Automated insect detection and biomass monitoring via AI and electrical field sensor technology

Abstract

Abstract Insects, vital for ecosystem stability, are declining globally necessitating improved monitoring methods. Trap-based approaches are labor-intensive, invasive, and limited in scope. This study therefore presents a novel, automated, non-invasive insect monitoring system that detects atmospheric electrical field modulations caused by flying insects. In-field sensors monitor insect activity and biomass without physical trapping, using differential electric field measurements and convolutional neural networks for detection and wing-beat frequency analysis. Furthermore, a biomass algorithm that estimates taxon-specific weights is introduced. To validate this method, paired sensor and Townes Malaise trap deployments were conducted at two sites in a Danish nature reserve. Results showed moderate to strong correlations between sensors and traps, particularly at one site (Spearman’s $$\rho =0.725$$ for counts; 0.644 for biomass), supporting the method’s viability. A discrepancy in biomass estimates between methods, greater than that of counts, suggests the need for further refinement of the sensor’s biomass estimation. For inter-method consistency, sensor-sensor correlations ( $$\rho =0.758$$ for counts; 0.867 for biomass) exceeded Malaise-Malaise correlations ( $$\rho =0.597$$ for counts; 0.641 for biomass), though not significantly so ( $$P=0.304$$ for counts; $$P=0.057$$ for biomass). Overall, the study concludes that while further work is needed, this innovative approach shows promise for future insect monitoring and ecological research.