Dataset underlying the publication: Machine learning for predicting spatially variable lateral hydraulic conductivity: a step towards efficient hydrological model calibration and global applicability

Two globally distributed maps of horizontal-to-vertical saturated hydraulic conductivity (fKh0) were generated using machine learning algorithms using random forest and boosted regression trees. Linking the calibrated benchmark of fKh0 achieved by Weerts et al. (2024) over 551 subbasins over the Great Britain to the structural soil properties from SoilGrids v1.0, we estimate pedo-transfer functions to predict fKh0 values globally at 250m spatial resolution.<br>Reference:Weerts, A. H. (2024). Dataset underlying the publication: Revealing spatial patterns of lateral hydraulic conductivity through sensitivity analysis. 4TU.ResearchData. Retrieved from https://doi.org/10.4121/6026ee8f-1e37-4760-abb6-b0a6251b3089.v2

本研究采用随机森林（Random Forest）与提升回归树（Boosted Regression Trees）两种机器学习算法，生成了两张水平向-垂向饱和导水率之比（fKh0）的全球分布图谱。本研究将Weerts等人（2024年）在英国境内的551个子流域中得到的校准后fKh0基准值，与土壤网格（SoilGrids）v1.0的结构性土壤属性进行关联，进而估算土壤转换函数（pedo-transfer functions），以250米空间分辨率在全球范围内预测fKh0数值。

参考文献：Weerts, A. H. (2024). 支撑该学术论文的数据集：通过敏感性分析揭示侧向导水率的空间分布格局. 4TU.ResearchData. 检索自：https://doi.org/10.4121/6026ee8f-1e37-4760-abb6-b0a6251b3089.v2