土壤水分特征曲线模型模拟性能评价

土壤水分特征曲线模型作为实验测定土壤水分特征数据的一种替代方法，因其具有计算方便快捷和便于嵌入数值模拟程序的优点，开始受到越来越广泛的关注。虽然文献中存在众多的土壤水分特征曲线模型，但是这些模型的适用范围及拟合性能尚不明确。为了获得更加准确适用的土壤水分特征曲线，在实际应用中通常需要花费大量时间和精力去测试各种模型。为了解决上述问题，在国内外研究成果的基础上收集整理了12种典型的土壤水分特征曲线模型，并利用包含不同质地、有机质含量及容重的8种土壤的实测土壤水分特征数据来评估比较这些模型的模拟性能。模型性能通过均方根误差（RMSE）、平均偏差（AD）、AIC准则（Akaike Information Criterion）和纳什效率系数（NSE）4个指标评估。研究结果表明：大部分的模型能够提供比较接近于实际的拟合结果，评价指标值也比较相近。其中，KCGS2006（包含3个参数）和K1999模型（包含2个参数）拟合效果最好，而Gregson1987（包含1个参数）的拟合效果最差。该研究可以深入了解各种土壤水分特征曲线模型的适用性与局限性，更好地为生态环境建设和农业可持续发展研究中土壤水力参数的选取提供依据和参考。

Performance evaluation of 12 models describing the soil water retention characteristics

Models for estimating soil water retention characteristic curves (SWRCs), as an alternative method for measuring soil moisture by direct experiments, are attracting more and more attention because of its convenience, rapidness and ease of calculation. But the applicability and fitting performance of these models have not fully understood in practical application in order to obtain accurate SWRC. In this study, 12 SWRC models were selected and evaluated with experimental measurements of SWRC data for eight soils in literature. Four indices including root mean square error (RMSE), average deviation (AD), Nash efficiency coefficient (NSE) and Akaike Information Criterion (AIC) were used to evaluate the performance of these models. The result showed that KCGS2006 with 3 parameters and K1999 with 2 parameters ranked the top among the 12 models, while the Gregson1987 model performed the worst. The performance or applicability of the SWRC models largely depend on the algorithms used for the expressions, the parameters (degree of freedoms), the specific objectives they were developed for problem solving (e.g., soil types or matric potential/water content range). This study could provide information and guide on SWRC selection for disciplines including earth and environmental science, ecology, agronomy or sustainable agricultural development.