Bias correction of satellite soil moisture through data assimilation

Soil moisture exhibits great spatio-temporal heterogeneity and plays a critical part in land surface energy and water cycles, being identified as a terrestrial essential climate variable. Thus, it is urgently needed in a wide variety of environmental processes such as hydrology, meteorology, agriculture, and ecology. Microwave remote sensing has the potential to provide near real-time soil moisture estimates on large spatial scales according to the distinctive contrast between dielectric properties of water and dry soils. Thus, many space-borne microwave sensors have been launched for retrieving soil moisture. Especially, SMOS and SMAP at L-band frequency (1.4 GHz) supply an unprecedented opportunity for retrieving surface soil moisture due to their deeper penetration than instruments at other bands. However, these satellite soil moisture products need bias correction before application such as data assimilation. Two common correction methods require reliable land surface soil moisture simulations. However, the quality of these simulations relies heavily on model parameters, such as soil porosity and texture, which are almost unavailable in remote regions such as the Tibetan Plateau. In this study, a dual-cycle assimilation algorithm is taken to make on-line bias correction when assimilating SMAP soil moisture products. During the assimilation, a linear bias correction scheme is regarded as the observation operator to link the simulated soil moisture values and the satellite retrievals. In the inner cycle, a sequentially based assimilation algorithm is run with both model parameters and bias correction coefficients, which are provided by the outer cycle. At the same time, both the analyzed soil moisture and the innovation are reserved at each analysis moment. In the outer cycle, the innovation time series kept by the inner cycle are fed into a likelihood function to adjust the values of both model parameters and correction coefficients through an optimization algorithm. A series of numerical experiments are designed and conducted, indicating that the soil moisture estimates by the presented algorithm are superior to those with the existing bias correction schemes.