高级检索

    基于Meta分析的黄土高原坡面典型水土保持措施减流减沙效应研究

    A Meta-Analysis of the Runoff and Sediment Reduction Effects of Typical Soil and Water Conservation Measures on Slopes of the Loess Plateau

    • 摘要: 背景黄土高原是我国生态环境最为脆弱、且水土流失最严重的区域。评估水土保持措施效益并明晰驱动机制对区域生态安全具有重要意义。方法本研究基于次降雨事件,采用 Meta 分析整合降雨量(P)、坡度(S)、坡长(L)和措施类型(M)等多源数据,结合相关性分析、冗余分析与机器学习预测模型,揭示了不同水土保持措施的减流减沙效果差异。结果(1)植被与工程措施中,减流效益最高的分别为灌木(50.07%)和鱼鳞坑(36.93%);减沙效益最高的则为灌木(76.41%)和梯田(83.28%)。农业措施中,减流和减沙效益最高的分别是垄作(78.62%)和等高耕作(40.63%)。(2)降雨量(P)与工程措施减沙效益呈极显著负相关(p < 0.001);坡度(S)与植被措施减流效益呈极显著正相关(p < 0.001),而与工程措施减流、农业措施减沙分别呈显著(p < 0.05)和极显著负相关(p < 0.001);坡长(L)与植被措施减沙、工程措施减流效益呈高度显著(p < 0.01)负相关。冗余分析表明,在减流效益方面,植被、工程和农业措施条件下贡献率最高的因子分别为措施类型(M)、坡长(L)和措施类型(M),而在减沙效益方面则分别是坡长(L)、降雨量(P)和坡度(S)。(3)基于机器学习模型预测各水土保持措施的减流减沙效益,其中XGBoost对植被措施减流预测精度最高(R²=0.74),多层感知机(MLP)对工程措施减沙预测最优(R²=0.69),随机森林对农业措施减流预测误差最小(R2=0.51)。结论本研究从相关性与因子贡献角度系统阐明了不同水土保持措施效益差异的主导驱动因素,为黄土高原水土保持措施优化配置提供了科学依据。

       

      Abstract: Background The Loess Plateau, widely acknowledged as one of the most ecologically fragile regions in China with the most severe soil erosion, serves as a critical area for soil and water conservation. Assessing the effectiveness and underlying driving mechanisms of conservation measures is essential for enhancing regional ecological security and sustainability. MethodsThis study was based on the observation of individual rainfall events. Meta-analysis was adopted to integrate multi-source data, including precipitation (P), slope gradient (S), slope length (L), and measure type (M). Furthermore, correlation analysis, redundancy analysis, and machine learning prediction models were combined to reveal the differences in the effects of different soil and water conservation measures and their dominant factors. Results 1) Among vegetation and engineering measures, shrubs (48.88%) and fish-scale pits (38.64%) had the highest runoff reduction efficiency (RR), while shrubs (76.41%) and terraces (72.31%) achieved the highest sediment reduction efficiency (SR). For agricultural measures, ridge tillage achieved the highest RR (78.62%), whereas contour tillage showed the highest SR (40.63%). 2) Correlation analysis showed P was extremely significantly negatively correlated with engineering SR (p < 0.001); S was extremely significantly positively correlated with vegetation RR (p < 0.001), but significantly (p < 0.05) and extremely significantly (p < 0.001) negatively correlated with engineering RR and agricultural SR, respectively; L had highly (p < 0.01) significant negative correlations with vegetation SR and engineering RR. Redundancy analysis indicated M, L, M were the top contributors to RR of the three measure types, and L, P, S dominated their SR. 3) Machine learning models predicted the RR and SR of soil and water conservation measures. XGBoost achieved the highest accuracy for vegetation RR (R² = 0.74), MLP (Multi-Layer Perceptron) for engineering SR (R² = 0.69), and Random Forest the smallest error for agricultural RR (R2 = 0.51). Conclusion This study systematically elucidates the differential effectiveness and key driving mechanisms of various soil and water conservation measures, thereby providing a scientific basis for their optimal allocation on the Loess Plateau.

       

    /

    返回文章
    返回