There are many approaches to improve hydrologic model predictions, including pre-processing to deal with input uncertainty, data assimilation to treat initial and boundary condition uncertainty, model calibration to reduce parametric uncertainty. Hydrologic post-processing is an approach for treating uncertainties from hydrologic model outputs propagated from all upstream sources. It works by relating model outputs (e.g., streamflow) to corresponding observations through a statistical model. This paper compares the effect of post-processing and model calibration in improving hydrologic forecasts under different hydroclimatic conditions and across different models.Observed and simulated daily streamflow data from the Second Workshop on Model Parameter Estimation Experiment (MOPEX) were used for the comparisons described above. The results from 7 hydrologic models showed that post-processing alone was better than the results from hydrologic model calibrations for 12 basins in the eastern United States. The predictive QQ plot indicates that the predictive distributions of post-processed ensemble streamflow simulations are reliable. Post-processed results were similar for different hydrologic models, but were quite different for different basins. In terms of ensemble prediction, post-processing results tended to be over-confident. In general, post-processing can improve hydrological forecasts and reduce uncertainty in wet basins, but caution should be taken when applying post-processing to dry basins where there are many zeros values in the data.
All Science Journal Classification (ASJC) codes
- Water Science and Technology
- Hydrologic model