Continental Water Dynamics Modeling
By Cédric H. David
June 13, 2007
The following narrative supports the corresponding PowerPoint.
The continental Water Dynamics Modeling project is a one-way atmosphere-land surface-hydrology model that aims at calculating river flow (with assimilation of gage measurements) on a continental scale. Several atmospheric model results or measurements (such as NARR or NEXRAD) can be used as forcing for the Noah land surface model. The recently enhanced National Hydrography Dataset (NHDPlus) serves as a hydrologic base for the Continental U.S.
The main purpose of a land surface model is to compute bottom boundary conditions (water and energy balances) at the land surface, for input to a weather or climate model. Noah was first developed in 1999 as a community effort and has been constantly improved (especially hydrologic components) since then. Noah was fully coupled with Eta when Eta was operational and is now fully coupled with the Weather Research and Forecast (WRF) model, operational at NCEP. Noah is compliant with netCDF files that are becoming a standard in atmospheric science.
Noah-distributed, a routing-capable enhancement of Noah, computes overland and subsurface runoff. The resolution used for the land-atmosphere processes (900 m in this study) and the routing processes (30 m) are different. Noah-distributed is used to determine the amount of runoff into the NHDPlus river network.
NHDPlus is the integration of the National Hydrography
Dataset, National Elevation Dataset and National Land Cover Dataset just
completed by EPA. It has 2.6 million
river reaches and catchments for the
The
Atmospheric models as well as land surface models use a sphere to represent the Earth (this simplifies Coriolis acceleration in atmospheric calculations). Hydrographic datasets use the more accurate spheroid as the shape of the Earth. This conceptual difference must be taken into account when "feeding" the land surface model with hydrological features, and appropriate projections need be done.
The shape of the Earth is not the only conceptual difference between hydrologic and atmospheric sciences. The objects used for calculations are boxes (or grids) in atmospheric science and vectors in hydrology. Therefore another translation between the two worlds is necessary. For each catchment in the basin, a pour point grid cell is determined and the total amount of runoff accumulated in it by the land surface model. This amount is used as the amount of water that runs off into the corresponding river reach, using unique identifiers for the catchment and its reach.
A preliminary model run simulates a storm in May 2006 using 3-hourly NARR data (precipitation, evaporation, radiation, wind, etc.) on the 900m land-atmosphere processing. NHDPlus elevation is used for the 30 m routing processing. NEXRAD precipitation data will be used as a replacement of NARR precipitation in the future.
Each reach in the basin is represented by the upstream flow Q flowing into it, the amount of water QNoah that runs off from the land surface, and the potential presence of a gage.
A 5-reach, 2-node, 2-gage river network was used for a theoretical study on data assimilation.
Conservation of mass at each node of a river network can be expressed by a matrix equation. The matrix equation is the constrain that the computation of flows has to verify in order to be acceptable.
Cost functions are used in optimization to determine how good a solution is. The two cost functions used here are very typical. The first one gives an estimate of how far a computed flow rate is from observations at all gages. The second one gives an estimate of how reasonable the computation is when compared to best estimates of the flow (could be seasonal average, prior calculation, etc).
With a large optimization problem (here a large river network), it is often preferable to use unconstrained optimization and to include the constrain as a cost function.
Observability ratios can be defined to give an idea of how well a basin is monitored. A first ratio compares the number of river reaches to the number of gages in a network. Since the presence of a gage on a reach doesn't always mean that a measurement is taken at the time considered, a second ratio can also be defined.
Conclusions:
The NHDPlus dataset can serve as the land base for a water
dynamics model of the continental
NHDPlus can be linked to the NOAH land surface model to produce watershed runoff into the river system
Subtle conceptual conversion issues affect the linkage
A framework for data assimilation has been determined and tested on a simple case.
Cédric H. David
Center for Research in Water Resources
email: cedric.david@mail.utexas.edu
These materials may be used for study, research, and education, but
please credit the authors and the Center for Research in Water Resources, The