China's Loess Plateau (CLP) has undergone substantial landscape transformation due to the widespread implementation of soil and water conservation (SWC) measures, resulting in a marked reductions in both sediment yield and soil organic carbon (SOC) loss. However, the extent to which these measures influence SOC loss remains insufficiently unexplored, limiting the transferability of these successes to other erosion-prone regions. Here, we first map the spatial distribution of the SWC measures in dam-controlled watersheds by integrating high-resolution satellite imagery, digital surface models, and Random Forest-based machine learning algorithms, achieving an average overall accuracy and F1-score of 0.94. We subsequently combine these spatial datasets with sediment yield and SOC loss records from 25 dam-controlled watersheds to investigate the impacts of SWC measure configurations on SOC loss across the watershed, slope-position, and landscape scales. We show that rainfall, topography, and vegetation exert substantial influences on SOC loss, with extreme rainfall events exhibiting particularly pronounced effects. Both vegetation-based SWC measures and terracing substantially reduce SOC loss, with terracing demonstrating maximum effectiveness at 0-40 % coverage. The effectiveness of SWC measures varied by slope position and watershed type, with mid-slope zones (15-25 degrees) exhibiting the strongest influence on SOC loss mitigation. Moreover, the landscape configuration of SWC measures exert a notable effect on SOC loss, with higher levels of spatial fragmentation associated with greater losses. These findings advance mechanistic understanding of the role of SWC measures in SOC dynamics and provide an empirical foundation for optimizing erosion control strategies and the spatial allocation of conservation interventions in the CLP and similar erosion-ecological systems.