The shape of sound: A new R package that crosses the bridge between Bioacoustics and Geometric Morphometrics

Abstract

Abstract The field of Bioacoustics has flourished in the past few decades, with numerous initiatives towards its application as a tool for biodiversity conservation. Despite the development of new methods involving computational programming and machine learning algorithms, higher accuracy is still needed for species identification, which is why we present herein the new SoundShape package for R platform. The new package focuses on implementing the eigensound method, a promising, and yet little explored protocol for bioacoustical analysis. The eigensound function is the main feature of SoundShape, allowing its user to convert sound waves into a dataset that can be analysed similar to coordinate sets from Geometric Morphometrics Methods, thus enabling the direct comparison between stereotyped calls from different species. Besides, SoundShape also features complementary functions for basic analysis of quantitative variation and illustration of hypothetical sound shapes representing the sample of sounds. Implementation of SoundShape is summarized through a workflow guide, using data from the new package. The sample study resulted in nearly 90% variation expressed in an ordination plot, therefore successfully summarizing complex sound waves previously described by large datasets of amplitude values. Moreover, we also introduce two addendums to eigensound: (a) three steps to prevent errors and biased results—aimed at securing a meaningful comparison between acoustic units from different species; and (b) the option of applying a logarithmic grid on the x‐axis (time)—which emphasizes short duration calls while also encompassing long ones. SoundShape package now provide the tools required for anyone to replicate and implement the eigensound method on the r platform. This will enable future studies to focus on further exploring the applications of sound shape analysis on various scientific areas, including (but not limited to) taxonomy, systematics, evolution, acoustic niche partitioning, soundscape ecology and machine learning algorithms focused on species identification for conservation biology.