基于机器学习与SHAP算法的胶东半岛土壤酸化影响因素分析

Analysis of factors influencing soil acidification in the Jiaodong Peninsula based on machine learning and the SHAP algorithm

  • 摘要: 土壤酸化已成为制约胶东半岛农业可持续发展的关键问题,但其空间分异的驱动因子尚不明确。为识别影响土壤酸化的关键因子及其作用,于2021年5月在胶东半岛采集144个农田耕层土壤样品,应用随机森林(RF)模型结合SHAP(SHapley Additive exPlanations)算法,对土壤pH及其13个影响因子进行了系统分析。结果表明,胶东半岛土壤pH呈现由西北向东南递减的空间分异特征,东南部区域酸化问题突出。RF模型实现了对土壤pH的准确预测(R2=0.9999),并通过SHAP分析进一步量化了各影响因子的贡献。胶东半岛土壤pH变化的主要驱动因素为硝态氮(NO3-N)与土壤自身缓冲能力(钙镁盐基离子),其中NO3-N贡献率(84.0%)显著高于交换性Ca2+、Mg2+及阳离子交换量(CEC),且NO3-N与Ca2+有强烈的正向交互作用,交换性Ca2+、Mg2+与CEC间存在正向协同作用。土壤pH与NO3-N存在显著的正相关关系(P<0.001),并且土壤NO3-N、Ca2+、Mg2+和CEC空间数值大小与土壤pH值的空间分布高度一致。高缓冲能力土壤对硝态氮和盐基离子的协同固持效应能够有效抑制胶东半岛西北部土壤pH下降;相反,胶东半岛东南部区域的严重酸化,主要是由于该地区低于土壤自身缓冲能力条件下的氮输入导致的。研究结果为该区域防控土壤酸化提供了关键科学依据。

     

    Abstract: Soil acidification has become a critical issue constraining the sustainable agricultural development of the Jiaodong Peninsula, yet the driving factors of its spatial differentiation remain unclear. To identify the key factors influencing soil acidification and clarify their roles, 144 farmland topsoil samples were collected from the Jiaodong Peninsula in May 2021. A Random Forest (RF) model combined with the SHAP (shapley additive explanations) algorithm was employed to systematically analyze soil pH and its 13 influencing factors. The results indicate that soil pH in the Jiaodong Peninsula exhibites a spatial differentiation pattern, decreasing from the northwest to the southeast, with prominent acidification issues in the southeastern region. The RF model achieves an accurate prediction of soil pH (R2=0.9999), while SHAP analysis further quantifies the contribution of each influencing factor. The primary drivers of soil pH variation are identified as nitrate-nitrogen (NO3-N) and the soil's intrinsic buffering capacity (calcium and magnesium base cations). Notably, the contribution of NO3-N (84.0%) is significantly higher than that of exchangeable Ca2+, Mg2+, and cation exchange capacity (CEC). Furthermore, a strong positive interaction is observed between NO3-N and Ca2+, along with a positive synergistic effect among exchangeable Ca2+, Mg2+, and CEC. A significant positive correlation between soil pH and NO3-N(P<0.001) is found, and the spatial distributions of soil NO3-N, Ca2+, Mg2+, and CEC values are highly consistent with that of soil pH. It is revealed that the synergistic retention effect of soils with high buffering capacity on nitrate and base cations effectively inhibits the decline in soil pH in the northwestern Jiaodong Peninsula. Conversely, the severe acidification in the southeastern region is primarily caused by nitrogen input into soils with low intrinsic buffering capacity. These findings provide a critical scientific basis for the prevention and control of soil acidification in the region.

     

/

返回文章
返回