Compared to afforestation on tidal flats, the transition from ponds-to-mangroves offers greater advantages in terms of mangrove functional restoration, especially in terms of cost-effectiveness for land transformation. It is anticipated that prioritizing ecological restoration in previously reclaimed mangrove aquaculture ponds is expected to be the primary method of mangrove restoration in the future. However, the distribution, types, and conditions of aquaculture ponds within mangroves have not yet to be fully assessed. This study proposes an object-oriented method for extracting nearshore aquaculture ponds based on Sentinel-2, integrating multidimensional features such as spectral, morphological, and textural characteristics to achieve precise extraction of aquaculture ponds within China’s mangroves in province levels. The results indicate that, 1) the overall classification accuracy of this method reached 87%, with a Kappa coefficient of 0.86, demonstrating high classification precision; 2) Guangdong has the largest area of coastal aquaculture ponds, totaling 37,853.88 ha and accounting for 53.79% of the total area, with significant clustering in Jiangmen and Zhanjiang; 3) the aquaculture ponds in Zhejiang (0.377), Taiwan (0.365), and Guangxi (0.364) exhibit a relatively high degree of fragmentation, with more scattered distribution. The P/A index (perimeter-to-area ratio) of aquaculture ponds in Hainan (0.063) and Fujian (0.055) is higher than that of other provinces, indicating that the shape of the ponds is relatively regular. The findings enable precise identification of nearshore aquaculture ponds, facilitating the selection of suitable sites for mangrove restoration as a method for prioritizing areas in mangrove blue carbon projects.

Object-oriented, Aquaculture ponds, Mangrove, Remote sensing identification, Blue carbon