| 基于XGBoost-SHAP模型的长江中游粮食主产区农业生产碳平衡时空分异及驱动因素 |
| 摘要点击 408 全文点击 38 投稿时间：2025-08-18 修订日期：2025-10-28 |
| 查看HTML全文
查看全文 查看/发表评论 下载PDF阅读器 |
| 中文关键词 农业生产碳平衡 影响因素 XGBoost模型 夏普利加性解释（SHAP） 长江中游粮食主产区 |
| 英文关键词 carbon balance of agricultural production influencing factors XGBoost model SHapley Additive exPlanations（SHAP） main grain producing areas in the middle reaches of the Yangtze River |
| DOI 10.13227/j.hjkx.202508169 |
|
| 中文摘要 |
| 辨识粮食主产区农业生产碳收支平衡体系是推动农业绿色转型的有效技术支撑. 基于“动-静”结合视角测算2005~2023年长江中游粮食主产区农业生产碳收支平衡水平，分别探究其静态平衡水平和动态平衡水平的时空分异特征. 在比较4种机器学习模型精度的基础上，采用XGBoost与SHAP模型识别农业生产碳收支动态平衡水平变化的主要影响因素与影响程度，为制定针对性的农业碳减排策略提供科学依据. 结果表明：①长江中游粮食主产区农业生产碳收支静态平衡水平整体呈缓慢下降趋势，区域内部呈现显著市际差异，41.935%的城市呈U型演变，38.710%呈倒U型演变. ②2023年农业生产碳收支动态平衡指数显著提升，其中仅有荆门、益阳和新余等6个城市FDBI<0（2020~2023年）. 在空间分区上，以2014年为分界点，空间格局从“东南部高平衡、西北部低平衡”演变为高平衡区域“南北双核”集聚的格局. ③农业生产碳收支动态平衡水平主要受：乡村人口数、农地规模、农业从业人员与第一产业结构影响. 其中，乡村人口数影响呈“V型”趋势；农地规模在0.160~0.231 hm2·人-1区间内具规模报酬递增效应；农业从业人员与一产结构则均在超临界值后呈现系统性负向效应. ④在农业从业人口数量较多与农地经营规模较小的组合情境下，对碳收支动态平衡值的预测贡献相对有限. 研究表明，长江中游粮食主产区部分城市仍存在碳收支失衡情况，需针对关键影响因素实施差异性的农业碳减排策略，以促进区域农业生产的绿色低碳发展. |
| 英文摘要 |
| Establishing a carbon balance system for agricultural production in major grain-producing regions serves as a crucial technical foundation for advancing green agricultural transformation. This study evaluates the carbon balance levels of agricultural production in the Yangtze River Midstream Grain Production Zone from 2005 to 2023 using a dynamic-static integrated perspective. It analyzes both static and dynamic equilibrium patterns across time and space， compares four machine learning models， and employs XGBoost and SHAP models to identify key factors influencing changes in dynamic equilibrium levels， thereby providing scientific support for targeted agricultural carbon reduction strategies. Key findings include： ① The static carbon balance in the Yangtze River Midstream Grain Production Zone showed an overall gradual decline， with significant regional disparities. Among 31 cities， 38.710% exhibited inverted U-shaped trends， while 41.935% followed U-shaped patterns. ② The dynamic carbon balance index demonstrated substantial improvement in 2023， with only six cities including Jingmen， Yiyang， and Xinyu maintaining FDBI values below 0 （2020-2023）. Spatially， the region transitioned from a “high balance in the southeast， low balance in the northwest” pattern to a “dual-core north-south cluster” configuration after 2014. ③ The dynamic equilibrium level was primarily influenced by rural population size， farmland scale， agricultural workforce， and primary industry structure. The rural population showed a “V-shaped” trend in its impact； agricultural land scale demonstrated increasing returns to scale within the range of 0.160-0.231 hectares per capita； both agricultural employment and primary industry structure exhibited systematic negative effects beyond critical thresholds. ④ The combination of large agricultural workforce and small-scale land operations contributed minimally to predicting dynamic carbon balance equilibrium. The research indicates that some cities in the Yangtze River Midstream Grain Production Zone still face carbon imbalance issues， necessitating differentiated agricultural carbon reduction strategies targeting key influencing factors to promote sustainable regional agricultural development. |
