Spatial Distribution and Environmental Variables Associated with Control Failures of Phthorimaea absoluta by Insecticides Determined by Machine Learning Algorithm

For pest control to be sustainable, the methods applied must be efficient and have a low environmental impact. Pest control failures bring economic and environmental problems. Phthorimaea absoluta is the main pest in tomato crops worldwide. Benzoylureas, diamides, and pyrethroids are among the insecticides with the highest reports of pest control failures, and Brazil is the country where this has been most observed. Machine learning models are suitable for predicting biological events. Thus, this study aimed to determine the risks of failures in the control of P. absoluta by insecticides in Brazilian biomes using the MaxEnt machine learning algorithm. The risks of pest control failures by benzoylureas and pyrethroids were higher in tomato crops located in the Cerrado and Atlantic Forest biomes, and annual precipitation was the critical variable associated with these failures. The risks of control failures by diamides were higher in crops located in the Caatinga, Cerrado, and Atlantic Forest, and temperature seasonality was the critical variable associated with these failures. In conclusion, the models determined in the study are robust to predict the regions with higher risks of P. absoluta control failures by insecticides, and they indicated the environmental variables associated with these risks.