Important notices for directory co2dat_4a:

  This directory is the repository of all co2 data processed (using a smoothing algorithm)
  at EMC for NCEP forecast models.

(1)
  a: Data files ended with '.txt' and without any suffix are from observations. Due to time
     delay when obtaining the data (usually one year or later), these data files are not
     suitable for real-time or future forecasts.  However, in addition for hindcast, the
     data are used for projection of next year's data and beyond.

    (For example: near the end of 2012, most of 2011 data are collected and used to produce
     observational 2011 data set, and to update the previous projection for 2012. Then this
     updated 2012 projection and 2011 or earlier observations are used to project 2013 data
     for the continuation of regular forecasts.)

  b: Data files ended with '_proj' are data projections from prior 5-year's observed data.
     For example, projected 2012 data were made in 2011, but prior five-year period of
     2006-2010 data were available at the time for the projection. The projected data are
     copied to the directory of '../fix_co2_proj' and the suffix '_proj' is removed in that
     directory.  Those projected data are suitable for real-time and future forecast.

    (note: data for _2009.txt_proj is the same as _2009.txt_proj_u)

  c: Data files ended with '_proj_u' are updated projections from prior 6-year's actual
     observations.  As explained in item a, those updated data sets are usually not produced
     early enough for actual forecasts use. Instead they are used for future data
     preparations.  Those data are copied into directory '../fix_co2_update' but the
     suffix '_proj_u' is cut off in the directory.

(2)
  Starting from 2008, a new data processing algorithm that includes a 9-point grid average
  is used to replace the previous algorithm to reduce noise in the global distribution map.
  All data in this directory are from the new smoothing method, including reprocessed data
  earlier than 2008.  This presents a little differences for the early data sets (1956-2007)
  which are from the original unsmoothing method (the early data are still kept in the
  CFS operational data directory for the purpose of preserving model's reproducibility).
  After that, all observation data sets later than 2008 are smoothed and are coming from
  this repository.