Ehsan Forootan (University of Bonn, Institute for Geodesy and Geoinformation, Cardiff University, School of Earth and Ocean Science, Germany, United Kingdom); Anne Braakmann-Folgmann (University of Bonn, Institute for Geodesy and Geoinformation, Germany); Jürgen Kusche (University of Bonn, Institute for Geodesy and Geoinformation, Germany)

Soil moisture represents an important component of the terrestrial water cycle, which drives evapotranspiration and vegetation growth. Therefore, knowledge of its variability is essential to better understand the hydrological water cycle. However, terrestrial measurements are sparse and and provide limited information on the spatial variability of soil moisture. Over the last two decades several specialized soil moisture satellite missions have been launched (e.g. ERS/SCAT, AMSR, SMOS or SMAP) which allow a global mapping of soil moisture in the top few centimeters of soil.

Satellite radar altimeters, which have been designed to monitor changes of sea surface height over the ocean, also provide backscatter measurement over land surfaces. The backscatter from Ku- and C-/S-Band over land is provided at high resolution (~300m) at 10 or 35 day intervals for Jason-2 and Envisat, respectively, and related to surface features, such as roughness, vegetation or soil moisture.

Here, we present an inversion framework that allows to retrieve soil moisture information from along-track Jason-2 and Envisat satellite altimetry backscatter observations. There are three main steps to our approach: (i) computing time-invariant spatial patterns (base-functions) by applying principal component analysis (PCA) to simulated soil moisture from an available land surface model. (ii) Estimating time-variable soil moisture evolution by fitting these base functions of (i) to the along-track retracked backscatter coefficients in a least squares sense. (iii) Combining the estimated time-variable amplitudes and the pre-computed base-functions to reconstruct altimetry based soil moisture information.
We validated our approach over the arid and semi-arid regions of Western Australia against an available local high resolution (0.05°x0.05°) model (AWRA-L), as well as against very fine resolved (200m) model data from the Wasim model over Western Germany.