==========
Section 1:  Parameters from which statistics are calculated
==========
 1) $fld       = field number, 1-11 (see description under Program
                 function) 
 2) $sd        = sum of differences (model-observed) for the given
                 $filetime
 3) $sa        = sum of the absolute value of the differences for the
                 given $filetime
 4) $sds       = sum of the squares of the differences
 5) $numobs    = number of observations for this $filetime.  If number
                 is greater than 1, a region has been chosen and not
		 just a single station
 6) $rsumsd    = running sum of $sd from the initially chosen $filetime to
                 the current $filetime
 7) $rsumsa    = running sum of $sa
 8) $rsumsds   = running sum of $sds
 9) $rsumnumobs= running sum of $numobs
10) $filetime  = model initialization time

From these parameters, the following statistics are obtained for each
date ($filetime):
==========
Section 2:
==========
      $runsumsd += $sd;
      $runsumsa += $sa;
      $runsumsds += $sds;
      $runsumnumobs += $numobs;
      $run_me  = $runsumsd / $runsumnumobs;
      $run_mae = $runsumsa / $runsumnumobs;
      $run_rms = sqrt ( $runsumsds / $runsumnumobs );
      $me = $sd / $numobs;
      $me2 = $rsumsd / $rsumnumobs;
      $mae = $sa / $numobs;
      $mae2 = $rsumsa / $rsumnumobs;
      $rms = sqrt ($sds / $numobs );
      $rms2 = sqrt ( $rsumsds / $rsumnumobs );
      $me_sum          = $me_sum + $me;
      $me_squaredsum   = $me_squaredsum + ($me)**2;
      $mae_sum         = $mae_sum + $mae;
      $mae_squaredsum  = $mae_squaredsum + ($mae)**2;
      $rms_sum         = $rms_sum + $rms;
      $rms_squaredsum  = $rms_squaredsum + ($rms)**2;
      $count++;  ## incremented for each date
      $me_bar  = $me_sum / $count;
      $mae_bar = $mae_sum / $count;
      $rms_bar = $rms_sum / $count;
      $sd_me  = sqrt ( ( $count * ($me_bar)**2 + $me_squaredsum - (2 * $me_bar * $me_sum) ) / ($count - 1) );
      $sd_mae = sqrt ( ( $count * ($mae_bar)**2 + $mae_squaredsum - (2 * $mae_bar * $mae_sum) ) / ($count - 1) );
      $sd_rms = sqrt ( ( $count * ($rms_bar)**2 + $rms_squaredsum - (2 * $rms_bar * $rms_sum) ) / ($count - 1) );

   $crm_me, $crm_mae, and $crm_rms are (in this case) 9-day centered
      means of $me, $mae, and $rms respectively
 

The files that do not start with "stdev." can be described as follows:
==========
Section 3:
==========
Filename Format: $fld.$model_$domain_$region_$ii$ff1001.ts
---------------
       $fld    =  {
                   temp = temperature at lowest sigma in Kelvin
                   t2   = average of surface and lowest-sigma level
                          temperature in Kelvin
                   rh   = relative humidity in percent
		   wsp  = wind speed at lowest sigma in knots
		   wsp2 = log-reduced wind speed at 10 meters in knots
		   wdir = wind direction in degrees
		   vwd  = vector wind difference
		   vwd2 = vector wind difference at 10 meters
		   slp  = mean sea-level pressure in millibars
                 pcp3_6 = precipitation in inches:
		          3-hour total at 3,9,15, and 21Z,
		          6-hour total at 0,6,12,and 18Z 
	          pcp24 = 24-hour precipitation in inches
		  }
       $model  = {mm5, eta, gfs, etc.}
       $domain = {d1 = 36-km or native grid,
                  d2 = 12-km,
		  d3 =  4-km}
       $region = {dom4, station, lat_lon, xy_coarse,
                  dom36, dom12, 36km, 12km, 4km}
       $ii     = {00, 12} forecast initialization
       $ff     = {any number 0 - 72 as 2 digits} forecast hour
       1001    = denotes surface
       .ts     = denotes time series file

Example filename:
----------------
       wsp.gfs_d1_dom4_00451001.ts => 
  contains 00Z forecast hour 45 lowest-sigma level wind speed in
  knots statistics for GFS model for all surface stations within 
  the 4-km domain

====================================
Section 4 "stdev.*" file description:
====================================
Column Headings (SPLUS format):
------------------------------
  missing values = NA
  where
Col                   Section 2
 #    Column Name     Parameter  Description
---   -------------   ---------  ---------------------------------------------
 1)   Mean_Error    = $me        mean error for the specific time in File_Date
 2)   RMean_E       = $me2       running mean of Mean_Error from initial File_Date 
                                 to File_Date
 3)   Ctr_Mean_E    = $crm_me    9-day (or other chosen interval) centered mean of
                                 Mean_Error centered on File_Date
 4)   Me_Ab_E       = $mae       mean absolute error for the specific time in File_Date
 5)   RM_Ab_E       = $mae2      running mean of Me_Ab_E from initial File_Date 
                                 to File_Date
 6)   Ctr_Me_Ab_E   = $crm_mae   9-day (or other chosen interval) centered mean of
                                 Me_Ab_E centered on File_Date
 7)   RMSE          = $rms       root mean square error for time specified in File_Date
 8)   Run_RMSE      = $rms2      running mean of RMSE from initial File_Date to File_Date
 9)   Ctr_RMSE      = $crm_rms   9-day (or other chosen interval) centered mean of
                                 RMSE centered on File_Date
10)   Count         = $numobs    Number of observations that went into Mean_Er 
                                 for File_Date
11)   Run_Count     =$rsumnumobs running mean of Count from initial File_Date to File_Date
12)   File_Date     = $filetime  model initialization time

For the "stdev*" files, these have the same filename structure as
described in Section 3 above except with "stdev." prepended.  The data
in these files are:

Mean Error      Mean Absolute Error              RMS
  $run_me              $run_mae             $run_rms    Running Values
  $me_bar              $mae_bar             $rms_bar    Means
  $sd_me               $sd_mae              $sd_rms     Standard Deviations
  $me_sum              $mae_sum             $rms_sum    Sums
$me_squaredsum   $mae_squaredsum      $rms_squaredsum   Sums of Squares
