Digital–Green Finance Synergy and Agricultural Carbon Emission Intensity in China: Machine Learning and Scenario Evidence

Agricultural carbon emission intensity (ACEI) is an important indicator of the low-carbon transformation of agriculture, yet the predictive roles of financial, technological, resource, and production conditions remain insufficiently understood. Using a balanced panel of 31 Chinese provinces from 2011 to 2023, this study constructs provincial ACEI through agricultural greenhouse-gas accounting and develops a temporally ordered explainable machine-learning framework combining multi-model comparison, Kernel SHapley Additive exPlanations (SHAP) interpretation, robustness analysis, and conditional scenario simulation. Provincial mean ACEI declined by 36.76% over the study period, although substantial regional heterogeneity persisted. Among the structural machine-learning models, the full-predictor support vector regression model achieved the best validation performance, whereas a persistence benchmark remained more accurate for short-horizon fixed-origin forecasting. Kernel SHAP identified digital–green finance synergy (DGI) as the leading source of predictive information in the main specification. Its importance weakened but remained among the leading predictors after removing persistent provincial differences and the common temporal trend, indicating that its predictive relevance is not entirely attributable to long-run trends. Scenario analysis further shows that DGI enhancement alone produces only limited changes in predicted ACEI, whereas the coordinated DGI–technology–irrigation pathway yields the largest supported reduction relative to business as usual by 2030 (5.22%). These findings suggest that digital–green finance is more meaningfully associated with agricultural low-carbon transformation when combined with technological upgrading and resource governance rather than considered as an isolated policy lever.