No abstract

The development of progressive failure in sensitive clay slopes is analysed using principles from fracture mechanics, where failure stress depends on a material property (toughness) and length of an existing weakness, unlike traditional soil mechanics where failure stress is generally assumed to be only a function of material strength. An existing weakness, or partially developed failure surface, can propagate due to sudden loads, as might be induced by seasonal erosion cycles, small local landslides at a river bank or earthquake shaking. The paper examines the growth of the failure surface over time due to such loads. The analysis shows that a large landslide in sensitive clay can occur after a number of seasonal erosion cycles with no obvious trigger. It also shows that a larger load due to either a small landslide or earthquake can cause a developing failure surface to propagate freely toward general collapse. Large landslides in sensitive clay are often preceded by a smaller landslide at the river bank, but can be triggered by large earthquakes or may occur for no obvious reason. The agreement between predictions and real behaviour suggests that the model for progressive failure of large landslides in sensitive clay deserves further consideration.

Large landslides are common in the gently sloping clay plains of the Saint Lawrence Lowlands of eastern Canada. These tend to occur along rivers carved into the marine soils deposited in the former Champlain Sea, which occupied the area roughly 10 000 years ago. This paper presents a landslide susceptibility model, developed at the regional scale using a bivariate statistical method: the weights of evidence method. The analysis considers the association of existing large landslides in a portion of the study area with key terrain features, such as ground elevation, flow accumulation in adjacent streams, soil type, soil thickness, and land use. The resulting model identifies three different levels of susceptibility: low, low to moderate, and moderate to high. These descriptors are related statistically to the probability of encountering existing large landslides within 500 m, 1 or 2 km, respectively. The model is tested along primary railway corridors and isolates 8% of the total length for further consideration of landslide hazard. Reconnaissance level air photo survey results further reduce the length of corridor with elevated susceptibility to 2% of the total length, thus focusing the application of additional resources to a very small proportion of the total inventory.

No abstract