A particle smoother with sequential importance resampling for radiative transfer parameter estimation

Correct parameterization of radiative transfer models is very important for high accuracy soil moisture retrievals from spaceborne L-band passive microwave sensors, such as the ESA Soil Moisture and Ocean Salinity (SMOS) Mission. In order to investigate the characteristics of radiative transfer parameters such as vegetation opacity and soil surface roughness, a dual state and parameter update data assimilation system has been developed. By assimilating SMOS brightness temperatures into the L-band Microwave Emission of the Biosphere (L-MEB) model, respective parameters can be estimated. The data assimilation system makes use of a temporal smoothing algorithm based on the Sampling Importance Resampling Particle Filter. The new approach, namely the Sampling Importance Resampling Particle Smoother, makes use of a particle weighting function valid not for a single time step, but for a specific time period. The resulting parameters estimated are more stable throughout the whole period under investigation, where the possibility to vary with time is still given. This is important to cover the seasonality of e.g. vegetation parameters.