THE USE OF NEURAL NETWORKS IN AGROECOLOGICAL MODELING

Starting from illuminating the role of variability and uncertainty in agroecological observations and measurements the paper discusses the p ossibilities of applying neural networks or neural networks in combina tion with fuzzy techniques in the held of agroecological modelling. Be cause of the lack of a consistent theoretical background on the one si de, but the availability of plenty of observations and subjective empi rical knowledge on the other side, the investigation of many scientifi c and the management of many practical questions is very data-driven i n agroecology. Therefore neural networks and other data driven modelli ng techniques seem to be adequate modelling tools. Two quite different applications form the main part of the paper: Neural networks for mod elling development and matter processes in agroecosystems and a combin ed neural network-fuzzy approach for modelling habitats of plants and animals (the EMU-NF model family). (C) 1997 Elsevier Science B.V.