Forewarning the seasonal dynamics of corn leafhopper and mollicutes through neural networks

Abstract

The corn leafhopper (CL), Dalbulus maidis (DeLong & Wolcott) (Hemiptera: Cicadellidae), has become the most important corn pest in Brazil and other corn-producing countries. This highly efficient insect vector transmits corn stunting pathogens resulting in significant yield losses in corn fields. This study aimed to investigate the relationship between CL abundance and pathogen infection in adult CL with weather variables, day of the year (DOY), and corn season in four Brazilian corn-producing areas using artificial neural networks (ANN). We developed three ANN models, using monitoring data from 2019 to 2023, for year-round forewarning of CL populations and infection of corn stunt spiroplasma (CSS) and maize bushy stunt phytoplasma (MBSP) in CL adults. The best-fit models demonstrated strong correlations in the validation set for CL abundance (0.71), and substantial classification agreement for both CSS (0.81) and MBSP (0.81). The final inputs for the models included relative humidity, air temperature, wind speed, DOY, corn season, and CL abundance. The presence of corn plants and DOY are manageable factors for achieving CL and mollicute control. This can be made by eliminating volunteer plants, reducing planting windows, and avoiding late-plantings. Our results are suitable for further predictions and offer essential guidance to be incorporated into the IPM of D. maidis and to better understand CSS and MBSP infection on a large-scale. Lastly, ANN is a reliable machine-learning algorithm to predict vector population dynamics and the infection of phytopathogens in D. maidis.