Projection pursuit nonparametric regression applied to field counts of the aphid Macrosiphum euphorbiae (Homoptera: Aphididae) on tomato crops in greenhouses.

Abstract

The number of plants that sampling requires in aphid population studies often exceeds one hundred. Thus, only quick and nondestructive methods can be used to sample this pest in a reasonable time interval. We propose a visual method for estimating the density of the aphid Macrosiphum euphorbiae (Thomas) on tomato plants reared in greenhouses. After approximately 1min of visual observation plants can be assigned to abundance classes, the boundaries of which are roughly the powers of square root 10. Precise counts were collected simultaneously on sets of reference plants from the same greenhouse. Projection pursuit nonparametric regression was then used to provide unbiased estimates of aphid densities from the abundance classes and several easily gathered explanatory variables. The robustness of the method was evaluated by testing the models on the complementary data sets from plants in which the aphid densities were precisely counted. In both single and twin-row cultural conditions, for the reference and complementary data sets, the order of magnitude of the error was less than one class rank per plant. The investigation time was reduced by approximately 10-fold compared with the exact counting method. This easy-to-teach field method could be useful in large-scale population surveys and for optimizing integrated pest management strategies.