Mississippi River Climate and Hydrology Conference

Tuesday, 14 May 2002
Precipitation Assimilation in NCEP Regional Reanalysis
Perry Shafran, NOAA/NWS/NCEP/EMC and SAIC, Camp Springs, MD; and W. Ebisuzaki, Y. Lin, Y. Fan, W. Higgins, K. Mitchell, E. Rogers, W. Shi, G. DiMego, E. Kalnay, and F. Mesinger
A precipitation assimilation system has been under development at NCEP for more than a decade, as a component of the Eta Data Assimilation System (EDAS). Its use in the NCEP Regional Reanalysis (RR) was at the forefront of the effort; the system has also been operationally implemented in July 2000. The system uses observed precipitation to adjust the model's latent heating, moisture, and cloud water fields during the free-run segments of the assimilation. To that end, at each time step, and for each grid point where precipitation observations are available, the model precipitation is compared against the observed, and adjustment of the model fields are made so they would be more consistent with the precipitation observations.

A variety of efforts have been mounted to collect the most comprehensive as well as appropriate input precipitation dataset. Over the contiguous United States (ConUS), CPC's "unified daily precipitation" is used. The version which will be employed in the RR production is gridded on the 1/8° resolution LDAS domain, using the "inverse distance" scheme, with mountain mapper (PRISM). Mexican and Canadian data are on a 1° grid. CMAP precipitation is used over the ocean, but only south of 50° N, and if not deemed affected by hurricane events. Disaggregation to hourly data is needed for all datasets; for ConUS it is done using the actual hourly observations, and for the remaining areas it is based on the NCEP/DOE "Reanalysis 2" forecasts.

January, April, and July 1998, and October 1987 pilot results, run at 80-km/38-layer resolution, have been examined. For all of these months, a very high degree of agreement with the observed/analyzed precipitation is obtained, in particular in the warmer months. It is felt that the procedure bodes well especially for the realism of the near-surface and land-surface components of the RR data which will be obtained in the production, about to begin at the time of this writing. 32-km/45-layer resolution will be used for the production runs, on the domain same as the current operational Eta Model domain.

Supplementary URL: