Predicting distribution pattern of the Mediterranean pine engraver, Orthotomicus erosus (Coleoptera: Curculionidae: Scolytinae), by geostatistics and artificial neural network

The Mediterranean pine engraver, Orthotomicus erosus (Wollaston) (Coleoptera: Curculionidae: Scolytinae), is one of the most important pests of pine trees in Kermanshah. Attack and feeding of this pest destroy the floem tissues under the bark of the infected trees and disrupt the plant sap flow, causing the death of infected trees. The purpose of this study was to predict and mapping the distribution of O. erosus using multi-layer perceptron neural networks combined with genetic and imperialist competitive algorithms in Kermanshah. The sampling of pine trees was done in 2015-2016 in Kermanshah. To evaluate the ability of the used neural networks to predict the distribution was used statistical comparison the parameters such as mean, variance and statistical distribution between actual and predicted values by multi-layer perceptron neural networks combined with genetic and imperialist competitive algorithms. Results showed that in training and test phases, was no significant differences between average, variance and statistical distribution of actual and predicted data that indicates the high accuracy and the ability of neural networks to map the distribution of this pest in Kermanshah. The R2 values revealed that imperialist competitive algorithm had a higher accuracy to estimate the density of O. erosus compared with the other two methods. In addition, the comparison of the coefficients of the R2 between different neural networks and geostatistics method showed that all three neural network models predicted the distribution pattern of O. erosus better than the geostatistics method. The maps drawn by all three neural networks showed that the distribution of this pest was cumulative. The results obtained from the geostatistics method represented the cumulative distribution of the pest.