The combined effectiveness of acoustic indices in measuring bird species richness in biodiverse sites in Cyprus, China, and Australia.

Human activities are accelerating biodiversity loss, necessitating tools capable of monitoring biodiversity patterns over large spatial and temporal scales. Passive acoustic monitoring methods, including acoustic indices, are emerging as a promising approach for surveying vocalizing animals. Numerous studies have assessed the effectiveness of acoustic indices in surveying animal communities, focusing mostly on birds and seven commonly used indices, yielding mixed results. Combining the indices has been proposed as a solution to produce more accurate predictions. In this study, we use data from 114 biodiverse sites in three countries, Cyprus, China, and Australia, to evaluate the combined effectiveness of sixty different acoustic indices in measuring bird species richness. Using the Boruta feature selection algorithm and random forest regressors, we find that the effectiveness of the indices varies considerably across study areas, and it is generally lower than what would be required to monitor bird species richness accurately (R2Cyprus = 0.06, R2China = 0.31, R2Australia = 0.52). Moreover, the most useful set of indices varied for each area; none of the sixty indices were useful in all three areas, and only three indices were useful in more than one area. Our findings, along with those of other recent studies, suggest that acoustic indices may not currently be an effective method for accurately monitoring bird species richness despite their utility in other applications, such as surveying broader biodiversity patterns. Moreover, we recommend that whenever researchers investigate the efficacy of acoustic indices, they consider all available indices to identify the most useful in their study region.