Ensemble Kalman Filter Data Assimilation with the ParFlow Hydrologic Model

Abstract:

Hydrometeorological research has shown that simulations of atmospheric processes benefit from sophisticated land surface formulations. Moisture and energy fluxes between the land surface and lower atmosphere are influenced strongly not only by atmospheric conditions, but by terrestrial hydrologic processes, soil moisture distribution in particular. By improving the representation of hydrologic processes, better predictive skill can be achieved in a fully-coupled weather forcasting model. Further improvements in the model can be realized by incorporating observed data values into the hydrologic model. This work applies the Ensemble Kalman Filter functionality included in the Data Assimilation Assimilation Research Testbed (DART), a collection of data assimilation tools maintained at the National Center for Atmospheric Research, to the ParFlow hydrologic model—the hydrologic component of the TerrSysMP fully coupled hydrologic – land surface – atmospheric model system. This generalized data assimilation tool allows observations of variables in the hydrologic component of the system to be incorporated into the overall error covariance matrix thus guiding the development of quantities that define the model state. Single dimension column tests, and a three-dimensional idealized catchment drainage and dry-out test were performed with the ParFlow-DART system to evaluate the effects of assimilating pressure head, soil moisture, and outflow observations on the development of the model through time. The data assimilation system was then applied to the hydrologic portion a fully-coupled (subsurface, land surface, and atmosphere) simulation over the North Rhine-Westphalia region in western Germany to demonstrate the utility of this system in a non-idealized and realistic forecasting situation. The success of these tests will allow the ParFlow-DART system to be developed into a complete data assimilation package for the TerrSysMP fully-coupled modeling system.