| 引用本文: | 王旭,朱金凤,STRAUSS Peter,等.土壤性质相似性对土壤转换函数模拟性能及不确定性的影响[J].灌溉排水学报,0,():-. |
| WangXu,ZhuJinfeng,STRAUSS Peter,et al.土壤性质相似性对土壤转换函数模拟性能及不确定性的影响[J].灌溉排水学报,0,():-. |
|
| 摘要: |
| 为了探究土壤属性不相似性对于土壤转换函数预测的不确定性及误差的影响，并识别土壤转换函数预测的不确定成分的主要来源，本文以田间持水量（θ-300）和凋萎含水量（θ-15000）为研究对象，基于不同性质（相似或不相似）的数据集对土壤转换函数的不确定性进行分析，使用MAE、ME和SDE等指标量化模型模拟性能；并采用拉丁超立方抽样和Bootstrap方法对不确定性来源贡献进行分析。结果表明，无论θ-300或θ-15000，基于相似数据集建立的PTFs模型模拟性能有所提升，但效果较不明显。从不确定成分的主要贡献来看，对于θ-300，无论是基于相似还是不相似数据集，模型输入引起的不确定性贡献较大；但对于θ-15000，基于不相似数据集建立的PTFs则相反，即模型参数不确定性对模拟预测结果的影响占比较大，原因主要在于θ-15000往往具有较小的空间变异特征，因此其输入不确定性通常都较小。研究指出，在针对农区构建PTFs预测田间持水量这一常数特征时，控制土壤样本理化性质的相似性有助于减少构建模型模拟结果的不确定性；而构建PTFs预测凋萎含水量时，则样本采集对模拟结果不确定性无明显影响，模拟结果的不确定性主要由构建模型的模型参数控制。 |
| 关键词: PTFs；相似性；不确定性；LHS；Bootstrap |
| DOI： |
| 分类号:S152.7 |
| 基金项目: |
|
| Impact of Soil Properties Similarity on Simulation Performance and Uncertainty |
|
WangXu1, ZhuJinfeng1, STRAUSS Peter2, ZhangHao1, WangShengping1
|
|
1.North China Electric Power University;2.Federal Agency for Water Management,Institute for Land and Water Management Research
|
| Abstract: |
| In order to explore the impact of soil properties dissimilarity on the uncertainty of PTFs, and to identify the main sources of uncertainty of PTFs, used field capacity water content (θ-300) and wilting point water content (θ-15000) as the research object, based on the data set of different properties (similar or dissimilar dataset) to analyze the uncertainty of the PTFs.And used MAE, ME and SDE indicators to quantify the model simulation performance; This study used the Latin Hypercube Sampling and Bootstrap methods to analyze the sources of uncertainty. The results showed that, regardless of θ-300 or θ-15000, the simulation performance of PTFs model based on similar data sets is improved, but it is not obvious. From the main sources of uncertainty, for θ-300, whether based on similar or dissimilar datasets, the uncertainty caused by input uncertainty was more important; but for θ-15000, the impact of parameter uncertainty is relatively larger than the input uncertainty. The main reason is that θ-15000 data set tends to have smaller spatial variability characteristics. We emphasized that when developing PTFs for predicting field capacity water content, it would be helpful to reduce modeling uncertainty by using similar soil samples only. While it might be worth putting much effort to model calibration and model construction when developing PTFs for predicting wilting point water content. |
| Key words: PTFs；similarity；uncertainty；LHS；Bootstrap |
