Predicting the seasonal dynamics of Dalbulus maidis (Hemiptera: Cicadellidae) in corn using artificial neural networks.

Abstract

This study addresses the challenge of predicting Dalbulus maidis (DeLong & Wolcott) (Hemiptera: Cicadellidae) density in cornfields by developing an artificial neural network (ANN). Over two years, we collected data on meteorological variables (atmospheric pressure, air temperature, dew point, rainfall, relative humidity, solar irradiance, and wind speed), plant age, and density of D. maidis in cornfields located in two Brazilian biomes (Atlantic Forest and Brazilian Tropical Savannah). Out of 1056 ANNs tested, the neural network featuring a 30-day time lag, six neurons, logistic activation, and resilient propagation demonstrated the lowest root mean squared error (0.057) and a high correlation (0.919) with observed D. maidis densities. This ANN exhibited an goodness of fit in low-density (Atlantic Forest) and high-density (Brazilian Tropical Savannah) scenarios for D. maidis. Critical factors influencing D. maidis seasonal dynamics, including corn plant age, rainfall, average air temperature, and relative humidity, were identified. This study highlights the potential of the ANN as a promising tool for precise predictions of pest seasonal dynamics, positioning it as a valuable asset for integrated pest management programs targeting D. maidis.