Monthly Means and Variances
                       ------- ----- --- ---------

                        Treatment of Missing Data


There are several fields that often have missing data.  For example, cloud 
temperatures are hard to define without clouds, and some of the isentropic 
layers (ex. 270K) can go underground.  To make the problem more difficult, 
GRIB does not provide a method to indicate the number observations used to 
compute the average at any grid point.  Possible methods of dealing with the 
problem are:


     A) Compute the mean if 100% of data are available

     B) Compute the mean if xx% of data are available

     C) Compute the mean if any data are available


The two main classes with missing data are (1) isentropic level data, and 
(2) everything else (mainly cloud data).  For class (1), I used method (A) and 
for class (2), I used method (C).




                                Variances
                                ---------


The monthly variances were computed by 


            X = (1/N) * (x(1)+x(2)+...+x(N))
            X = monthly mean
            N = number of points in mean
            Var = (1/N) * ((x(1)-X)**2 + (x(2)-X)**2 +...+ (x(N)-X)**2)