Neural network modelling of NO3- time series from small headwater catchments

A variety of different processes is known that determine water and solute f luxes in headwater catchments. Water resources management of these systems, however, relies in most cases on empirical experience with respect to its overall response. A promising method to bridge the gap between comprehensiv e scientific investigations and the need to manage the systems on the basis of limited data sets seems to be the application of artificial neural netw orks (ANN). Here, time series of NO, concentrations in the runoff of two fo rested headwater catchments in south Germany are investigated. Furthermore, the application of nonlinear methods presented here reveals a rather intri cate behaviour also on the temporal scale, and considerable differences bet ween the two catchments. This demonstrates the validity of ANN as universal descriptive tools.