Sensitivity of the Model LISEM to Variables Related to Agriculture

Abstract

A sensitivity analysis of the model LISEM was performed for those variables that are directly related to arable farming in Northern France. LISEM is a physically based spatial erosion model integrated into a raster GIS. The variables are: saturated hydraulic conductivity, random roughness, Manning's n, cohesion and aggregate stability. The sensitivity analysis was done on an artificial dataset, first by a direct comparison of input and output, seCond by calculating a normalised sensitivity factor, and third by coupling the responses of the main algorithms in LISEM to analyse the chain of events. It appeared that changing one variable and comparing it with a baseline value is not sufficient to understattd the model response. Combinations of different input values gave the same output. LISEM is sensitive to changes in Ksat in all processes, but the sensitivity varies with different levels of surface roughness. Discharge and net erosion are very sensitive to random roughness, but only when Ksat is relatively high and there is not much water for overland flow. Flow defachment is moderately sensitive to Manning's n (through velocity and transport capacity) an