Air-Based Monitoring of Honey Bee Colony Diseases Using In-Hive Sensing and Machine Learning: System Overview and Laboratory Validation

Parasites and pathogens are an increasing threat to honey bee colonies, and practical noninvasive tools for continuous remote bee colony monitoring are lacking. We present an integrated framework for continuous, air-based monitoring of honey bee colony health, combining in-nest sensing of volatile organic compounds, oxidizing gases, gas resistance, temperature, humidity and pressure with machine learning classification, and we report its validation on three separate controlled laboratory cage experiments: Vairimorpha/Nosema ceranae infection (20 cages, 397 cage-days), exposure to Ascosphaera apis (20 cages, 460 cage-days), and Varroa destructor infestation (13 cages, 504 cage-days). Every model was evaluated by cage-disjoint stratified group cross-validation, in which all daily observations of a cage belong to the same fold. Calendar variables were removed, and a single fixed estimator with balanced class weights was used. Exposure to A. apis was separated from control cages at macro F1 0.961 ± 0.047 against a baseline of 0.333, above that baseline on every fold. The nosemosis detection task reached 0.762 ± 0.195 against a baseline of 0.442, above that baseline on four of the five folds, with the best separation in fully populated cages. The varroosis recognition did not exceed its own baseline (0.365 against 0.426), mainly due to the low number of control cages. The results demonstrate that machine learning analysis of in-cage air and microclimate measurements can separate cages with both brood and adult worker bees exposed to A. apis from unexposed ones under laboratory conditions, with temperature features ranked highest. The suitability of the solution for different diseases, its field effectiveness and the importance of other features under field conditions require further study.

Authors

Institutions

Publication Details

Journal
Sensors
Published
2026-11-13
DOI
https://doi.org/10.3390/s26227162
Primary Topic
Insect and Pesticide Research
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Air-Based Monitoring of Honey Bee Colony Diseases Using In-Hive Sensing and Machine Learning: System Overview and Laboratory Validation

Michał Wierzbiński, Maciej Sylwester Bryś, Aneta Strachecka, Krzysztof Olszewski et al.
3 links
Sensors
Insect and Pesticide Research
article

Air-Based Monitoring of Honey Bee Colony Diseases Using In-Hive Sensing and Machine Learning: System Overview and Laboratory Validation

Michał Wierzbiński, Maciej Sylwester Bryś, Aneta Strachecka, Krzysztof Olszewski, Patrycja Skowronek, Bartosz Kądziołka, Tomasz Dziubak, Dagmara Stasiowska, Milena Jaremek, Anna Gryboś, Sebastian Cibor
article en

Abstract

Parasites and pathogens are an increasing threat to honey bee colonies, and practical noninvasive tools for continuous remote bee colony monitoring are lacking. We present an integrated framework for continuous, air-based monitoring of honey bee colony health, combining in-nest sensing of volatile organic compounds, oxidizing gases, gas resistance, temperature, humidity and pressure with machine learning classification, and we report its validation on three separate controlled laboratory cage experiments: Vairimorpha/Nosema ceranae infection (20 cages, 397 cage-days), exposure to Ascosphaera apis (20 cages, 460 cage-days), and Varroa destructor infestation (13 cages, 504 cage-days). Every model was evaluated by cage-disjoint stratified group cross-validation, in which all daily observations of a cage belong to the same fold. Calendar variables were removed, and a single fixed estimator with balanced class weights was used. Exposure to A. apis was separated from control cages at macro F1 0.961 ± 0.047 against a baseline of 0.333, above that baseline on every fold. The nosemosis detection task reached 0.762 ± 0.195 against a baseline of 0.442, above that baseline on four of the five folds, with the best separation in fully populated cages. The varroosis recognition did not exceed its own baseline (0.365 against 0.426), mainly due to the low number of control cages. The results demonstrate that machine learning analysis of in-cage air and microclimate measurements can separate cages with both brood and adult worker bees exposed to A. apis from unexposed ones under laboratory conditions, with temperature features ranked highest. The suitability of the solution for different diseases, its field effectiveness and the importance of other features under field conditions require further study.

SensorsVol. 26(22)
University of Life Sciences in Lublin (PL), AGH University of Krakow (PL)
Insect and Pesticide Research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon