资源环境科技发展态势分析平台

The spatial and temporal variability in soil moisture modulates runoff generation and the degree of land-atmosphere coupling. Numerous statistical and modeling approaches have been implemented to characterize soil moisture spatial heterogeneity at fine spatial resolution using data from sparse observational networks or distributed model simulations. This characterization has been subsequently employed to translate coarse model simulations (of the order of a few hundred meters or kilometers) to finer spatial scales for a range of ensuing applications that rely on high-resolution characterization of soil moisture. One common feature of these disaggregation methods is that the impact of soil moisture memory is ignored. This results in both spatial and temporal persistence being poorly simulated, leading to poorer specifications of cropping and irrigation plans. To overcome this shortcoming, we developed a hybrid disaggregation method that uses the first-order autoregressive model (AR1) constructed from fine-resolution (60m) soil moisture simulations to disaggregate catchment mean soil moisture obtained from remote sensing or semidistributed model simulations. Soil moisture simulations from an integrated land surface-groundwater model, ParFlow-Common Land Model in Baldry subcatchment, Australia, are used as virtual observations. We examined the AR1 method performance against topographic wetness index-based methods and those developed from temporal stability method. Results illustrate that the disaggregation schemes calibrated to a 10-day fine-scale model simulation perform better than the topographic-based methods in approximating soil moisture distribution at a 60-m resolution in the catchment. Furthermore, the AR1 model is the best model (Nash-Sutcliffe efficiency [NSE]>0.45) among various alternatives explored here. Applying the hybrid univariate AR1 model is promising for disaggregating semidistributed models' soil moisture simulations while significantly reducing the computational time.