Detection of Honeybee Disease: Varrosis using a Semiconductor Gas Sensor Array

Abstract

The presented study was focussed on the detection of Varroa destructor infestation of honeybee colonies, based on gas sensor measurements of beehive air. The detection consisted in determination whether the colony infestation rate was 0% or different. An array of partially selective gas sensors was used in measurements. It included the following semiconductor gas sensors: TGS832, TGS2602, TGS823, TGS826, TGS2603 and TGS2600. The sensors were exposed in dynamic conditions. The infestation detection problem was solved using a classification approach. The basis for classification were feature vectors. They were composed of responses of sensors, elements of the gas sensor array. The utilised responses were associated with various parts of the sensor signal recorded during dynamic exposure and regeneration. As a reference, we used the V. destructor infestation rate of bee colonies estimated using a flotation method. The smallest misclassification error was 17% and it was achieved with the k-NN classifier. The experimental study was performed in field conditions. It included honeybee colonies of various kinds, settled in beehives made of various materials, differently located, examined in various atmospheric conditions, at different times of the day. Taking this into consideration, the detection error at the level of 17 % is a promising result. It demonstrates the possibility to detect varroosis using an array of partially selective sensors.