Verification case studies within the 12km North American land data assimilation system (NLDASE) project

A continental scale land data assimilation system project (NLDAS-E) featuring a 12km resolution and operating on the Arakawa E grid is currently ongoing at the NASA Goddard Space Flight Center. Led by NASA GSFC with support from NOAA NCEP, the NLDAS-E project centers on the initialization of the land surface fields of NCEP's mesoscale 12km coupled workstation Eta model. Featuring multiple LSMs and assimilating multiple land surface quantities, this system will serve to supply the Eta model with accurate, unbiased and uncoupled initial land surface conditions on its native Arakawa E grid. NLDAS-E project goals include: 1) Generation of land surface states over the North American domain, with and without application of land data assimilation techniques, 2) Initialization of the NCEP workstation Eta model with uncoupled NLDAS-E states and internally cycled Eta land surface states, 3) Execution of ensemble forecasts using the NLDAS-E and Eta modeling systems, 4) Verification of workstation Eta model forecasts using NCEP's forecast verification system to determine impact of initialization with NLDAS-E conditions. The system will make use of the Noah, CLM and Mosaic LSMs, and will utilize standard workstation Eta as well as hi-resolution soil and vegetation fields in order to explore the impact that the use of such fields has on NWP forecasts. NLDAS-E research will improve understanding of the interaction between the land surface and the atmosphere from short- to medium-length time scales, will create accurate, high resolution land surface data sets useful for future research and applications, will improve the initialization of the Eta model in a way that may be adapted to additional NWP models in the future, will improve forecast accuracy, and will continue NASA GSFC's efforts in the collaborative NLDAS project.

Recent progress has included the generation of full-domain, 12km forcing files which will be used to drive the offline NLDAS-E system. These forcing files use 12km EDAS/Eta model output data as a data backbone, and include observed precipitation (CPC gage/Stage II Merged, CMAP, and CMORPH) and observed radiation (AGRMET and GOES) products when available. Generation of workstation Eta forecasts over the entire North American domain used by NCEP's operational Eta model, utilizing NASA GSFC's Compaq-Alpha supercomputing platform has also been completed. A 13 day benchmarking effort (May 1-13, 2003), involving the verification of ETA forecasts, centered on the extended tornado outbreak of May 3-11, 2003 has been performed to assess the impact of the use of uncoupled NLDASE land surface states. The use of the uncoupled land surface states is found to have a significant impact on the resulting ETA forecasts. Throughout the 10-day benchmarking period, some verification regions benefited from the use of NLDASE land surface states while others did not. Outputs from several individual forecasts are compared against available observations to address the following questions: How did the ETA model perform for specific cases with and without the use of NLDASE land surface states? Where and how did the use of the offline NLDASE land surface states improve the forecast in some cases and degrade it in others? Which model output fields improved the most and why? Where and when might the traditional forecast verification metrics (e.g. skill scores, root-mean-square error and the like) fail to indicate forecast improvement due to small timing or placement errors? Examples of observational data used in verifying the individual cases include standard surface synoptic/metar reports, upper air profiles from the Integrated Global Radiosonde Archive, and NEXRAD Level II/III radar products.

This presentation is intended to compliment the general NLDASE overview presentation given by Brian Cosgrove in the same session and aims to provide specific insights into the qualities of the NLDASE land surface states from select case studies and their associated impacts on the ETA model forecasts. Project data and additional information concerning NLDAS-E can be obtained from http://ldas.gsfc.nasa.gov.