土壤饱和导水率空间预测的不确定性分析

Uncertainty analysis for spatial prediction of soil saturated hydraulic conductivity

当土壤转换函数应用于土壤水力性质估计时,对于预测值的不确定性往往容易被忽视。为了有针对性地提出减少这种不确定性的方法和措施,提高土壤转换函数的实际应用能力,以两种现有的土壤转换函数(Vereecken和HYPRES模型)为例,将其应用于山东省平度市土壤饱和导水率的空间预测,并利用拉丁超立方抽样(LHS)方法对预测结果的不确定性进行了分析。结果表明,饱和导水率空间预测的不确定性主要来源于土壤基本性质的空间插值误差和土壤转换函数自身的预测误差。当Vereecken模型应用于饱和导水率空间预测时,预测结果的不确定性主要由土壤基本性质空间插值误差所决定,土壤转换函数预测误差的影响较小,而HYPRES模型则是受二者的双重影响。

Pedotransfer functions are commonly used to predict soil hydraulic characteristics.However,the analysis of prediction uncertainty is often ignored by researchers.In an attempt to reduce the prediction uncertainty and to improve the applicability of pedotransfer functions,two existing pedotransfer functions are used to predict the spatial distribution of soil saturated hydraulic conductivity in the Pingdu city of Shandong Province.The two functions are the Vereecken＇s pedotransfer function and the database of HYdraulic PRoperties of European Soils（HYPRES）.The Latin hypercube sampling method is applied to estimate the uncertainty in predicting soil saturated hydraulic conductivity.Results show that there are two major sources of uncertainty,which are：（1） the uncertainty in the spatial interpolation of basic soil properties using Kriging;and（2） the uncertainty associated with the use of pedotransfer functions to predict soil saturated hydraulic functions.The Kriging error is the major source of uncertainty in predicting soil saturated hydraulic conductivity using the Vereecken＇s pedotransfer function.While,the prediction results are affected by both HYPRES and Kriging.