Monitoring insect diversity with AI enabled devices

Monitoring insect biodiversity is essential for management of agricultural and natural ecosystems. Historically Malaise traps have been used for efficient biodiversity monitoring and sampling tools for many types of flying invertebrates. These traps provide standardized collection methods that allow researchers to assess biodiversity across a wide range of spatial and temporal scales. However, traditional approaches to processing Malaise trap samples, which often involve sorting, pinning, and morphological or molecular identification, are labor-intensive, costly, and time-consuming. These challenges create a bottleneck that limits the scalability and responsiveness of biodiversity assessments. Indeed, identification is most commonly performed at the order or family level, where taxonomic resolution is feasible given the volume and diversity of specimens. To overcome these limitations, we have developed an artificial intelligence-powered non-lethal monitoring system capable of automatically detecting and identifying invertebrates in real time as they enter the trap. This system integrates computer vision and machine learning algorithms trained on diverse image datasets to recognize distinct phenotypes and classify them taxonomically with high accuracy. By strategically training these automated units to function remotely across a range of habitats and taxa, we aim to enable high output, continuous, and cost-effective monitoring of invertebrate communities. This technological advancement will not only enhance our capacity to detect shifts in biodiversity and community composition in response to environmental change but will also support applied objectives in conservation biology, ecosystem health monitoring, and agricultural pest and beneficial invertebrate management.