Hydrologic Modeling

The research questions we are currently pursuing are: (1) how can we provide improved estimations of basin initial conditions (e.g., soil moisture, snow-pack) at the start of a forecast period; and (2) how do we characterize hydrologic model uncertainties. The first question is considered under the topic of Data Assimilation, and the second question is addressed through the Multi-model Super-ensemble techniques in hydrology.

Data Assimilation

We are currently working to implement the Ensemble Kalman Filter (EnsKF) in the SNOW-17 model used in the NWSRFS (National Weather Service River Forecasting System). Snow data assimilation using the Extended Kalman Filter (EKF) is available in the form of SNOW?43 module, but is not used operationally because of computing complexities.

The EnsKF is an elegant solution to the challenges posed by the EKF, and we are in the process of developing and testing its implementation first with a snow module.

Multi-model Super-ensemble

Our multi-model super-ensemble technique mixes and matches various methods for modeling hydrologic processes to allow the construction of multiple models, all with different structure. We are using the USGS Modular Modeling System [http://wwwbrr.cr.usgs.gov/mms/] as an integrating framework. For any given model automated parameter estimation methodology (e.g. MOCOM-UA) will be used to determine parameter sets. The potential benefits of this approach are:

• The super-ensemble provides probabilistic information content.
• The spread of the super-ensemble provides an estimate of forecast uncertainty.
• The approach allows automated configuration of hydrological models over multiple river basins with minimal human effort.
• The multi-model system reduces the commitment of operational agencies to their own model, and thus may allow more rapid transfer of new ideas from the research community to the operational setting.

Both these research topics are in their early stages and currently are works in progress.