000 kg/m3. The value for the soil transmissivity was obtained from Montgomery and Dietrich (1994). Dietrich and Montgomery (1998) found that slopes identified as unconditionally unstable commonly correspond to sites of bedrock outcrop. These observations are consistent with the results achieved in this project, whereby comparison of satellite imagery to the landslide susceptibility map shows that many slopes classified as unconditionally unstable correspond to exposed slope areas (Refer to 3D Visualization Diagram). From field observations, Dietrich and Montgomery (1998) came to the conclusion that areas classified as unconditionally stable are environments that can support saturated overland flow without failing. Essentially, slopes classified as unconditionally stable are too low in angle for saturation to cause instability. The model lacks predictive capabilities because it assumes many parameters which affect the stability of the hillslopes. For a model to accurately predict areas of instability, it would require the knowledge of many distinct soil properties such as; soil depth, saturated soil depth, and the hydrological conditions of the slope. Without this very specific data, which varies greatly over the area of a simple channel, the model lacks predictive capability. With this being said, the model does provide a good indication of the areas which may become unstable, but the results need to be overlaid by some historical data of past landslides. Thus, the analysis conducted in this project can be considered to be introductory, with results being indicative of areas of potential high landslide susceptibility requiring further detailed investigation and analysis. *Effects of Increased Steady State Rainfall on Slope Stability* Analysis of the effects of increasing steady state rainfall with varied friction angles on slope stability was conducted in Microsoft Excel using area data derived from histograms produced in ArcGIS. Graph 1 below shows the variation in slopes classified as unstable for friction angles of 40 degrees, 45 degrees, and 50 degrees. As can be seen, higher friction angles generally result in less area classed as unstable, however, the trend for all friction angles is very similar, as can be expected. Visualization of these results can be seen in the animated images for friction angles of 40 degrees, 45 degrees, and 50 degrees in the results section. Graph 2 shows the obvious relationship between friction angle and areas of unconditionally stable and unconditionally unstable slopes, that is, unconditionally stable slope area increases as friction angle increases, and unconditionally unstable slope area decreases. It should be noted that the increase in unconditionally stable area, as friction angle increases, does not equal the decrease in unconditionally unstable area. That is, unconditionally stable areas increase at a greater rate than unconditionally unstable areas of land decrease, as can be seen in Graph 2. Graph 1 - Unstable Area for Varied Friction Angle Graph 2 - Unconditionally Stable and Unconditionally Unstable Areas for Varied Friction Angles Increasing steady state rainfall and the area of unstable and stable slopes has a logarithmic relationship as can be seen in Graphs 3, 4, and 5 below. Increasing the friction angle has the effect of decreasing the area of unstable slopes. Although this is a basic and expected observation, the graphs below provide a useful visual tool for the comparison of the effects of friction angle on areas classified as stable or unstable. Graph 3 - Slope Stability as a function of Steady State Rainfall for a Friction Angle of 40 Degrees Graph 4 - Slope Stability as a function of Steady State Rainfall for a Friction Angle of 45 Degrees Graph 5 - Slope Stability as a function of Steady State Rainfall for a Friction Angle of 50 Degrees Analysing the effects of varying the friction angle on areas classified as stable or unstable is essentially a basic attempt at a sensitivity analysis. The graphs give an indication to the amount of variation that could be expected if the friction angle that we adopted (45 degrees) was inappropriate. The results that we have obtained using the parameters adopted are generally consistent with previous knowledge of landsliding along the Sea to Sky highway, including the location of control structures and historical occurences of landslides. This suggests that a friction angle of 45 degrees is an appropriate value, however, it can only be considered an estimate, and thus results must be considered to be relatively low accuracy. *Spatial Data and Analysis Concerns* The major spatial data concern in this project is with the spatial resolution of the DEMs used. This project relied upon DEMs for the creation of both the hydrological and slope models, and as such the resolution of the DEMs is of fundamental importance to the precision and accuracy of the results. The DEMs used had a spatial resolution of 30m, which we consider an appropriate resolution at the scale of the study. However, it must be recognized that a 30m resolution DEM will not provide the detail to identify very high angle slopes, for example in small gullies, which may be of concern in the identification of susceptible slopes. That is, some susceptible slopes may be missed with a DEM spatial resolution of 30m, as detailed topography is generalized, or 'smoothed'. Montgomery and Dietrich (1994) use a DEM with a 5m spatial resolution, however, their study area is also much smaller. Higher resolution DEMs would enable more precise susceptibility maps due to being able to identify highly localised changes in topography, and flow accumulation to greater precision and accuracy, however, a 30m spatial resolution DEM is appropriate for a generalized shallow slope landsliding model for the size of study attempted in this project. There is some rule uncertainty related to the method of flow direction used to calculate flow accumulation. The method used in this project is the default method of ArcGIS 9, that is, the D8 algorithm, which uses a 3x3 neighbourhood operation on a DEM to approximate the flow direction into units of 45 Degrees. Other methods and variations of the D8 exist, which can provide more precise flow direction rasters including the Rho8 algorithm, a statistical version of the D8 algorithm, and the FD8 and FRho8 algorithms which are modifications of the D8 algorithim that allow for flow dispersion to be modelled (Burrough and McDonnell, 1998). It is difficult to suggest which algorithm would provide the most accurate results, and as the other algorithms were not used in this project a comparison of the differences is not available. A brief mention of the classification scheme used in the landslide susceptibility maps is appropriate. A quantile classification scheme was used, as out of other options including natural breaks, and standard deviation classifaction schemes, the quantile classification provided the greatest variation in the lower values of required critical steady state rainfall. Variation in the lower values of critical steady state rainfall required for slope instability are the most important in determining landslide susceptibility. For example, it is much more useful to provide classes which show slopes that will potentially fail at 50mm and 75mm of steady state rainfall, compared with classes that show slopes that will potentially fail at 50mm and 150mm of steady state rainfall, as lower amounts of steady state rainfall will occur more frequently, thus slopes potentially failing at lower steady state rainfalls are much more susceptible to shallow landsliding. *Improvements and Future Work* There are a number of improvements and alterations which could be done to provide a more comprehensive analysis, better precision and accuracy, and possibly more useful results than have been provided in this project. Firstly, acquisition of higher resolution DEMs would allow for more topographic detail and thus more precise and accurate flow accumulation and slope layers, and consequently more precise and accurate landslide susceptibility mapping. A climatic model could be incorporated into the model, with statistical analysis on precipitation to predict the likelihood of particular slopes to fail. In combination with data on return periods of debris flows of particular magnitudes and frequencies, a probabalistic risk map could be produced, which would be more useful in zoning applications and in consideration of construction of defense structures. A more comprehensive analysis could have been achieved in this project had different methods of flow direction been applied in the creation of the hydrological model, to analyse differences obtained using the various algorithms. A useful addition to the analysis conducted in this project might also include some analysis of debris flow routing to determine areas that would likely be affected by channelized debris flows initiated on unstable slopes at high elevations. Other improvements to the model used in this project are centred around the availability of data. If high resolution and accurate soil data, surficial geology, and vegetation cover were obtained, the accuracy of estimated parameters in this model could be improved. However, given the very localised variation of factors such as soil and slope cohesion as a result of vegetation cover, a reasonably high degree of uncertainty will always be present. Back to the Top Home

Discussion

Assumed values and Limitations of the Model Used

The model used is essentially a physically-based digital terrain model for mapping the relative slope stability potential across a landscape. By using digital elevation models, the simple equations which solve for slopes that are likely to be unstable can be modeled. This allows for an inexpensive and relatively easy method for modeling the stability of hillslopes. The model does require some background knowledge of the physical processes which act on the study area. This is where the model requires fairly large assumptions to run. Without the knowledge of acceptable values for the assumed parameters, the outcome of the model may be significantly altered. The values we used were taken from both the paper by Montgomery and Dietrich (1994) and from the advice of Dr. Brett Eaton, Instructor of Hillslope Geomorphology at the University of British Columbia. We came up with the following list of assumptions:

Assuming the soil has no cohesion is acceptable for the case of saturated soils, but this does not take into account the role of cohesion from the roots of vegetation. To account for the cohesion provided by vegetation we increased the friction angle from the typical values of 30-40 degrees to 45 degrees. According to Montgomery and Dietrich (1994) this is acceptable in accounting for the role of cohesion from vegetation. The assumed value of the soil bulk density was taken from the advice of Dr. Brett Eaton, where typical values are anywhere from 1600 to 2000 kg/m3. The value for the soil transmissivity was obtained from Montgomery and Dietrich (1994).

Dietrich and Montgomery (1998) found that slopes identified as unconditionally unstable commonly correspond to sites of bedrock outcrop. These observations are consistent with the results achieved in this project, whereby comparison of satellite imagery to the landslide susceptibility map shows that many slopes classified as unconditionally unstable correspond to exposed slope areas (Refer to 3D Visualization Diagram). From field observations, Dietrich and Montgomery (1998) came to the conclusion that areas classified as unconditionally stable are environments that can support saturated overland flow without failing. Essentially, slopes classified as unconditionally stable are too low in angle for saturation to cause instability.

The model lacks predictive capabilities because it assumes many parameters which affect the stability of the hillslopes. For a model to accurately predict areas of instability, it would require the knowledge of many distinct soil properties such as; soil depth, saturated soil depth, and the hydrological conditions of the slope. Without this very specific data, which varies greatly over the area of a simple channel, the model lacks predictive capability. With this being said, the model does provide a good indication of the areas which may become unstable, but the results need to be overlaid by some historical data of past landslides. Thus, the analysis conducted in this project can be considered to be introductory, with results being indicative of areas of potential high landslide susceptibility requiring further detailed investigation and analysis.

Effects of Increased Steady State Rainfall on Slope Stability

Analysis of the effects of increasing steady state rainfall with varied friction angles on slope stability was conducted in Microsoft Excel using area data derived from histograms produced in ArcGIS. Graph 1 below shows the variation in slopes classified as unstable for friction angles of 40 degrees, 45 degrees, and 50 degrees. As can be seen, higher friction angles generally result in less area classed as unstable, however, the trend for all friction angles is very similar, as can be expected. Visualization of these results can be seen in the animated images for friction angles of 40 degrees, 45 degrees, and 50 degrees in the results section.
Graph 2 shows the obvious relationship between friction angle and areas of unconditionally stable and unconditionally unstable slopes, that is, unconditionally stable slope area increases as friction angle increases, and unconditionally unstable slope area decreases. It should be noted that the increase in unconditionally stable area, as friction angle increases, does not equal the decrease in unconditionally unstable area. That is, unconditionally stable areas increase at a greater rate than unconditionally unstable areas of land decrease, as can be seen in Graph 2.

Graph 1 - Unstable Area for Varied Friction Angle

Graph 2 - Unconditionally Stable and Unconditionally Unstable Areas for Varied Friction Angles

Increasing steady state rainfall and the area of unstable and stable slopes has a logarithmic relationship as can be seen in Graphs 3, 4, and 5 below. Increasing the friction angle has the effect of decreasing the area of unstable slopes. Although this is a basic and expected observation, the graphs below provide a useful visual tool for the comparison of the effects of friction angle on areas classified as stable or unstable.

Graph 3 - Slope Stability as a function of Steady State Rainfall for a Friction Angle of 40 Degrees

Graph 4 - Slope Stability as a function of Steady State Rainfall for a Friction Angle of 45 Degrees

Graph 5 - Slope Stability as a function of Steady State Rainfall for a Friction Angle of 50 Degrees

Analysing the effects of varying the friction angle on areas classified as stable or unstable is essentially a basic attempt at a sensitivity analysis. The graphs give an indication to the amount of variation that could be expected if the friction angle that we adopted (45 degrees) was inappropriate. The results that we have obtained using the parameters adopted are generally consistent with previous knowledge of landsliding along the Sea to Sky highway, including the location of control structures and historical occurences of landslides. This suggests that a friction angle of 45 degrees is an appropriate value, however, it can only be considered an estimate, and thus results must be considered to be relatively low accuracy.

Spatial Data and Analysis Concerns

The major spatial data concern in this project is with the spatial resolution of the DEMs used. This project relied upon DEMs for the creation of both the hydrological and slope models, and as such the resolution of the DEMs is of fundamental importance to the precision and accuracy of the results. The DEMs used had a spatial resolution of 30m, which we consider an appropriate resolution at the scale of the study. However, it must be recognized that a 30m resolution DEM will not provide the detail to identify very high angle slopes, for example in small gullies, which may be of concern in the identification of susceptible slopes. That is, some susceptible slopes may be missed with a DEM spatial resolution of 30m, as detailed topography is generalized, or 'smoothed'.

Montgomery and Dietrich (1994) use a DEM with a 5m spatial resolution, however, their study area is also much smaller. Higher resolution DEMs would enable more precise susceptibility maps due to being able to identify highly localised changes in topography, and flow accumulation to greater precision and accuracy, however, a 30m spatial resolution DEM is appropriate for a generalized shallow slope landsliding model for the size of study attempted in this project.

There is some rule uncertainty related to the method of flow direction used to calculate flow accumulation. The method used in this project is the default method of ArcGIS 9, that is, the D8 algorithm, which uses a 3x3 neighbourhood operation on a DEM to approximate the flow direction into units of 45 Degrees. Other methods and variations of the D8 exist, which can provide more precise flow direction rasters including the Rho8 algorithm, a statistical version of the D8 algorithm, and the FD8 and FRho8 algorithms which are modifications of the D8 algorithim that allow for flow dispersion to be modelled (Burrough and McDonnell, 1998). It is difficult to suggest which algorithm would provide the most accurate results, and as the other algorithms were not used in this project a comparison of the differences is not available.

A brief mention of the classification scheme used in the landslide susceptibility maps is appropriate. A quantile classification scheme was used, as out of other options including natural breaks, and standard deviation classifaction schemes, the quantile classification provided the greatest variation in the lower values of required critical steady state rainfall. Variation in the lower values of critical steady state rainfall required for slope instability are the most important in determining landslide susceptibility. For example, it is much more useful to provide classes which show slopes that will potentially fail at 50mm and 75mm of steady state rainfall, compared with classes that show slopes that will potentially fail at 50mm and 150mm of steady state rainfall, as lower amounts of steady state rainfall will occur more frequently, thus slopes potentially failing at lower steady state rainfalls are much more susceptible to shallow landsliding.

Improvements and Future Work

There are a number of improvements and alterations which could be done to provide a more comprehensive analysis, better precision and accuracy, and possibly more useful results than have been provided in this project.

Firstly, acquisition of higher resolution DEMs would allow for more topographic detail and thus more precise and accurate flow accumulation and slope layers, and consequently more precise and accurate landslide susceptibility mapping. A climatic model could be incorporated into the model, with statistical analysis on precipitation to predict the likelihood of particular slopes to fail. In combination with data on return periods of debris flows of particular magnitudes and frequencies, a probabalistic risk map could be produced, which would be more useful in zoning applications and in consideration of construction of defense structures.

A more comprehensive analysis could have been achieved in this project had different methods of flow direction been applied in the creation of the hydrological model, to analyse differences obtained using the various algorithms. A useful addition to the analysis conducted in this project might also include some analysis of debris flow routing to determine areas that would likely be affected by channelized debris flows initiated on unstable slopes at high elevations.

Other improvements to the model used in this project are centred around the availability of data. If high resolution and accurate soil data, surficial geology, and vegetation cover were obtained, the accuracy of estimated parameters in this model could be improved. However, given the very localised variation of factors such as soil and slope cohesion as a result of vegetation cover, a reasonably high degree of uncertainty will always be present.

Back to the Top

Home