Combination of statistical and conceptual approaches for debris-flow susceptibility modelling at a regional scale, British Columbia, Canada

• Debris-flow susceptibility maps were produced for Canadian Rocky Mountains area. • Channelized and hillslope debris-flows should be studied separately. • Combination of statistical and conceptual approaches can provide satisfactory results. • Integration of the models yields effective classification of runout susceptibility. This paper describes the data and methodological approaches used to assess the initiation and runout susceptibility of debris-flows in the Valemount area, east-central British Columbia, Canada. Debris-flows are frequent in this area and have impacted the roads and dwellings. The study area covers about 1200 km 2 . A landslide inventory for this area delineates past debris-flows, including their source areas and deposits. This inventory includes hillslope and channelized debris-flows, enabling the development of separate models for each type of event. For hillslope debris-flows, a supervised multivariate regression technique was used to identify possible initiation zones. Subsequently, a conceptual model was trained and applied to simulate runout and classify areas according to runout susceptibility. Modeled hillslope debris-flow deposits reaching the main valley channels were considered as a proxy for potential source areas for channelized debris-flows, even though source sediments may also result from other processes, including gradual erosion or mass movements from adjacent slopes. Conceptual modelling was then applied to this second type as well. The results of the two models were combined to classify the area according to its predisposition to debris-flow runout. Debris-flow datasets other than those used to train the models, were used to optimize and validate the models. Results indicate that, considering both debris-flow types, there is a 75 % of agreement between the modeled susceptible areas and the validation debris-flow fans. This suggests that the models can effectively distinguish between potential debris-flow fan areas and non-debris-flow areas.