Ecological Indicators (Oct 2021)
The relationship between acoustic indices, elevation, and vegetation, in a forest plot network of southern China
Abstract
An emerging method of monitoring biodiversity is through the use of audio recordings, often made by autonomous recording units. Acoustic indices have been developed to estimate animal diversity, especially across human disturbance gradients, and have been shown to often correlate with manual counts of animals. Less work has examined whether acoustic indices can detect finer-scale habitat gradients, such as differences in vegetation within mature forests, especially using long-term vegetation plots. We used autonomous recorders to capture sound predominantly produced by birds in the morning and insects at night across a network of 27 1-ha forest vegetation plots, located between 300 and 1850 m asl in southern China, collecting 38,200 min of sound recordings. Based on the animal diversity literature, we hypothesized that acoustic indices would have strongest relationships with elevation, intermediary relationships with vertical structure (vertical heterogeneity, canopy height and tree density), and the weakest relationships with tree species diversity. Generalized linear mixed models, followed by model averaging, showed that elevation was indeed the strongest of the predictor variables, with the highest mean importance factor. In 6 of 14 models the coefficients had confidence intervals that did not cross zero, and in all such significant relationships acoustic diversity decreased with elevation. Vertical heterogeneity had the second highest mean importance factor, but one of its two significant relationships was negative. The only other significant relationship was a positive one between tree species diversity and the H index. Morning and evening recordings had similar results, despite representing biophony made by different taxa with different acoustic characteristics. However, the seven acoustic indices gave dissimilar results, with the H index having three significant relationships, compared to the ACI, BIO and AR indices that had none. We conclude that autonomous recorders, if analyzed with multiple acoustic indices, can be used to investigate landscape or vegetation-related gradients that may influence animal diversity on long-term vegetation plots.
