[ncl-talk] Strange stippling behaviour when valid points are few and scattered apart

Wed Jul 8 10:04:11 MDT 2020

Hi Adam and Dennis,

Thanks for the follow-up and routines. In the past I resorted to deleting _FillValue, but I had not realized that this was a graphics bug.

Cheers,
-Jonathan

On Jul 8, 2020, at 5:51 PM, Adam Phillips via ncl-talk <ncl-talk at mailman.ucar.edu<mailto:ncl-talk at mailman.ucar.edu>> wrote:

Hi all,
My reply was accidentally not attached in the previous reply, so I have added it here.
----------------------
Hi all,
Building upon what Rick said (offline), this is obviously a graphics bug of some kind. I believe this bug exists in other similar situations as well. I think what is going on is that the valid values all fall between 95 and 100 (which are the set contour levels), and the rest of the values are set to _FillValue. This leads to a situation where all valid values fall between two contours, which is a situation that NCL admittingly does not handle well when filling, and can lead

Another thing to keep in mind is that when stippling, NCL does not apply stipple at each grid box. If a minimal number of points are spaced apart from one another, stippling might not be shown at all. Setting cnFIllMode = "RasterFIll" does not alter this behavior at all. This is the downside of using stippling on sparse spatially inhomogeneous data.  The solution to this issue is to not use any of the gsn_csm_* plotting routines, but to use gsn_add_polymarker to add a dot (or stipple) at each significant grid box.

I have attached a script that will color fill the significant grid boxes green (done only to highlight where the boxes are), and then will apply a polymarker to each valid grid box.

Hope that helps!
Adam

On Tue, Jul 7, 2020 at 6:48 PM Dennis Shea via ncl-talk <ncl-talk at mailman.ucar.edu<mailto:ncl-talk at mailman.ucar.edu>> wrote:
This was looked at offline.

Likely, there is a "bug of some kind." See Adam's response below. He also provided a sample script
.

Adam Phillips provided the attached script.

Stay Healthy
==========

On Thu, Jul 2, 2020 at 12:57 PM Buzan, Jonathan via ncl-talk <ncl-talk at mailman.ucar.edu<mailto:ncl-talk at mailman.ucar.edu>> wrote:
You need to delete _FillValue and set the number to some arbitrary value.

-Jonathan

On Jul 2, 2020, at 8:55 PM, Rashed Mahmood <rashidcomsis at gmail.com<mailto:rashidcomsis at gmail.com>> wrote:

I do not think that we should use 0 as missing value, please also see this page:
http://www.ncl.ucar.edu/Document/Language/error_messages.shtml#ZeroMissingValue
and no, setting _FillValue to some arbitrary number does not work.

I am guessing that there may be a workaround for this, which I haven't  found yet!

On Wed, Jul 1, 2020 at 7:06 PM Buzan, Jonathan <jbuzan at purdue.edu<mailto:jbuzan at purdue.edu>> wrote:
I think it’s interpolating between the data.
Have you tried deleting the _FillValue and set it to 0 (or some arbitrary number)?

-Jonathan

On Jul 1, 2020, at 6:25 PM, Rashed Mahmood <rashidcomsis at gmail.com<mailto:rashidcomsis at gmail.com>> wrote:

Jonathan,
Please look at the if statement that was introduced just to play with it a bit. Please have a look at the plot named *_Original.pdf. This is what the question is about.

Rashed

On Wed, Jul 1, 2020 at 6:03 PM Buzan, Jonathan <jbuzan at purdue.edu<mailto:jbuzan at purdue.edu>> wrote:
Hi Rashed,

Looking at your script, it looks like the stipple is doing exactly what you said it to do:

      do n=0,20
         sig({n},:) = 95.01
      end do

You set  resS at cnLevels             =(/ 95.,100. /)

The {n} in sig({n},:) means to set the latitudes 0-20 to 95.01. And the stipple is greater than 95. But lower than 100. So it stippled from 0-20 latitude.

-Jonathan

> On Jul 1, 2020, at 5:57 PM, Rashed Mahmood via ncl-talk <ncl-talk at mailman.ucar.edu<mailto:ncl-talk at mailman.ucar.edu>> wrote:
>
> Hi NCL,
>
> I encountered strange stippling behaviour when valid data points are very few and are scattered apart from each other. At best I would expect no stippling overlaid on a filled map, however, it seems that NCL somehow covers the whole map with stippling.
>
> To explain this, I created a small example script (attached) which reads data from the attached file. To play with it a bit, I just added some "fake data" to see what happens. In the script we could set it using "add_bogus_data = True".
> Attached are two plots after using original data and adding some fake data. I am not sure what is going on here, anyone? Ignore the file names!!
>
> Cheers
> Rashed
>
> <funny_stiple.ncl><tst_diff.nc<http://tst_diff.nc/>><Mask_TEST_Original.pdf><Mask_TEST_Bogus.pdf>_______________________________________________
> ncl-talk mailing list
> ncl-talk at mailman.ucar.edu<mailto:ncl-talk at mailman.ucar.edu>
> List instructions, subscriber options, unsubscribe:
> https://mailman.ucar.edu/mailman/listinfo/ncl-talk

_______________________________________________
ncl-talk mailing list
ncl-talk at mailman.ucar.edu<mailto:ncl-talk at mailman.ucar.edu>
List instructions, subscriber options, unsubscribe:
https://mailman.ucar.edu/mailman/listinfo/ncl-talk
_______________________________________________
ncl-talk mailing list
ncl-talk at mailman.ucar.edu<mailto:ncl-talk at mailman.ucar.edu>
List instructions, subscriber options, unsubscribe:
https://mailman.ucar.edu/mailman/listinfo/ncl-talk

--
Adam Phillips
Associate Scientist,  Climate and Global Dynamics Laboratory, NCAR
www.cgd.ucar.edu/staff/asphilli/<http://www.cgd.ucar.edu/staff/asphilli/>   303-497-1726

<http://www.cgd.ucar.edu/staff/asphilli>
_______________________________________________
ncl-talk mailing list
ncl-talk at mailman.ucar.edu<mailto:ncl-talk at mailman.ucar.edu>
List instructions, subscriber options, unsubscribe:
https://mailman.ucar.edu/mailman/listinfo/ncl-talk

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.ucar.edu/pipermail/ncl-talk/attachments/20200708/c6fa975c/attachment.html>