Soil erosion by water is one of today’s most important environmental problems, in great part due to changes in agricultural land use and management. This paper illustrates the formulation, calibration, sensitivity and validation analysis of a hybrid model of expert decision trees and artificial neural networks (named ImpelERO) to evaluate the soil erosion process A total of 237 field units were selected, which represent 34 major land resource areas (MLRAs)for five traditional crops in the Andalusian region, southern Spain. The field units observed cover the whole range of erosion events, from what was considered very small to extreme erosion. Seventy-six per cent of the fields suffered small, moderate or large erosion problems. However, only 14% of the fields suffered very small, and 10% of the fields suffered very large or extreme erosion problems. Because of the complexity of the soil erosion process, and the interrelationships of the parameters, ImpelERO was developed as an Universal Soil Loss Equation (USLE) type model following traditional land evaluation analysis and advanced empirical modelling techniques. Using expert-decision trees, soil survey information and expert knowledge of the soil erosion process were combined through land and management qualities. An artificial neural network approach was then applied to capture the interactions between the land and management qualities and one output: vulnerability index (Vi) to soil erosion. The neural network was trained using the Correlation-cascade algorithm. The trained network estimated the output with a high degree of accuracy (maximum deviation 14%), and also had a good generalisation capacity. By means of correlation analysis, observed erosion vulnerability data were compared with predicted data using a previously developed model and using the ImpelERO model. The latter model gave more accurate results (r=0.91) than the previous approach (r=0.66). Along with the prediction of soil loss by water erosion, ImpelERO could be used as a optimisation tool for selecting the land use and management practices which satisfy the optimum environmental protection including reduction of soil erosion.

EXPLOR_STEP=$WICRI_ROOT/Wicri/Agronomie/explor/SisAgriV1/Data/Main/Exploration

HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 001391 | SxmlIndent | more

HfdSelect -h $EXPLOR_AREA/Data/Main/Exploration/biblio.hfd -nk 001391 | SxmlIndent | more

{{Explor lien
   |wiki=    Wicri/Agronomie
   |area=    SisAgriV1
   |flux=    Main
   |étape=   Exploration
   |type=    RBID
   |clé=     ISTEX:8D61DF7A6EFD7A3B8ADC01DB749742B3B638C369
   |texte=   An expert system/neural network model (ImpelERO) for evaluating agricultural soil erosion in Andalucia region, southern Spain
}}