Re: [ferret_users] Embedded expressions and @LOC

[Jaison Kurian <jaison@xxxxxxxxxxxxxxxxxx>]

Hi Steve,
          In summary, the errors you faced occurs just because of the 
"rounding off" when the value returned from @MAX is used with @LOC.

How to fix it ? (a) Use a very high precision like @LOC:`max_val,p=-10` 
                (b) Use the idea of masking --> may not be easier for
                         practical purposes.

Details :

When does the bad value comes ?? Suppose the maximum value is 29.53875.
@MAX (when used as @LOC:`max_val`) will return an approximated value 
29.539. Since this value is precisely higher than the maximum value in 
the array, @LOC:`max_val` will retrun a bad value.

Where does the decimal places come from ?? Suppose the maximum value is
29.62250. @MAX (when used as @LOC:`max_val`) will return an approximated 
value 29.622. Now it is clear that why @LOC returns a cooridnate (Y) value
slightly less than the actual, with decimal points.

Example for (a) and (b) :

     use coads_climatology
     let xx = 161 ! X=60  for "bad value" case
                  ! X=161 for "decimal"   case

     let max_val   = SST[l=1,x=`xx`,j=@MAX]
     let max_loc   = SST[l=1,x=`xx`,j=@LOC:`max_val`]
     let max_loc_p = SST[l=1,x=`xx`,j=@LOC:`max_val,p=-10`]
     let max_mask  = IF SST[l=1,x=`xx`] EQ max_val THEN Y[GY=SST]
 
     say " max_val   = `max_val`"
     say " max_loc   = `max_loc`"
     say " max_loc_p = `max_loc_p`"
     set list/precision=7
     list SST[l=1,x=`xx`], max_mask

Hope this helps,

Jaison


On Tue, 12 Sep 2006, Stephen R. Guimond wrote:

> Hello,
>   I want to find the maximum value for each time step in an array of 
> data, and then find the coordinate position that the maximum value 
> occurs at.  I have used a loop in a script like this:
> 
> repeat/range=1:18:1/name=g (let max_val = lf[l=`g`,j=@MAX];let max_rad 
> = lf[l=`g`,j=@LOC:`max_val`];say `max_val`;say `max_rad`;list/l=`g` lf)
> 
> However, "max_val" is not finding the precise locations of my maximum 
> value.  I thought this had to do with precision, so I used 
> `max_val,p=-3` in the @LOC statement to specify 3 decimal places, but 
> there is still a problem.  See the output below.  For output 1, the 
> @LOC value should be 150, but it comes up as 149.97.  For output 2, the 
> @LOC value should be 50, but it comes up "bad" even though it has the 
> exact same number.  Any ideas??
> 
> Thanks,
> Steve
> 
> OUTPUT.....
> 
> result of `max_val,p=-3` = 5.309
> result of @LOC = 149.97
>              VARIABLE : LF
>              FILENAME : quad_rain_trop.dat
>              SUBSET   : 20 points (Y)
>              TIME     : 01-AUG-2004 18:00
> 50     /  1:  3.298
> 100    /  2:  4.598
> 150    /  3:  5.309
> 200    /  4:  3.017
> 250    /  5:  1.740
> 300    /  6:  1.111
> 350    /  7:  0.669
> 400    /  8:  1.114
> 450    /  9:  0.663
> 500    / 10:  0.781
> 550    / 11:  0.900
> 600    / 12:  1.161
> 650    / 13:  1.535
> 700    / 14:  1.620
> 750    / 15:  1.194
> 800    / 16:  1.335
> 850    / 17:  0.660
> 900    / 18:  0.266
> 950    / 19:  0.682
> 1000   / 20:  0.547
> 
> result of `max_val,p=-3` = 5.331
> result of @LOC = bad
>              VARIABLE : LF
>              FILENAME : quad_rain_trop.dat
>              SUBSET   : 20 points (Y)
>              TIME     : 02-AUG-2004 18:00
> 50     /  1:  5.331
> 100    /  2:  2.335
> 150    /  3:  2.186
> 200    /  4:  0.892
> 250    /  5:  0.329
> 300    /  6:  0.778
> 350    /  7:  1.785
> 400    /  8:  1.896
> 450    /  9:  1.289
> 500    / 10:  0.792
> 550    / 11:  1.213
> 600    / 12:  1.934
> 650    / 13:  1.697
> 700    / 14:  0.743
> 750    / 15:  0.828
> 800    / 16:  0.818
> 850    / 17:  0.999
> 900    / 18:  1.374
> 950    / 19:  1.599
> 1000   / 20:  2.458
> 
> 
> =======================================================
> Stephen R. Guimond
> Graduate Research Assistant
> Center for Ocean-Atmospheric Prediction Studies (COAPS)
> Tallahassee, FL 32304
> =======================================================
> 
> ----------------------------------------------------------------
> 
>