An official website of the United States government
Coral reefs comprise some of the most biodiverse habitats on the planet. These ecosystems face a range of stressors, making quantifying community assemblages and potential changes vital to effective management. To understand short- and long-term changes in biodiversity and detect early warning signals of decline, new methods for quantifying biodiversity at scale are necessary. Acoustic monitoring techniques have proven useful in observing species activities and biodiversity on coral reefs through aggregate approaches (i.e. energy as a proxy). However, few studies have ground-truthed these acoustic analyses with human-based observations. In this study, we sought to expand these passive acoustic methods by investigating biological sounds and fish call rates on a healthy reef, providing a unique set of human-confirmed, labeled acoustic observations. We analyzed acoustic data from Tektite Reef, St. John, US Virgin Islands, over a 2 mo period. A subset of acoustic files was manually inspected to identify recurring biotic sounds and quantify reef activity throughout the day. We found a high variety of acoustic signals in this soundscape. General patterns of call rates across time conformed to expectations, with dusk and dawn showing important and significantly elevated peaks in soniferous fish activity. The data reflected high variability in call rates across days and lunar phases. Call rates did not correspond to sound pressure levels, suggesting that certain call types may drive crepuscular trends in sound levels while lower-level critical calls, likely key for estimating biodiversity and behavior, may be missed by gross sound level analyses.
more »
« less
This content will become publicly available on March 1, 2026
Rapid detection of fish calls within diverse coral reef soundscapes using a convolutional neural network
The quantity of passive acoustic data collected in marine environments is rapidly expanding; however, the software developments required to meaningfully process large volumes of soundscape data have lagged behind. A significant bottleneck in the analysis of biological patterns in soundscape datasets is the human effort required to identify and annotate individual acoustic events, such as diverse and abundant fish sounds. This paper addresses this problem by training a YOLOv5 convolutional neural network (CNN) to automate the detection of tonal and pulsed fish calls in spectrogram data from five tropical coral reefs in the U.S. Virgin Islands, building from over 22 h of annotated data with 55 015 fish calls. The network identified fish calls with a mean average precision of up to 0.633, while processing data over 25× faster than it is recorded. We compare the CNN to human annotators on five datasets, including three used for training and two untrained reefs. CNN-detected call rates reflected baseline reef fish and coral cover observations; and both expected biological (e.g., crepuscular choruses) and novel call patterns were identified. Given the importance of reef-fish communities, their bioacoustic patterns, and the impending biodiversity crisis, these results provide a vital and scalable means to assess reef community health.
more »
« less
- Award ID(s):
- 2024077
- PAR ID:
- 10652110
- Publisher / Repository:
- Acoustical Society of America
- Date Published:
- Journal Name:
- The Journal of the Acoustical Society of America
- Volume:
- 157
- Issue:
- 3
- ISSN:
- 1520-8524
- Page Range / eLocation ID:
- 1665 to 1683
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Coral reefs are biodiverse marine ecosystems that are undergoing rapid changes, making monitoring vital as we seek to manage and mitigate stressors. Healthy reef soundscapes are rich with sounds, enabling passive acoustic recording and soundscape analyses to emerge as cost-effective, long-term methods for monitoring reef communities. Yet most biological reef sounds have not been identified or described, limiting the effectiveness of acoustic monitoring for diversity assessments. Machine learning offers a solution to scale such analyses but has yet to be successfully applied to characterize the diversity of reef fish sounds. Here we sought to characterize and categorize coral reef fish sounds using unsupervised machine learning methods. Pulsed fish and invertebrate sounds from 480 min of data sampled across 10 days over a 2-month period on a US Virgin Islands reef were manually identified and extracted, then grouped into acoustically similar clusters using unsupervised clustering based on acoustic features. The defining characteristics of these clusters were described and compared to determine the extent of acoustic diversity detected on these reefs. Approximately 55 distinct calls were identified, ranging in centroid frequency from 50 Hz to 1,300 Hz. Within this range, two main sub-bands containing multiple signal types were identified from 100 Hz to 400 Hz and 300 Hz–700 Hz, with a variety of signals outside these two main bands. These methods may be used to seek out acoustic diversity across additional marine habitats. The signals described here, though taken from a limited dataset, speak to the diversity of sounds produced on coral reefs and suggest that there might be more acoustic niche differentiation within soniferous fish communities than has been previously recognized.more » « less
-
Acoustic enrichment can facilitate coral and fish larval settlement, offering a promising method to rebuild degraded reefs. Yet it is critical to understand sound propagation in complex shallow-water coral reefs to effectively apply this method over large restoration-scale areas. In this field-based study, we quantified propagation features of multiple sound types emitted through a custom playback system over varying coral reef habitat. Sound levels were computed at different distances from the source in both pressure and particle motion, the latter being detected by marine invertebrates. Detection distances were primarily determined by source levels, and depth-dependent transmission losses. Transmission losses and detection distances were similar for sound pressure and particle acceleration measurements. Importantly, broadband particle acceleration levels could be closely estimated at distances >10 m using a single hydrophone and a plane wave approximation. Using empirically determined coral larvae sound detection thresholds, we found that low frequency sounds (<1 kHz) such as fish calls from healthy coral reef soundscapes may be detectable by larvae hundreds of meters away. These results provide key data to help design standardized methods and protocols for scientists, managers and restoration practitioners aiming to rebuild coral reef ecosystems over reasonably large spatial scales using acoustic enrichment.more » « less
-
The_Royal_Society_Publishing (Ed.)Coral reefs, hubs of global biodiversity, are among the world’s most imperilled habitats. Healthy coral reefs are characterized by distinctive soundscapes; these environments are rich with sounds produced by fishes and marine invertebrates. Emerging evidence suggests these sounds can be used as orientation and settlement cues for larvae of reef animals. On degraded reefs, these cues may be reduced or absent, impeding the success of larval settlement, which is an essential process for the maintenance and replenishment of reef populations. Here, in a field-based study, we evaluated the effects of enriching the soundscape of a degraded coral reef to increase coral settlement rates.Porites astreoideslarvae were exposed to reef sounds using a custom solar-powered acoustic playback system.Porites astreoidessettled at significantly higher rates at the acoustically enriched sites, averaging 1.7 times (up to maximum of seven times) more settlement compared with control reef sites without acoustic enrichment. Settlement rates decreased with distance from the speaker but remained higher than control levels at least 30 m from the sound source. These results reveal that acoustic enrichment can facilitate coral larval settlement at reasonable distances, offering a promising new method for scientists, managers and restoration practitioners to rebuild coral reefs.more » « less
-
Ecologists interested in monitoring the effects caused by climate change are increasingly turning to passive acoustic monitoring, the practice of placing autonomous audio recording units in ecosystems to monitor species richness and occupancy via species calls. However, identifying species calls in large datasets by hand is an expensive task, leading to a reliance on machine learning models. Due to a lack of annotated datasets of soundscape recordings, these models are often trained on large databases of community created focal recordings. A challenge of training on such data is that clips are given a "weak label," a single label that represents the whole clip. This includes segments that only have background noise but are labeled as calls in the training data, reducing model performance. Heuristic methods exist to convert clip-level labels to "strong" call-specific labels, where the label tightly bounds the temporal length of the call and better identifies bird vocalizations. Our work improves on the current weakly to strongly labeled method used on the training data for BirdNET, the current most popular model for audio species classification. We utilize an existing RNN-CNN hybrid, resulting in a precision improvement of 12% (going to 90% precision) against our new strongly hand-labeled dataset of Peruvian bird species.Jacob Ayers (Engineers for Exploration at UCSD); Sean Perry (University of California San Diego); Samantha Prestrelski (UC San Diego); Tianqi Zhang (Engineers for Exploration); Ludwig von Schoenfeldt (University of California San Diego); Mugen Blue (UC Merced); Gabriel Steinberg (Demining Research Community); Mathias Tobler (San Diego Zoo Wildlife Alliance); Ian Ingram (San Diego Zoo Wildlife Alliance); Curt Schurgers (UC San Diego); Ryan Kastner (University of California San Diego)more » « less
