MSPB: a longitudinal multi-sensor dataset with phenotypic trait measurements from honey bees

Area of Science:

• Apiculture and Animal Science
• Environmental Monitoring
• Machine Learning Applications

Background:

• Honey bee colonies face significant threats, including Varroa destructor infestations and winter mortality.
• Effective hive management requires continuous monitoring of colony health and environmental conditions.
• Existing monitoring methods often lack the integration of multi-sensor data and detailed phenotypic measurements.

Purpose of the Study:

• To present a novel, year-long multi-sensor dataset from honey bee colonies (Apis mellifera).
• To combine extensive phenotypic measurements with continuous sensor data for comprehensive hive analysis.
• To demonstrate the utility of this dataset for developing advanced hive monitoring applications using machine learning.

Main Methods:

• Collected non-stop audio, temperature, and humidity data from 53 honey bee hives over one year (April 2020-April 2021).
• Recorded detailed phenotypic data including population, brood cells, Varroa destructor infestation, behaviors, honey yield, and winter mortality.
• Employed expert annotation from apicultural science for phenotypic measurements.
• Performed data pre-processing and analysis of sensor and phenotypic data.

Main Results:

• Established a rich, integrated dataset of honey bee colony health and environmental parameters.
• Visualized phenotypic data distribution and sensor data patterns.
• Demonstrated successful machine learning applications for predicting winter mortality, estimating hive population, and detecting queen status.

Conclusions:

• The presented dataset is a valuable resource for advancing honey bee research and apiculture.
• Integrated multi-sensor and phenotypic data facilitate a broader scope of analysis for hive monitoring.
• Machine learning models applied to this data show promise for real-time hive management and health assessment.