Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
5771461 | Journal of Hydrology | 2016 | 59 Pages |
Abstract
The Extended Kalman Filter (EKF) is used to assimilate in situ surface soil moisture and streamflow observation at the outlet of an experimental watershed outlet into a semi-distributed SWAT (Soil and Water Assessment Tool) model. Watershed scale, instead of HRU scale soil moisture was used in state vector to reduce computational burden. Numerical experiments were designed to select the best state vector which consists of streamflow and soil moisture in all vertical soil layers. Compared to open-loop model and direct-insert method, the estimate of both soil moisture and streamflow has been improved by EKF assimilation. The combined assimilation of surface soil moisture and streamflow outperforms the assimilation with only surface soil moisture or streamflow especially in the estimate of full profile soil moisture. The NSC has been improved to 0.63 from â4.45 and the RMSE has been reduced to 12.34Â mm from 47.44Â mm in open-loop. Such improvement is also reflected in the short term forecast of soil moisture. The improvement of streamflow prediction is relatively moderate in both simulation and forecast mode compared to quality of the soil moisture prediction. The quantification of the model error, especially the error covariance between different state variables, was found to be critical to the estimate of the state variable corresponding to the error covariance.
Related Topics
Physical Sciences and Engineering
Earth and Planetary Sciences
Earth-Surface Processes
Authors
Leqiang Sun, Ousmane Seidou, Ioan Nistor, Kalifa Goïta, Ramata Magagi,