Three-dimensional slope stability analysis using limit equilibrium methods is a time-consuming procedure. Using traditional sampling methods such as Monte Carlo or Latin Hypercube may take days of computation, especially when there are multiple random variables and complicated geometries that require advanced analysis methods. Stochastic response surface (SRS) method is a very fast and effective approach for probabilistic analysis of 3D complicated geometries which reduces the number of simulations and simulation time dramatically. The stochastic response surface method takes a fraction of spread out samples that cover the parameter space and uses those to train the model. Any number of samples can then be plugged into this model and will result in the estimated factor of safety values for each sample.
In this study, an SRS algorithm using third order Hermite polynomial expansion, developed for complex 3D probabilistic analysis is presented. A multilayered slope with several random variables has been investigated using the SRS method and the results are compared with Latin Hypercube simulation results. The results using both methods are in good agreement. However, the SRS method computation time is a fraction of that of the Latin Hypercube simulation.

The 2017 Elephant Hill wildfire burned 190,000 hectares in British Columbia. In August 2018, intense rainfall over the burned area initiated many debris flows that swept across Highways 97 and 99, which endangered people and properties and caused a fatality. Areas of high burn severity continued to generate debris flow activity in 2019, with highways blocked numerous times by debris and mudflow deposits. Two mudflows in a small valley-side watershed were triggered by intense rainfall in June 2019. These mudflows blocked Loon Lake road and affected a private residence located on a fan deposit at the Bonaparte River. Debris flow hazards are widely acknowledged to increase after a wildfire causes significant burn severity in a watershed. Increased awareness of the hazard level is necessary, but this is often insufficient to reduce the risk for homeowners with structures built on old fans. Mitigative measures are required. This paper uses the June 2019 mudflows to illustrate how relatively low-cost deflection berms can be designed and implemented to mitigate geohazards. A key design element is the use of a detailed topographic map containing 0.5 m contours of the fan and the locations of structures requiring protection. For this purpose, a Remotely Piloted Aircraft System was used to collect aerial images. In this case study, images were taken several days after the mudflows occurred. However, the goal of this paper is to highlight a hazard mitigation workflow that can be completed following a wildfire, yet before small debris flows have the opportunity to affect developments on a fan. The aerial images were processed in Structure-for-Motion software to generate a digital elevation model, fan topography, and orthophotos. These provide the basis for locating deflection berms and channels to direct future debris away from vulnerable structures.

A low-cost datalogger was built in the lab and installed in place of conventional, commercially available datalogging systems. The units have been deployed at three sites across western Canada in British Columbia, Alberta, and Saskatchewan. The project involved the modification of open source software and programming readily available on the Internet. Data collected with conventional units and the constructed datalogger were compared to determine the accuracy and precision of the lower cost units. Testing of the low-cost datalogging units have shown them to be highly adaptable, with the ability to measure negative pore water pressure (matric suction), volumetric water content, and temperature from SDI-12 sensors as well as positive pore water pressure and temperature from vibrating wire piezometers. Telemetry has been built into the more remote set of dataloggers to transmit occasional data points and periodically verify that the unit and sensors are logging as expected. Assembly, installation, and ongoing monitoring using the low-cost datalogging system has occurred over the past two years as these units have proven durability and continue to operate at each location. Field application of these dataloggers to date has encountered some challenges that are addressed, documenting some limitations to the development of a low-cost datalogger for geotechnical instrumentation.

Delineating the soil properties and the associated uncertainties are critical in geotechnical risk assessment, particularly in a regional urban area and for infrastructural development projects. Modeling the spatial distribution of the topsoil morphology (i.e. thickness) and properties (sand, clay, till, etc.) is the first step of a regional geotechnical risk analysis. This is particularly challenging in areas with highly variable soil properties and limited soil sampling. The assessment of soil properties and associated uncertainties need a proper estimation procedure to depict a realistic variability in a design model. We propose a probabilistic approach to model the soil types at a regional scale, and to produce a thickness map of topsoil and consequently of the bedrock topography, assisted by kriging interpolators. The highly clustered sampling pattern results in highly skewed data distribution with weak spatial stationarity; the estimation process included in the proposed approach alleviates this cause of uncertainty. The methods comprise Transformed-Gaussian kriging (via Box-Cox method), kriging with external drift, and Empirical Bayesian Kriging (EBK) in addition to Triangulated Irregular Network. The results show that the approach of locally varying mean and variance in EBK outcomes a more accurate method in regional studies involving extensive data. After indicator transformation of soil types, the spatial dependency, and the geological continuity of domains are appraised by indicator variograms. Then the soil types and associated probability of occurrence are delineated within the space between bedrock and the ground surface topography using sequential indicator simulations. The predicted soil types and their probabilities improve key geological points for probabilistic geotechnical risk evaluation.

Snow avalanches differ from other geohazards such as debris flows and landslides in ways that affect mapping and mitigation. In contrast to other geohazards: snow avalanches consist of a material that exists and fails near its melting point; have more frequent occurrences in a specified track or path; and the deposits melt within months.

Snow avalanches, especially those of small magnitude, occur frequently in same path. Where observations are available for a decade or more, this often results in better occurrence and runout records. Runouts can be linked to return periods, often for return periods up to 10 or 30 years. Runouts can then be extrapolated up to a return period of ~100 years. However, because snow avalanche deposits melt within months, subsurface sampling cannot detect avalanche runouts with a long return periods, e.g. 1000 years.

Snow avalanches start as a result of failure in a bonded granular material in which the bonds are usually within 10°C of the melting point. (This temperature threshold cannot be used to predict avalanche release since most snow in a temperate climate exists in this temperature range.) Periods of instability are often limited to hours or days after which an increase in bonding restores stability. Explosives can be effective triggers of unstable snow, thereby shortening periods of instability and allowing recently threatened terrain to be re-opened for human activity quickly.

As a result of these differences: snow avalanche mapping and land-use guidelines in most jurisdictions do not refer to return periods longer than 300 years; short term closures of human activity in snow avalanche runout zones are often practical; and human activity with moving elements at risk such as roads and ski operations are more often located in areas threatened by snow avalanches than by other geohazards.

Slope stability is a critical terrain characteristic for safe rail operation and a focus of geotechnical engineers globally. Recent studies indicate that the impacts of landslides within the Canadian Prairies cost Canada’s two major rail service providers between $10 and $18 million dollars annually in direct damages and prevention funding. The Assiniboine River Valley is a critical transportation corridor located within the Canadian Prairies which links the east to the west. X-ray computer tomography (CT) is a powerful non-destructive analysis tool that has been in use since the early 1970’s where internal structures of objects can be observed based primarily on variations in density and atomic composition. The original applications were primarily reserved for the medical industry to make qualitative observations such as dark areas in human lungs. More recently, it’s gained increasing popularity in the geoscience community due to an increase in availability and affordability of scanners, particularly medical grade and desktop scanners capable of micro CT imaging. This coupled with the increase in computing power and post processing software development now allows quantitative analysis. This technology, however, still appears to be under utilized as a core analysis tool for slope stability. This paper aims to present available techniques, explore the reasons for under-utilization of this technology and assesses its overall validity as an analysis tool for slope stability purposes using an active slow-moving landslide in the Assiniboine River Valley that is affecting Canadian Nationals Mainline.