Multidimensional Host Distribution and Nonrandom Parasitism: A Case Study and a Stochastic Model

Abstract

In order to examine the spatial distribution of parasitism I recorded the tridimensional position of each leaf harboring tentiform mines of the leafminer Phyllonorycter blancardella Fabr. (Lepidoptera, Gracillariidae) in different apple trees. The great majority of mined leaves had only one mine. I also recorded which of the hosts had been parasitized and determined each parasitoid species. Using the nearest neighbor (NN) distance between hosts, I tested the hypothesis of spatial randomness of parasitism with Monte Carlo tests for nine parasitoid species. This procedure of characterizing and classifying spatial distributions is borrowed from the statistics on spatial point processes, and it relies on extensive simulations. The distribution of parasitism differs between parasitoid species and between trees, even for the same species. Aggregation of parasitism at low NN distances is by far the most common nonrandom spatial distribution of parasitism. The aggregation of parasitized hosts by the most common parasitoid species, the hymenopteran Cirrospilus vittatus, diminishes with the number of hosts parasitized. I also developed a stochastic model for the spatial distribution of C. vittatus. The model is based (i) on results obtained in the first phase of the work, (ii) on independent information on the oviposition rate, and (iii) on theoretical considerations. They key idea is that a single female produces an aggregated pattern of parasitism (i.e., cluster) during her visit to an apple tree. The combination of spatially independent clusters produces the final distribution of parasitism. The model is evaluated with Monte Carlo tests on data other than those used for its development. Finally, I discuss the statistical assessment, the biological realism, and the generality of the model.