Comparing Assimilation of Synthetic Soil Moisture Versus C-Band Backscatter for Hyper-Resolution Land Surface Modeling

Abstract

This paper presents the assimilation of synthetic surface soil moisture retrievals and C-band backscatter signal in a 100-meter resolution version of the Canadian Land Data Assimilation System (CaLDAS) on footprint scale soil moisture with a time interval of three hours. The synthetic surface soil moisture map was generated by extrapolating a regression relationship between in situ measurements and open loop land surface model outputs based on the soil texture of a given pixel. The surface roughness was inverted from RADARSAT-2 imagery using a modified Integral Equation Model (IEM) model. Three hourly synthetic backscatter maps were created from this surface roughness and the synthetic soil moisture. The Ensemble Kalman filter (EnKF) with bias correction was applied to mitigate the impact of nonlinear errors introduced by multi-sourced perturbations. Both time series and spatial maps were examined in the evaluation of the assimilation experiments. Results show that the assimilation of backscatter is as effective as assimilating soil moisture retrievals although the later has slightly better temporal statistics. Compared to the open loop, both approaches improved the analysis of surface and root zone soil moisture with significantly lower bias. For the later, the open loop average bias was reduced from −1.25 vol/vol to 0.59 vol/vol by backscatter assimilation and to 0.41 vol/vol by soil moisture assimilation.