Abstract: Mapping the distribution of shallow sea aquaculture and understanding its spatial expansion changes is the prerequisites of implementing dynamic marine monitoring, regional aquaculture management, layout optimization, and ecological protection. Our study chooses the Qinzhou Bay as the research area and introduces an adaptive threshold algorithm for identifying aquaculture raft frames to characterize their spatiotemporal changes in the bay. Significantly, a Raft Index is used into the Landsat series dataset. Then, the fuzzy clustering algorithm helps to enhance the raft frame structure features, and a chessboard segmentation algorithm is applied to partition the clustering Raft index images. The raft frames in the images are discerned based on the Otsu threshold method. Our results indicate that since 2000, aquaculture raft frames in the Qinzhou Bay have expanded significantly, with the maximum area reaching to 78,609.6km² in 2022. The distribution of raft frames has extended from the waters near Longmen Town toward the outer areas of the Qinzhou Bay and Maowei Sea. A notable correlation is established between the area of aquaculture raft frames and seawater quality in the Qinzhou Bay. The correlations between raft frame area and active phosphate, comprehensive pollution index, and eutrophication index are R² = 0.73, 0.5, and 0.36, respectively. The findings of this study provide valuable insights for understanding the spatiotemporal evolution of aquaculture raft frames in Qinzhou Bay and their relationship with the marine environment.

Key words: Raft extraction, adaptive threshold algorithm, Landsat images, Qinzhou Bay