利用机器学习预测饱和水电导率及其对土壤结构扰动的敏感性

重点：
基于土壤转换函数建立了高精度机器学习模型来预测饱和水力传导度
变量重要性测量被用来识别和排序对预测重要的土壤属性
用容重和有机碳含量作为预测土壤结构变化的指标

Key Points:
High accuracy machine learning based pedotransfer function models are developed to predict saturated hydraulic conductivity;
Variable importance measures are used to identify and rank soil properties that are most important for predictions;
Bulk density and organic carbon content are used as proxies to describe the effect of soil structural alterations in predictions

饱和水力传导度(Ks)是土壤的一种基本性质，它决定着土壤中水分的去向，然而，它的测量是繁琐的，通常使用土壤转换函数(PTFs)来估计它。尽管这些年取得了很大的进步，但是目前通用的PTFs估计Ks的性能仍然很差。利用机器学习、高性能计算和超过18000个土壤的大型数据库，我们开发了新的PTFs来预测Ks。我们比较了四种机器学习算法和不同预测集的性能。我们评估了土壤性质在解释Ks中的相对重要性。基于改进回归树算法的PTF模型产生了最佳模型，其均方根转换误差在0.4到0.3 (log10(cm/day))之间。第十百分位粒径(d10)被发现是仅次于粘土含量、容重(ρb)和有机碳含量(C)的重要的预测因子。使用ρb和C表征土壤结构来研究Ks对土壤结构的敏感性。观测到ρb与Ks之间存在相反的关系，对于大多数结构，在1.8g/cm3时敏感性最强。土壤C与Ks呈复杂的关系，细和中等结构总体上呈正相关，粗结构土壤总体上呈负相关。本研究试图从大型数据库中最大限度地提取信息，以开发基于通用机器学习的PTF来估计Ks。这里开发的模型已经公开，可以很容易地用于预测Ks。

Saturated hydraulic conductivity (Ks) is a fundamental soil property that regulates the fate of water in soils. Its measurement, however, is cumbersome and instead pedotransfer functions (PTFs) are routinely used to estimate it. Despite much progress over the years, the performance of current generic PTFs estimating Ks remains poor. Using machine learning, high performance computing, and a large database of over 18,000 soils, we developed new PTFs to predict Ks. We compared the performances of four machine learning algorithms and different predictor sets. We evaluated the relative importance of soil properties in explaining Ks. PTF models based on boosted regression tree algorithm produced the best models with root mean squared log transformed error in ranges of 0.4 to 0.3 (log10(cm/day)). The 10th percentile particle diameter (d10) was found to be the most important predictor followed by clay content, bulk density (ρb), and organic carbon content ©. The sensitivity of Ks to soil structure was investigated using ρb and C as proxies for soil structure. An inverse relationship was observed between ρb and Ks, with the highest sensitivity at around 1.8 g/cm3 for most textural classes. Soil C showed a complex relationship with Ks with an overall positive relation for fine textured and midtextured soils but an inverse relation for coarse textured soils. This study sought to maximize the extraction of information from a large database to develop generic machine learning based PTFs for estimating Ks. Models developed here have been made publicly available and can be readily used to predict Ks.

[1]. Araya, S.N. and T.A. Ghezzehei, Using Machine Learning for Prediction of Saturated Hydraulic Conductivity and Its Sensitivity to Soil Structural Perturbations. Water Resources Research, 2019. 55(7): p. 5715-5737.