Article révisé par les pairs
Résumé : The analysis and explanation of the spatial distribution of organisms within a locality are problematic. This study uses a combination of standard analytical methods (Generalized Linear Interactive Modeling [GLIM]) with recently developed spatial statistical techniques (geostatistical analysis), on a model system of insect-plant interactions in an urban setting. Infestations of the horse chestnut scale, Pulvinaria regalis, were mapped on three tree species in Oxford, United Kingdom. Various tree parameters were measured, as were aspects of the site in which each tree was growing. Using general linear modeling and geostatistics, the distribution and intensity of scale populations were investigated in relation to these parameters. The trees were separated into those that showed no symptoms of lack of vigor and those that were clearly unhealthy. In both cases, the only parameter that explained much of the variance in the scale egg densities on trees was the impermeability of the substrate surface under the trees, such that as substrates became more impermeable to water and nutrients (for example, as a result of concrete or roadways), the higher were the pest densities on those trees. For the vigorous trees alone, an extra parameter, that of building distance, was also found to be significant, so that trees very close to buildings also showed high pest densities. The spatial dependence of scale insect eggs on trees was found to be anisotropic through the sample area, in the southwest/northeast direction, attributable to wind speed, direction, and canyon effects.