Real-time variational assimilation of streamflow and radar-based precipitation data into operational hydrologic forecasting
Abstract
To deal with various sources of error on the initial and boundary conditions, and in model parameters and structure, some form of state updating is necessary in operational forecasting that makes use of real-time streamflow observations. Here we analyze the benefit of variational assimilation as an automatic updating technique in an operational setting. Compared to state space-based techniques (e.g. Kalman filtering), variational assimilation (VAR)-based techniques offer at least two important advantages in state updating of operational hydrologic models that are, in particular, driven by radar-based precipitation input. 1) Because VAR does not require the hydrologic model to be rendered into a state-space form, no modifications are necessary to the model. Hence, the model parameters are completely transferable between calibration and state updating/assimilation. 2) Because VAR is a smoother, as opposed to a filter, VAR is very effective in assimilating data that are subject to significant biases, such as radar-based precipitation estimates. Following long-term off-line evaluation, a prototype VAR algorithm has been implemented recently at the National Weather Service West Gulf River Forecast Center (NWS/WGRFC). In this presentation, we describe the operational experience to date and the issues identified, and offer directions for further improvement.
- Publication:
-
EGS - AGU - EUG Joint Assembly
- Pub Date:
- April 2003
- Bibcode:
- 2003EAEJA....14671S