North American Regional Reanalysis (NARR)

A plot of NARR data zoomed-in on Hurricane Isabel making landfall on September 18, 2003, at 18 UTC. This plot shows the total amount of water (in a vertical column of air) being zonally transported in the atmosphere. The northern bull's-eye shows a strong westward motion of water, while the southern bull's-eye shows a strong eastward motion of water--indicating the strong cyclonic motion of the hurricane. Data contours were generated with the Grid Analysis and Display System (GrADS) and visualized with Google Earth.

In addition to the above list of input data to NARR, a major component of the NARR is the assimilation of precipitation. The precipitation dataset comes from a variety of sources. The data over the continental United States come from a 1/8-degree gauge dataset analyzed using PRISM and a least-squares distance-weighting algorithm. Over Canada and Mexico, the precipitation comes from 1-degree gauge datasets. Much of the rest of the domain's precipitation comes from the Climate Prediction Center (CPC) CMAP (CPC Merged Analysis of Precipitation), a merged combination of satellite and gauge precipitation.

Other datasets include winds and precipitable water from TOVS (TIROS [Television Infrared Observations Satellite] Operational Vertical Sounder) satellite radiances, wind and moisture from hourly and 3-hourly surface stations, and ship and buoy data. Snow depth comes from the 512x512 Air Force snow dataset. Sea surface temperatures (SSTs) contain a 1-degree Reynolds dataset, including the Great Lakes. Sea ice data come from a satellite dataset used for the NCAR/NCEP Reanalysis-I. Canadian lake ice comes from the Canadian Ice Center.

NARR fields use earth-relative winds, which do not need adjustment when used with the Gridded Analysis and Display System (GrADS) and the Grads Data Server (GDS).

Naming Convention

For each analysis time, GRIB files are named as below.

narr-a_YYYYMMDD_HH00_000.grb
narr-b_YYYYMMDD_HH00_000.grb

YYYYMMDDHH is the date code which corresponds to the analysis time or the starting time of the forecast for model predicted quantities such as flux or precipitation. The possible HH are 0,03, 06, 09, 12, 15, 18, and 21. All times are in UTC.

One GrADS control file is used for each day.

narr-a_YYYYMMDD_0000_000.ctl
narr-b_YYYYMMDD_0000_000.ctl

The eight 3-hour time increments covered by the specific day (YYYYMMDD), are defined as forecast hours 1 through 8. However, these files actually contain no forecast data.

GrADS and the File Differences

The data for each analysis time is split into two files so that the data will be compatible with the software program, GrADS. The "narr-b" file is much smaller and contains fields that are similar to those in the larger file except they have some differences that GrADS does not recognize. For example, the "narr-b" file contains a 3-hour forecast of the sensible heat flux whereas the larger file contains the average from the 00-03 hour forecast. GrADS has only one time variable.

Example:

File A: Analysis at 2004-01-01-00Z
File B: 3-hour forecast valid at 2004-01-01-00Z

GrADS allows you to combine datasets together so you can make time series. If we combined all the "A" files together, one could set the time to 2004-01-01-18Z and make a plot. On the other hand, if the A and B files were combined together, GrADS would get confused if you set the time to 2004-01-01-18Z. GrADS would ask, do you mean the analyses, the forecast valid at that time or even the forecast started from that time?

To avoid this GrADS limitation, all the data were put into the "A" files except the forecasts that would cause the timing confusion. The average person would only need the "A" files and only people who want to know some hydrological analyses increments would need the "B" files.