Depth-Resolved Soil Organic Carbon Stocks of Monsoon Riverine Wetlands: Digital Soil Mapping in the Han, Geum, and Yeongsan–Seomjin River Basins, South Korea

Inland wetlands store disproportionate soil organic carbon (SOC), yet spatially explicit information is scarce for temperate monsoon riverine wetlands, where recurrent erosive flooding shapes carbon accumulation. To map depth-resolved SOC stocks (kg C m−2) and evaluate satellite covariates, we combined stocks measured at 494 sites in the Han, Geum, and Yeongsan–Seomjin river basins, South Korea, with multi-temporal Sentinel-2 and terrain covariates in random forest models at 20 m for the organic layer, 0–30 cm, and 0–100 cm. Cross-validated R2 (squared correlation) was low (0.21, 0.18, and 0.15, respectively) but within the range reported for wetland SOC mapping. Satellite covariates raised R2 by 0.02–0.07 and reduced the root mean square error (RMSE) relative to a non-satellite baseline; dormant-season reflectance added no information. The April normalized difference vegetation index (NDVI) ranked highest for both mineral-soil layers, while hydrology and parent material gained importance at 0–100 cm. NDVI showed a threshold-type partial-dependence response despite weak rank correlations; one plausible interpretation is that vegetation cover indicates the geomorphic stability of the ground surface. The model-based regional stocks are 3.11 and 6.08 Mt C at 0–30 and 0–100 cm (27.7 and 54.1 t C ha−1), at the low end of published values for Asian monsoon riverine wetlands. The maps provide a baseline for carbon assessment.