Conclusions :
§EKF data assimilation is superior to direct insertion of
observation data because
EKF can use the soil moisture correlation between the surface layer and deeper layer to
update the deeper layer soil moisture.
§Model error variance has impact on the efficiency of EKF, but
all EKF assimilation
results will eventually align with the observation if observation error variance is small (e.g.
0.005fc).
§For the Mosaic model, the impact of wrong soil moisture
initialization can not be removed in the simulation for many days if
observations are not efficiently assimilated. However, increased (20% higher)
solar radiation did not put large impact on soil moisture simulation.
§Observation error variance has significant impact on the results
of EKF data
assimilation. If observation error variance is known to be large and model error variance small,
then EKF assimilation may be used to single out less trustable
observations. Thus, the EKF assimilation is useful for validation of satellite
observation products.