Learning Hub
Overview of Marine Conservation Bioacoustics
Bioacoustics is the study of how living organisms produce, transmit, and receive sounds. In marine environments, acoustics play a crucial role because sound travels faster and farther underwater than in air. For many marine species, sound is essential for communication, navigation, finding food, and avoiding predators.
However, in order to appreciate the significance of sounds underwater, the first step is understanding which species make which sounds. This remains a fundamental challenge in the field.
Functions of Sound
-
Communication: Many fish and marine mammals use sound to communicate with each other, especially for mating and territoriality (Tricas & Boyle, 2014).
-
Navigation and Foraging: Toothed whales use echolocation, emitting sounds that bounce back from objects, helping them navigate and find food. This is a form of active sonar, like that used by navies. While not yet known, some fish may use a similar method to find food hidden in the substrate.
-
Predation and Defense: Sounds can also be used to scare off predators or, for whales, to coordinate hunting. And the sounds that animals make can even be used by silent predators like sharks, which can listen for prey calls.
Dolphins use echolocation to navigate and hunt for food.
Applications in Conservation
By listening to the sounds of the ocean, scientists can monitor biodiversity, assess the health of ecosystems, and study animal behaviors without intrusive methods (Rice et al., 2023). More and more, monitoring levels of sound pollution from humans is becoming a part of marine conservation.
Monitoring Biodiversity
Scientists use hydrophones (underwater microphones) to record sounds in marine environments. By analyzing these recordings, they can identify the presence and absence of different species, the density of populations, and resilience within the marine community.
Ecosystem Health
Scientists can gauge the health of an ecosystem from its soundscape. Healthy coral reefs, for example, are more complex than degraded ones. Some studies calculate “ecoacoustic indices” of these soundscapes, helping conservationists get an idea about the general state of habitats (Williams et al., 2022). FishEye Collaborative is working on the classification and identification of specific fish species’ sounds, which will give us more precise data.
Behavioral Studies
Understanding how marine animals use sound can provide insights into their activity, social structures, and interactions (Tricas & Boyle, 2014). This information is vital for creating effective conservation strategies.
Hydrophones help scientists identify and monitor fish species in coral reefs
Additional Considerations
Impact of Climate Change on Marine Acoustics
Climate change is increasing ocean temperatures, which alters how sound travels underwater. These changes can impact marine animals that rely on sound for communication and navigation.
Marine Noise Pollution
Human activities, such as shipping and drilling, cause noise pollution in the ocean (Lynch, 2022). Even recreational motor boats and Jet Skis are extremely loud. This can interfere with the natural sounds of marine life, leading to stress and disorientation. Efforts are being made to reduce noise pollution and protect marine habitats.
Case Studies
Identifying Fish Species in Coral Reefs in the Caribbean
Scientists have used hydrophones to identify various fish species by their unique sounds. This non-invasive method will help monitor fish populations and the health of coral reefs. Until recently, the most effective efforts to assign specific sounds to species have relied on rebreather divers with specialized cameras, or on hydrophone and camera arrays with a limited field of view (Mouy et al., 2023). FishEye Collaborative’s recent results show that species can be identified more quickly and effectively using new technologies. In our first studies in Curaçao we’ve identified more species-specific natural sounds than were previously known across the entire Caribbean.
Back to top
Overview of Passive Acoustic Monitoring (PAM)
Passive acoustic monitoring (PAM) is a method used to listen to and record underwater sounds over long periods. PAM devices are usually small self-contained units that have a hydrophone, recorder, and a lot of batteries. These are placed out in nature for days to months. Unlike active sonars, like fish-finders, which send out sound waves and listen for echoes, PAM involves only listening. This makes it a non-invasive and cost-effective tool for studying marine environments.
How PAM Works
Equipment
PAM relies on hydrophones, which are specially designed to pick up underwater sounds. Scientists can deploy them in various configurations, from single units to complex arrays, depending on the research needs.
Data Collection
Researchers place PAM devices in strategic locations to record sounds over extended periods. With other techniques, biologists are only able to count fishes when they are physically in the field, giving only intermittent numbers. Remote cameras can be used in some situations, but they have a limited view and usually short duration. Long-term passive acoustic monitoring provides an around-the-clock, continuous stream of data that can better reveal patterns in marine soundscapes.
Data Analysis
Analyzing PAM data involves identifying and interpreting sound patterns. Basic techniques include:
-
Frequency Analysis: Breaking down sounds into their component frequencies to identify specific calls.
-
Time-Domain Analysis: Examining the timing of sounds to detect patterns and rhythms.
Frequency analysis helps identify specific marine species by their unique sound signatures.
When we know which sounds come from which species, as we do for whales, we can do far more.
Applications in Conservation
PAM has numerous applications in marine conservation:
-
Species Identification: Traditionally, species identification using PAM involved comparing recorded sounds to known species calls. This method, while useful, is limited by the number of known species sounds.
-
Habitat Monitoring: By analyzing the soundscape of a habitat, scientists can detect changes over time. For example, a decrease in fish sounds might indicate a decline in population or habitat degradation.
-
Impact of Human Activities: PAM can also be used to monitor the effects of human activities on marine environments. For instance, increased noise levels from shipping lanes can be detected and assessed for their impact on marine life.
Comparison with Other Monitoring Techniques
Technique | Pros | Cons |
|---|---|---|
Direct Visual Surveys | Direct observation | Limited by visibility, time-consuming, labor-intensive, requires expertise |
Camera deployments | Non-invasive, natural behavior, more data | Limited field of view, limited duration, |
Sonar | Effective for some counting | Limited species specification, time-consuming, expensive |
Tagging | Detailed individual data | Invasive, labor-intensive |
eDNA | Non-invasive, effective for some measures like presence/absence | Requires genomic experience, limited ability to assess temporal patterns or or abundance |
PAM | Non-invasive, continuous data, less labor-intensive | Requires expertise in sound analysis |
Comparison of pros and cons of various marine monitoring techniques.
PAM uses hydrophones to record underwater sounds without disturbing marine life.
Case Studies
Whistling dolphins and singing humpback whales immediately come to mind when people think about underwater communication. The recognition of the significance of whale sounds contributed to the modern-day ocean conservation movement and strategies to protect marine biodiversity.
We can apply PAM techniques to fish and invertebrates as well.
Example 1: Using PAM To Protect Endangered Whale Species
Researchers have used hydrophones towed by research vessels to monitor North Atlantic right whale migrations. By tracking their calls, scientists can identify where they are and what they’re doing. This helps alert fishing vessels and other ships to stay clear (there are fewer than 340 right whales left).
Example 2: Using PAM To Reduce Environmental Risks
PAM is being used for offshore operations and developments like wind farms. When technicians notice the presence of endangered species in the area, they can take preventative measures to avoid disturbing them during construction (RPS, n.d.).
Construction of offshore wind farms can cause underwater noises and vibrations that alter marine wildlife behavior. Companies can use PAM to reduce the risk.
Overview of Artificial Intelligence for PAM
Artificial intelligence involves developing algorithms to learn from data and make predictions or decisions. In bioacoustics, AI can build informative patterns from large volumes of acoustic data, making it easier to detect, analyze, and interpret underwater sounds.
Machine learning algorithms can analyze large sets of data to identify patterns in marine sounds.
AI Techniques
Types of Training Data
-
Supervised Learning: The algorithm is trained on labeled data (e.g., fish sounds that have been correctly identified), learning to predict outcomes based on input features. This is commonly used for species identification.
-
Unsupervised Learning: The algorithm identifies patterns in data without prior labels, useful for discovering new sound patterns or when the identity of sounds are not known.
-
Semi-supervised Learning: This method combines both supervised and unsupervised algorithms, using a small amount of labeled data and a larger set of unlabeled data to enhance pattern recognition
Training Models
Training an AI model involves one or more of these steps:
-
Data Preprocessing: Cleaning and organizing data to ensure normalization.
-
Feature Extraction: Identifying relevant features in the data that the model will use for learning.
-
Model Training: Having the model learn from training data to recognize patterns.
“Sea of Sound” Documentary
Contact us for the Teacher’s Guide PDF and Files
Challenges in Artificial Intelligence for PAM
While AI—more specifically, Machine Learning (ML)—has shown promise in bioacoustics applications, its success depends on several critical factors, including well-designed input-feature preprocessing, appropriate labels, choice of ML models, effective training strategies, and the availability of ample training and validation datasets. Here are some frequent challenges:
-
Data Quality and Preprocessing: High-quality data is essential for accurate ML models. Underwater recordings often contain noise from various sources, such as waves, boats, and other marine life. Preprocessing this data to filter out irrelevant sounds is complex and time-consuming.
-
Labeling and Ground Truthing: Accurate labeling of sounds is crucial for supervised learning models. However, identifying and labeling sounds in vast datasets is labor-intensive and requires expert knowledge. This process is often a bottleneck in developing effective ML models.
-
Model Selection and Training: Choosing the right ML model and training strategy is vital for success. Different models have varying strengths and weaknesses, and selecting the most appropriate one depends on the specific characteristics of the acoustic data. Additionally, training these models requires significant computational resources and expertise.
-
Data Availability: ML models require large datasets for training and validation. In bioacoustics, there is only a limited number of labeled recordings available, which hinders the development of robust models. The scarcity of species-specific labeled data remains a significant challenge.
-
Species Identification: Traditional methods of species identification in PAM rely on matching sounds to known species calls. However, many marine species have not been well-documented acoustically, making it difficult to accurately identify them using existing databases.
To address these challenges, FishEye Collaborative has developed the UPAC-360, an innovative 360° underwater passive acoustic camera. This device allows researchers to correctly identify and label fish-sound data by providing a comprehensive view of the soundscape and visually confirming the sources of sounds. With the FishEye UPAC-360, our technicians can create a set of species-specific labeled recordings that can be used to train ML models, considerably enhancing the accuracy and efficiency of bioacoustic monitoring and marine conservation efforts. FishEye Collaborative has already produced a large increase in the number of fish species with enough recordings to train AI on. As we publish our findings, we will upload our recordings to our Fish Sound Library.
Other Marine Conservation Bioacoustic Resources:
More Great Resources:
Applications in Conservation
Automated Species Identification
AI models can be trained to recognize species-specific sounds, automating the process of species identification (Rubbens et al., 2023).
Soundscape Analysis
AI can analyze entire soundscapes to detect patterns and anomalies. This
is useful for monitoring ecosystem health and detecting changes over time, and it reduces reliance on labor-intensive surveys by experts (Williams et al., 2022).
Predictive Modeling
By analyzing past data, AI models can predict future trends in marine populations and behaviors. This “ecological inference” could help conservationists make informed decisions about protecting marine habitats (Rubbens et al., 2023).
References:
Lynch, C. (2022, May 2). Fish Are Chattier Than Previously Thought.The Scientist. https://www.the-scientist.com/fish-are-chattier-than-previously-thought-69938
Mouy, X., Black, M., Cox, K., Qualley, J., Dosso, S., & Juanes, F. (2023). Identification of fish sounds in the wild using a set of portable audio-video arrays. Methods in Ecology and Evolution, 14(8), 2165-2186. https://doi.org/10.1111/2041-210x.14095
Rice, A. N., Garcia, M. L., Symes, L. B., & Klinck, H. (2023). Conservation Bioacoustics: Listening to the Heartbeat of the Earth. Acoustics Today, 19(3), 46-53. https://doi.org/10.1121/at.2023.19.3.56
RPS. (n.d.). Marine life mitigation - PSOs, MMOs and PAM. https://www.rpsgroup.com/services/oceans-and-coastal/marine-life-mitigation-psos-mmos-and-pam/
Rubbens, P., Brodie, S., Cordier, T., Destro Barcellos, D., Devos, P., Fernandes-Salvador, J. A., Fincham, J. I., Gomes, A., Handegard, N. O., Howell, K., Jamet, C., Kartveit, K. H., Moustahfid, H., Parcerisas, C., Politikos, D., Sauzède, R., Sokolova, M., Uusitalo, L., Van den Bulcke, L., … Irisson, J.-O. (2023). Machine learning in marine ecology: An overview of techniques and applications. ICES Journal of Marine Science, 80(7), 1829-1853. https://doi.org/10.1093/icesjms/fsad100
Tricas, T. C., & Boyle, K. S. (2014). Acoustic behaviors in Hawaiian coral reef fish communities. Marine Ecology Progress Series, 511, 1-16. https://doi.org/10.3354/meps10930
Williams, B., Lamont, T. A. C., Chapuis, L., Harding, H. R., May, E. B., Prasetya, M. E., Seraphim, M. J., Jompa, J., Smith, D. J., Janetski, N., Radford, A. N., & Simpson, S. D. (2022). Enhancing automated analysis of marine soundscapes using ecoacoustic indices and machine learning. Ecological Indicators, 140, 108986. https://doi.org/10.1016/j.ecolind.2022.108986