Multimodel Inference for the Prediction of Disease Symptoms and Yield Loss of Potato in a Two-Year Crop Rotation Experiment

Abstract

The second order Akaike information criterion was used for the assessment of 139 regression models for three responses of potato test crops: (a) incidence of Spongospora subterranea on the harvested tubers, (b) percentage of haulms infected with Verticillium dahliae, and (c) tuber yield. Six variables that are likely related to the response variables were taken into consideration: soil infestations of the fungus Verticillium dahliae and of three nematode species (Globodera pallida, Trichodoridae, and Meloidogyne spp.) and, furthermore, soil pH and water soluble phosphor (P). Interactions between V. dahliae and the three nematode species were included as well. Based on multimodelling, predictors are given a weight from which one may decide about the need to include them in a prediction of crop yield. The most important predictors were soil infestation levels of V. dahliae and G. pallida and soil pH. The outcome also showed that tubers suffered more from S. subterranea for higher soil pH values. Finally, yield reduction from the presence of V. dahliae was enhanced by the presence of higher densities of G. pallida.