Enhancing automated analysis of marine soundscapes using ecoacoustic indices and machine learning

Abstract

Historically, ecological monitoring of marine habitats has primarily relied on labour-intensive, non-automated survey methods. The field of passive acoustic monitoring (PAM) has demonstrated the potential of this practice to automate surveying in marine habitats. This has primarily been through the use of ‘ecoacoustic indices’ to quantify attributes from natural soundscapes. However, investigations using individual indices have had mixed success. Using PAM recordings collected at one of the world’s largest coral reef restoration programmes, we instead apply a machine-learning approach across a suite of ecoacoustic indices to improve predictive power of ecosystem health. Healthy and degraded reef sites were identified through live coral cover surveys, with 90–95% and 0–20% cover respectively. A library of one-minute recordings were extracted from each. Twelve ecoacoustic indices were calculated for each recording, in up to three different frequency bandwidths (low: 0.05–0.8 kHz, medium: 2–7 kHz and broad: 0.05–20 kHz). Twelve of these 33 index-frequency combinations differed significantly between healthy and degraded habitats. However, the best performing single index could only correctly classify 47% of recordings, requiring extensive sampling from each site to be useful. We therefore trained a regularised discriminant analysis machine-learning algorithm to discriminate between healthy and degraded sites using an optimised combination of ecoacoustic indices. This multi-index approach discriminated between these two habitat classes with improved accuracy compared to any single index in isolation. The pooled classification rate of 1000 cross-validated iterations of the model had a 91.7% 0.8, mean SE) success rate at correctly classifying individual recordings. The model was subsequently used to classify recordings from two actively restored sites, established >24 months prior to recordings, with coral cover values of 79.1% (±3.9) and 66.5% (±3.8). Of these recordings, 37/38 and 33/39 received a classification as healthy respectively. The model was also used to classify recordings from a newly restored site established <12 months prior with a coral cover of 25.6% (±2.6), from which 27/33 recordings were classified as degraded. This investigation highlights the value of combining PAM recordings with machine-learning analysis for ecological monitoring and demonstrates the potential of PAM to monitor reef recovery over time, reducing the reliance on labour-intensive in-water surveys by experts. As access to PAM recorders continues to rapidly advance, effective automated analysis will be needed to keep pace with these expanding acoustic datasets.

Full text

Ecological Indicators 140 (2022) 108986
1470-160X/© 2022 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Original Articles
Enhancing automated analysis of marine soundscapes using ecoacoustic
indices and machine learning
Ben Williams
a
,
b
,
*
, Timothy A.C. Lamont
a
,
c
, Lucille Chapuis
a
, Harry R. Harding
d
,
Eleanor B. May
a
, Mochyudho E. Prasetya
e
, Marie J. Seraphim
f
, Jamaluddin Jompa
g
,
David J. Smith
h
,
i
, Noel Janetski
e
, Andrew N. Radford
d
, Stephen D. Simpson
a
,
d
a
College of Life and Environmental Sciences, University of Exeter, Exeter EX4 4PS, UK
b
Centre for Biodiversity and Environment Research, University College London, UK
c
Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK
d
School of Biological Sciences, University of Bristol, Bristol BS8 1TQ, UK
e
MARS Sustainable Solutions, Makassar, Indonesia
f
School of Health and Life Sciences, University of the West of Scotland, PA1 2BE, UK
g
Graduate School, Hasanuddin University, 90245 Makassar, Indonesia
h
Coral Reef Research Unit, School of Life Sciences, University of Essex, Colchester, Essex CO3 4SQ, UK
i
Mars Incorporated, 4 Kingdom Street, Paddington, London W2 6BD UK
ARTICLE INFO
Keywords:
Passive acoustic monitoring
Ecoacoustics
Restoration
Coral reef
Marine
Machine learning
ABSTRACT
Historically, ecological monitoring of marine habitats has primarily relied on labour-intensive, non-automated
survey methods. The eld of passive acoustic monitoring (PAM) has demonstrated the potential of this practice to
automate surveying in marine habitats. This has primarily been through the use of ‘ecoacoustic indices’ to
quantify attributes from natural soundscapes. However, investigations using individual indices have had mixed
success. Using PAM recordings collected at one of the world’s largest coral reef restoration programmes, we
instead apply a machine-learning approach across a suite of ecoacoustic indices to improve predictive power of
ecosystem health. Healthy and degraded reef sites were identied through live coral cover surveys, with 90–95%
and 0–20% cover respectively. A library of one-minute recordings were extracted from each. Twelve ecoacoustic
indices were calculated for each recording, in up to three different frequency bandwidths (low: 0.05–0.8 kHz,
medium: 2–7 kHz and broad: 0.05–20 kHz). Twelve of these 33 index-frequency combinations differed signi-
cantly between healthy and degraded habitats. However, the best performing single index could only correctly
classify 47% of recordings, requiring extensive sampling from each site to be useful. We therefore trained a
regularised discriminant analysis machine-learning algorithm to discriminate between healthy and degraded
sites using an optimised combination of ecoacoustic indices. This multi-index approach discriminated between
these two habitat classes with improved accuracy compared to any single index in isolation. The pooled clas-
sication rate of 1000 cross-validated iterations of the model had a 91.7% 0.8, mean SE) success rate at correctly
classifying individual recordings. The model was subsequently used to classify recordings from two actively
restored sites, established >24 months prior to recordings, with coral cover values of 79.1% (±3.9) and 66.5%
(±3.8). Of these recordings, 37/38 and 33/39 received a classication as healthy respectively. The model was
also used to classify recordings from a newly restored site established <12 months prior with a coral cover of
25.6% (±2.6), from which 27/33 recordings were classied as degraded. This investigation highlights the value
of combining PAM recordings with machine-learning analysis for ecological monitoring and demonstrates the
potential of PAM to monitor reef recovery over time, reducing the reliance on labour-intensive in-water surveys
by experts. As access to PAM recorders continues to rapidly advance, effective automated analysis will be needed
to keep pace with these expanding acoustic datasets.
* Corresponding author at: Centre for Biodiversity and Environment Research, University College London, WC1H 0AG, UK.
E-mail address: ben.williams.20@ucl.ac.uk (B. Williams).
Contents lists available at ScienceDirect
Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind
https://doi.org/10.1016/j.ecolind.2022.108986
Received 14 January 2022; Received in revised form 13 May 2022; Accepted 15 May 2022

Ecological Indicators 140 (2022) 108986
2
1. Introduction
Ecological monitoring of marine habitats is key to understanding
these ecosystems and successfully measuring the outcomes of conser-
vation and restoration programmes happening in our oceans. This kind
of ecological monitoring often relies on visual census surveys. However,
these come with limitations which include the requirement of expert
data collectors, logistical complexities, are often expensive, are typically
poor at monitoring cryptic organisms within the ecological community,
and, only collect a snapshot in time of the target site, rather than con-
tinous long-term data (Mooney et al., 2020; Munger et al., 2022).
Moreover, conservation and restoration programmes are typically
limited by time and resources, making it a challenge to sufciently
report on their progress using these methods (Bostr¨
om-Einarsson et al.,
2020; Rilov et al., 2020).
Automated passive acoustic monitoring (PAM) of whole soundscapes
has the potential to address many of these limitations (Lindseth and
Lobel, 2018; Mooney et al., 2020, Lamont et al., 2021). Low-cost
acoustic recording technology capable of recording continuously for
several days, or longer with duty cycling, is becoming available (Chapuis
et al., 2021; Lamont et al., 2022). These devices can be deployed rapidly
by non-expert data collectors and left to record autonomously. These can
collect data on cryptic organisms that disproportionately rely on
acoustic communication compared to non-cryptic species (Lamont et al.,
2022). A growing number of studies have found relationships between
the soundscapes of marine habitats and traditional ecological metrics
such as benthic cover, sh communities and overall habitat quality using
automated approaches (Nedelec et al., 2015; Butler et al., 2016;
Freeman and Freeman, 2016; Harris et al., 2016; Gordon et al., 2018;
Elise et al., 2019). As well as being useful for the tracking of habitat
characteristics, soundscapes also constitute important components of
ecosystem functioning, especially for orientation and recruitment of a
variety of organisms (Simpson et al., 2005; Lecchini et al., 2018; Gordon
et al., 2019). Surveying reefs using acoustics can therefore provide new
understandings that may explain both ecological and behavioural
processes.
Given recent innovations in autonomous hydrophone technology,
our ability to capture large databases of long-term soundscape re-
cordings is expanding (Sousa-Lima et al., 2013; Lamont et al., 2022).
However, analytical approaches must keep pace with this if the potential
of these data is to be maximised. Early analysis was based on aural as-
sessments or visual inspection of spectrograms to score characteristics
such as the frequency of occurrence and diversity of acoustic events like
sh vocalisations (Putland et al., 2017; Archer et al., 2018; McWilliam
et al., 2017, 2018; Bertucci et al., 2020; Lamont et al., 2021). However,
these approaches can be slow and labour intensive, introducing a severe
limit on the speed at which PAM data can be analysed.
Computationally generated ecoacoustic indices are becoming a
popular approach to overcome this limitation (Gibb et al., 2019).
Ecoacoustic indices have primarily been developed for terrestrial habi-
tats (Sueur et al., 2014) where they are used to quantify soundscape
attributes including variability across time and/or frequency bands
(Stowell and Sueur, 2020). These indices can be automatically derived
from long-term acoustic recordings, enabling extended recordings to be
analysed. Several indices have been tested recently in the marine envi-
ronment, revealing relationships between these and attributes of the
ecological community, habitat quality and ecological functioning of
marine habitats (Harris et al., 2016; Lindseth and Lobel, 2018; Gordon
et al., 2018; Elise et al., 2019b; Mooney et al., 2020).
These studies have primarily used individual ecoacoustic indices to
test for differences between sites (e.g., low- and high-quality habitats,
reefs pre- and post-bleaching on reefs), or relationships with other
ecological metrics (e.g., biodiversity, abundance etc) so far. However,
these indices do not perform consistently across all marine in-
vestigations (Kaplan et al., 2015; Bertucci et al., 2016; Dimoff et al.,
2021). Results from any individual index can also be biased by
individual components of the soundscape, such as a high density of
snapping shrimps or repetitive sh chorusing, limiting their utility to
characterise the wider community (Staaterman et al., 2013; Bolgan
et al., 2018; Dimoff et al., 2021).
Some terrestrial soundscape ecology investigations have attempted
to overcome similar performance issues through combining several
ecoacoustic indices to generate multivariate representations of acoustic
recordings known as ‘compound indices’ to generate a more holistic
representation of the soundscape (Eldridge et al., 2018). These com-
pound indices can then be input into machine learning algorithms which
are able to identify relationships between this ‘feature set’ of indices and
the task asked of the algorithm through nding patterns and interactions
between the indices. (Eldridge et al., 2018; Bradfer-Lawrence et al.,
2019; Gibb et al., 2019; Sethi et al., 2020).
Such tasks include supervised approaches such as classication
problems which group soundscape recordings into categories specied
by the researchers (e.g logged or unlogged forest) or regression tasks
which can place recordings along a gradient (e.g avian diversity) (Sethi
et al,. 2020, 2021). Alternatively, unsupervised approaches can be used
such as clustering which can be used to identify key groups from re-
cordings, or, anomaly detection can be used to identify recordings which
signicantly deviate from the majority (e.g containing noise pollution or
a rare animal chorus) (Sethi et al., 2020).
In this study, we use whole soundscape recordings to test whether a
machine learning algorithm could be trained using a compound index to
accurately classify recordings from two different ecostates of a marine
habitat. We used coral reef soundscapes due to the diversity of sounds
present on these ecosystems and known relationships between their
soundscape and ecological attributes (Kaplan et al., 2015; McWilliam
et al., 2017; Nedelec et al., 2015). We then trialled this model in a real
world application using recordings from neighbouring reefs which had
been actively restored by one of the world’s largest reef restoration
programmes, located in Sulawesi, Indonesia. Coral restoration pro-
grammes typically fail to collect adequate monitoring data (Bayraktarov
et al., 2019; Bostr¨
om-Einarsson et al., 2020). Through demonstrating the
utility of this approach, we intend to highlight a more efcient and cost-
effective means of eld data collection for the monitoring of restoration
and other marine conservation projects alike.
2. Methods
2.1. Study sites
Recordings were made at sites in the Spermonde Archipelago (South
Sulawesi, Central Indonesia; 4◦56.9′S, 119◦18.1′E; Fig. 1C) around
Pulau Badi (Fig. 1A) and Pulau Bontosua (Fig. 1B). In 2006, the Mars
assisted reef restoration system (MARRS) (buildingcoral.com) was
developed to provide a novel methodology to re-establish coral cover at
sites that have been degraded to rubble elds by historical coral mining
and dynamite shing (Williams et al., 2019). Coral fragments are
attached to ‘reef stars’, interlinked metal frames, which provide a stable
substrate. This prevents the smothering of new coral recruits through the
turn over of rubble by waves and tides which typically slows the re-
covery process to a decadal timescale or longer (Fox et al., 2003). Be-
tween 2013 and 2017 this practice increased coral cover from
approximately 10% to 60% across 7,000 m
2
of reef (Williams et al.,
2019). Recordings were made at seven sites which encompassed four
distinct types of reef habitat: healthy (Badi & Bontosua), degraded
(Bontosua & Salisi), mature restored (Badi & Bontosua) and newly
restored (Salisi). Coral cover was assessed to provide a measure of reef
health (methodology in Supp. 1). The two healthy sites exhibited
naturally high coral cover (Badi: 91.2% ±2.0; Bontosua: 93.1% ±2.6;
mean ±SE) whereas the degraded sites exhibited low coral cover (Salisi:
2.1% ±0.9; Bontosua: 17.6% ±4.6) (Fig. 2). The two mature restored
sites were established >24 months previously and exhibited increased
coral cover (Badi: 79.1% ±3.9; Bontosua: 66.5% ±3.8) compared to
B. Williams et al.

Ecological Indicators 140 (2022) 108986
3
the newly restored site (25.6% ±2.6), which was established <12
months previously.
2.2. Acoustic recordings
The recordings were made across the seven sites during
August–September 2018 as part of the MARRS monitoring programme.
We used a regime which sampled one hour blocks from sites ve days
either side of the full moon (26th August 2018) and three days either
side of the following new moon (10th September 2018) during daylight
(09:00–15:00), twilight (half an hour either side of sunrise and sunset)
and night time (half an hour either side of midnight) periods. Three
SoundTraps (SoundTrap 300STD, 48 kHz sampling rate, Ocean
Instruments, Auckland, NZ) were used, with one suspended 0.5 m above
the seabed for each deployment at a site. In each new round of de-
ployments, SoundTraps were assigned randomly to recording sites
within a counterbalanced blocking design, in order to control for po-
tential instrument error.
We sub-sampled ve non-overlapping one-minute segments from
each of the hour-long periods at random. Only samples recorded during
calm conditions (wind speed <20 km h
−1
) were used. These samples
were also screened for motorboat noise and any recordings with this
disturbance were removed, resulting in 262 recordings in the nal
sample set. This sample set was previously used in Lamont et al. (2021)
to compare sh sound diversity between sites.
Fig. 1. Location and habitat class of the seven reef sites, present within the broader Spermonde Archipelago, Indonesia (A) where soundscape recordings were
collected. Fringing reefs from two nearby islands: Bontosua (B) and Badi (C) were used. Modied from Lamont et al. (2021).
Fig. 2. Representative habitat and coral cover images from the four habitat classes at which soundscape recordings were made. (A) Degraded, (B) healthy, (C) newly
restored and (D) mature restored.
B. Williams et al.

Ecological Indicators 140 (2022) 108986
4
2.3. Processing recordings
Each recording was ltered using a short-term Fourier transform
band-pass lter into three frequency bands: low-frequency (0.05–0.8
kHz), medium-frequency (2–7 kHz) and a broadband (0.05–20 kHz).
The low-frequency band covered all the known sh vocalisations within
the dataset (Lamont et al., 2021), while the medium-frequency band
comprised invertebrate (primarily snapping shrimp) sound (Elise et al.,
2019a). The broad-frequency band was used to encompass the full
spectrum of potentially relevant frequencies, as previously used in coral
reef soundscape investigations (Kaplan et al., 2015; Lyon, 2018). Fre-
quencies below 0.05 kHz were excluded from the low- and broad-
frequency band recordings to remove geophonic noise and self-noise
from the recording system (Curtis et al., 1999). All processing was
performed in R (v3.4.2. R Development Core Team, 2020): audio les
were read and written using tuneR (v.1.3.3) (Ligges et al., 2018) and the
lters were implemented using Seewave (v2.1.6) (Sueur et al., 2008).
2.4. Ecoacoustic indices
Twelve ecoacoustic indices were chosen from a range of previous
soundscape studies (Table 1). Each index was calculated for all three
frequency bands, with two exceptions: Snap rate was only calculated for
the middle- and broad-frequency bands, because snapping shrimp
cavitation bubbles are not audible at lower frequencies (Bohnenstiehl
et al., 2016), and the normalised difference soundscape index (NDSI)
was only calculated for the broad-band recordings. NDSI is typically
used to quantify discrepancies in amplitude between an anthropogenic
noise band up to 1 kHz and a biophonic noise band at selected higher
frequencies (Kasten et al., 2012). We instead used this index to quantify
differences in the 1 kHz band where sh noise dominates, and, a higher
2–7 kHz frequency band where snapping shrimp sound dominates (Au
and Banks, 1998). Thus, we established a feature set of 33 index values
across 12 indices and three frequency bands for each of the 262 one-
minute recordings. All indices were calculated using the R package
Seewave (Sueur et al., 2008) where possible, all remaining indices were
calculated in Soundecology R package (v.1.3.3) (Villanueva-Rivera et al.,
2018) other than SPL which was calculated in paPAM (Nedelec et al.,
2016) and snap rate which calculated using a custom MATLAB script
(Gordon et al., 2018).
2.5. Selection of indices to differentiate healthy and degraded habitats
All 33 indices were examined for separation between recordings
from healthy (n =81) and degraded (n =71) habitats using Mann-
Whitney U tests. Since the purpose was to explore the potential of
each candidate index for model development, we did not control for
spatial and temporal pseudoreplication within the library of recordings,
nor control for cross-correlation between indices with similar trends.
Violin plots were used to visualise the degree of overlap between the
distributions for indices with indicative differences. Where minimal
overlap was observed between the two habitats, the index could be
considered likely to provide a promising measure with which to differ-
entiate between healthy and degraded habitats.
2.6. Machine-learning approaches to develop a compound index
Following analysis of individual indices, we developed a supervised
machine-learning model to assign recordings to either healthy or
degraded habitat classes. A regularised discriminant analysis (RDA) al-
gorithm was selected to account for the high level of collinearity re-
ported between indices (Supp. 1). An optimised set of indices was
selected in a ‘feature selection’ stage, using recursive feature elimination
(RFE) and a multivariate adaptive regression spline (MAR) (Kuhn and
Johnson, 2019) (Supp. 1). The RFE highlighted increases in model ac-
curacy with the multi-index approach as additional indices were added
sequentially (Supp. 1, Fig. S1). Predictive accuracy was greatest with
eight indices, followed by a gradual decline as the addition of further
indices introduced noise and/or caused model overtraining. The list of
suggested features from RFE included the following index/frequency
band combinations: broadband ACI, H, NDSI and H
t
; and medium-
frequency band ACI, BI, H and H
t
. This was highly congruent with
rankings obtained from the relative importance scores using the MAR
(Fig. 3).
Following RFE, further manual feature selection was conducted by
systematic removal and addition of indices whilst executing the full
model, to select a nal feature set with the lowest misclassication rate.
This led to discarding H
t
in both the broad-range and middle-frequency
bands, and the introduction of low-frequency band ACI and middle-
frequency band AR. Thus, the nal set was: low-frequency band ACI,
medium-frequency band ACI, AR and BI, broadband ACI, H and NDSI.
Feature selection was performed using the R packages mlbench (v2.1.1)
(Leisch and Dimitriadou, 2010) and Caret (v.6.0–86) (Kuhn, 2020).
2.7. Constructing the nal model
An RDA model was constructed using the healthy and degraded
datasets, using the R packages MASS (v.7.3-53) (Venables and Ripley,
Table 1
Twelve ecoacoustic indices calculated from recordings with summary description of the mechanistic principle, software used and respective settings employed.
Index Mechanism Software Settings Origin
Acoustic complexity index
(ACI)
Measures variability in intensity of frequencies across
time
Seewave in R Window size =512; type =Hamming; overlap =
0
(Pieretti, 2011)
Acoustic entropy (H) Measures randomness across temporal and spectral
domains
Seewave in R Window size =512; envelope =Hilbert (Sueur, 2008)
Acoustic eveness index
(AEI)
Measures diversity across frequency bands Soundecology in
R
Max freq =upper bound of band in use; freq step
=max freq/10; threshold =-50 dB
(Villanueva-Rivera,
2011)
Amplitude index (M) Measures median of amplitude envelope Seewave in R Envelope =Hilbert (Sueur, 2008)
Acoustic richness (AR) Ranks recordings based on amplitude multiplied by
randomness across the temporal domain
Seewave in R Envelope =Hilbert (Depraetere, 2012)
Bioacoustic index (BI) Measures cumulative intensity across frequency
bands
Soundecology in
R
Min and max frequency matched to band in use;
window size =512
(Boelman, 2007)
Normalised mean
difference index (NDSI)
Measures amplitude difference between two selected
frequency bands
Seewave in R Min and max frequency matched to band in use;
window size =512
(Kasten, 2012)
Number of peaks Number of major frequency peaks obtained from a
mean spectrum
Seewave in R Window size =512; type =Hanning; overlap =0 (Sueur, 2008)
Spectral entropy (Ht) Measures randomness across the frequency domain Seewave in R No settings required (Sueur, 2008)
Temporal entropy (Hf) Measures randomness across the temporal domain Seewave in R No settings required (Sueur, 2008)
Snap rate Measures rate of snapping shrimp snaps MATLAB Custom script from Gordon et al., (2018) Widely used
Sound pressure level (SPL) Calibrated measure of root mean squared sound
pressure level
paPAM in
MATLAB
Window length =1024; type =Hamming;
Overlap =50%
Widely used
B. Williams et al.

Ecological Indicators 140 (2022) 108986
5
200) and KlaR (v.0.6-15) (Weihs et al., 2005). Model accuracy was
assessed using k-fold cross validation (10 folds), whereby the dataset
was partitioned into ‘training’ and ‘test’ sets to prevent overestimation
of model accuracy when presented with new data (Supp. 1). To account
for variation in the RDA, 1000 repeats of the cross-validated model
construction were performed to assess accuracy (Rao et al. 2008).
A principal component analysis (PCA) and a pairs plot comparing
each combination of the eight selected indices were generated for all
recordings. These were used to test whether soundscape properties from
restored reefs diverged from the healthy and degraded classes, which
would lead to potentially inappropriate classications using the RDA
trained on the healthy and degraded recordings. Both tests were con-
ducted in R using inbuilt functions (Supp. 1, Fig. S4).
3. Results
3.1. Comparing indices between healthy and degraded sites
Exploratory Mann-Whitney U tests revealed signicant differences
between healthy and degraded habitat index scores for 12 of the 33
indices (Fig. 4). Bonferroni corrections were also used to reduce the
likelihood of false positives in the search for a signicant difference; the
original alpha value of 0.05 was therefore divided by 33 to provide a
new value of 0.00152. Using this more conservative approach no longer
reveals AI in the broad and medium-frequency bands, as well as AEI in
the medium-frequency band, to be signicantly different.
Violin plots of the three most signicantly different index results
Fig. 3. Relative importance rankings of indices obtained from the multivariate adaptive regression (MAR) analysis used for feature selection. The eight recom-
mendations obtained from the recursive feature elimination (RFE) analysis are indicated by the black lines. The top eight indices of the MAR analysis were congruent
with the eight recommendations from RFE, though the order was not conserved. Black dots to the right of bars indicate features which were selected for the nal
model after further manual feature selection.
B. Williams et al.

Ecological Indicators 140 (2022) 108986
6
between the healthy and degraded sites revealed large zones of overlap
between values for these indices between habitat classes (Fig. 5). The
strongest signicant difference was reported for H in the 2–7 kHz band,
here 71 of the 152 (47 %) of recordings reported results that did fall in
range of both habitat types.
3.2. Comparing indices to phonic richness
We searched for a correlation between each ecoacoustic index and
the diversity of sh sounds present in each recording using the ‘phonic
richness’ method which counts the number of unique sound types
audible in each recording (Lamont et al., 2021) (Supp. 1). This revealed
no strong relationships (Pearson correlations) between phonic richness
and any of the 33 indices trialled (Supp. 1, Fig. S3). The strongest
relationship was a weak negative correlation with the acoustic entropy
index (H) for the broad-frequency band (Pearson correlation; rho =
-0.43; p <0.001), with all other indices reporting weaker correlations
than this.
3.3. Regularised discriminant analysis
From the 1000 repeated constructions of the cross-validated model
using the 152 recordings taken across healthy and degraded sites, the
pooled mean misclassication rate was 8.27% (0.84, mean SE). Across
these model constructions, of the 81 recording samples taken from the
two healthy sites, 73.0 (0.1) of these were correctly classied as healthy,
with 8.0 (0.1) misclassied as degraded. Of the 71 recordings taken from
the two degraded sites, 67.2 (0.1) of these were classied as degraded,
with 3.7 (0.1) misclassied as healthy (see Fig. 6 for individual results
for each recording sample).
Cluster analysis using the principal component analysis (PCA; Fig. 7)
and a pairs plot (Supp. 1, Fig. S4) were used to examine whether the 110
samples taken from recordings of the three restored sites overlapped
with recordings from the control sites. If they deviated then it would be
likely that the model would provide inappropriate classications to
these sites. For the mature restored and newly restored sites, 70/81 and
70/71 samples respectively fell within one or both of the predictive el-
lipses for the two existing classes. This indicates that the soundscapes of
the restored sites did not diverge from the soundscape present at the
other two habitat types when using the properties investigated here.
This supports the inputting of restored samples into the model as the
data were not divergent from the original training data. Additionally,
the PCA showed that 61/81 samples from the mature restored sites fell
within the ellipse that could be used to predict healthy sites, whereas
24/27 samples of recordings from the newly restored site fell within the
ellipse that could be used to predict degraded sites. However, it is
important to note that there was a large region of overlap between the
healthy and degraded class when using only the two dimensions shown
by the PCA, with most of the ellipse of the degraded classes encompassed
by that of the healthy class.
Analysis of the restored site samples revealed that the majority of
samples from mature restored sites were classied as healthy, but
Fig. 4. Heat map displaying results from Mann-Whitney U tests between the ecoacoustic index scores calculated from recordings of healthy (n =81) and degraded (n
=71) sites in low-, medium- and broad-frequency bands. The habitat class with the higher mean is indicated by the letter in the bottom right corner of each cell (H =
Healthy; D =Degraded). Blank cells indicate indices for which values from the corresponding frequency band were not calculated (see Methods).
Fig. 5. Violin plots of the three indices with the most signicant differences between healthy (n =81) and degraded (n =71) habitat. (A) Medium-frequency band
Entropy Index (H) (Mann-Whitney U; U =1.98, p <0.001), (B) Broad-frequency band Acoustic Complexity index (ACI) (U =1.78, p <0.001), (C) Medium-frequency
band Temporal Entropy (H
t
) (U =1.63, p <0.001).
B. Williams et al.

Ecological Indicators 140 (2022) 108986
7
samples from the newly restored site were mainly classied as degraded
(Fig. 8). The Bontosua mature restored site was classied more clearly
than the Badi mature restored site, with 37/38 and 33/39 samples
classied as healthy respectively. The six samples classied as degraded
from the Badi mature restored site occurred consecutively on the new
moon at night. At the newly restored site, 27/33 samples were classied
as degraded, and all of these were during the full moon (though only
four new moon samples were available) and ve of these were at night.
The model trained on the 2018 recordings was also tested on a
smaller number of recordings taken at the same sites 10 months later
(June/July 2019). Here, the model provided similar predictions for six
of the seven sites; while one site (Healthy Bontosua) exhibited a change
in prediction between 2018 and 2019, changing from primarily being
classied as healthy to degraded (Table 2, full results in Fig. S5).
4. Discussion
This study compared the value of individual ecoacoustic indices and
a machine-learning model trained on a compound index to discriminate
between coral reef ecostates. While no single ecoacoustic index could
reliably discriminate between healthy and degraded reefs, a supervised
machine-learning approach more accurately predicted habitat class
from randomly drawn acoustic samples. This highlights the potential of
combining PAM with machine learning for monitoring the health of
marine ecosystems.
Twelve individual ecoacoustic indices were calculated in up to three
frequency bandwidths, totalling 33 values, of which 12 indices were
signicantly different between healthy and degraded reefs (Fig. 4).
There were no strong correlations between any of these indices and
Fig. 6. Machine learning classication of
acoustic samples from the healthy and
degraded sites. Each cell indicates a single
one-minute recording from the 152 taken in
healthy and degraded habitats. The model
was executed 1000 times on the dataset,
generating a new habitat class prediction
each time for every recording. Values within
cells represent the proportion of these 1000
iterations in which the recording was pre-
dicted as originating from a healthy site,
with the remaining being predicted as
degraded (green shading: >0.5; pink
shading: <0.5). Recordings on the left of the
partition were made during the day and re-
cordings to the right were made during
crepuscular or nighttime periods. Although frequent gaps were present in the sampling regime, the order with which cells are presented within their respective blocks
conserves the overall order with which they were sampled across time. (For interpretation of the references to colour in this gure legend, the reader is referred to the
web version of this article.)
Fig. 7. Plot from the principal component analysis of PC1 and PC2 scores for the Healthy and Degraded site recording samples. Samples from recordings of Restored
sites are overlaid on this to help determine whether these conform with either of the two existing classes or whether the properties of their soundscape are distinct.
Ellipses indicate the zone within which a new sample can be assigned to a class using the two principal components presented in this gure. Overlapping areas
indicate ambiguous recordings which cannot be differentiated by PCA.
B. Williams et al.

Ecological Indicators 140 (2022) 108986
8
phonic richness (Supp. 1), indicating that sh sound diversity alone is
not the dominant driver of these results; rather alternative aspects of the
soundscape are responsible. A metric that combines abundance along-
side diversity of sh vocalisations may reveal more about the role these
play in driving index values. Other contributors may originate from
alternative biotic or abiotic sources. Invertebrates are well documented
contributors to reef soundscapes including snapping shrimp (Bohnen-
stiehl et al., 2016), urchin feeding activities (Radford et al., 2008) and
the movement of hard shelled organisms (Freeman et al., 2014). The
photosynthetic process of macro algae has also been reported as a sound
producer (Freeman et al., 2018). Though understudied, abiotic attri-
butes may also inuence the soundscape as ecostates change (Duarte
et al. 2021). Though recordings were taken in calm conditions, low level
geophonic noise produced by waves and wind may propagate differently
through a rubble eld compared to a more structurally complex reef, or
the regular movements of unconsolidated coral rubble due to hydro-
dynamic forcing might also contribute to the soundscape (Kenyon et al.
2020).
The distribution of values within indices showing signicant differ-
ences between healthy and degraded reefs exhibited substantial overlap
between the two habitat classes. The best performing individual index
was H in the 2–7 kHz band. For this index, 71 of the 152 (47 %) of re-
cordings had non-overlapping values, meaning they could not be
correctly classied as healthy or degraded, all other recordings were
ambiguous as the values fell within the range reported for both habitat
types. This means that the ability to distinguish between habitat classes
from a single recording using an individual index is low, as any given
value from one class has a high chance of being reported from a
recording in the other class. Violin plots of the three most signicant
results demonstrate this large overlap between the values of each class
(Fig. 5). In isolation, these indices could discriminate between habitats if
extensive sampling is achievable for all sites of interest to build up a
dataset that can be tested for statistical signicance, as demonstrated
here. However, their potential to deliver reliable results from short
‘snapshot’ recordings is lower and their ability to deliver insights into
more complex tasks, beyond a coarse healthy-degraded comparison,
may be limited.
In contrast to bivariate analyses, combining multiple indices with
regularised discriminant analysis (RDA) gave a strong predictive ability
to classify habitats based on single recordings. Recursive feature elimi-
nation (RFE) highlights the increase in accuracy attainable through
constructing an optimised set of multiple indices (Supp. 1, Fig. S1)
compared to using individual indices (Fig. 5). The misclassication rate
of the nal RDA model was 8.27% (0.84, mean SE) when applied to
recordings from the same season. The model made accurate predictions
despite being kept blind to diel and lunar period, which are both known
to inuence marine soundscapes (Staaterman et al., 2014); this high-
lights its robustness to temporal changes in soundscapes. The model also
reliably delivered the same classication for recordings from six of the
seven sites taken ten months later. The feature-selection stage of this
approach is specic to the data and questions considered in this study.
However, indices within the nal feature set may offer a useful starting
place for similar investigations elsewhere. To produce optimised
models, investigations at new locations addressing new questions should
carry out an independent feature selection process on their own data.
Following the successful classication of healthy and degraded
habitats, our compound index based model was used to examine
soundscape recordings taken from nearby coral reef habitats that had
been restored (Williams et al., 2019). This tested the ability of this
approach to perform a rapid assessment of these restored sites using one-
minute soundscape recordings. The model was able to detect differences
between the two mature restored sites and the newly restored site. Of the
recording samples from the two mature restored sites, 33/39 and 37/38
were primarily classied as healthy, whereas 27/33 samples from the
newly restored sites were classied as degraded (Fig. 8). The mature
restored sites were more than twice as old as the newly restored site
(restoration started >24 months prior to recordings on mature restored
sites, compared to <12 months for the newly restored site), and had
approximately three times more live coral cover (79.1% ±3.9 and
66.5% ±3.8 for the mature restored sites, 25.6% ±2.6 for the newly
Fig. 8. Machine-learning classication of acoustic samples from the restored sites. Each cell indicates a single one-minute recording from the 110 taken from restored
sites. The model was executed 1000 times on the dataset, generating a new habitat class prediction each time for every recording. Values within cells represent the
proportion of these 1000 iterations in which the recording was predicted as originating from a healthy site, with the remaining being predicated as degraded (green
shading: >0.5; pink shading: <0.5). Recordings on the left of the partition were made during the day and recordings to the right were made during crepuscular or
nighttime periods. Despite gaps in the sampling regime, the order within blocks conserves the overall order with which they were sampled across time. (For
interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)
Table 2
Results from the application of the 2018 model when tested on recordings taken at the same sites in 2019.
Badi
Healthy
Bontosua
Healthy
Salisi
Degraded
Bontosua
Degraded
Badi Mature
Restored
Bontosua Mature
Restored
Salisi Newly
Restored
Recordings classied as
Healthy
9/9 2/12 0/5 5/12 9/9 12/12 8/9
Proportion classied as
Healthy
1.0 0.17 0 0.42 1.0 1.0 0.89
B. Williams et al.

Ecological Indicators 140 (2022) 108986
9
restored site; values all % live coral cover mean ±SE; full data in Supp.
1). Restoration progress was clearly detected in the soundscape, with
better classication made possible by using a machine-learning-driven
approach, suggesting PAM can be a useful tool for monitoring restora-
tion against reference sites. More generally, this study highlights the
potential for using machine-learning approaches to explore PAM data to
provide greater analytical power in coral reef monitoring programmes.
Further improvements include considering sources of observed error
in the model, which could be due to several factors working in isolation
or in combination. The RDA approach operates best when input features
have Gaussian distributions (Wu et al., 1996), but some features used in
this study exhibited sub-Gaussian distributions. This is likely due to
inclusion of samples from various times of day and from multiple sites.
Diel trends are frequently detected in reef soundscapes across a range of
ecoacoustic indices (Kaplan et al., 2015; Bertucci et al., 2020b; Carriço
et al., 2020). Additionally, reef soundscapes can differ over small spatial
scales (Putland et al., 2017). In this study, samples were taken from
spatially separated sites to reduce pseudoreplication, thus differences
within habitat classes are likely. Both these factors may have skewed the
distributions of the feature sets. Furthermore, the dataset used to train
the model will also contain natural outliers through ecological
randomness that cannot be resolved at the sampling resolution
employed. Longer periods of recording can be used to minimise impacts
of this natural variation (Bradfer-Lawrence et al., 2019), though we
show here that short periods of recording can still be used to identify
accurate classications between signicantly different habitat states.
Six of the seven sites studied in 2018 retained similar classications
when resampled 10 months later in 2019. The outlier was Bontosua
Healthy, for which 10/12 recordings were incorrectly classied as
degraded. Recordings at this site were only collected during the day in
2019, and 9/12 of these were taken during the new moon period. The
soundscape may therefore have been inadequately sampled. Alterna-
tively, this could be an early indicator of a changing state of reef health
at this site, not yet seen in the coral cover data, which was similar in both
years (Supp. 1), or, this could demonstrate the specicity of this model
to the time it was taken.
Future investigations could build on the present study by considering
a more nuanced approach to classifying ecostate. For example, this study
employed a binary classication of reef health but, in reality, marine
habitats occur across gradients of ecostates (Downs et al., 2005; Smith
et al., 2008). Sampling across these gradients, and using regression-
based algorithms such as logistic regression, random forests or neural
networks could support models that can predict on a continuous scale.
Additionally, within the context of coral reefs, although live coral cover
may be a strong indicator of overall reef health (Smith et al., 2016;
Dietzel et al., 2020), other attributes of interest could be incorporated to
better determine the ecostate of a site. For example, soundscape-based
machine-learning models could be trained to predict metrics which
are effort and training-intensive, such as sh or invertebrate abundance
or diversity, and other habitat attributes could be explored such as
structural complexity or ecosystem stability. Similar approaches could
also be applied to other kinds of marine habitats where soundscape
research has so far been limited (Pieretti and Danovaro, 2020). By
drawing comparisons with a wider range of traditional metrics used in
marine monitoring, the potential for machine-learning-based analyses of
ecoacoustic recordings can be further developed.
5. Conclusion
Given the increasing availability of hydrophone technology (Chapuis
et al., 2021; Lamont et al., 2022), acoustic datasets from the marine
environment are set to rise.
Automated analyse is needed to efciently process and analyse large
acoustic datasets so that insights from the information held within can
be maximised. However, this is so far underdeveloped. Our investigation
presents an automated approach, through the use of a compound index
and machine-learning, that improves upon existing approaches used in
the marine environment. We rst demonstrate the use of this to classify
coral reef habitats into healthy or degraded ecostates based on short-
term recordings. We then demonstrate this in an applied setting, high-
lighting the utility of this approach when assessing areas of restored reef,
revealing that restoration progress is detectable in the soundscape. This
investigation provides the rst evidence that compound indices and
machine learning are able to outperform the use of single ecoacoustic
indices on a tropical reef and that this approach should be considered for
use in other marine and terrestrial habitats applications.
Funding
This work was funded by a Natural Environment Research Coun-
cil–Australian Institute of Marine Science CASE GW4+Studentship NE/
L002434/1 (to Timothy A.C. Lamont); Swiss National Science Founda-
tion Early Postdoc Mobility fellowship P2SKP3–181384 (to Lucille
Chapuis); a University of Exeter Education Incubator Research-Inspired
Learning grant (to Timothy A.C. Lamont, Lucille Chapuis and Stephen D.
Simpson); the University of Exeter Global Challenges Research Fund; a
Natural Environment Research Council Research Grant NE/P001572/1
(to Stephen D. Simpson and Andrew N. Radford); and MARS Sustainable
Solutions, part of Mars, Inc.
CRediT authorship contribution statement
Ben Williams: Conceptualization, data curation, formal analysis,
visualization writing (original draft). Timothy A.C. Lamont: Investi-
gation, data curation, writing (review & editing), funding acquisition.
Lucille Chapuis: Investigation, writing (review & editing), funding
acquisition. Harry R. Harding: Investigation, writing (review & edit-
ing). Eleanor B. May: Investigation. Mochyudho E. Prasetya: Inves-
tigation. Marie J. Seraphim: Investigation. Noel Janetski: project
administration, resources. Jamaluddin Jompa: project administration,
resources. Dave Smith: project administration, resources. Andrew N.
Radford: writing (review & editing), supervision. Stephen D. Simpson:
writing (review & editing), funding acquisition and supervision.
Declaration of Competing Interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Acknowledgments
Data were collected with assistance from the Mars Coral Reef
Restoration Project monitoring programme, in collaboration with
Hasanuddin University; we thank Lily Damayanti, Saipul Rapi, Alicia
McArdle, Freda Nicholson, Jos van Oostrum and Frank Mars for their
advice and logistical support. We thank the Department of Marine Af-
fairs and Fisheries of the Province of South Sulawesi; the Government
Ofces of the Kabupaten of Pangkep, Pulau Bontosua and Pulau Badi;
and the communities of Pulau Bontosua and Pulau Badi for their
support.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.ecolind.2022.108986.
References
Archer, S.K., Halliday, W.D., Riera, A., Mouy, X., Pine, M.K., Chu, J.W.F., Dunham, A.,
Juanes, F., 2018. First description of a glass sponge reef soundscape reveals sh calls
and elevated sound pressure levels. Mar. Ecol. Prog. Ser. 595, 245–252. https://doi.
org/10.3354/meps12572.
B. Williams et al.

Ecological Indicators 140 (2022) 108986
10
Au, W.W.L., Banks, K., 1998. The acoustics of the snapping shrimp Synalpheus
parneomeris in Kaneohe Bay. J. Acoust. Soc. Am. 103, 41–47. https://doi.org/
10.1121/1.423234.
Bayraktarov, E., Stewart-Sinclair, P.J., Brisbane, S., Bostr¨
om-Einarsson, L., Saunders, M.
I., Lovelock, C.E., Possingham, H.P., Mumby, P.J., Wilson, K.A., 2019. Motivations,
success, and cost of coral reef restoration. Restor. Ecol. 27, 981–991. https://doi.
org/10.1111/rec.12977.
Bertucci, F., Guerra, A.S., Sturny, V., Blin, E., Sang, G.T., Lecchini, D., 2020a.
A preliminary acoustic evaluation of three sites in the lagoon of Bora Bora, French
Polynesia. Environ. Biol. Fishes 103 (8), 891–902. https://doi.org/10.1007/s10641-
020-01000-8.
Bertucci, F., Maratrat, K., Berthe, C., Besson, M., Guerra, A.S., Raick, X., Lerouvreur, F.,
Lecchini, D., Parmentier, E., 2020b. Local sonic activity reveals potential
partitioning in a coral reef sh community. Oecologia 193, 125–134. https://doi.
org/10.1007/s00442-020-04647-3.
Bertucci, F., Parmentier, E., Lecellier, G., Hawkins, A.D., Lecchini, D., 2016a. Acoustic
indices provide information on the status of coral reefs: an example from Moorea
Island in the South Pacic. Sci. Rep. 6, 1–9. https://doi.org/10.1038/srep33326.
Bertucci, F., Parmentier, E., Lecellier, G., Hawkins, A.D., Lecchini, D., 2016b. Acoustic
indices provide information on the status of coral reefs: an example from Moorea
Island in the South Pacic. Sci. Rep. 6, 33326. https://doi.org/10.1038/srep33326.
Bohnenstiehl, D.R., Lillis, A., Eggleston, D.B., Radford, C.A., 2016. The curious acoustic
behavior of estuarine snapping shrimp: temporal patterns of snapping shrimp sound
in sub-tidal oyster reef habitat. PLoS One 11 (1), e0143691. https://doi.org/
10.1371/journal.pone.0143691.
Bolgan, M., Amorim, M.C.P., Fonseca, P.J., Di Iorio, L., Parmentier, E., 2018. Acoustic
complexity of vocal sh communities: a eld and controlled validation. Sci. Rep. 8,
1–11. https://doi.org/10.1038/s41598-018-28771-6.
Bostr¨
om-Einarsson, L., Babcock, R.C., Bayraktarov, E., Ceccarelli, D., Cook, N., Ferse, S.C.
A., Hancock, B., Harrison, P., Hein, M., Shaver, E., Smith, A., Suggett, D., Stewart-
Sinclair, P.J., Vardi, T., McLeod, I.M., Bianchi, C.N., 2020. Coral restoration – a
systematic review of current methods, successes, failures and future directions. PLoS
One 15 (1), e0226631. https://doi.org/10.1371/journal.pone.0226631.
Bradfer-Lawrence, T., Gardner, N., Bunnefeld, L., Bunnefeld, N., Willis, S.G., Dent, D.H.,
Zamora-Gutierrez, V., 2019. Guidelines for the use of acoustic indices in
environmental research. Methods Ecol. Evol. 10 (10), 1796–1807. https://doi.org/
10.1111/2041-210X.13254.
Butler, J., Stanley, J.A., Butler, M.J., 2016. Underwater soundscapes in near-shore
tropical habitats and the effects of environmental degradation and habitat
restoration. J. Exp. Mar. Bio. Ecol. 479, 89–96. https://doi.org/10.1016/j.
jembe.2016.03.006.
Carriço, R., Silva, M.A., Vieira, M., Afonso, P., Menezes, G.M., Fonseca, P.J., Amorim, M.
C.P., 2020. The use of soundscapes to monitor sh communities: meaningful
graphical representations differ with acoustic environment. Acoustics 2, 382–398.
https://doi.org/10.3390/acoustics2020022.
Curtis, K.R., Howe, B.M., Mercer, J.A., 1999. Low-frequency ambient sound in the North
Pacic: long time series observations. J. Acoust. Soc. Am. 106, 3189–3200. https://
doi.org/10.1121/1.428173.
Dietzel, A., Bode, M., Connolly, S.R., Hughes, T.P., 2020. Long-term shifts in the colony
size structure of coral populations along the Great Barrier Reef. doi: 10.1098/
rspb.2020.1432.
Dimoff, S.A., Halliday, W.D., Pine, M.K., Tietjen, K.L., Juanes, F., Baum, J.K., 2021. The
utility of different acoustic indicators to describe biological sounds of a coral reef
soundscape. Ecol. Indic. 124, 107435 https://doi.org/10.1016/j.
ecolind.2021.107435.
Downs, C.A., Woodley, C.M., Richmond, R.H., Lanning, L.L., Owen, R., 2005. Shifting the
paradigm of coral-reef “health” assessment. Mar. Pollut. Bull. 51, 486–494. https://
doi.org/10.1016/j.marpolbul.2005.06.028.
Duarte, C.M., Chapuis, L., Collin, S.P., Costa, D.P., Devassy, R.P., Eguiluz, V.M., Erbe, C.,
Gordon, T.A.C., Halpern, B.S., Harding, H.R., Havlik, M.N., Meekan, M.,
Merchant, N.D., Miksis-Olds, J.L., Parsons, M., Predragovic, M., Radford, A.N.,
Radford, C.A., Simpson, S.D., Slabbekoorn, H., Staaterman, E., Van Opzeeland, I.C.,
Winderen, J., Zhang, X., Juanes, F., 2021. The soundscape of the Anthropocene
ocean. Sci 371 (6529). https://doi.org/10.1126/science.aba4658.
Eldridge, A., Guyot, P., Moscoso, P., Johnston, A., Eyre-Walker, Y., Peck, M., 2018.
Sounding out ecoacoustic metrics: Avian species richness is predicted by acoustic
indices in temperate but not tropical habitats. Ecol. Indic. 95, 939–952. https://doi.
org/10.1016/J.ECOLIND.2018.06.012.
Elise, S., Bailly, A., Urbina-Barreto, I., Mou-Tham, G., Chiroleu, F., Vigliola, L.,
Robbins, W.D., Bruggemann, J.H., 2019a. An optimised passive acoustic sampling
scheme to discriminate among coral reefs’ ecological states. Ecol. Indic. 107, 105627
https://doi.org/10.1016/j.ecolind.2019.105627.
Elise, S., Urbina-Barreto, I., Pinel, R., Mahamadaly, V., Bureau, S., Penin, L.,
Adjeroud, M., Kulbicki, M., Bruggemann, J.H., 2019b. Assessing key ecosystem
functions through soundscapes: A new perspective from coral reefs. Ecol. Indic. 107,
105623 https://doi.org/10.1016/j.ecolind.2019.105623.
Fox, H.E., Pet, J.S., Dahuri, R., Caldwell, R.L., 2003. Recovery in rubble elds: long-term
impacts of blast shing. Mar. Pollut. Bull. 46, 1024–1031. https://doi.org/10.1016/
S0025-326X(03)00246-7.
Freeman, S.E., Rohwer, F.L., D’Spain, G.L., Friedlander, A.M., Gregg, A.K., Sandin, S.A.,
Buckingham, M.J., 2014. The origins of ambient biological sound from coral reef
ecosystems in the Line Islands archipelago. J. Acoust. Soc. Am. 135, 1775–1788.
https://doi.org/10.1121/1.4865922.
Freeman, L.A., Freeman, S.E.S., 2016. Rapidly obtained ecosystem indicators from coral
reef soundscapes. Mar. Ecol. Prog. Ser. 561, 69–82. https://doi.org/10.3354/
meps11938.
Freeman, S.E., Freeman, L.A., Giorli, G., Haas, A.F., Radford, C.A., 2018. Photosynthesis
by marine algae produces sound, contributing to the daytime soundscape on coral
reefs. PloS one 13 (10), e0201766. https://doi.org/10.1371/journal.pone.0201766.
Gibb, R., Browning, E., Glover-Kapfer, P., Jones, K.E., B¨
orger, L., 2019. Emerging
opportunities and challenges for passive acoustics in ecological assessment and
monitoring. Methods Ecol. Evol. 10 (2), 169–185. https://doi.org/10.1111/2041-
210X.13101.
Gordon, T.A.C., Harding, H.R., Wong, K.E., Merchant, N.D., Meekan, M.G.,
McCormick, M.I., Radford, A.N., Simpson, S.D., 2018. Habitat degradation
negatively affects auditory settlement behavior of coral reef shes. Proc. Natl. Acad.
Sci. 115, 5193–5198. https://doi.org/10.1073/pnas.1719291115.
Gordon, T.A.C., Radford, A.N., Davidson, I.K., Barnes, K., McCloskey, K., Nedelec, S.L.,
Meekan, M.G., McCormick, M.I., Simpson, S.D., 2019. Acoustic enrichment can
enhance sh community development on degraded coral reef habitat. Nat. Commun.
10, 1–7. https://doi.org/10.1038/s41467-019-13186-2.
Harris, S.A., Shears, N.T., Radford, C.A., Reynolds, J., 2016. Ecoacoustic indices as
proxies for biodiversity on temperate reefs. Methods Ecol. Evol. 7 (6), 713–724.
https://doi.org/10.1111/2041-210X.12527.
Kaplan, M., Mooney, T., Partan, J., Solow, A., 2015. Coral reef species assemblages are
associated with ambient soundscapes. Mar. Ecol. Prog. Ser. 533, 93–107. https://doi.
org/10.3354/meps11382.
Kasten, E.P., Gage, S.H., Fox, J., Joo, W., 2012. The remote environmental assessment
laboratory’s acoustic library: an archive for studying soundscape ecology. Ecol.
Inform. 12, 50–67. https://doi.org/10.1016/j.ecoinf.2012.08.001.
Kenyon, T.M., Doropoulos, C., Dove, S., Webb, G.E., Newman, S.P., Sim, C.W., Arzan, M.,
Mumby, P.J., 2020. The effects of rubble mobilisation on coral fragment survival,
partial mortality and growth. J. Exp. Mar. Biol. Ecol 533, 151467. https://doi.org/
10.1016/j.jembe.2020.151467.
Kuhn, M., Johnson, K., 2019. Feature Engineering and Selection: A Practical Approach
for Predictive Models. CRC Press.
Lamont, T.A.C., Williams, B., Chapuis, L., Prasetya, M., Seraphim, M.J., Harding, H.R.,
May, E.B., Janetski, N., Jompa, J., Smith, D.J., Radford, A.N., Simpson, S.D., 2022.
The sound of recovery: Coral reef restoration success is detectable in the soundscape.
J. Appl. Ecol 59 (3), 742–756. https://doi.org/10.1111/1365-2664.14089.
Lamont, T.A., Chapuis, L., Williams, B., Dines, S., Gridley, T., Frainer, G., Fearey, J.,
Maulana, P.B., Prasetya, M.E., Jompa, J. and Smith, D.J., 2022. HydroMoth: Testing
a prototype low-cost acoustic recorder for aquatic environments. Remote. Sens. Ecol.
Conserv. doi: 10.1002/rse2.249.
Lecchini, D., Bertucci, F., Gache, C., Khalife, A., Besson, M., Roux, N., Berthe, C.,
Singh, S., Parmentier, E., Nugues, M.M., Brooker, R.M., Dixson, D.L., H´
edouin, L.,
2018. Boat noise prevents soundscape-based habitat selection by coral planulae. Sci.
Rep. 8, 9283. https://doi.org/10.1038/s41598-018-27674-w.
Lindseth, A., Lobel, P., 2018. Underwater soundscape monitoring and sh bioacoustics: a
review. Fishes 3 (3), 36. https://doi.org/10.3390/shes3030036.
Lyon, R.P., 2018. Fish biodiversity, habitat complexity, and soundscape characteristics of
patch reefs in a tropical, back-reef nursery 42.
McWilliam, J.N., McCauley, R.D., Erbe, C., Parsons, M.J.G., 2017. Patterns of biophonic
periodicity on coral reefs in the Great Barrier Reef. Sci. Rep. 7, 1–13. https://doi.
org/10.1038/s41598-017-15838-z.
McWilliam, J.N., McCauley, R.D., Erbe, C., Parsons, M.J.G.G., 2018. Soundscape
diversity in the Great Barrier Reef: Lizard Island, a case study. Bioacoustics 27,
295–311. https://doi.org/10.1080/09524622.2017.1344930.
Mooney, T.A., Di Iorio, L., Lammers, M., Lin, T.-H., Nedelec, S.L., Parsons, M., Radford,
C., Urban, E., Stanley, J., 2020. Listening forward: approaching marine biodiversity
assessments using acoustic methods. doi: 10.1098/rsos.201287.
Munger, J.E., Herrera, D.P., Haver, S.M., Waterhouse, L., McKenna, M.F., Dziak, R.P.,
Gedamke, J., Heppell, S.A., Haxel, J.H., 2022. Machine learning analysis reveals
relationship between pomacentrid calls and environmental cues. Mar. Ecol. Prog.
Ser. 681, 197–210. https://doi.org/10.3354/MEPS13912.
Nedelec, S.L., Campbell, J., Radford, A.N., Simpson, S.D., Merchant, N.D., Fisher, D.,
2016. Particle motion: the missing link in underwater acoustic ecology. Methods
Ecol. Evol. 7 (7), 836–842. https://doi.org/10.1111/2041-210X.12544.
Nedelec, S.L., Simpson, S.D., Holderied, M., Radford, A.N., Lecellier, G., Radford, C.,
Lecchini, D., 2015. Soundscapes and living communities in coral reefs: temporal and
spatial variation. Mar. Ecol. Prog. Ser. 524, 125–135. https://doi.org/10.3354/
meps11175.
Pieretti, N., Danovaro, R., 2020. Acoustic indexes for marine biodiversity trends and
ecosystem health. Acoustic indexes for marine biodiversity trends and ecosystem
health. 375 (1814), 20190447. https://doi.org/10.1098/rstb.2019.0447.
Putland, R.L., Constantine, R., Radford, C.A., 2017. Exploring spatial and temporal
trends in the soundscape of an ecologically signicant embayment. Sci. Rep. 7, 1–12.
https://doi.org/10.1038/s41598-017-06347-0.
Radford, C., Jeffs, A., Tindle, C., Montgomery, J.C., 2008. Resonating sea urchin
skeletons create coastal choruses. Mar. Ecol. Prog. Ser. 362, 37–43. https://doi.org/
10.3354/meps07444.
Rao, R.B., Fung, G., Rosales, R., 2008. On the dangers of cross-validation. An
experimental evaluation, in: Proceedings of the 2008 SIAM International Conference
on Data Mining. SIAM, pp. 588–596.
Rilov, G., Fraschetti, Simonetta, Gissi, Elena, Pipitone, Carlo, Badalamenti, Fabio,
Tamburello, Laura, Menini, Elisabetta, Goriup, Paul, Mazaris, Antonios D.,
Garrabou, Joaquim, Benedetti-Cecchi, Lisandro, Danovaro, Roberto, Loiseau,
Charles, Claudet, Joachim, Katsanevakis, Stelios, Rilov, C., Fraschetti, S., Gissi, E.,
Pipitone, C., Badalamenti, F., Tamburello, L., Menini, E, Goriup, P. Mazaris, A.D.,
Garrabou, J., Benedetti-Cecchi, L. Danovaro, R., Loiseau, C., Claudet, J.,
Katsanevakis, S., 2020. A fast-moving target: achieving marine conservation goals
under shifting climate and policies. Ecol. Appl. 30, e02009. doi: 10.1002/EAP.2009.
B. Williams et al.

Ecological Indicators 140 (2022) 108986
11
Sethi, S.S., Ewers, R.M., Jones, N.S., Sleutel, J., Shabrani, A., Zulkii, N., Picinali, L., n.d.
Soundscapes predict species occurrence in tropical forests. https://doi.org/10.1101/
2020.09.24.311381.
Sethi, S.S., Jones, N.S., Fulcher, B.D., Picinali, L., Clink, D.J., Klinck, H., Orme, C.D.L.,
Wrege, P.H., Ewers, R.M., 2020. Characterizing soundscapes across diverse
ecosystems using a universal acoustic feature set. Proc. Natl. Acad. Sci. U. S. A. 117,
17049–17055. https://doi.org/10.1073/pnas.2004702117.
Simpson, S.D., Meekan, M., Montgomery, J., McCauley, R., Jeffs, A., 2005. Homeward
Sound. Science (80-.). 308, 221. doi: 10.1126/science.1107406.
Smith, J.E., Brainard, R., Carter, A., Grillo, S., Edwards, C., Harris, J., Lewis, L.,
Obura, D., Rohwer, F., Sala, E., Vroom, P.S., Sandin, S., 2016. Re-evaluating the
health of coral reef communities: baselines and evidence for human impacts across
the central Pacic. Proc. R. Soc. B Biol. Sci. 283, 20151985. https://doi.org/
10.1098/rspb.2015.1985.
Smith, T.B., Nemeth, R.S., Blondeau, J., Calnan, J.M., Kadison, E., Herzlieb, S., 2008.
Assessing coral reef health across onshore to offshore stress gradients in the US
Virgin Islands. Mar. Pollut. Bull. 56, 1983–1991. https://doi.org/10.1016/j.
marpolbul.2008.08.015.
Sousa-Lima, R.S., Norris, T.F., Oswald, J.N., Fernandes, D.P., 2013. A Review and
Inventory of Fixed Autonomous Recorders for Passive Acoustic Monitoring of Marine
Mammals. Aquat. Mamm. 39. https://doi.org/10.1109/RIOAcoustics.2013.6683
984.
Staaterman, E., Paris, C.B., DeFerrari, H.A., Mann, D.A., Rice, A.N., D’Alessandro, E.K.,
2014. Celestial patterns in marine soundscapes. Mar. Ecol. Prog. Ser. 508, 17–32.
https://doi.org/10.3354/meps10911.
Staaterman, E., Rice, A.N., Mann, D.A., Paris, C.B., 2013. Soundscapes from a Tropical
Eastern Pacic reef and a Caribbean Sea reef. Coral Reefs 32, 553–557. https://doi.
org/10.1007/s00338-012-1007-8.
Stowell, D., Sueur, J., 2020. Ecoacoustics: acoustic sensing for biodiversity monitoring at
scale. Remote Sens. Ecol. Conserv. 6, 217–219. https://doi.org/10.1002/rse2.174.
Sueur, J., Aubin, T., Simonis, C., 2008. Equipment review: Seewave, a free modular tool
for sound analysis and synthesis. Bioacoustics 18, 213–226. https://doi.org/
10.1080/09524622.2008.9753600.
Sueur, J., Farina, A., Gasc, A., Pieretti, N., Pavoine, S., 2014. Acoustic indices for
biodiversity assessment and landscape investigation. Acta Acust. United Acust. 100,
772–781. https://doi.org/10.3813/AAA.918757.
Villanueva-Rivera, L.J., Pijanowski, B.C., Villanueva-Rivera, M.L.J., 2018. Package
‘soundecology.’.
Williams, S.L., Sur, C., Janetski, N., Hollarsmith, J.A., Rapi, S., Barron, L., Heatwole, S.J.,
Yusuf, A.M., Yusuf, S., Jompa, J., Mars, F., 2019. Large-scale coral reef rehabilitation
after blast shing in Indonesia. Restor. Ecol. 27, 447–456. https://doi.org/10.1111/
rec.12866.
Wu, W., Mallet, Y., Walczak, B., Penninckx, W., Massart, D.L., Heuerding, S., Erni, F.,
1996. Comparison of regularized discriminant analysis, linear discriminant analysis
and quadratic discriminant analysis, applied to NIR data. Anal. Chim. Acta 329,
257–265. https://doi.org/10.1016/0003-2670(96)00142-0.
B. Williams et al.