Using acoustic indices to estimate wolf pack size

摘要

Acoustic indices were recently developed for biodiversity evaluation by measuring the acoustic heterogeneity generated by animals in natural environments. Some of these indices focus on the species diversity in a community by studying frequency and temporal variations in acoustic signals. We explored the possibility of using acoustic indices to estimate the population size of a specific species. More precisely, the objective was to estimate the size of grey wolf packs by testing six acoustic indices: H, Hf, Ht, M, AR, and ACI.The relationship between the averaged values of the indices and the number of howling wolves was studied based on artificial solos and choruses created from howls extracted from wolf choruses recorded in captivity. Then, 16 real choruses were used to test the size predictions based on index values calculated previously and considered as references. Finally, we explored three biases that might influence the acoustic index values and thus the chorus size estimates.All of the acoustic indices were positively correlated with chorus size, although large standard deviations were observed. Moreover, H, Hf, and Ht reached a plateau at 7-8 wolves. The size predictions based on real choruses were overestimated or underestimated. However, ACI was the most accurate with chorus size predictions close to the actual value. M and AR also had good predictive power, especially for choruses made by a relatively small number of howling wolves. The overestimates may be explained by several sources of bias related to the natural composition of real choruses. Indeed, the acoustic indices were influenced by the audio file duration, signal-to-noise ratio (SNR), and temporal overlap of the wolf howls, but not in the same manner for each index. In particular, H, Ht, and M were significantly influenced by the audio file duration and their values decreased as the duration increased. Excluding AR, all of the indices were affected by adding background noise. The H and Hf values decreased as the SNR decreased, but the opposite trend occurred for ACI. Only Hf and AR were not influenced by the temporal overlap of howls and the values of the four other indices decreased to a greater extent when more wolf howls overlapped.The most promising indices were ACI, AR, and Hf, and they may provide an innovative census tool for estimating wolf pack size. Our results are encouraging although further research is needed to obtain a more effective and accurate tool. Several recommendations and directions for further studies are discussed.