Re: strange difference btw area avg. and zonally avg.

deni boz wrote:
> Hi all,
>
> I was not able to find why there occurred a difference between area
> averaged value of a certain region and average value of zonally
> averaged values of that region?
>
> I have an array t2max(time,lat,lon) with dimension sizes ntime=248
> (3-hour maximum values in a month), nlat=52, nlon = 148.
>
> In the first step, I calculated the area averaged value for each time
> according to certain latitude and longitude values. By applying do
> loop to this 1-d array with 248 elements and selecting the maximum
> value of each day, I got daily area averaged values and finally I got
> monthly area averaged value by averaging the daily values.
>
> In the second step, I tried to determine the maximum value over all
> time at each lat/lon grid point in the t2max (time,lat,lon) array by
> using dim_max function. I got an array with 2-d (52,148) after this
> step and with the aid of dim_avg_Wrap function, I finally got an array
> consisting of zonally averaged values for the same region with a
> dimension size of 52.
>
> The problem is that when I average these 52 values, the result is not
> same as in the first step. Actually both of them must be same. Isn't it?
>
> Any comment will be appreciated. Thanks a lot!
>
> Deniz
=========================================
I have taken a quick look.

[1]
Since you have set the weights to 1.0, the following
is an arithmetic average, not an area average:

avg_temp_oct_max = wgt_areaave(oct_max,1.0,1.0,0)

Given the small area of interest, the difference between the
arithmetic and area average is not significant.

[2]
     [a] It looks like you have taken an average over a region.
          Then taken the max of these averages
     [b] Selected the max for each time period then computed the
           average over a region.

     This sequence is not 'symmetric'. Order does matter in this case.

     I speculate that
                      max_zonally_averaged_oct > max_area_oct

_______________________________________________
ncl-talk mailing list
ncl-talk_at_ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Mon Dec 24 2007 - 10:30:46 MST