Context: The global decline in honeybee populations poses a major threat to pollination, food security, and ecological balance. Among the most harmful threats was the Varroa destructor mite—a parasitic agent that weakens bee immunity and spreads viral infections. Traditional detection methods were invasive, slow, and often unreliable, creating a need for smarter, automated alternatives.

Objective: This study presents a real-time, low-cost, and non-invasive monitoring system that integrates sensor data and image capture to enable early detection and thermal control of Varroa infestations within beehives.

Methods: A physical prototype was developed using ESP32-CAM and DHT22 sensors, with thermal regulation provided by an STC-1000 thermostat. Sensor and image metadata were transmitted over Message Queuing Telemetry Transport (MQTT) to a digital twin modeled on Eclipse Ditto. The system was evaluated through lab tests using one physical hive and 20 simulated virtual twins.

Results: The system reliably maintained hive temperatures near the 42 °C treatment threshold and demonstrated consistenet communication performance with minimal latency. The image classification model was trained on the HoneyBee Annotated Images dataset. However, it was executed off-device because the ESP32 lacked sufficient memory. Detection results were integrated into the twin model manually.

Conclusions: The proposed architecture effectively demonstrates how combining WSN-based sensing with digital twins can enable early detection and thermal treatment of Varroa mite infestations. The system’s use of low-cost, energy-efficient hardware and lightweight MQTT communication aligns with the principles of green communication, making it suitable for scalable deployment in resource-constrained environments. Future work should focus on embedded AI deployment, energy-efficient edge computing, and field trials across distributed hives.