[ncl-talk] hovmoller plotting issue using high-resolution data

Wed Feb 24 14:00:11 MST 2016

Thanks Alan for the reply and input, as well as to Jack. 

I indeed only have 12 time elements available, and can create more or less spatial elements (currently there are 167). However it appears that, regardless of the plotting program (I've tried Python as well with the same result), that the small time dimension (which I can't add more available times) makes it difficult to get this signal without the result that NCL is plotting in the top right. And I do not think I have the energy to manipulate NCL's plotting procedure to change how it does the contour filling for my case.

If anyone does think of a solution to interpolate the steps in-between the positive sensitivity values in the top right, please let me know. Otherwise I will try and think of alternative methods to get the signal. 

But I will say: Thank you to all on NCL-Talk for your attention and input. I appreciate your insights.

Best Wishes,
Jeremy

On Feb 23, 2016, at 1:18 PM, Alan Brammer <abrammer at albany.edu> wrote:

> Let’s try and actually attach an image this time. 
> <var2_all in combined_fields.png>
> 
>> On 23 Feb 2016, at 13:16, Alan Brammer <abrammer at albany.edu> wrote:
>> 
>> I don’t have a solution but to show that it’s not really an NCL thing here is the same data in Panoply.  I think it’s really a problem with the big mismatch in dimension sizes.  I can’t think of a way to interpolate the steps in the top right without getting the result NCL is plotting.  
>> 
>> Higher temporal resolution would smooth out the step function.  
>> 
>> 
>> 
>>> On 23 Feb 2016, at 13:10, Jack Ogaja <jack.ogaja at gmail.com> wrote:
>>> 
>>> Hi Jeremy,
>>> 
>>> Yes, your variables readings are correct, my earlier suggestion was wrong. However, I don’t see any way of selectively filling the areas between the contour levels using “gsn_csm_contour”. I think “cnFillon = True” will always fill all areas between the contour levels. Probably the only way is to use “gsn_contour_shade” just exactly as you have done for the statistical significance.
>>> 
>>> Or maybe somebody else have an idea…
>>> 
>>> Jack
>>> 
>>>> On 23 Feb 2016, at 17:12, Jeremy Berman <jdberman at albany.edu> wrote:
>>>> 
>>>> Hi Jack,
>>>> 
>>>> Thanks for the response. I double-checked, but the variables read in my script are correct: var1_all (ensemble-mean values) is stored correctly in cntfld (for contour lines), var2_all (sensitivity values) is stored correctly in shdfld (for the color fill or shading), and for that matter tscore_all is stored correctly in stsfld (for the stippling). I should mention that the color fill and contouring in the previously attached image for the most part is what I expected --- with the exception being the issue I brought up.
>>>> 
>>>> To reiterate my issue: I don't understand how or why NCL is doing the color fill of low positive sensitivity between the "boxes" of high positive sensitivity (the color fill) (e.g., at 1400 km at 21 h, 1600 km at 24 h, 1800 at 27 h, etc.), rather than just having the signal be persistent between those boxes. Again, there is only data every 3 hours (as labeled on the y-axis) so there are no values in-between those "boxes" but NCL appears to be treating those regions as some other value.
>>>> 
>>>> Thus my problem still persists. If you or anyone have any ideas to address this issue, that would be much appreciated. 
>>>> 
>>>> Thanks,
>>>> Jeremy
>>>> 
>>>> On Feb 23, 2016, at 6:13 AM, Jack Ogaja <jack.ogaja at gmail.com>
>>>> wrote:
>>>> 
>>>>> It looks like you are just mixing up the variables. If you switch var1_all and var2_all and re-evaluate your level limits you will probably see what you expect.
>>>>> 
>>>>> Jack
>>>>> 
>>>>>> On 22 Feb 2016, at 21:51, Jeremy Berman <jdberman at albany.edu> wrote:
>>>>>> 
>>>>>> Hello,
>>>>>> 
>>>>>> I am creating a Hovmoller using data that was computed based off of a WRF simulation (15 km grid, output every 3 hours). The data consists of an ensemble-mean isentropic level of potential vorticity, sensitivity values, and a student's t-test; all as a function of time (12 different forecast hours) and distance along the hovmoller (~2400 km). My NCL script reads in this data, and overlays these three variables on a single plot, as a function of forecast hour and distance --- the resulting plot is attached.
>>>>>> 
>>>>>> The issue is focused on the shading and how NCL is plotting it. The shading corresponds to the sensitivity values, with positive (negative) values in the warm (cool) colors. However, NCL seems to be doing something weird with interpolating between the data points. For example: there are boxes of high positive sensitivity at 1400 km at 21 h, 1600 km at 24 h, 1800 at 27 h, etc.
>>>>>> 
>>>>>> Thus my question is: Is there something wrong with how I am plotting the data in NCL? Such as with my resources? I am using Rasterfill for example, which has worked in previous applications, but have tried other ones with the same result. Or is there something inherently wrong with using NCL for this application?
>>>>>> 
>>>>>> Has anyone experienced this plotting problem before, and have any suggestions? Attached is the .nc file (combined_fields.nc), the main NCL script (hovmoller_clean_pv.ncl), and the outputted image (NCL_hovmoller_335K_PV.png). I am using NCL version 6.1.2.
>>>>>> 
>>>>>> Thank you for your time and consideration.
>>>>>> 
>>>>>> Many Thanks,
>>>>>> 
>>>>>> Jeremy Berman
>>>>>> PhD Graduate Student
>>>>>> University at Albany/SUNY
>>>>>> 
>>>>>> <NCL_hovmoller_335K_PV.png><hovmoller_clean_pv.ncl><combined_fields.nc>
>>>>>> 
>>>>>> _______________________________________________
>>>>>> ncl-talk mailing list
>>>>>> ncl-talk at ucar.edu
>>>>>> List instructions, subscriber options, unsubscribe:
>>>>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>>>> 
>>>> 
>>> 
>>> _______________________________________________
>>> ncl-talk mailing list
>>> ncl-talk at ucar.edu
>>> List instructions, subscriber options, unsubscribe:
>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20160224/d0f8dc3d/attachment.html 