The SHETRAN model for simulating the sediment yield arising from shallow landslides at the scale of a river catchment was applied to the 45-km 2 Ijuez catchment in the central Spanish Pyrenees, to investigate the effect of loss of forest cover on landslide and debris flow incidence and on catchment sediment yield. The application demonstrated how such a model, with a large number of parameters to be evaluated, can be used even when directly measured data are not available: rainfall and discharge time series were generated by reference to other local records and data providing the basis for a soil map were obtained by a short field campaign. Uncertainty bounds for the outputs were determined as a function of the uncertainty in the values of key model parameters. For a four-year period and for the existing forested state of the catchment, a good ability to simulate the observed long term spatial distribution of debris flows (represented by a 45-year inventory) and to determine catchment sediment yield within the range of regional observations was demonstrated. The lower uncertainty bound on simulated landslide occurrence approximated the observed annual rate of landsliding and suggests that landslides provide a relatively minor proportion of the total sediment yield, at least in drier years. A scenario simulation in which the forest cover was replaced by grassland indicated an increase in landsliding but a decrease in the number of landslides which evolve into debris flows and, at least for drier years, a reduction in sediment delivery to the channel network.

Abstract. The SHETRAN model for determining the sediment yield arising from shallow landsliding at the scale of a river catchment was applied to the 180-km 2 Valsassina basin in the Italian Southern Alps, with the aim of demonstrating that the model can simulate long term patterns of landsliding and the associated sediment yields and that it can be used to explore the sensitivity of the landslide sediment supply system to changes in catchment characteristics. The model was found to reproduce the observed spatial distribution of landslides from a 50-year record very well but probably with an overestimate of the annual rate of landsliding. Simulated sediment yields were within the range observed in a wider region of northern Italy. However, the results suggest that the supply of shallow landslide material to the channel network contributes relatively little to the overall long term sediment yield compared with other sources. The model was applied for scenarios of possible future climate (drier and warmer) and land use (fully forested hillslopes). For both scenarios, there is a modest reduction in shallow landslide occurrence and the overall sediment yield. This suggests that any current schemes for mitigating sediment yield impact in Valsassina remain valid. The application highlights the need for further research in eliminating the large number of unconditionally unsafe landslide sites typically predicted by the model and in avoiding large overestimates of landslide occurrence.

Abstract:This paper outlines a physically based model developed for prediction of both arid region¯ash¯oods and associated transmission loss. It is based on coupling numerical solutions of the St Venant equations and the kinematic wave equation for channel and overland¯ow routing, respectively, with the Richards' equation for the determination of in®ltration. These hydrological components are integrated in one comprehensive model for the above-mentioned purposes. The robustness of the model and its components have been tested and it has then been applied to a medium-size catchment in an arid region. The results indicate that reasonable simulations can be achieved with a minimum of calibration, but point to problems of representing input rainfall conditions with the necessary time and space resolutions. #

No abstract