Evaluation of models to predict take-all incidence in winter wheat as a function of cropping practices,soil, and climate

The incidence and severity of take-all, caused by Gaeumannomyces graminis var. tritici (Ggt), in susceptible crops depend on climate, soil characteristics and cropping practices. Take-all can be controlled by modifying crop rotation, crop management and fungicide treatment. When available, fungicides are used as a seed treatment and are partially effective. There is currently no reliable method for helping farmers to optimise their choice of cropping system to improve take-all control. In this study, we defined 16 models, based on various mathematical functions and input variables, for predicting disease incidence in a wheat crop as a function of soil characteristics, climate, crop rotation and crop management. The parameters of these models were estimated from field experiments carried out at six sites in the north of France over a ten-year period. The root mean squared error of prediction (RMSEP) values of the models were estimated by cross validation and compared. RMSEP was in the range 16.34–65.93% and was higher for the models based on multiplicative functions. The lowest RMSEP value was obtained for a dynamic model simulating disease incidence during the crop cycle and which included among input variables the percentage of diseased plants determined at GS30.