NARR (North American Regional Reanalysis) Data

The NCEP North American Regional Reanalysis (NARR) products are on a Lambert conformal conic grid (32km) at 29 pressure levels. They were produced using Eta 32km/45-layer model. The input data includes all observations used in NCEP/NCAR Global Reanalysis project, and additional precipitation data, TOVS-1B radiances, profiler data, land surface and moisture data, etc. The output analyses are 3-hourly with additional 9 variables in the 3-hour forecasts to reflect accumulations or averages.

NARR data is distributed on a native lambert conformal grid with 2D latitudes and longitudes. For a description of native lambert conformal grids and how to plot them accurately in NCL, click here, or refer to any of the examples below.

NARR data is available from NCAR's Data Support Section in GRIB format, a subset of NARR data is also available from NOAA's Earth Systems Research Laboratory (formerly, the Climate Diagnostics Center [CDC]) in netCDF format. See Example 2.

Important note: in the corners of the images on this page, you may notice some small slivers of missing data. The original NARR Eta-12 model grid is regridded to standard NCEP grid 221 for public distribution, Lambert conformal conic, which you see here.

There is a slight mismatch between the chosen grid boundaries, resulting in the slivers of missing data. This is discussed on pages 39, 41, 42 of this early Powerpoint summary from NCEP (2005). Page 42 is the picture that speaks a thousand words:

http://www.emc.ncep.noaa.gov/mmb/rreanl/narr.ppt

narr_1.ncl: This script demonstrates how to quickly read a NARR GRIB file into NCL, and shows how to plot the data on the NARR native grid. The file used in the example was obtained from NCAR's Data Support Section.

Note that NCL can read GRIB files directly. In addition, NCL provides additional information [eg, geographical coordinates] that are not on the original GRIB formatted file. The ncl_filedump utility can be used to view any GRIB [nc, hdf-sds, hdf-eos] file as seen via NCL.

narr_2.ncl: This script demonstrates how to read in a NARR netCDF file into NCL. In this case, the data is stored as 16-bit integers (referred to as type "short"). In general, data of type short must be transformed prior to calculations and plotting . The contributed function short2flt may be used to directly read the data. Note that short2flt will automatically apply the add_offset and scale_factor attributes during the conversion from short to float.

NOTE: Some of the NARR files from NOAA/PSD contain a _FillValue attribute that differs from the missing_value attribute. This problem has been reported. If the values look bad, add the following fix suggested by David Allured.

Replace:

           x       = short2flt( f->air(10,:,:,:) )

with

           xShort  = f->air(10,:,:,:)                
           if (isatt(xShort,"_FillValue") .and. isatt(xShort,"missing_value")) then
               if (xShort@_FillValue .ne. xShort@missing_value) then
                   xShort@_FillValue = xShort@missing_value   ; use alt. missing value
               end if                                         ; only if needed
           end if

           x       = short2flt( xShort) 
           delete(xShort)                                     ; no longer needed

narr_3.ncl: This script uses the nearest neighbor function triple2grid to interpolate from the NARR Lambert Conformal grid to a 2 degree by 2 degree grid. The original grid is plotted in the top panel, while the regridded data is shown in the bottom panel. The sample script outputs a netCDF file containing the interpolated 2x2 grid. An ncl_filedump of the netCDF file is here.

narr_4.ncl: This script uses the same files used in Example 1. The data are on a Lambert Conformal projection. The narr_4.ncl script plots the region of the lower 48 United States using two different map projections: Cylindrical Equidistant and Robinson. Further, the mpLimitMode= "LatLon" is used to manually set the appropriate limits for each projection.