Using lower trophic level factors to predict outcomes in classical biological control of insect pests

Importation of exotic natural enemies for biological control of insect pests entails risks to the environment. Pre-release estimates of the likelihood of achieving successful control would be helpful in avoiding ineffective importations. Based on strong evidence of multi-trophic level interactions in terrestrial ecosystems, we tested whether variation in ecological and biological factors found at the plant and herbivore trophic levels (levels one and two) could be used to create a simple, empirically based formula, capable of estimating the probability of successful biological control against holometabolous insect pests. We constructed a database consisting of 828 records of biological control attempts against 91 pest insect species and used stepwise logistic regression to test whether five basic features of the ecosystem, crop, and pest (habitat type, crop use, pest order, pest feeding niche, and damage severity) were correlated with rates of successful control. Natural enemy characteristics were not included in the model. The final model included 10 significant independent variables, nine of which were two-way interactions; all five basic ecosystem features appeared in significant interactions. The model provided good estimates of historical success rates against pest species in the data set. In a further test, the model was able to correctly rank amenability to biological control for 10 pest species not included in the original data set. These results provide evidence that lower trophic level factors can be useful in the search for a predictive formula for biological control.