Are agronomic models useful for studying farmers' fertilisation practices?

Nitrogen fertilisation is a source of potential groundwater pollution and is a key issue in the current debate about the environmental impacts of agricultural production. It is also a key element in the management of cropping systems by farmers. Therefore, cropping system design entails the understanding and evaluation of farmers' fertilisation practices. Biophysical models describing the soil-plant system can serve this purpose. A comparison between model outputs and farmers' practices was made of a set of 128 apple (Malus domestica Borkh.) plots from 31 members of a farmers' co-operative in south-eastern France. Farmers' fertilisation practices were compared with theoretical practices generated by a series of soil-plant system models of increasing complexity, each model giving the amount of nitrogen that should be applied to the plot according to the knowledge included in the model. The model that reproduced farmers' fertilisation practices most closely was the most complex, taking all plant requirements, soil organic matter and residue mineralisation, denitrification and irrigation supply into account. A Monte Carlo method showed that the differences between farmers' practices and model outputs were not random. Spatial analysis showed a strong spatial organisation of these differences, mainly due to three farms. This congruence between farmers' practices and model outputs suggests the existence of some indicators that depict the N nutrition status of the orchard as a basis for rules indicating how much nitrogen should be applied. The spatial analysis suggests the existence of farmer and neighbourhood effects, which need to be explained. Models appear to be useful tools to study farmers' practices by removing biophysical effects (soil, variety, etc.). This raises new questions concerning agricultural research at the interface between the biophysical and social sciences.