Using bioacoustics to understand species-specific patterns of Native Hawaiian forest bird call density across elevation and forest types of Hakalau

Abstract

Accurate monitoring of native species is essential for detecting ecological changes and informing conservation, particularly as climate changes accelerates the spread of avian malaria into higher elevation forests of Hawaiʻi. This study evaluated the use of passive acoustic monitoring (PAM) and machine learning to estimate the call density of eight native Hawaiian forest bird species across the elevational gradient of Hakalau Forest National Wildlife Refuge. Acoustic data were collected using SM4 song meters deployed at nine stations spanning open and closed forest types between 2022 and 2023. Perch, a machine learning classifier, was used to identify species-specific vocalizations, and validated detections were used to estimate call density. Generalized linear and generalized additive models were used to assess the effects of elevation and forest types on call density, while LiDAR-derived forest structural metrics were explored for their relationships with species call density. Results revealed species-specific responses to elevation and forest types, with elevation influencing call density for some and closed forest supporting higher call densities for several species. LiDAR-derived structural metrics showed limited association with species call density. These findings demonstrate the value of bioacoustics, combined with machine learning, as a scalable and non-invasive approach for long-term monitoring of native Hawaiian forest birds and for informing conservation and management in Hawaiʻi.