Dynamic Monitoring of Marine Floating Raft Aquaculture in Jiangsu Province Based on Multi-source SAR Imagery

Wen Li¹Jia Xu^1*Yuanyuan Chen²Chongbin Liu³

• ¹Hohai University, Nanjing, China
• ²Nanjing Forestry University, Nanjing, Jiangsu Province, China
• ³Artificial Intelligence and Digital Economy Guangdong Provincial Laboratory (Shenzhen), Shenzhen, China

Long-term change monitoring of marine floating raft aquaculture (MFRA) in Jiangsu Province is urgently needed to support aquaculture restructuring and protect the regional marine environment. Optical remote sensing images have been widely used in the extraction of MFRA but are usually limited by cloudy and rainy weather conditions. Compared with optical images, Synthetic Aperture Radar (SAR) can acquire images even under dense cloud cover , offering a reliable alternative. However, existing studies mainly focused on limited geographic areas and didn't consider different types of MFRA. This study proposes an automatic framework for monitoring of MFRA based on multi-source SAR imagery. To achieve a better extraction of MFRA, two key enhancements were introduced. First, the SDWI and SDRI indices were utilized to effectively differentiate MFRA from seawater Second, a deep learning framework termed Boundary-Enhancing Swin Transformer (BE-Swin) was developed for extracting different types of MFRA. Results showed that (1) Combining ALOS-1 and Sentinel-1 data enables the mapping the long-term dynamics of the MFRA in Jiangsu province, China. (2) The BE-Swin model enhances boundary extraction, reducing errors, omissions, and adhesion issues. Compared with other deep neural network models, the BE-Swin model improved the extraction accuracy by incorporating 3 key modules within the Swin Transformer. (3) MFRA in Jiangsu Province is distributed in Meizhou Bay of Lianyungang, the radial ridge group region of Yancheng, and the northern coast of Nantong. From 2008 to 2022, raft aquaculture in the coastal region of Jiangsu Province experienced rapid expansion followed by gradual contraction. (4) MFRA in Jiangsu Province is can be categorized into three types, pole-pixed, semi-floating, and full-floating raft aquaculture. The area of pole-fixed raft aquaculture steadily increased, reaching 171.38 km2 in 2022. Semi-floating raft aquaculture peaked at 318.36 km2 in 2015 but was reduced by nearly half in 2022. Full-floating raft aquaculture was initially absent but has shown a trend toward large-scale adoption, reaching 78.42 km2 in 2022.