Annual rainfall erosivity in the universal soil loss equation is the number of rainfall erosion index units (EI30) for a particular location usually recognized as a tool for describing soil erosion by water. It is also a basic input to simple and widespread soil erosion prediction models like USLE and RUSLE. However, its calculation on the basis of original rainfall records is a very laborious operation and is completely impossible for many locations without a precise and detailed rainfall data. The aim of this study was to develop a new simple method of estimation of annual rainfall erosivity on the basis of rainfall data. Efforts were made to see the possibility of applying artificial neural networks (ANN) for estimation of annual rainfall erosivity, on the basis of the available annual rainfall values. The study was done with the use of annual rainfall erosivity values of 38 years data of Chandigarh. As a result of the study radial basis function network with network architecture as 1-6-1 was found best in prediction of annual rainfall erosivity for Chandigarh with input as annual rainfall data. Model efficiency of observed versus predicted rainfall erosivity was found to be higher with ANN than with that obtained for the simple regression models. The study results suggested the possibility of application of neural networks for estimation of rainfall erosivity on the basis of annual rainfall totals instead of simple statistical relationships.
Artificial neural networks, annual rainfall, rainfall erosivity
