Search

In cold regions, the coexistence and interconversion of liquid water and ice in frozen soils have important implications for energy partitioning and surface runoff at the Earth's surface. Passive microwave remote sensing is crucial for the global monitoring of soil moisture (SM). However, current research on SM focuses mainly on unfrozen soil conditions. Limited studies have been conducted on variations in soil liquid water content throughout the freezing season. This article investigated the potential use of brightness temperature observations from the Soil Moisture Active Passive (SMAP) satellite for retrieving all-season liquid SM. The single-channel algorithm and the Zhang-Zhao dielectric model, which was specifically developed for freezing and thawing soils, achieved successful retrieval of liquid SM in both frozen and thawed soils, even when snow cover was present. The results indicate improved spatial coverage (during winter) and consistent spatial patterns in SM compared with the SMAP products. Validation at 17 SM networks suggests that the retrieved all-season liquid SM effectively captures the dynamic characteristics of each region with an average bias of 0.011 m3/m3, an average unbiased root mean square error of 0.056 m3/m3, and an average correlation coefficient of 0.76. The additional retrieval of unfrozen water content during the freezing season would enhance the monitoring and understanding of the hydrological cycle and energy balance in cold regions.

Fusion of thermal infrared (TIR) and passive microwave (PMW) observations are important techniques for estimating all-weather land surface temperature (LST) of relatively higher resolutions. For such fusion techniques under cloudy weather, PMW observations are especially useful in correcting the cloud-shadowing bias of the real cloudy LST against a theoretically clear-sky one in the daytime. However, PMW observations have very limited transit times in a day, and cannot be fused with asynchronous TIR data as LST can vary significantly during the daytime. To overcome this flaw, we developed a model for simulating the variation of cloud-shadowing bias based on cloud total amount (CTA) information observed by geostationary satellites. As constant coefficients of the model were found to have dependence on space and time, they were dynamically fitted for each pixel in each month using cloud-shadowing bias calculated by TIR and synchronous PMW data in the early-afternoon (1:30 P.M.). The model was then used to provide middle-morning (10:30 A.M.) cloud-shadowing bias estimates, based on which, all-weather LST at 1 km were produced at that specific moment. Validation results show that reconstructed cloudy LST in the middle-morning, with no synchronous PMW observations, has its overall RMSE values against ground measurements and real remote sensing data under cloud between 2.4 and 4.0 K. This accuracy result is generally comparable to that obtained through fusing information from synchronous PMW data in the early-afternoon moment. Therefore, it is prospective to be further applied for generating high-resolution all-weather LST with much shorter re-visit cycles than the satellite PMW data.