Using niche models to understand the occurrence of Helicoverpa armigera outbreaks in South America

The invasive H. armigera triggered a major phytosanitary crisis in Brazil during the 2012/2013 harvest, especially in soybeans. Species distribution modeling (SDM) can help to better understand the recent outbreaks of this pest in South America. A total of 269 unique occurrences for H. armigera were collected. Seven principal components, obtained through PCA analyses of 19 raw environmental variables from WorldClim were used to obtain the potential distribution of this species. The grid resolution used was 2.5 arc-min (0.041 ≈ 4 km). Both distance-based (Euclidean distance-EUC; Envelope score-ENV; Mahalanobis distance-MHL) and artificial intelligence methods (GARP with best subsets; Support Vector Machines-SVM; Maximum Entropy-MAX) were used. The True Skilled Statistics was used to assess the models performance, and a mean consensus served to show all areas which were predicted as more and less prone to be infested. All models presented TSS values higher than 0.5, except ENV. The three best algorithms were MHL, MAX, and SVM. EUC, ENV, and MHL showed wider potential distributions, particularly in Bolivia and Paraguay. Considering the threshold applied, the most suitable areas for the occurrence of H.armigera outbreaks were found in the Central-West and Southeast of Brazil. These coincided with the Brazilian States for which phytosanitary emergency status has been declared recently. These results corroborate the high dispersive ability and skill to survive in adverse conditions of this pest.