Modeling spatiotemporal distribution of yellow rust wheat pathogen using machine learning algorithms: Insights from environmental assessment

The yellow rust pathogen (Puccinia striiformis Westend) poses a significant threat to wheat production in the world, necessitating a comprehensive understanding of its spatiotemporal distribution and the influence of climatic factors. In this study, we employed an ensemble of four prominent machine learning algorithms to assess the impact of various environmental and remote sensing variables on the spread of yellow rust at a national scale. Our analysis incorporated 55 climatic parameters, including monthly temperature, precipitation, solar radiation, and wind speed. The results demonstrated that the RF algorithm yielded robust predictions, with a Receiver Operator Characteristic (ROC) of 0.916 and True Skill Statistic (TSS) of 0.748. Furthermore, the study identified key influencing variables for wheat disease modeling, such as annual precipitation, temperature seasonality, and isothermality. Projections based on the model indicate a potential decrease in disease spread by 2050 in specific regions. The findings underscore the efficacy of ensemble modeling in predicting the spatiotemporal distribution of yellow rust on a large scale, offering valuable insights for the development of robust agricultural management strategies in the face of evolving climate conditions. [Display omitted] •Four ML algorithms were evaluated to predict yellow rust.•Machine learning algorithms identified the most important variables affecting disease epidemics.•Yellow rust on wheat in Iran is increasing and climate change impact needs understanding.•The ensemble model efficiently predicts yellow rust spatiotemporal distribution at a large scale.