Prediction of soil detachment in agricultural loess catchments: Model development and parameterisation

The objective of this paper is to derive and parameterise a detachment approach that balances simplicity and necessary process complexity to be implemented in a process-based erosion model applicable at the catchment scale. We proposed a semi-empirical model approach that relates the potential detachment rate by means of a bi-linear combination to the attacking forces of rainfall (characterised as momentum flux of rainfall) and overland flow (characterised as shear stress). The resisting forces against detachment are characterised by two empirical parameters: i) erosion resistance fcrit, which is a threshold of the attacking forces and ii) erodibility parameter p1, which represents the linear increase of detachment once the erosion resistance is overcome. The parameter P2 weights the momentum flux of rainfall against shear stress. The empirical parameters were determined for conventionally tilled loess soils using data sets of rainfall simulation experiments performed in the laboratory and in the field. At first, a data set of laboratory experiments carried out under varying conditions of rainfall and overland flow (published by Schmidt, 1996) but using resembling loess soil samples was used to determine the parameters p1 and P2. Secondly, data from 58 rainfall simulation experiments performed on 24 m2 plots in the Weiherbach catchment (Southwest Germany) were used to parameterise the erosion resistance fcrit under field conditions. We found evidence that cultivation was the first order control of the erosion resistance on conventionally tilled loess soils: crops that are cultivated in rows were strongly susceptible to detachment because runoff is channelled along the intermediate areas of plant rows. For bare soils we found a strong correlation between the erosion resistance, surface roughness and clay content. These results can be used to predict the erosion resistance and regionalise it to the catchment scale as input parameter in a soil erosion model.