Insect Population Forecast: A Shiny app for predicting pest population dynamics

Background : Invasive and polyphagous insect pests pose an increasing threat to Moroccan agriculture, highlighting the need for forecasting-based decision-support tools to strengthen sustainable Integrated Pest Management (IPM). This study focuses on two economically important pests, the spotted wing drosophila ( Drosophila suzukii ) and the green leafhopper ( Jacobiasca lybica ), which exhibit contrasting population dynamics and management challenges. Methods : We developed and operationalized pest population forecasting approaches combining degree-day based phenological modeling and Random Forest machine-learning methods using multi-year field monitoring data and meteorological variables. Model performance was evaluated using species-specific temporal validation strategies, and the outputs were implemented within an interactive Shiny web application designed to support advisory risk windows. Results : The Random Forest model achieved higher short-term predictive accuracy under temporal hold-out validation, while the phenological approach provided a highly robust and coherent representation of seasonal population timing across years. To facilitate practical use, the forecasting outputs were integrated into a Shiny web application providing location-based visualization of high-risk windows, geolocation support, and linkage to the updated Moroccan phytosanitary index. Conclusions : This work demonstrates the complementary value of phenological and data-driven modeling approaches for pest forecasting and presents a flexible, extensible digital platform to support advisory decision-making within IPM programs. The application emphasizes cautious interpretation outside calibrated reference contexts, as the models were developed using data from a limited number of sites, including a single orchard, and provides a foundation for future expansion through additional field validation and environmental data integration.