Simulation of the effects of land cover, soil degradation, and rainfall on runoff from a small tropical catchment using a minimally calibrated distributed model

• A barely calibrated model was used successfully to simulate runoff processes for a small catchment. • Degradation increased stormflow by ≥25% and decreased deep percolation by ≥15% • Eucalypt stands reduced deep percolation, masking the other effects of land degradation on stormflow. • Stormflow and deep percolation depend more on annual rainfall than land cover. Catchments undergoing shifting agriculture typically have a mosaic of land covers, with active agricultural fields, fallows in different stages of regrowth, remnant forest, and degraded grasslands. Although the rainfall-runoff responses differ for each of these land covers, there is little quantitative information on how this affects the catchment-scale runoff response. The Marolaona catchment (31.7 ha) in eastern Madagascar has been subject to shifting cultivation for at least 70 years. We used the spatially-distributed physically-based RoGeR_Dyn-model to investigate the sensitivity of catchment-scale runoff to land degradation and annual rainfall. Only the parameter related to the permeability of the subsoil was optimized within the range of observed values; all other soil parameters were based on field measurements or set to default values. The minimally calibrated model reproduced the observed hydrological responses and was used to compare the simulated water budgets and runoff responses for different land covers, namely the mixed land cover in 2015/16, complete tree fallow cover without soil degradation (1960 s baseline), and a hypothetical case with tree fallows and degraded soils. The simulated stormflow is 20–30% higher and deep percolation is ∼22% less for the current land cover than the scenario with a uniform fallow cover without soil degradation (1960 s baseline). The high water use of eucalypts dominated evapotranspiration and masked the effects of soil degradation on stormflow. Furthermore, the effects of land cover and soil degradation on stormflow and deep percolation are highly dependent on, and smaller than, the effects of interannual variation in rainfall, making it difficult to detect these effects from streamflow data alone.