Using passive acoustic monitoring and machine learning analysis to investigate katydid ecology and behavior

Passive acoustic monitoring (PAM) can provide detailed information on the spatial and temporal distribution of sound producing insects. When combined with machine learning approaches for extracting data from multiple sites and multiple years, PAM can provide exceptionally detailed information about the ecology of the calling insect community. We placed recording devices in the forest canopy on Barro Colorado Island in Panamá and used a combination of manual annotation and machine learning analysis in Koogu (an open source python package) to test the following hypotheses in Neotropical forest katydids: (1) The forest canopy species assemblage will consist disproportionately of katydid species with high flight and dispersal ability (reflected by low wing-loading coefficients), (2) katydids aggregate on an individual tree during the short window when a tree flushes new leaves (resulting in short concordant peaks of signaling activity across multiple katydid species), and (3) in species with relatively short calling seasons, males will have little time to accumulate nuptial gifts for females and will instead invest in mate searching (reflected by high male:female sex ratios in insects captured at lights). In changing forests, consistent approaches for insect sampling will be key for understanding insect ecology and generating interpretable and actionable data.