[ncl-talk] problem removing missing values from obs dataset

[Dave Allured - NOAA Affiliate]

This is not valid:

   *VAR = where (VAR.eq.VAR at _FillValue, 0., VAR)*

... because the array result of the *".eq."* operator, with fill values
present, can contain three values:  True, False, and fill value.  The
*"where"* function must receive a conditional mask (first argument) that
contains only *True* and *False*.

This construction will form the desired missing value mask, and do what you
intended:

   *VAR = where (ismissing (VAR), 0., VAR)*

Next, if *VAR at missing_value* is different than the original *VAR at _FillValue*,
then you should keep the second statement the way you had it, to change
*"missing_values"* to zero.


On Thu, Feb 27, 2020 at 4:50 PM Adam Herrington via ncl-talk <
ncl-talk at ucar.edu> wrote:

> Hi All,
>
> I have a variable that I am trying to remove the missing values from. Here
> are the attributes from printVarSummary call:
>
> Variable: VAR
> Type: float
> Total Size: 8220672 bytes
>             2055168 values
> Number of Dimensions: 3
> Dimensions and sizes: [time | 223] x [lat | 32] x [lon | 288]
> Coordinates:
>             time: [9504..10170]
>             lat: [-14.60732984293194..14.60732984293194]
>             lon: [   0..358.75]
> Number Of Attributes: 9
>   remap : remapped via ESMF_regrid_with_weights: Bilinear
>   _FillValue : -9999.9
>   missing_value : -9999.9
>   time_statistic : instantaneous
>   units :
>   long_name : precipitation
>   comments : Unknown1 variable comment
>   delta_t : 0000-00-00 03:00:00
>   avg_period : 0000-00-00 01:00:00
>
> So there are two missing value attributes. To set all missing values to
> zero, I try:
>
>   VAR = where(VAR.eq.VAR at _FillValue,0.,VAR)
>   VAR = where(VAR.eq.VAR at missing_value,0.,VAR)
>
> But then subsequently looking for missing values: nMsg =
> num(ismissing(VAR)), it says there are still all the missing values:
>
> (0) missing values present: not allowed: nMsg=17913
>
> Does anyone see why this is occurring, why my where statements are not
> resulting in num(ismissing(VAR))=0?
>
> Thanks,
>
> Adam
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20200227/29e4565f/attachment.html>