表征全吸力范围的土壤水分特征曲线模型评估及其转换函数构建

基于非饱和土壤水力性质数据库（UNSODA）中选取的从砂土到黏土共256个土壤样本,系统性地评价了表征全吸力范围的土壤水分特征曲线模型（LIAO模型）的适用性,并构建和验证了预测LIAO模型参数的土壤转换函数（PTFs）。结果表明：（1）与传统的van Genuchten模型（仅适用于描述毛管水运动）相比,LIAO模型对不同质地土壤水分特征曲线的预测精度更高,均方根误差（RMSE）降低了约45%;（2）LIAO模型参数与土壤基本性质（如砂粒、粉粒、黏粒、有机质含量和容重）之间存在不同程度的相关性,其中参数θ_s与容重的相关性（r=-0.783,P<0.01）最强,而其余参数与粉粒的相关程度最高;（3）基于逐步回归方法构建的PTFs能够解释LIAO模型参数总变异的31%~65%,其中对θ_s的预测精度最高,经双交叉验证表明PTFs稳定性较好。研究成果可为区域（尤其是干旱和半干旱地区）土壤水文模型提供参数支持。

Evaluation of soil water retention curve model from saturation to oven-dryness and development of pedotransfer functions for predicting model parameters

Capillary water and film water are two important forms of soil water. In humid and semi-humid areas, capillary water is the water that moves quickly and is easily absorbed and utilized by plants. However, in arid and semi-arid areas, film water becomes the key factor limiting soil and water conservation and vegetation growth, and nutrient migration and utilization in soil due to the low soil water content. Soil water characteristic curve (SWRC) is an important soil hydraulic property, which represents the functional relationship between soil water content and pressure head, and is an essential parameter of the soil water movement model. Therefore, a model is urgently needed to describe the SWRC with consideration of the characteristics of capillary water and film water simultaneously. Recently, LIAO model has been proposed to represent the SWRC due to both capillary water and film water. However, this model has not been systematically evaluated. In this study, a total of 256 soil samples from coarse- to fine-textured were selected from unsaturated soil hydraulic property database (UNSODA). These samples were then applied to evaluate the soil water retention curve model from saturation to oven-dryness (LIAO model). Finally, pedotransfer functions (PTFs) for predicting LIAO model parameters were developed and validated. The results showed that: (1) Compared with the classic van Genuchten model which can only describe soil water retention characteristics due to capillary flow, LIAO model can better predict water retention curves for different textured soils with the decrease in root mean squared error by 45%. (2) LIAO model parameters were found to have different degrees of correlation with basic soil properties such as sand, silt, clay, organic matter contents and bulk density. θ_s had the strongest correlation (r=-0.783, P<0.01) with bulk density, while other parameters had the strongest correlation with silt. (3) Stepwise regression based-PTFs can explain 31%-65% of the variance in LIAO model parameters. The highest accuracy was found for θ_s. The signs of the regression coefficients and the determination coefficients were stable by the double cross-validation method. The results provided supports for soil water movement simulation at regional scale, especially in arid and semi-arid environments.