Comparison and validation of population models for cereal aphids

Models used to simulate changes in populations of cereal aphids need to be compared and validated if they are ever to be improved sufficiently for wide-scale use in crop protection. With this in mind, simulation runs using the models SIMLAUS, LAUS, and GETLAUS01 were tested over several years using data collected in different regions of Germany. The numbers of cereal aphids and aphid antagonists were recorded in fields of winter wheat and winter barley, by the adapting number of samples to aphid density. Simulation runs with the model SIMLAUS predicted accurately (96% of case studies, e.g. location and years) the type of hibernation of Sitobion avenae and Rhopalosiphum padi in fields of both winter wheat and winter barley. In eight of 52 case studies, from different locations and different years, the model predicted anholocyclic hibernation, though overwintering aphids were not found in the field. The model SIMLAUS was not suitable for predicting the number of aphids found in cereal crops in the autumn. In 12 of 35 case studies, the model LAUS predicted accurately (R² > 0.44,P < 0.05), from à priori data, the populations of S. avenae in cereal fields during spring and early summer. The intercepts (a) and the slopes (b) of the regressions between the observed and predicted cereal aphid populations differed (P = 0.05) from zero and one in 3% and 83% of case studies, respectively. After adjusting the starting values, by using data (à posteriori) collected in the field, the simulations were improved greatly in 90% of the case studies. In total (82%) of the simulation runs done with the model GETLAUS01 (à posteriori), produced close relationships (R² > 0.44,P < 0.05) between the observed and predicted populations of cereal aphids found during the summer in crops of winter wheat. Although the slopes of regressions differed from one in 12 of the case studies, all of the intercepts passed through the origin (a = 0). Correction of systematic errors, standardization of case studies, and possible model extensions are discussed with a view to using the revised models as an integral part of pest management.