We present the design of a predictive digital twin for large-scale Varroa management in honey bee apiaries. The system couples a wireless sensor network for continuous hive sensing with remote actuation of treatment commands. It captures and analyzes images from beehives for automated mite detection. It also uses generative time-series models to forecast colony dynamics, reveal environmental and operational drivers, and conduct counterfactual analyses. Inter-colony spread is modeled with a statistical network model that, together with the time-series models, enables prediction and what-if evaluation of treatment strategies. The digital twin and its embedded models are updated as new measurements arrive. To address limited field data at deployment, we pretrain the models by creating synthetic datasets from bee–mite population-dynamics models and network-level spread. We also propose a dynamic monitoring strategy that adjusts surveillance intensity to current Varroa transmission conditions. Feasibility is assessed with a node-level energy budget for sensing, communication, and in-beehive treatment applicators. This digital twin forecasts infestation trajectories, supports what‑if planning, and enables timely, targeted interventions. The system improves over time through continuous online updates from field data. Overall, the results show a scalable path to real-time Varroa management across apiary networks from regional to national and cross-border scales.
Eivazzadeh et al. (Mon,) studied this question.