Since the 50 last years, the rapid development of modern agriculture in industrialised countries has considerably affected the quality of water resources, up to the point to jeopardise the capacity of rural territories to produce drinking water. Hence, agronomy has been interested in the complex nitrate biogeochemical interactions for a long time. While agronomists are able to produce very accurate physical models of nitrate propagation at different scales, their tools have a limited relevance if the information regarding geology or agriculture is missing. Consequently, in many cases, it prevents the specialists of being affirmative about the prediction of their current actions on the water quality. By opposition, a system identification methodology is here presented to predict nitrate concentration in water. It has the advantage of being applicable even when very little knowledge is available. It will be shown how external variables such as rainfall and temperature can play an important role in modelling water pollution systems. The efficiency of the approach, both in terms of prediction and physical insight, is discussed on a real life dataset.
展开▼
