Evaluation of the soil-crop model STICS over 8 years against the “on farm” database of Bruyeres catchment

The European Water Framework Directive (2000) offers a new challenge for farmers and water policy makers. It requires the establishment of quantitative environmental diagnosis of water quality. This can be done using crop models properly tested in the short and long-terms and taking into account current farming practices. The aim of this paper is to check the robustness of a crop model (STICS) for predicting the nitrogen uptake and nitrate leaching in various fields during 8 successive years. The model was evaluated on the soil–crop database of a small catchment in northern France. It includes data of crop production and N uptake, water and mineral nitrogen contents in soil measured three times a year at 36 sampling sites representative of crops (wheat, sugar beet, pea, barley, oilseed rape) and soil parent materials (loam, loamy clay and rocks, sandy loam and limestone, sand). A few crop parameters of STICS were recalibrated on independent databases in order to improve the predictions obtained with the standard parameterization. STICS was then evaluated either by resetting simulations each year (RS) or during continuous simulations (CS) over the 8 years. A reasonable agreement was obtained between observed and simulated values, except for soil N mineral content at harvest and N content in crop residues. The model efficiencies using CS mode were 0.53, 0.94 and 0.38 for N uptake, soil water and mineral N in late autumn, respectively. The mean calculated drainage was 192 mm y−1 and N leaching was estimated at 20 kg N ha−1 y−1, respectively. Averaging the outputs according to soil and crop type improved the quality of fit. The outputs of CS simulations were close to those obtained with RS. The mean nitrate concentration in water drainage was estimated as 46 and 45 mg NO3 l−1 with RS and CS, respectively. Continuous simulations did not induce a substantial drift in mineral N in late autumn and can be trusted for predicting nitrate leaching. However, the leaching predictions were more sensitive to spatially variable parameters such as maximal rooting depth and potential mineralisation rate for CS than for RS. This emphasizes the difficulty in extrapolating the model over the long-term for large spatial areas. Another important uncertainty concerns fertiliser use efficiency, which had a small effect on leaching but a marked influence on gaseous N losses. Further assessments of the model will concern the whole N balance prediction over the long-term