基于单一和集合土壤转换函数模型对土壤含水量的模拟性能分析

Performance Analysis of Simulating Soil Moisture Content Based on Single and Ensemble Pedotransfer Function

以海河流域3个实验场地(密云站、大兴站、馆陶站)为研究区,采用7种常用的单一土壤转换函数(soil pedotransfer function,PTF)模型预测土壤水力参数作为HYDRUS-1D的模型参数,求解Richards方程获得土壤含水量,并与实测土壤含水量进行比较,评价了常用单一PTF模型预测的土壤水力参数对土壤含水量的模拟性能。此外,采用3种方法构建集合PTF模型,评价了集合PTF模型对土壤含水量的模拟性能。结果表明:基于van Genuchten方程构建的单一PTF作为模型参数模拟土壤含水量的均方根误差最小;而其中Rosetta3模型表现更优。在集合PTF模型中,基于贝叶斯平均法构建的模型表现最好。集合PTF模型预测土壤水力参数可以较好地捕捉多个单一PTF预测土壤水力参数的整体趋势,弥补单一PTF在某些情况下模拟误差较大的不足。

In this study,soil hydraulic parameters estimated by 7 commonly used soil pedotransfer functions(PTFs)were used as the parameters fed into the HYDRUS-1D model,and soil moisture was obtained by solving Richards equation,and was compared with the measured dataset at Miyun,Daxing and Guantao stations in Haihe River Basin.The simulation performance of soil hydraulic parameters estimated by single PTF on the estimated soil moisture was evaluated.In addition,soil hydraulic parameters were also estimated by three newly developed ensemble PTF,and the influence of ensemble PTF models on the simulation performance of soil moisture was evaluated.Results show that RMSE of single PTF model based on van Genuchten equation is the lowest,and the Rosetta3 PTF shows a better performance.The ensemble PTF model developed based on Bayesian averaging model has the best performance in terms of characterizing the overall trend of soil hydraulic parameters estimated by multiple single PTF.In addition,the ensemble model can overcome the deficiency of large error resulted from the single PTF in some cases.