A Large-Scale IoT Telemetry Framework for Edge-Aware Forecasting and Diagnostics in Precision Apiculture

Precision apiculture increasingly relies on Internet of Things (IoT) telemetry to support colony healthmonitoring, productivity assessment, and early anomaly detection. However, existing approaches typicallyfocus on either long-term forecasting or short-term diagnostics, while deployment constraints associatedwith edge computing are often overlooked. This paper presents a large-scale IoT telemetry framework forresource-efficient forecasting and diagnostics in precision apiculture. The proposed framework integratesmultimodal telemetry collected from 116 instrumented beehives deployed across multiple European countries between 2022 and 2025, comprising 1,645,074 observations acquired at sampling intervals rangingfrom 15 to 60 minutes, with a median interval of 16.28 minutes. The system combines hive weight, internaltemperature, humidity, and acoustic measurements within a unified analytics pipeline supporting bothoperational forecasting and diagnostic monitoring. A decoupled architecture is introduced in which longhorizon forecasting models are separated from short-horizon state-space diagnostics, enabling improvedinterpretability and deployment flexibility. In addition to predictive performance, the framework evaluatesedge-oriented metrics including model size, inference latency, and computational efficiency. Experimentalresults demonstrate that lightweight regression-based models provide stable forecasting performance whileremaining suitable for deployment in resource-constrained IoT environments. The proposed frameworkoffers a scalable and transferable approach for smart beehive monitoring and may be adapted to otheragricultural and environmental sensing applications