Comparison of CERES, WOFOST and SWAP models in simulating soil water content during growing season under different soil conditions

A lysimeter experiment conducted on three soil types in a main agricultural production region of Austria in Marchfeld (latitude 48degrees 12'N, longitude 16degrees34'E and altitude 150 m above sea level), was used to test the performance of the three widely used crop models, CERES, SWAP and WOFOST. The soils included chernozem, sandy chernozem and fluvisol with a 2.0 m profile depth. Daily measurements of the soil water content were taken using TDR probes (one per 0.3 m of depth) in six replicates for each soil type. The analysis was carried out for winter wheat and spring barley grown on the site during seasons 2000 and 2001 and included a detailed comparison of the simulated and measured soil water contents as well as an analysis of seasonal soil water balances, root front velocities and an evaluation of the modeled crop yields. CERES and SWAP, in contrast to WOFOST, simulated the grain yield of barley and wheat well. All three models simulated soil water content in the profile with similar results. The root mean square error (RMSE) range of soil water content was 0.71-4.67% for barley and 2.32-6.77% for wheat, depending on the model and soil type. None of the models simulated total soil water content in the profile significantly better, but there was a general tendency for the models to overestimate soil water depletion. Both CERES and SWAP mimicked the soil water content dynamics well in the top 0.3 m of the soil. The study shows that the multiple layer approach models (SWAP or CERES) including more sophisticated estimation methods for root growth and soil water extraction should be preferred in comparable environments. Further adjustments of evapotranspiration subroutines to the local conditions should be considered prior to the model use for drought impact assessment, yield forecasting or climate change impact studies. (C) 2003 Elsevier B.V. All rights reserved.