Using bioacoustics to explore the ecology of neotropical katydids

Insects are vital to terrestrial ecosystems, yet their populations are notoriously difficult to monitor, especially in challenging habitats like forest canopies. Despite their ecological importance, research on insects has historically relied on limited geographic sampling and often focused on specific taxonomic groups. Advances in bioacoustic methods, particularly passive acoustic monitoring (PAM) and machine learning, offer promising techniques for studying spatial and temporal distributions of sound-producing insects, as well as for asking more complex questions about insect behavior and ecology. By pairing PAM with machine learning techniques for automating sound detection, researchers can achieve high-resolution data on the dynamics of calling insect communities across larger spatial and temporal scales. This innovative approach bridges significant gaps in species monitoring and population analysis, yielding actionable knowledge needed to address ecological changes.

To illustrate this, we conducted multi-year acoustic monitoring on Barro Colorado Island in Panamá, focusing on katydids. Our research employed focal acoustic recordings, as well as long-term canopy recordings (PAM) across several years. A custom-engineered neural network was developed to analyze the sounds, which enabled us to assess spatial, temporal, and seasonal patterns, investigate environmental factors like moonlight and habitat structure, and explore links between insect calls and bat activity. Automated sampling methods are crucial for tracking insect population trends in changing ecosystems through time, especially as we try to understand patterns in insect communities across larger areas. They provide ecological insights critical for conservation and guide informed actions to mitigate biodiversity loss.