Performance of four ecoacoustic indices as proxies of bird species richness in a Neotropical forest

Abstract

Ecoacoustic indices derived from sound recordings are increasingly used to assess ecosystem health and biodiversity, but their meaningfulness remains debated, particularly in tropical forests. We evaluated the effectiveness of four commonly used ecoacoustic indices (ACI, BIO, NP, NDSI) in predicting bird species richness in a Mexican tropical forest, examining whether combining indices or selecting specific sampling periods enhances predictive power. We collected acoustic data over 11 days at 25 sampling locations, and used BirdNET for automated bird species detection. We calculated mean and standard deviation values for three time periods (using full-day recordings and two peak activity periods) and applied general linear models to evaluate the predictive power of ecoacoustic indices on observed bird species richness. The highest-ranking models explained up to 50% of data variability, with SD NDSI emerging as the sole important predictor. When the four indices were pooled into principal components, the models explained less variability (31–35%) compared to those using ecoacoustic indices as separate predictors. Our study highlights the limitations in the generalizability of widely used ecoacoustic indices and challenges previous research that emphasizes the utility of pooling information from multiple indices. The predictive power of ecoacoustic indices remained consistent across time periods, showing no advantage in focusing sampling on peak vocal activity periods. Our findings emphasize the importance of carefully considering site-specific factors, sampling design, and index selection in ecoacoustic studies.