Using landscape indicators to predict high pest infestations and successful natural pest control at the regional scale

Designing multifunctional landscapes requires accurate indicators to assess the effect of landscape structure on the provision on ecosystem services. Biological pest control relying on natural enemies is an important ecosystem service considered as a sustainable alternative to chemical control. The aim of this study was to measure and compare the accuracy of landscape indicators computed at various spatial scales to predict pollen beetle infestations and successful biological control in northwestern France. The sensitivity, specificity, and probability of correctly ranking fields were estimated for each indicator based on a survey of 42 fields using the receiver operating characteristic procedure. For pest infestation, the proportion of woodland and the proportion of semi-natural habitats were found to be informative indicators with good discriminatory abilities. For biological control, the proportion of woodland, the proportion of semi-natural habitats and the proportion of the previous year's oilseed rape fields with reduced soil tillage were found to be informative indicators with good discriminatory abilities. By using indicator values and optimal thresholds we were able to compute maps of areas at risk for pest infestation and those displaying successful biological control at the regional scale. This study provides tools that could help extension services, landscape planners, and policy makers in optimizing landscape structure according to the provision of a key ecosystem service. The results of this study also provide new grounds for understanding trophic interactions at the regional scale as well as the ambivalent effect of landscape complexity on pest and natural enemy populations. (C) 2011 Elsevier B.V. All rights reserved.