Honey Bee Colony Population Daily Loss Rate Forecasting and an Early Warning Method Using Temporal Convolutional Networks

Abstract

The population loss rate of a honey bee colony is a critical index to verify its health condition. Forecasting models for the population loss rate of a honey bee colony can be an essential tool in honey bee health management and pave away to early warning methods in the understanding of potential abnormalities affecting a honey bee colony. This work presents a forecasting and early warning algorithm for the population daily loss rate of honey bee colonies and determining warning levels based on the predictions. Honey bee colony population daily loss rate data were obtained through embedded image systems to automatically monitor in real-time the in-and-out activity of honey bees at hive entrances. A forecasting model was trained based on temporal convolutional neural networks (TCN) to predict the following day’s population loss rate. The forecasting model was optimized by conducting feature importance analysis, feature selection, and hyperparameter optimization. A warning level determination method using an isolation forest algorithm was applied to classify the population daily loss rate as normal or abnormal. The integrated algorithm was tested on two population loss rate datasets collected from multiple honey bee colonies in a honey bee farm. The test results show that the forecasting model can achieve a weighted mean average percentage error (WMAPE) of 17.1 ± 1.6%, while the warning level determination method reached 90.0 ± 8.5% accuracy. The forecasting model developed through this study can be used to facilitate efficient management of honey bee colonies and prevent colony collapse.

A Framework for Better Sensor-Based Beehive Health Monitoring

“…Internal hive condition such as temperature [Braga et al, 2020, Seritan et al, 2018, Meikle et al, 2017, Murphy et al, 2016], CO 2 concentration [Seritan et al, 2018,Murphy et al, 2016], humidity [Paffhausen et al, 2021,Murphy et al, 2016], audio [Sharif et al, 2020,Pérez et al, 2016] and weight [Paffhausen et al, 2021,Hladun et al, 2016] have also been used extensively as proxies for hive health. Overwintering colony loss rate [Meikle et al, 2017,Becher et al, 2013, Ngo et al, 2021b] and pathogen infestation rate are also proxies for hive health [Jara et al, 2020]. As hive health assessment plays such a critical role in hive management, and underpins many developments in hive monitoring across Operational, Investigative and Predictive monitoring approaches, we devote a specific section (discussion Sec.…”

“…Forager workforce size and/or dynamics: Population loss rate [Ngo et al, 2021b]; activity at the hive entrance [Paffhausen et al, 2021]; size of the overwintering colony [Becher et al, 2013]; traffic at the hive entrance [Tashakkori et al, 2021]; bee behaviour at the hive entrance [Schurischuster et al, 2018]; overwintering loss rate [Meikle et al, 2017].…”

“…Internal hive condition such as temperature[Braga et al, 2020,Seritan et al, 2018,Meikle et al, 2017,Murphy et al, 2016, CO 2 concentration[Seritan et al, 2018,Murphy et al, 2016, humidity[Paffhausen et al, 2021,Murphy et al, 2016, audio[Sharif et al, 2020,Pérez et al, 2016 and weight[Paffhausen et al, 2021,Hladun et al, 2016 have also been used extensively as proxies for hive health. Overwintering colony loss rate[Meikle et al, 2017,Becher et al, 2013, Ngo et al, 2021b and pathogen infestation rate are also proxies for hive health[Jara…”

Towards intelligent and integrated pest management through an AIoT‐based monitoring system

“…An exponential distribution function was selected since it can sufficiently explain the trend in daily increase of the insect pest count in each site, as found in our previous work 25 . First, a Severe threshold level, which represents the highest threat of insect pest infestation, was determined from the frequency distribution using isolation forest, an adaptive anomaly detection algorithm 26,27 . The values below the Severe threshold level were subdivided into High , Moderate , and Guarded threshold levels based on the 0.05, 0.25, and 0.50 percentiles of the fitted exponential distribution, respectively; the Low level was determined based on the values below the Guarded threshold level.…”

IoT Monitoring and Prediction Modeling of Honeybee Activity with Alarm

“…In recent years, beekeepers have encountered problems with mass bee deaths [1,2] and bee migrations due to climate change, and the impact of weather conditions on flowering disorders in nature in periods when bees should collect pollen. The use of various pesticides and herbicides in plant protection, and spraying at a time when bees are active in periods of disturbed climatic conditions due to high temperatures and humidity cause bees to be active in the later hours when spraying is performed [3].…”