A priori daily soil moisture estimates are obtained from the PC Raster Global Water Balance (PCR-GLOWB) conceptual hydrological model. Use of the PCR-GLOWB model in this data-sparse region is advantageous because the only forcing data required are daily precipitation from TRMM and an actual evapotranspiration product derived from VIS and TIR observations. Model error is introduced by perturbing the forcing data as well as key model parameters, notably the saturated hydraulic conductivity and thickness of the first soil layer.
While there is a strong correlation between ERS scatterometer data and the relative saturation in the first layer of the model, the relationship is non-trivial. This is accounted for in the measurement operator to reduce the risk of introducing a bias in the estimation process.
In the absense of direct ground truth observations of soil moisture, the impact of the ERS1/2 observations is quantified through comparison of simulated and observed streamflow at the outflows of the Volta and Senegal Basins with and without assimilation.
Continuing work is concerned with refining the assimilation process in preparation for the availability of observations from the SMOS and SMAP soil moisture missions. The specification of model uncertainty is also being improved to correctly account for spatial correlation in both the model and observation error terms.
Supplementary URL: