Re: [ferret_users] ferret inconsistency/errors in masking

[Ansley Manke <ansley.b.manke@xxxxxxxx>]

Hi Lev,
It's a bit hard to tell without seeing what the grids of the different 
variables are, and so on, but I imagine that this may have to do with 
the order of operations as the expression is being evaluated. You have 
some regridding operations going on in conjunction with the IF test. In 
the incorrect result you show below, is it showing H*(1.-0.91/1.028)?

To test that, try some different ways to make the definitions. Try:

   let hd  = if ifloat[g=h] GT 0.9 THEN H*(1.-0.91/1.028) ELSE HDG-SLEV
or
   let hd  = if ifloat[g=h] GT 0.9 THEN H*(1.-0.91/1.028) ELSE 
HDG[g=h]-SLEV[g-h]

Also does "cancel mode interpolate" make a difference?

Finally, looking at the output when "set mode diagnostic" is turned on 
may give us an idea.  It'll show variables being read or loaded into 
memory and grids being reconciled with one another as the expression is 
evaluated.

Ansley

On 2/25/2014 6:36 PM, Lev Tarasov wrote:
> I'm trying to merge some data sets, and get inconsistent
> results:
>
> For instance, variable hd, defined as follows:
> yes?show var hd
>  HD = IF IFLOAT GT 0.9 THEN H*(1.-0.91/1.028) ELSE HDG-SLEV
> should equal hdg-slev for the following region, since ifloat=0:
>
> yes? stats ifloat[g=h]
>              IF H*0.91/1.028 LT SLEV+HP THEN 1 ELSE 0
>              LONGITUDE: 89W (interpolated)
>              LATITUDE: 36.5N (interpolated)
>              Z:  N/A
>              T: -121 to -33
>              E:  N/A
>              F:  N/A
>              DATA SET: ./idHept.nn9894
>
>  Total # of data points: 44 (1*1*1*44*1*1)
>  # flagged as bad  data: 0
>  Minimum value: 0
>  Maximum value: 0
>  Mean    value: 0 (unweighted average)
>  Standard deviation: 0
>
> but it doesn't:
>
> yes? stats hd[g=h]-(hdg[g=h]-slev[g=h])
>
>              HD[G=H]-(HDG[G=H]-SLEV[G=H])
>              LONGITUDE: 89W (interpolated)
>              LATITUDE: 36.5N (interpolated)
>              Z:  N/A
>              T: -121 to -33
>              E:  N/A
>              F:  N/A
>              DATA SET: ./idHept.nn9894
>
>  Total # of data points: 44 (1*1*1*44*1*1)
>  # flagged as bad  data: 0
>  Minimum value: -84.7
>  Maximum value: 92.7
>  Mean    value: -29.884 (unweighted average)
>  Standard deviation: 46.089
>
>      FERRET v6.84
>      Linux 2.6.9-101.ELsmp 64-bit - 07/19/13
>
> How do I fix this or is it a bug?
>
> Cheers,
>
> lev
>
> Lev Tarasov -   Dept of Physics and Physical Oceanography,
>         Memorial University of Newfoundland.
>                 email: lev@xxxxxx
>                 http://www.physics.mun.ca/~lev/
>         Tel (709)-864-2675
>         Fax (709)-864-8739