EVALUATION OF CECOL, A MODEL OF WINTER RAPE (BRASSICA-NAPUS L.)

The CECOL model is a dynamic model that describes the growth and devel opment of winter rape. Here the model is evaluated using a number of d ifferent approaches, including sensitivity analysis, investigation of behavior in extreme conditions, graphical analysis and estimation of t he mean-squared error of prediction (MSEP) and its components. The MSE P value gives an overall value of model predictive accuracy. For CECOL , the value is 0.22 (t ha(-1))(2). This is somewhat smaller than MSEP of the very simple model that uses the average of past yields to predi ct future yield. The other evaluation techniques are complementary to MSEP in that they help identify specific types of situation where the model is a poor predictor. Here it is found that the CECOL model predi cts flowering dates that are too early for early sowing and too late f or late sowing, does not correctly describe the differences between ir rigated and nonirrigated treatments, and does not give reasonable valu es for very low planting densities. In practical terms, these last two errors are not very important. (C) 1998 Elsevier Science B.V.