The NASA/GIFTS Fast Model is available in two versions, providing coefficients in either little-endian or big-endian format.  These coefficients are based on LBLRTM-6.01 and HITRAN-2000,  and were generated by Leslie Moy and Dave Tobin, on November 15, 2002. Both versions have executables produced from FORTRAN code written by Hal Woolf, current as of May 13, 2002.

To download the fast model and coefficients:

1. Choose a format:

    Little-endian Version 1.20030116 Download TAR (gzip) file
    Big-endian Version 1.20030116 Download TAR (gzip) file

2. Unzip the file: gzip -d filename.

3. Untar the file: tar -xvf filename.

There should now be 2 sub-directories:

i) bin (contains the scripts used to run the gifts model (giftsrun), the executable (giftsfrtemat), and Matlab code to run the model and display the results (prof.mat, writeGIFTS.m, readGIFTS.m)).

ii) coeff (contains the coefficients).

The coefficients were created using 32 profiles at 6 satellite zenith angles. The profiles are available in a netcdf file here. The brightness temperatures from the LBLRTM and Fast Model runs are available in a netcdf file here. The dependent set statistics based on these profiles' brightness temperatures are provided in an ascii file here.

The main difference between this model and Hal Woolf's is in how the coefficients were derived. We used optical depth weighting and SVD analysis; Hal Woolf used a ridge regression. Hal Woolf's fast model can be found here.

The first step is to set a variable GIFTSFM_HOME for your environment.

For example, if you downloaded the tar file in the directory /home/gifts/, then set GIFTSFM_HOME = $HOME/gifts/  in your .bash_profile file.

From within the bin/ directory, run Matlab.  A sample file, prof.mat, is available.

From the Matlab prompt, type load prof.

This loads a structure prof into memory that contains 41 sample profiles.
When you want to make your own equivalent structure for GIFTS profiles, it should have the following fields:
 
 

Once the prof structure is loaded into memory, you can now generate the corresponding binary file that'll be used as input to the fast model. This is accomplished by typing writeGIFTS(prof, 'test.bin',{'radiance','optical_depth'}) at the Matlab prompt.

writeGIFTS is a Matlab function that takes three arguments: the profile structure, the name of the binary file to be output for use in the GIFTS Fast Model, and a cell array specifying the desired GIFTS fast model output using pre-defined keywords. Currently, these keywords can be 'radiance', 'optical depth' or both.  In the latter case, both radiance and optical depth will be output from the model run.

At this point, two new files should be present in the bin/ directory: parameters.txt and test.bin.

Now, the giftsrun script can be executed. From the Matlab prompt, type !giftsrun test.bin testout/.

The first parameter must be the name of the input file to the Fast Model (this should be the same as that produced by writeGIFTS).
The second parameter is an optional field representing the directory where results are output.  If this parameter is not specified, a results directory with username and timestamp information will be produced in the bin/ directory.

Once giftsrun has completed, 4 files will be available in your output directory: outLW.bin, outSMW.bin (containing radiance and/or optical depth information for the longwave and short-midwave regions, respectively) and wvnLW.bin, and wvnSMW.bin. which contain the corresponding wavenumbers for those regions.

From the Matlab prompt, type data=readGIFTS('testout'). readGIFTS takes one argument corresponding to the output directory and returns a structure (here called data) that contains the following fields:
 
 

These are the output results from the GIFTS Fast Model.  Using Matlab, plots can be made at this point. For example, the command plot(data.wvnLW,data.radLW) plots wavenumber versus radiance for the longwave band.

To test the if the download and model run was correct, a check file is provided (check.mat). Load this file in, and type load check. The structure data should be equal to check.

Acknowledgements:
The GIFTS fast model was developed under NASA and MURI funding.

Source code availability:
Source code for Hal Woolf's model is located at /abyss/gifts/giftsfrte20020501/src.
Source code for generating the coefficients is located at /deluge.ssec.wisc.edu/cgi-bin/cvsweb.cgi/devel/MURI/FastModel/

For further information, please contact leslie.moy@ssec.wisc.eduor kevinb@ssec.wisc.edu.

Last updated: 27 May 2003