[A-CG51] Coastal Ecosystems-2. Coral reefs, seagrass and macroalgal beds, and mangroves

[ACG51-P02] Automated Detection of Drifting Seaweed in the East China Sea Using Multispectral Satellite Imagery

*KAI HSIANG NG¹, Shinichiro Kako¹, Eisuke Tsutsumi², Hirohiko Nakamura², Toru Kobari² (1.Graduate School of Science and Engineering, Kagoshima University, 2.Faculty of Fisheries, Kagoshima University)

Keywords:Drifting Seaweed, Satellite, East China Sea, Multi-Spectral Instrument, Machine Learning

Drifting seaweed in the East China Sea (ECS) is a vital resource for aquaculture, particularly for harvesting juvenile yellowtail. Fishermen along Japan's west coast rely on catching wild yellowtail juveniles, which feed on drifting seaweed. Various seaweed species provide essential habitats and trophic support for marine organisms (Rosenblatt et al., 2022); however, their precise origins remain uncertain (Lin et al., 2023). Satellite imagery is used to study the ecological impact of drifting seaweed, offering advantages such as cost-effectiveness and wide-area monitoring (Zhu et al., 2023). Drifting seaweed exhibits spectral properties similar to common vegetation, absorbing sunlight and reflecting in the red and near-infrared (NIR) wavelengths. The Normalized Difference Vegetation Index (NDVI) is calculated as the ratio between red and NIR reflectance, with seaweed-containing pixels displaying higher NDVI values than surrounding waters. This study develops an automated system to improve seaweed detection efficiency using high-resolution Multi-Spectral Instrument (MSI) satellite imagery to analyze its spatiotemporal distribution. However, tracking drifting seaweed via satellite imagery is highly dependent on weather conditions, particularly cloud interference and satellite orbital constraints. To mitigate atmospheric interference, image processing techniques are applied to remove cloud interference and enhance detection accuracy. As a next step, we will implement machine learning (ML) to automatically identify seaweed based on NDVI data, enhancing detection efficiency and accuracy. In the next phase, we plan to integrate a high-resolution ocean surface numerical model to forecast seaweed drift pathways, improving our understanding of seaweed transport dynamics and ecological evolution.