National mariculture mapping in China at 10-m spatial resolution based on EMA-UNet and Sentinel-1/2 imagery

Abstract

Accurately measuring mariculture’s area and spatial distribution is vital for the sustainable development of marine resources. However, existing research primarily focuses on the regional scale, while national mariculture studies remain limited, with significant differences in extracted areas. This study utilized time-series Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 Multi-Spectral Instrument (MSI) imagery and developed the Efficient Multi-scale Attention U-Net (EMA-UNet) to generate a fine-grained map of China’s mariculture. The spatial distribution characteristics of mariculture areas were revealed through spatial analysis methods. The results showed that: (1) The EMA-UNet proposed in this study achieved higher accuracy than comparative models, with an overall accuracy of 95.9 %, an F1 score of 94.3 %, and an Intersection over Union (IoU) of 89.4 %, with improved extraction effects. (2) China’s mariculture was estimated to cover 2,941.7 km², with northern provinces accounting for 56.46 % of the total. Fujian, Shandong, and Liaoning were the three provinces with the largest mariculture areas, contributing 27.36 %, 25.85 %, and 20.87 % respectively. (3) China’s mariculture density showed a “dense in the east and sparse in the west”, “dense in the north and sparse in the south” pattern, with high-intensity aggregations in Fujian and central-southern Guangdong, and low-intensity aggregations in Bohai Bay and nearshore Guangxi, reflecting the influence of natural-socioeconomic factors. This study provides a fine-grained map of China’s mariculture distribution, serving as a data foundation for relevant studies. It also investigated the effects of extraction methods, data sources, and band features, offering insights for optimizing mariculture extraction methods.