The quality of land surface model (LSM) output is closely tied to the quality of the meteorological forcing data used to drive the model. With this in mind, the N-LDAS Project, a joint research effort between NASA GSFC, NOAA NCEP, NOAA NESDIS, NOAA OHD, Princeton University, Rutgers University, the University of Washington and the University of Maryland, has sought to construct quality controlled, spatially and temporally consistent forcing data sets from the best available observations and model output to support it’s multi-LSM modeling activities. The N-LDAS project has both retrospective and real-time components, and as such has dealt with the creation and use of two different forcing data sets. Both the real-time and retrospective data sets contain a common set of variables in an 1/8th degree, hourly format on a continental US domain, but draw on different raw data sources according to what data was available at the time of their creation.

The real-time data set is updated once per day using a set of cshell scripts and Fortran 90 programs. Three-hourly 40 km output from the Eta Data Assimilation System (EDAS) is used as a base for the data set, providing such quantities as 2 meter temperature and specific humidity, surface pressure, wind speed, shortwave and longwave radiation, total and convective precipitation and convective available potential energy. This basic set of data is augmented by hourly, observed data from GOES satellites (0.5° shortwave radiation, photosynthetically active radiation and skin temperature) and WSR88D Doppler radar sites (4km precipitation), and by data from daily rainfall gauges (0.25° precipitation). Forcing data are interpolated from their native resolution to a 1/8th degree, hourly resolution and are stored in compressed GRIB format. A merged precipitation product, based on all available precipitation forcing data, is also included in this data set. To create the merged product, Doppler radar rainfall estimates (or EDAS estimates when the radar product is missing) are used to derive hourly rainfall weighting factors, which are then applied to daily rainfall gauge totals to arrive at hourly gauge-based values. Production of forcing data sets is automated, as is the suite of quality control checks that the data must pass through before being distributed to N-LDAS participants.

The 5-year N-LDAS retrospective forcing data set (9/96-9/01) was produced in a fashion similar to that of the real-time forcing data. Differences exists, however, as production of retrospective forcing is not subject to the same time constraints as is production of the real-time forcing. This added time insulates the retrospective forcing from ftp and computer-related problems in data set availability and production, and allows for more extensive quality control checking and for the use of data sets that are not available in real-time. In particular, the retrospective forcing includes data from Wayne Higgins’ Unified Precipitation Atlas, from Rachel Pinker’s reprocessed GOES satellite-based radiation dataset and from daily rainfall gauge stations that the real-time forcing is unable to include.

Both the real-time and retrospective N-LDAS forcing datasets are undergoing extensive validation using Oklahoma Mesonet data.