A food web approach to evaluating the effect of insecticide spraying on insect pest population dynamics in a Philippine irrigated rice ecosystem.
Data from a 645-taxa Philippines-wide food web and from multiple regression models were used to predict population fluctuations of insect pests in a rice field. Independent variables of pest models included the biomass of rice plants in the field, the abundance of each pest and the abundances of 5 highly correlated enemies of the pest, all as functions of time. To test the ability of the models to reveal effects of insecticide spraying, a rice cv. IR1917-3-17 field in Laguna Province, Philippines, was divided into plots sprayed with deltamethrin and unsprayed plots. Data on the abundance of Nephotettix virescens, Recilia dorsalis, Sogatella furcifera, Nilaparvata lugens, Hydrellia philippina, Nephotettix nigropictus and Cofana spectra and their natural enemies were collected during the dry season of 1990. Spraying insecticide disrupted the population dynamics of the insect species. Multiple regression models were less able to forecast the population fluctuations of pest species on the basis of various numbers and combinations of independent variables in the sprayed plot than in the unsprayed plot; e.g. current pest abundance alone was a significant predictor of future pest abundance for 4 of the 7 pests in the unsprayed plot, but significant fits were found for only 2 pests in the plot sprayed with insecticide. In the unsprayed plot, independent variables were significant predictors of future pest abundance in 4 of 7 initial models compared with 1 of 7 models in the sprayed plot. Step-wise removal of independent variables in the models enhanced their forecasting power in both the sprayed and unsprayed plots, but significant models in the unsprayed plot nearly always outnumbered those in the sprayed plot. In the unsprayed plot, N. virescens retained 5 of 7 independent variables as significant predictors, compared with 4 for S. furcifera, 3 for C. spectra, and 1 for the remaining pests. Classical models that contain 1 or 2 species as independent variables may not be sufficient to forecast future abundances of some Philippine rice pests in unsprayed and sprayed plots. In general, models that included interaction terms and either the presence or absence of sprays among the independent variables did not improve the forecasting power of models in either the sprayed or unsprayed plot. It is suggested that the methods developed here could be applied to other interventions besides spraying (such as the introduction of genetically engineered cultivars), other biotic communities besides arthropods, and other crops besides rice.