ALBERT

Description: This paper presents an approach to model runoff and erosion risk in a context of data scarcity, whereas the majority of available models require large quantities of physical data that are frequently not accessible. To overcome this problem, our approach uses different sources of data, particularly on agricultural practices (tillage and land cover) and farmers' perceptions of runoff and erosion. The model was developed on a small (5 ha) cultivated watershed characterized by extreme conditions (slopes of up to 55 %, extreme rainfall events) on the Merapi volcano in Indonesia. Runoff was modelled using two versions of STREAM. First, a lumped version was used to determine the global parameters of the watershed. Second, a distributed version used three parameters for the production of runoff (slope, land cover and roughness), a precise DEM, and the position of waterways for runoff distribution. This information was derived from field observations and interviews with farmers. Both surface runoff models accurately reproduced runoff at the outlet. However, the distributed model (Nash–Sutcliffe = 0.94) was more accurate than the adjusted lumped model (N–S = 0.85), especially for the smallest and biggest runoff events, and produced accurate spatial distribution of runoff production and concentration. Different types of erosion processes (landslides, linear inter-ridge erosion, linear erosion in main waterways) were modelled as a combination of a hazard map (the spatial distribution of runoff/infiltration volume provided by the distributed model), and a susceptibility map combining slope, land cover and tillage, derived from in situ observations and interviews with farmers. Each erosion risk map gives a spatial representation of the different erosion processes including risk intensities and frequencies that were validated by the farmers and by in situ observations. Maps of erosion risk confirmed the impact of the concentration of runoff, the high susceptibility of long steep slopes, and revealed the critical role of tillage direction. Calibrating and validating models using in situ measurements, observations and farmers' perceptions made it possible to represent runoff and erosion risk despite the initial scarcity of hydrological data. Even if the models mainly provided orders of magnitude and qualitative information, they significantly improved our understanding of the watershed dynamics. In addition, the information produced by such models is easy for farmers to use to manage runoff and erosion by using appropriate agricultural practices.