Soundscape Ecology Through Automated Acoustic-Based Biodiversity Indices
Soundscape Ecology Through Use of Automated Acoustic-Based Biodiversity Indices: A Test Using RCA Broadband Hydrophone Data. Adapted by OOI from Ferguson et al., 2023.
Ferguson’s et al., 2003 paper explores the use of myriad biodiversity indices, generated by automated acoustic classifications, using data from three of the Regional Cabled Array (RCA) broadband hydrophones. As the authors point out, human in-the-loop evaluation of marine species and anthropogenic noise from very large data volumes generated by passive acoustic sensors is formidable. Yet, identification of marine organisms and anthropogenic noise is increasingly important for biodiversity conservation and ecosystem monitoring. Automated biodiversity indices have been utilized in terrestrial environments, but only limited studies have used machine learning to study soundscape ecology in marine systems. This study used broadband hydrophone (HYDBBA) data from Slope Base on the Shallow Profiler Mooring (200 m water depth and ~ 100 km offshore), at Oregon Offshore (580 m water depth and ~ 72 km offshore), and the Oregon Shelf site (80 m water depth and ~ 16 km offshore) (Figure 28a) to examine seven diversity indices. Note, these study sites are valuable to making progress in soundscape ecology because the Cascadia Margin is characterized by very high biological productivity impacted by the California current, it is the site of intense shipping lanes, and because of the availability of continuous, real-time acoustic data streams provide by the RCA.
In this initial study, Ferguson et al., evaluated one month of data from the three sites: January 2017 for HYBDDA 103 and 106 and April 2018 for HYBDDA 105. Five minute files were used with 7,101 files for Slope Base, 4,725 files for Oregon Offshore, and 6,410 files for the Oregon Shelf. Data from these instruments had been previously annotated, providing ground truthing for machine learning results. Periods of vocalization of marine mammals occurred less frequently at Slope Base. Three months of data were reviewed to examine periods of mammal vocalizations and anthropogenic sounds.
Identifying the relationship between numerous acoustic indices and species characteristics is complex and requires attention to a significant number of factors and computation of multiple tests, as described in detail in this paper. The Acoustic Complexity Index (ACI, Figure 28b), is generated from an algorithm to quantify biological sounds based on intensity, it is the most commonly used index to assess acoustic indices in marine systems, and has been demonstrated useful in identifying species diversity. Results from this work show that ACI measurements increased during vocalizations by dolphins and sperm whales. However, evaluation of the seven indices show that biodiversity cannot be explicitly determined from any single acoustic index. A significant finding from this study is that true assessment of large-scale ecosystem health and changes in indicator species, which may be due to differences in seasonal and interannual variability, requires co-located physical and chemical oceanographic data. The authors note that the RCA and Endurance Array instrumentation provides “an ideal scenario for accurately monitoring system health”.
_______________________________________________
Ferguson, E.L., H.M. Clayton, and T. Sakai (2023) Acoustic Indices Respond to Marine Mammal Vocalizations and Sources of Anthropogenic Noise. Frontiers in Marine Science. 10:1025464; doi: 10.3389/fmars.2023.1025464.