Adaptive Multi-Level Scale Segmentation Method for Detecting Land and Marine Coastal Aquaculture Areas

Coastal zones provide favorable geographical conditions for aquaculture. Accurate information on land and marine aquaculture (LMA) activities is essential for managing their spatial distribution and estimating blue-food production. However, complex coastal landscapes and the heterogeneity of aquaculture facilities make it difficult to extract LMA areas accurately and completely at large scales. To address this challenge, we developed an adaptive multi-level scale segmentation (AMLSS) method for the precise extraction of LMA areas and modified a radar dual-polarization water index to reduce interference from the seawater background. We then constructed a class-specific hierarchical decision-tree model using multi-source features to map LMA areas at the national scale. AMLSS increased extraction accuracy by 19.7% for pond aquaculture and 22.9% for marine aquaculture compared with single-scale segmentation and was particularly effective in northern China. In 2022, landward pond and marine cage/raft aquaculture within China’s coastal zone covered approximately 7237 km2 and 1152 km2, respectively; the resulting map achieved an overall accuracy of 90.8%. This study provides a methodological basis for the long-term monitoring and spatially explicit mapping of LMA areas.