Add Fourier Decomposition option to use prior to computing Global Statistics

[dwfncar]

Currently, the EMC Global VX includes a section where a Fourier Decomposition is applied to fields like Geopotiential Height prior to the computation of continuous statistics. Similar to Wavelet-Stat, it allows for investigation of which scales are contributing the most error. We will need to add this to MET to accomplish our NGGPS deliverables.

From an email from Mallory Rowe on 5/9/2017:
Results can be found on http://www.emc.ncep.noaa.gov/gmb/emc.glopara/vsdb/gfs2017/ (for example). Look at the Anomaly Correlation for height and then click 'Dieoff' for any pressure level or any region will display the corresponding results for the wave decomposition.

2) Some examples of the output can be found in /scratch4/NCEPDEV/global/noscrub/stat/vsdb_data/anom/00Z/gfs with the variables HGT_WV1/(number)-(number). Not sure if there is any other output, but I can investigate more if you would like.

3&4) I think the option to define an array would be nice. Perhaps something specifying the range, like "0-20", "4-9", something kind of like what you see in the output files. I think this would be useful for the Series-Analysis tool as well as Grid-Stat. [MET-804] created by jensen

[dwfncar]

NGGPS FY16 by jensen

[dwfncar]

I added support to Grid-Stat for the user requesting fourier decomposition and Grid-Stat computing it. It writes output for CNT, SL1L2, SAL1L2, VL1L2, and VAL1L2.

Still need to do the following:

(1) Randy will change fitwav1d function to return a boolean. It returns false is any bad data values are present.
(2) John will updated Grid-Stat to check the return status, print a debug(2) message about the presence of bad data, and skip that output.
(3) Add info the MET User's Guide and config README files about this option.
(4) Add a unit test to exercise this option.
(5) Still need to compare MET's output to NCEP's to make sure its doing what we expect. by johnhg

[dwfncar]

I changed fitwav_1d to return a boolean indicating the presence of bad data and Grid-Stat to skip over instances when bad status is returned with the following Debug(2) message:

DEBUG 2: Skipping Fourier decomposition for waves 0 to 3 due to the presence of bad data values.
DEBUG 2: Skipping Fourier decomposition for waves 10 to 20 due to the presence of bad data values.

Still need to do the documentation, testing, and run it by NCEP folks. by johnhg

[dwfncar]

Added new unit test for grid_stat to exercise new Fourier functionality. by johnhg

[dwfncar]

Randy fixed the 1d decomposition, and John added a new test for grid_stat to exercise the Fourier functionality.

Still should add more info the the user's guide but the development work is finished. by johnhg