Measuring artisanal fisheries using remote sensing: A deep-learning model piloted in Hadramawt

An innovative remote sensing study piloted in two sites (Mukalla and Qusayir) in Hadramawt Governorate, Yemen, reveals the following key findings: Deep learning using very high resolution satellite imagery can accurately detect small artisanal fishing boats. The YOLO11 Oriented Bounding Boxes model performs best, with an overall accuracy of 96.9 percent, and enables measuring boat size and identifying boat types. Mukalla’s fleet numbered around 800 operational boats in 2021–2024, most of which were houris, while sanbuqs accounted for 6.9 percent. Qusayir’s fleet numbered 450 boats, all of which were houris. The composition of the combined Mukalla and Qusayir fleets by boat type and length class is similar to the composition of Hadramawt’s entire fleet, as shown by a recent landing site survey. Approximated fish-catch capacities of the Mukalla and Qusayir fleets indicate that current boat sizes are not a notable constraint to increasing the productivity of these artisanal fisheries. The proposed remote sensing approach is an important first step toward development of a cost-effective tool for monitoring and analyzing artisanal fishing activities in Southern Yemen and beyond.