| 徐祥明,张心怡,覃灵华,纪玲玲,李瑞.未来气候情景下中国2030年土壤阳离子交换量预测——基于机器学习模型[J].农业环境科学学报,2025,44(11):2852-2863. |
| 未来气候情景下中国2030年土壤阳离子交换量预测——基于机器学习模型 |
| Prediction of soil cation exchange capacity in China based on machine learning and future climate scenarios of CMIP6, 2030 |
| 投稿时间：2025-05-20 |
| DOI：10.11654/jaes.2025-0470 |
| 中文关键词: 环境变量 XGBoost SHAP 不确定性分析 阳离子交换量 |
| 英文关键词: environmental variables XGBoost SHAP uncertainty analysis cation exchange capacity |
| 基金项目:国家自然科学基金项目（42261015）；江西省教育厅科技项目（GJJ211426，GJJ201440） |
|
| 摘要点击次数: 1027 |
| 全文下载次数: 763 |
| 中文摘要: |
| 为了精准预测未来气候情景下的土壤阳离子交换量（CEC），本研究基于多源环境协变量数据集，系统比较了5种机器学习模型在我国全域土壤CEC预测中的表现，并定量揭示了在SSP2-4.5和SSP3-7.0未来气候情景下，2030年我国土壤CEC含量的空间分异特征及其预测结果的空间不确定性。研究结果表明，XGBoost模型在各项指标预测中均表现最优。土壤有机碳（贡献率为18.41%）和黏粒含量（贡献率为18.03%）是CEC的主要驱动因子，其次是温度、海拔及pH（贡献率均>10%）。其中，土壤有机碳、黏粒、总氮、降雨、坡度和人为活动指数对CEC具有正向影响，其余因子则呈抑制作用。未来气候情景预测表明，在SSP2-4.5情景下，2030年我国土壤CEC含量总体保持稳定；而在SSP3-7.0情景下，CEC含量将增至17.48 cmol·kg-1。从空间分布来看，东北地区CEC含量最高，其次为华南、西南、华中和华北地区，西北地区最低。两种情景下的预测不确定性总体相近，但存在区域差异：SSP2-4.5情景下，华东和华南地区不确定性最大，华北最小；SSP3-7.0情景下，华中地区置信区间最宽。由此可见，XGBoost模型是未来气候情景下土壤CEC预测的最优模型，2030年我国土壤CEC含量的空间分布及不确定性分析具有区域差异。 |
| 英文摘要: |
| In order to make an accurate prediction of soil cation exchange capacity（CEC）under future climate scenarios, this study systematically compared the performance of five machine learning models in predicting soil CEC across China. These models were based on a multi- source environmental covariate dataset. Additionally, the study quantified the spatial differentiation characteristics of soil CEC content in China in 2030 under the SSP2-4.5 and SSP3-7.0 future climate scenarios, along with the spatial uncertainty of the prediction results. The findings suggest that the XGBoost model demonstrates optimal performance across all metrics. The primary drivers of CEC are thus determined as soil organic carbon（contribution rate of 18.41%）and clay content（18.03%）, followed by temperature, elevation, and pH（all with contribution rates exceeding 10%）. Among these factors, soil organic carbon, clay content, total nitrogen, precipitation, slope, and human activity index have been shown to positively impact on CEC, while the remaining factors exhibit inhibitory effects. Projections indicate that under the SSP2-4.5 scenario, soil CEC content will remain stable by 2030; under the SSP3-7.0 scenario, CEC content will increase to 17.48 cmol·kg-1. From a spatial distribution perspective, The Northeast Region has the highest CEC content, followed by The South China Region, The Southwest China Region, The Central China Region, and The North China Region, with The Northwest Region having the lowest. The overall uncertainty in predictions under the two scenarios is similar, but regional differences exist. Under the SSP2-4.5 scenario, the highest uncertainty is observed in The East China Region and The South China Region, with the lowest in The North China Region. Under the SSP3-7.0 scenario, The Central China Region has the widest confidence interval. In summary, the XGBoost model is the optimal model for predicting soil CEC under future climate scenarios. The spatial distribution and uncertainty analysis of soil CEC content in China in 2030 demonstrate significant regional variations. |
| HTML
查看全文
查看/发表评论 下载PDF阅读器 |
|
|
|
