Using a discrete scattering model to constrain water cloud model for simulating ground-based scatterometer measurements and retrieving soil moisture

Abstract

Potential of constraining semiempirical model with physically based scatter model simulations has long been recognized. This study contributes to this topic through the assessment of backscattering coefficient (σ°) simulations and soil moisture retrieval using the water cloud model (WCM) constrained by a discrete scattering model (i.e., Tor Vergata) under both frozen and thawed soil conditions. The WCM is coupled with Oh (hereafter 'WCM+Oh') and Dubois (WCM+Dubois) surface scattering models, respectively. The soil permittivity is obtained using the four-phase dielectric mixing model. One year of C-band copolarized σ° observations are collected by a ground-based scatterometer deployed in the seasonally frozen Tibetan meadow ecosystem. It is found that: the calibrated Tor Vergata (hereafter 'TVG') model simulates well the seasonal dynamics and magnitudes of scatterometer measurements, and the simulated scattering components and vegetation transmissivity agree well with the seasonal vegetation dynamics; the total scattering simulated by the TVG constrained WCMs shows a good consistency with the scatterometer measurements, and the simulated soil and vegetation scattering components are in line with the TVG simulations; and the retrieved soil moisture based on the constrained WCMs captures well the seasonal variability noted in the in situ measurements. An additional experiment is performed to calibrate the WCMs directly, and the results show that the calibrated WCMs achieve comparable results with the calibrated TVG model and the constrained WCMs in terms of the total σ° and soil moisture retrieved. However, the direct calibration of the WCMs leads to unrealistic characterization of individual soil and vegetation scattering contributions, of which an underestimation of the vegetation contribution at VV polarization is most notable. These findings demonstrate that usage of a physically based scatter model to constrain semiempirical models leads to results that provide a more robust representation of reality, which is needed for developing worldwide soil moisture monitoring from active microwave remote sensing.