Transfer Function‐Noise Modeling Using Remote Sensing Data to Characterize Soil Moisture Dynamics: a Data-driven Approach.
Transfer Function‐Noise Modeling Using Remote Sensing Data to Characterize Soil Moisture Dynamics: a Data-driven Approach.
复制标题
使用遥感数据表征土壤湿度动态的传递函数噪声模型：一种数据驱动的方法。
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
S. Hulscher
中科院分区：
文献类型：
--
作者：
M. Pezij;D. Augustijn;D. Hendriks;S. Hulscher
虽然许多人认为土壤水分作为一个重要的水文变量，土壤水分模型的应用业务水管理是有限的。高分辨率遥感土壤湿度数据的提供为探索数据驱动的建模方法带来了新的机会。在这项研究中，我们评估了传递函数噪声模型（TFN）描述土壤水分动态的适用性。TFN已被应用于地下水模拟描述长期动态。我们表明，TFN是有用的土壤水分建模。TFN应用自回归移动平均法（阿尔马），将观测到的时间序列与使用脉冲响应函数的输入应力相关联。输入应力与脉冲响应函数卷积，以确定由于该应力引起的土壤水分响应。TFN的优点是它不需要对系统过程进行预先假设。在这项研究中，Python3包Pastas用于建立TFN模型。此外，降水和参考作物蒸散应力被用来解释土壤水分动态。SMAP L3增强型表层土壤水分产品用作训练数据集。结果表明，TFN能较准确地描述土壤水分动态.特别是，时间序列模型可以用来表征土壤水分对降水和蒸散胁迫的响应。降水量和蒸散量的脉冲响应函数描述了土壤水分对这些应力的绝对变化和反应时间的敏感性。这些属性的空间分布描述了水系特征。这项研究的结果使水资源管理人员能够采取更稳健的决策，因为水资源管理人员可以深入了解非饱和带动态。进一步的研究可能集中在TFN预测短期土壤水分动态的适用性。
Although many consider soil moisture as an important hydrological variable, the application of soil moisture modelling for operational water management is limited. The availability of high-resolution remotely sensed soil moisture data leads to new opportunities for exploring data-driven modelling methods. In this study, we assessed the applicability of transfer function‐noise modeling (TFN) for describing soil moisture dynamics. TFN has been applied in groundwater modelling for describing long-term dynamics. We show that TFN is useful for soil moisture modelling. TFN applies an autoregressive-moving-average method (ARMA) to relate observed time series to input stresses using impulse-response functions. The input stress is convoluted with the impulse-response function to determine the soil moisture response due to that stress. The advantage of TFN is that it does not require prior assumptions on system processes. In this study, the Python 3 package Pastas is utilized for setting up the TFN model. Furthermore, precipitation and reference crop evapotranspiration stresses are used to explain soil moisture dynamics. The SMAP L3 Enhanced surface soil moisture product is used as a training dataset. We found that the TFN can accurately describe soil moisture dynamics. In particular, the time series model can be used to characterize the response of soil moisture to precipitation and evapotranspiration stresses. The individual impulse-response functions for precipitation and evapotranspiration describe the sensitivity of soil moisture in terms of absolute change and reaction time to these stresses. The spatial distribution of these attributes describes water system characteristics. The results of this study enable water management to take more robust decisions, as water managers get insight in unsaturated zone dynamics. Further research may focus on the applicability of TFN for predicting short-term soil moisture dynamics.