Abstract:Honey bees (Apis spp.) are often used as biological indicators of environmental changes. Recently, bees have been explored to monitor air contaminants by listening to the beehive sound. The beehive sound is believed to encode information on bee responses to chemicals outside their hives. Here we conducted an experiment to address this. First, we randomly fed colonies with pure syrup (PS), acetone-laced syrup (AS), or ethyl acetate-laced syrup (ES) in front of the beehives and collect the beehive sound. Based on the audio data, we build machine learning (ML) models to identify the types of syrup. The result shows that ML models achieved over 90% accuracy for identifying syrup types, indicating that the bees inside their hives emitted the sound associated with the chemicals outside their hives. Then, we sequentially fed the colonies in the order of PS, ES, and AS (the first session) and again in the reverse order (the second session), but did not remove the accumulated ES or AS in the alternative feeding experiment. Based on the audio data, the identification accuracy of both ES and AS by the ML model built on the randomly feeding experiment was different, indicating that the accumulated chemical residuals could confuse the ML models. Therefore, the beehive sound–based environmental monitoring should simultaneously consider the responses of bees to the chemicals outside their hives and their responses to those accumulated inside their hives, which may act simultaneously.

A matter of the beehive sound: Can honey bees alert the pollution out of their hives?