Concept and initial realization of PHYSER — A PHYsics-based framework for SpacebornE GNSS-R soil moisture retrieval with accurate soil reflectivity

Abstract

Abstract Global soil moisture (SM) observation using the spaceborne Global Navigation Satellite Reflectometry (GNSS-R) is becoming an effective supplement and enhancement to traditional microwave remote sensing observations. The state-of-the-art SM retrieval frameworks for spaceborne GNSS-R are based on empirical or semi-empirical modeling, which relies on reference SM data from other sources (e.g., microwave radiometer or in situ SM products) to eliminate the effects of land surface random errors (e.g., surface roughness and vegetation). This study defines a generic framework for PHYsics-based SpacebornE GNSS-R SM retrieval, namely PHYSER, and proposes initial strategies to realize the framework. The framework concept devotes to deriving accurate soil reflectivity and retrieving SM by estimating soil permittivity from Fresnel reflection coefficients, thus wholly independent of external SM products. It assumes that GNSS-R surface reflectivity and its related soil reflectivity are affected by observing system errors and land surface random errors. The framework is initially realized by deriving accurate soil reflectivity from empirical corrections to avoid the grand challenge of building a forward scattering model under complex land surface conditions. Accurate soil reflectivity is derived through two steps: 1) Surface Reflectivity CALibrating (SuR-CAL), aiming to calibrate the system errors using the reflectivity of inland water bodies, and 2) Soil Reflectivity CORrecting (SoR-COR), aiming to correct the random errors mainly from surface roughness and vegetation using the zeroth-order radiative transfer (τ–ω) model. The framework is validated using one-year data from BuFeng-1 A/B (BF-1) twin satellites. The findings and conclusions mainly include: 1) PHYSER reveals that independent spaceborne GNSS-R SM retrieval without reference SM products is achievable through deriving accurate soil reflectivity. 2) Land surface random errors play a more significant role in influencing soil reflectivity than system errors. The SuR-CAL and SoR-COR steps improve the correlation coefficient (R) between BF-1 reflectivity and the SMAP SM up to ~ 7% and ~ 36%, respectively. 3) The BF-1 SM estimates agree well with the SMAP SM and ERA5 SM (ubRMSD = 0.067 m3m− 3 and MAE = 0.073 m3m− 3 against SMAP; ubRMSD = 0.079 m3m− 3 and MAE = 0.088 m3m− 3 against ERA5). The BF-1 SM also agrees well with the in-situ measurements with mean ubRMSE = 0.055 m3m− 3 and MAE = 0.066 m3m− 3. The proposed framework provides a promising physics-based concept to independently retrieve SM for the GNSS-R community, which is expected to considerably support the in-orbit and next-generation GNSS-R missions to promote operational SM retrieval and applications.