A Continental Divide Hydrometeorological Observatory for Integrated Hydrologic Data Assimilation and Prediction Development

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 5 January 2015: 2:00 PM
127ABC (Phoenix Convention Center - West and North Buildings)
David J. Gochis, NCAR, Boulder, CO; and J. McCreight, A. Dugger, R. M. Rasmussen, M. Clark, A. W. Wood, and D. Yates

Data assimilation (DA) will remain critical to hydrologic forecasting applications. DA keeps models tethered to reality, aiding in model physics assessments, and improves forecasts by improving model initial conditions. While the hydrologic DA community has called for common tools and platforms for modeling, DA, and verification (Liu et al., 2012), a common verification component lags behind. Furthermore, integrated sets of important water cycle state and flux measurements are also lacking which leads to poor constraint in both model state adjustment during DA cycles and in a priori model parameter estimation. However, in practice, integrated sets of hydrometeorological and terrestrial hydrological measurements are not available to sufficiently develop and test much needed, new multi-variate, integrated DA and verification methods.

Ensemble DA methods combine probabilistic forecasts with data to improve forecasts and can help address issues related to uncertainty in measurements and in their spatial representativeness. For this reason ensemble DA has and will continue to play a central role in the future of hydrologic forecasting applications. Several grand challenges stand in the way of advancing ensemble DA and probabilistic hydrologic forecasting. Bias in the mean probabilistic forecast results in suboptimal assimilation and suboptimal probabilistic forecasts. Forecast bias arises in part from improper specification of error relative its sources: forcing error, parameter error, model structural error and observational error. With a coordinated approach to quantifying these sources of error and their interactions, the community is likely to advance the science much more quickly.

In this talk we introduce the community to a new hydrologic observatory located along the North American continental divide in Colorado which has been designed with an explicit purpose to support integrated hydrological model evaluation and hydrologic data assimilation. The goal of the observatory is to expand hydrologic modeling and data assimilation research in multiple directions including: physics development, observations, assimilation frameworks techniques, parameter calibration techniques, and integrated benchmarking metrics. The observatory possesses an unprecedented amount of radar and ground observation coverage for a complex terrain region and spans a range of elevations as well as physiographic and ecohydrologic regimes. A unique focus of the observatory is to design and evaluate targeted, multi-variate observations based on analysis of uncertainty in probabilistic forecasts. Specifically, observing system simulation experiments and actual observing system experiments will be conducted by the observatory to help quantify the interaction of model uncertainty, observations, and integrated metrics.

The observatory is complemented by several different community hydrological modeling systems, including the community WRF-Hydro system, the NASA-LIS modeling system and the NCAR Data Assimilation Research Testbed (DART) system which have each been implemented over the observatory. Combined, these forecast models, observations, and data assimilation techniques will be publically available and contributors are welcomed to add new experiments to explore the model uncertainty and to develop robust, integrated metrics. Data from the observatory is being converted into common hydrologic data standard formats as specified by the Open Geospatial Consortium (OGC) and operational weather and water prediction entities as well as the NSF EarthCube program. Examples of initial observational, data assimilation and verification experiments will be presented and plans for future observational and prediction campaigns will be discussed.