[Met_help] Use of the ASCII2NC utility in MET

[John Halley Gotway]

Ed,

To answer your questions...

(1) Yes, you can build MET without BUFRLIB.  Edit the top-level Makefile (METv1.1/Makefile) and set "DISABLE_PB2NC = 1".
Then when MET is built, it will skip the step for PB2NC which is the only tool that interfaces with BUFRLIB.

When you run the test scripts, the tests that rely on the output of PB2NC will fail, but that's fine.

(2) You'll probably have trouble interfacing with those NetCDF files.  METv1.1 can read NetCDF files but only a specific format of them. We're working to make it easier to use NetCDF files in MET in
future releases, but that's not in version 1.1.  In particular, MET expects the NetCDF variables to have specific names based on the corresponding GRIB code abbreviation.  The other issue is that MET
expects the grid to be defined in the global attributes of the NetCDF file.  When using GRIB data, MET extracts the grid definition from the GRIB file itself.  In NetCDF, we expect that it's encoded
in the global attributes.

Using GRIB files would be easiest for METv1.1.  However, if you're stuck with the NetCDF files, you'll probably need to write a program to reformat them slightly to add that grid definition info and
possibly modify the variable names.

Take a look at a PCP-Combine NetCDF output file for an example of the NetCDF format MET expects:
ncdump METv1.1/out/pcp_combine/sample_fcst_12L_2005080712V_12A.nc (this is generated by the "test_all.sh" test script)

(3) To my knowledge, we have not tested MET with the Intel compilers.
Here's a list of the compilers that are known to work:
http://www.dtcenter.org/met/users/support/online_tutorial/compilation/index.php

I don't believe that we have the Intel compilers available for testing within RAL.  But I know they're on wjet over at NOAA.  So we could probably test with them over there.

Building GSL and NetCDF really isn't very difficult.  I've done it on my own machine several times.  So you could try building everything with the GNU compilers.

If you do decide to try compiling MET with the Intel compilers, please let me know how it goes.

Thanks,
John

Ed Tollerud wrote:
> John,
> 
> Thanks for the very helpful responses. Here's yet another set of
> questions: If I don't have an immediate need for BUFRLIB, is it possible
> to build MET without it? My initial set of data will be ascii lists of
> gage observations. Second, we have access to WRF data files that are
> already in NETCDF format, but otherwise appear to be the same as the raw
> GRIB files of model output that MET assumes (same fields, etc.). Is this
> a problem for MET? Third, do you know if using intel compilers (ifort,
> etc.) will work? This was suggested by the systems team here because it
> is default on the machine we will be using and because the GSL was built
> using it.
> 
> Thanks for your help!
> 
> Ed
> John Halley Gotway wrote:
>> Ed,
>>
>> Another thing occurred to me that I wanted to mention.  When you run Point-Stat, you can select what interpolation method(s) you'd like to use to match the gridded forecast values to the value at the
>> observation location.  You may want to consider using several interpolation methods to see how the verification scores change depending on the scale you're using.
>>
>> This is controlled in the Point-Stat config file by setting the "interp_method" and "interp_width" parameters.  "interp_width" defines the size of the neighborhood to look at, e.g. a value of 1 means
>> use the nearest neighbor, 2 means look in a 2x2 box at the 4 closest grid points, and N mean look in an NxN at the N*N closest grid points.  "interp_method" defines what operation you perform on those
>> N*N forecast values before matching it to the observation values.  The options for "interp_method" are listed in the config file.  Personally, I like "DW_MEAN" for distance-weighted mean or "LS_FIT"
>> for least-squared fit.
>>
>> As long as you're running Point-Stat to generate matched pairs, you might as well get as much value out of it as possible.
>>
>> Just a suggestion.
>>
>> John
>>
>> Edward Tollerud wrote:
>>   
>>> I plan to use the ASCII2NC tool to do point verification of WRF
>>> forecasts, but I don't have elevations for my gage sites. Will this
>>> cause a problem in the reformatting and/or verification steps?
>>>
>>> Thanks for your help!
>>>
>>> Ed Tollerud
>>>
>>>
>>>
>>> _______________________________________________
>>> Met_help mailing list
>>> Met_help at mailman.ucar.edu
>>> http://mailman.ucar.edu/mailman/listinfo/met_help
>>>     
>