Computer vision-based population monitoring of immobile synchronous firefly swarms

Abstract

Southeast Asia hosts a firefly species, Pteroptyx malaccae, that holds incredible value both scientifically and culturally. These insects exhibit synchronous swarm flashing behavior for long periods while remaining immobile. This offers a unique opportunity for scientific study: it is possible to film the fireflies and label each individual with a unique virtual ID that is preserved over the course of a video. From early March 2023 until the end of February 2024, we gathered data in the form of short one- to two-minute videos on 16 distinct dates, recording up to sixty videos in one night for a total of 281 videos. To our knowledge, such precise long-term monitoring of insect wildlife is unprecedented. We use this data and custom computer vision techniques to gain insight into fluctuations in the firefly population over the course of that year. The data gathered can also be readily analyzed to obtain further insight: we present typical flashing frequencies and how that varies throughout the year—measures that are hard to acquire with the non-automated methods that have been used previously. By comparing the observed trends with various local weather and environmental conditions, we suggest possible explanations for their relative abundance over time. This work may allow for future automated monitoring of other firefly species, as well as help in long-term (i.e., multi-year) monitoring to accurately assess the severity of population decline.