Acoustic indicators of spatial and temporal variation in coral reef soundscapes

Abstract

Soundscape ecology has emerged in recent years as a powerful approach to investigate
and monitor ecosystems. On coral reefs, the diverse and abundant organisms present
create an underwater cacophony that reflects ecosystem health and can be used to
address ecological questions. In this thesis I employ novel statistical and deep-learning
(DL) approaches to explore spatiotemporal patterns in coral reef soundscapes in Kimbe
Bay, Papua New Guinea.
In the introduction, I review some of the key findings and approaches from coral reef
acoustics. I describe the importance of sound at the organismal and ecosystem levels,
as well some of the methods that have been used.
The first data chapter relates DL feature embeddings to acoustic indices, creating a
hybrid methodology that combines the power of DL to capture soundscape differences
with the greater interpretability of acoustic indices. My findings indicate that the acoustic
complexity index (ACI) and the entropy index (H) are the most important indices for
distinguishing sites, which likely reflects their ability to capture the activity of fish and
snapping shrimp.
The second data chapter models temporal periodicities and environmental drivers in the
soundscape of a small island over a two-month duration. Coral reef soundscapes are
known to be highly dynamic, yet many studies use short ‘snapshot’ recordings. Using
generalised additive mixed models (GAMMs), I discovered distinct diurnal and lunar
cycles in amplitude, as well as environmental effects, underlining the dynamic nature of
reef soundscapes.
Overall, my thesis demonstrates the potential of acoustics for developing our
understanding of coral reefs. A key aim was to explore how these tools can be used to
keep a pulse on the health of coral reef ecosystems. However, the findings also reveal the
spatiotemporal variability of soundscapes and the need for care in both data collection
and automated methods.

Date of Award	10 Dec 2024
Original language	English
Awarding Institution	University of Bristol
Supervisor	Steve Simpson (Supervisor) & Hugo B Harrison (Supervisor)