Enhance MET statistics tools for climatological data.

[dwfncar]

Enhance MET statistics tools for climatological data by doing the following:

• Update GRIB library to support TR = 51.
  
• Remove the "-climo" flag from Point-Stat since we're re-doing this functionality.
  
• Add config file options for Point-Stat, Grid-Stat, and Ensemble-Stat to specify climatology means and standard deviations (optional). Not currently using stdev but will in the future for computing binned scores.
  
• Just like we have "fcst" and "obs" dictionaries, add one for "climo" that can either be empty (i.e. no climo) or of the same length as "fcst" and "obs". Include the path the climatology mean and stdev data. Read these arrays of file names from each "field" entry to give the user the most flexibility in organizing climo data.
  
• Add NCEP's 2.5 degree climo data in GRIB format to the MET website and include instructions for user's to grab it an use it.
  
• When evaluating a forecast, find the matching climatology fields (mean and stdev), searching for the nearest month/hour of day before and after the forecast time. For example, for 03Z data, find the 00Z and 06Z data for the same month. If only one is found, use it.
    - Add option for time interpolation method:
      - NEAREST for the closest one in time
      - UW_MEAN for the before/after average
      - DW_MEAN for linear interpolation in time (default)
  
• Interpolate linearly in time between the data before and after the forecast time.
  
• Automatically regrid this data to the verification grid.
  
• For each forecast/observation matched pair, compute the corresponding climo value and use it when computing statistics.
  
• Add the SAL1L2 output line type for Grid-Stat.
  
• Add the following statistics to existing line types:
     - AC (Anomaly Correlation) (or PAC for pattern anomaly correlation?) to the CNT line type.
        - Figure out if this can be derived from the SAL1L2 line type and add that functionality to STAT-Analysis.
     - BSS (Brier Skill Score) to the PSTD line type.
     - CRPSS (Continuous Ranked Probability Skill Score) to the RHIST line type.
     - RPSS (Ranked Probability Skill Score) to the RHIST line type.
     - Consider defining a new line type for Ensemble-Stat for continuous ensemble statistics (CRPS, IGN, CRPSS, RPSS).
  

Also need to determine how Economic Value should be calculated and where it should go in the output.

Questions...

(1) The NCEP climo data is monthly. When searching for a match, only consider records where the climo month matches the forecast month. What is users have climo data that's seasonal or daily? How to instruct MET to match those? [MET-549] created by johnhg

[dwfncar]

I just took a look at the 1.0 degree climatology data and it's daily GRIB files instead of monthly as in the 2.5 degree data:
  dakota:/d3/projects/MET/NCEP_unification/climdata_1deg

To handle that, we'll need a flag in the config file to specify "daily" vs "monthly" to control the matching logic. Should we support "seasonal" as well, or not until it comes up and we'd have something with which to test? by johnhg

[dwfncar]

The link to Yuejian's ensemble verification description can be found at: http://www.emc.ncep.noaa.gov/gmb/yzhu/naefs/ENS_VERIFICATION_short.pdf
This defines the calculation of RPSS, CRPSS, and to some extent Economic Value (which follows Murphy 1977) by jensen