In this presentation, we discuss a new approach to retrieve soil moisture from AMSR2 brightness temperature (TB) data. The approach accumulates a set of successive overpass observations at a given location to constrain the unknowns of the forward model. These unknowns include model coefficients that are otherwise difficult to determine globally and geophysical parameters such as land surface oil moisture that are of interest to hydrological applications. When the number of known observations exceeds the number of unknowns after a certain data accumulation period, it becomes possible to solve for the unknown quantities (model coefficients and geophysical parameters) iteratively, allowing a way to determine the spatial and temporal variability of soil moisture on a global basis. Because the same forward radiative transfer model is used in the formulation to determine the unknown quantities, the resulting soil moisture retrieval is less prone to non-convergence due to inconsistency between observations and ancillary data.
Time series of soil moisture retrieval based on AMSR2 data (2015-2016) using this approach will be compared with the corresponding ground soil moisture measurement from in situ data as well as soil moisture retrieval from applicable satellites. The comparison results will be discussed and the accuracy of the corresponding retrieval will be evaluated using metrics such as unbiased root-mean-squared error and correlation. Insights of extending this technique to other satellites operating at other frequencies will also be discussed.