Data-driven optimization of county-scale straw management for low-carbon and income-enhancing rice production in...

Returning straw to soil after harvest is not a simple binary decision for agricultural stakeholders. Optimizing straw management under site-specific conditions is crucial to balance the trade-off between climatic changes mitigation and food security in rice production. Here, we developed an integrated framework to determine the optimal ratios of straw return and removal across major rice-cropping regions in China. Using machine learning (ML) models, we identified site-specific straw return ratios that minimize greenhouse gas (GHG) emissions without compromising rice yield. Additionally, a complementary straw logistics model was proposed to manage excess straw instead of burning. Among five ML models evaluated, XGBoost demonstrated superior performance in predicting the impact of straw return on CH₄ and N₂O emissions, soil organic carbon sequestration rate and rice yield. In single rice systems, a high straw return rate of 73 % resulted in an average yield increases of 26.5 % and a net reduction in GHG emissions by 105 %. In contrast, our model recommends prioritizing straw removal in double rice systems due to difficulties in balancing yield-emission trade-offs. The net ecosystem economic benefit (NEEB) analysis revealed that economic gains from increased yields and reduced nitrogen fertilizer use outweighed the costs of straw collection and off-field transportation. Notably, the current social cost of carbon contributed minimally to total economic benefits, despite substantial GHG mitigation benefits realized through ML-guided strategies. Overall, this study highlights the potential of site-specific, data-driven straw management to promote sustainable agricultural development in China.