When it comes to assessing fisheries stock, machine learning models do a good job of capturing the non-linear relationships present in biological processes and can process large datasets to identify complex patterns and make predictions that improve the understanding of population dynamics. Such insights are key to understanding the main reasons behind the collapse of fisheries so that workers and researchers can take proactive management measures.
An interesting application of this technology includes identifying and characterizing the sounds fish make, which promotes less invasive methods of learning about fish communication, behavior, spawning location, and biodiversity. Machine learning, deep learning in particular, has also made some advances in species identification. Recent studies show success in identifying tuna species in vessels equipped with digital systems with about a 70 percent accuracy rate. Real-time species identification will allow fisheries management authorities to control and monitor fish quotas. Fish species identification and counting with underwater devices that employ computer vision techniques can also be used in biodiversity studies. Species identification is essential for sustainable biodiversity conservation, scientific research, and catch tracking and monitoring.
AI and machine learning are also being used to monitor the behavior of maritime vessels. For example, AI use along with IUU (illegal, unreported, and unregulated) fishing detection technologies helps to analyze vessel behavior and assess whether a vessel is engaged in any illegal activities by collecting data on vessel location (AIS, VMS), satellite imagery (VIIRS, SAR), and coastal radars (which are now being replaced with drones). Machine learning can also assist in overseeing and mitigating the ecological consequences of maritime operations and help forecast ship emissions, monitor oil spill incidents, and assess the well-being of marine ecosystems.
Other applications can include helping to make predictions about fishing activities. Historical data and current trends can be processed by machine learning models to forecast where and when illegal fishing activities are likely to occur. Methods with such predictive capabilities can be integrated into a decision support system to aid authorities in taking the necessary steps to combat illegal, unreported, and unregulated fishing. Global Fishing Watch is even pioneering a project to leverage machine learning to analyze satellite data to help end illegal fishing and safeguard the ocean and its critical biodiversity.
Click here to learn more about OEF’s work in fisheries management.
