Uncovering the heterogeneous effects of China's Organic-Substitute-Chemical-Fertilizer policy: The role of local conditions in shaping distinct policy responses

China's Organic-Substitute-Chemical-Fertilizer (OSCF) policy aims to reduce chemical fertilizer dependence, yet its county-level effectiveness may vary with local production, capacity, and implementation conditions. Using panel data for 2726 counties from 2012 to 2022, we estimate the average and heterogeneous effects of OSCF on chemical fertilizer application intensity with double machine learning and SHAP-based interpretation. The policy is associated with a statistically significant average reduction in standardized chemical fertilizer application intensity, with an average treatment effect of −0.0469 that remains stable across robustness checks. Cross-county heterogeneity is expressed mainly in response magnitude. The estimated reduction is 0.0291 larger in high- than low-target-relevance counties and 0.0338 larger in non-grain-oriented than grain-dominant counties. Gross Output Value of Agriculture and Gross Regional Product rank highest in feature importance, followed by broadband access, non-food cropping, and village committees. Among representative fitted county profiles, the high Gross Output Value of Agriculture profile has a fitted response 0.0256 closer to zero than the low profile, whereas the high Gross Regional Product profile is 0.0335 more negative than the low profile. Contributions of key conditions vary across observed levels. Descriptive joint profiles also show configuration-specific differences, with high target relevance combined with high broadband access having a fitted response about 0.041 more negative than the low-low profile. These results motivate differentiated county screening that begins with policy-scope relevance and then assesses production feasibility, farmer/local adaptive capacity, and policy transmission and implementation capacity to identify where local assessment and policy-addressable support should focus.