Reducing parameter uncertainty for operational ensemble streamflow forecasting using automated calibration
Abstract
Operational streamflow forecasting systems are increasingly used by different agencies to improve water management, hydro-power production, navigation and flood awareness. Such systems generally employ the use of hydrologic models in conjunction with numerical weather prediction models to produce streamflow forecasts. Parsimonious hydrologic models conceptualize the physical surface and sub-surface hydrological processes using tunable empirically derived parameters. The reliability of the simulations obtained using observed meteorological conditions depends on obtaining optimal parameter sets by calibration such that the simulated flows match historic observations. However, in a forecasting context, the skill of the streamflow forecasts hinges on the reliability of the meteorological inputs and the optimal set of model parameters that are representative of the underlying hydrological processes during the forecast initiation phase.
In this study we attempt to capture the time varying nature of the hydrologic model parameters rather than assuming a constant optimal value or stationary distribution. We create an adjustment table that encompasses the optimal parameter ranges and the corresponding hydrologic conditions defined by the previous periods' runoff volume and the future precipitation conditions. This table is created by running the semi-distributed hydrologic model for a representative hydrologic period using gridded precipitation obtained by reanalysis and verified with several hydrologic events. In the forecast mode the future precipitation conditions represented using ensemble forecasts and the previous period's runoff are used to select the optimal parameter sets from this criteria table using the concept of data depth. The results from this study show that the time varying model parameters chosen according to the above criteria helps to reduce parameter uncertainty compared to the time invariant parameters which have a constant optimal value.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.H33H2176W
- Keywords:
-
- 1816 Estimation and forecasting;
- HYDROLOGYDE: 1821 Floods;
- HYDROLOGYDE: 1847 Modeling;
- HYDROLOGYDE: 1860 Streamflow;
- HYDROLOGY