Surface soil moisture (SM) plays a fundamental role in energy and water partitioning in the soil–plant–atmosphere continuum. A reliable and operational algorithm is much needed to retrieve regional surface SM at high spatial and temporal resolutions. Here, we provide an operational framework of estimating surface SM at fine spatial resolutions (using visible/thermal infrared images and concurrent meteorological data) based on a trapezoidal space defined by remotely sensed vegetation cover (Fc) and land surface temperature (LST). Theoretical solutions of the wet and dry edges were derived to achieve a more accurate and effective determination of the Fc/LST space. Subjectivity and uncertainty arising from visual examination of extreme boundaries can consequently be largely reduced. In addition, theoretical derivation of the extreme boundaries allows a per-pixel determination of the VI/LST space such that the assumption of uniform atmospheric forcing over the entire domain is no longer required. The developed approach was tested at the Tibetan Plateau Soil Moisture/Temperature Monitoring Network (SMTMN) site in central Tibet, China, from August 2010 to August 2011 using Moderate Resolution Imaging Spectroradiometer (MODIS) Terra images. Results indicate that the developed trapezoid model reproduced the spatial and temporal patterns of observed surface SM reasonably well, with showing a root-mean-square error of 0.06 m3·m−3 at the site level and 0.03 m3·m−3 at the regional scale. In addition, a case study on 2 September 2010 highlighted the importance of the theoretically calculated wet and dry edges, as they can effectively obviate subjectivity and uncertainties in determining the Fc/LST space arising from visual interpretation of satellite images. Compared with Land Surface Models (LSMs) in Global Land Data Assimilation System-1, the remote sensing-based trapezoid approach gave generally better surface SM estimates, whereas the LSMs showed systematic underestimation. Sensitivity analyses suggested that the trapezoid method is most sensitive to field capacity and temperature but less sensitive to other meteorological observations and parameters.
Read full abstract