Test of the SHETRAN technology for modelling the impact of reforestation on badlands runoff and sediment yield at Draix, France

Physically based distributed models, such as SHETRAN, have the ability to predict the impacts of land management changes in advance of any change taking place. It needs to be shown, though, that they can deliver practical results while accounting for uncertainty in parameter evaluation. As a demonstration, SHETRAN was used to simulate the impact of reforestation on runoff and erosion in a badlands environment in south-east France. A model was first constructed for the 86-ha Laval basin, severely affected by gully erosion. It was then altered to represent this basin as if it were equivalent to the neighbouring 108-ha Brusquet basin, rescued from erosion by reforestation. Simulation uncertainty was quantified as a function of uncertainty in parameter evaluation. A good reproduction was obtained of the observed difference in runoff/rainfall ratio and the two orders of magnitude difference in sediment yield between the basins, as a function of their different vegetation covers and rainfalls, for a 5-year period. The results are encouraging for modelling land use change impacts but indicate also the need to reduce further the uncertainty in model parameter evaluation.