Suhana Shrivastava

Investigating the Acoustic Similarities of Auditory Elephant Deterrents to Optimize Currently Used Techniques

Started Oct. 29, 2021

Abstract or project description

One of the primary reasons elephants are endangered is human-elephant conflict (HEC), the opposition that occurs between elephants and the humans living nearby. The violence that erupts in settings of HEC, such as crop fields, often results in both human and elephant deaths as both species struggle to coexist. Many methods are being researched to mitigate human-elephant conflict, including playing audio playbacks that deter elephants away from crop fields and reduce chances of contact and destruction. Habituation to these stimuli creates the demand for a greater number and more types of auditory deterrents, but it would be unethical and inefficient to immediately jump to field tests without verifying these playbacks are at least somewhat effective. Thus, this project aims to analyze currently used auditory deterrents to determine if any acoustic similarities exist between them, and create a generalization for what characteristics make up an effective auditory deterrent.

To perform analysis, we used techniques in R and toolboxes such as soundgen for retrieving acoustic characteristics. Through performing principal component analysis (PCA), we were able to find that playbacks with low effectiveness on elephants are significantly different in terms of their acoustic characteristics. This contrasts the playbacks with high effectiveness, which were shown to be clustered together in the PCA. Through linear regression and calculation of p-values, we were also able to deduce that characteristics such as lowest dominant frequency band, loudness, the 25% quartile of the sound, the peak frequency, depth of amplitude modulation, and flux were all positively correlated towards the effectiveness of an auditory deterrent. We used a weighted average approach to calculate a recommended value for effectiveness of an auditory deterrent for each of these characteristics. Overall, we hope that these results will help optimize current playbacks and help create a threshold of characteristics to use before future testing, to reduce habituation and human-elephant conflict.