The North American Regional Reanalysis (NARR) is a reprocessing of the historical meteorological observations using a modern data assimilation/forecast model which in this case is the EDAS and ETA models (the NCEP regional atmospheric data assimilation and forecast model). The products of NARR will be a new set of meteorlogical analyses covering the North American domain with a 32 km horizontal resolution, 3 hour temporal resolution and 50 hPa vertical resolution for the years 1979 and onwards. The NARR will produce a much more detailed view of the historical weather and climate over North America than previously available. However, this detail comes with a heavy cost; the complete NARR archive is approximately 80 TB.

We expect to be able to support large majority of users' data requests with a 7 TB subset. However, even a 7 TB subset is still unmanageable by most users and may take an archive center weeks to handle. To make the NARR accessible to the users, one must worry about data volumes, peoples ability to read GRIB and the needs of the heavy and casual user. To address these problems, we have developed interactive web pages that allow downloads of custom subsets of the GRIB data, production of plots and of time series for importing into spreadsheets. The heavy user can use a RPC-like interface to perform these same tasks from within a program or script. We will describe the NARR subset, and the techniques used to perform interactive output data processing efficiently. In addition, we will describe the user and RPC-like interfaces.