Assimilating Soil Moisture Data into a Hydrological Model of the Volta Basin

Soil moisture is important for the Volta Basin. However, in-situ measurements of soil moisture in the Volta Basin are scarce, only remote sensing observations are available. Remote sensing observations are also provided with limited temporal and spatial resolutions. Also capabilities of the sensors are limited (e.g measuring soil moisture limited in shallow layer), and observations from different instruments do not totally agree. Hydrological model can predict soil moisture. But model predictions rely on assumptions and suffer from input uncertainties. Therefore, data assimilation is used to combine the information from observation and model to achieve optimal estimation of soil moisture. In this thesis, AMSR-E data of volumetric water content and ERS data of relative saturation for the Votla Basin were compared, and TRMM precipitation data were also used to compare with the two sets of soil moisture data. The dataset of ERS relative saturation was further assimilated into the PCR-GLOBWB model for the period of January 1st 2003 to December 31st 2006 in the Volta Basin. An ensemble size of 100 was applied. A key objective of the thesis is to address spatial correlation of data. Uncertainties of the model parameter (storage capacity) and input forcings (evapotranspiration and precipitation) were represented with spatial correlation. Correlation lengths of the parameter and forcings were thus calculated. The outcome of perturbation managed to maintain the spatial pattern of data. In the next part, four EnKF schemes (EnKF (Ind), EnKF (Blk), EnKF (DistMx) and EnKF (NonObserved)) were applied and compared. Results show that the block size is consistent with the correlation length and that schemes which account for spatial correlation produce better and more realistic results than the one without spatial correlation.Water ResourcesWater ManagementCivil Engineering and Geoscience