Development and Evaluation of a Logistic Risk Model: Predicting Frit Fly Infestation in Oats

Abstract

Forecasting insect pest damage allows for an adjustment of pesticide use to the actual need, and it is important to establish the potential benefits of pest forecasts in terms of profitability and reduction of pesticide use. Here I calibrate and evaluate a risk algorithm for frit fly infestation in spring-sown oats, aiming to improve damage forecasts and assess the profitability of different control strategies with regard to regional infestation prevalence. Oat fields in two regions in Sweden with high and low frequencies of infestation were surveyed for 11 and 13 yr, respectively. The surveyed fields were classified either as infested or noninfested, depending on whether or not the infestation level exceeded the economic threshold. The relation between various risk factors and frit fly damage was examined by logistic regression, and risk points were assigned to significant factors. Regional action thresholds (risk point sums) resulting in the lowest average costs were calculated.In regions where pest outbreaks are common, a control strategy based on the proposed risk algorithm was more profitable than prophylactic or no sprays. The percentage of sprayed fields was reduced to Ͻ50%, compared with routine use of pesticides. In low-risk regions, the profitability of the forecasting strategy and the no-control alternative were similar. The predictive accuracy of the risk algorithm was higher than that of a commonly used degreeday model, especially in regions where the frit fly is a minor pest. Based on these results, I suggest that use of the proposed risk algorithm can enhance the farmer's understanding of the biology of this insect pest, increase profitability, and reduce unnecessary use of pesticides. Long-term monitoring of pest and environmental data can facilitate not only on-farm pest control decisions, but can also provide essential information for developing regional pest control strategies.