Data masking issue when interpolating to Grid

[lauratomkins]

• Py-ART version: 1.19.2
• Python version: 3.9.20
• Operating System: Windows 10

Description

I'm working on interpolating some tropical cyclone data and noticed that the data seems to get masked after interpolating to a Grid object. This is also evident in some of the cases that I put together for this example

I'm not sure what's going on here and haven't been able to dig much into it, but it seems a little similar to issues #957 and #942 . I was able to replicate the masking and fuzzy edges with both nearest neighbor and cressman weighting methods.

This can be replicated with the examples using NEXRAD data on this page. Considering the non-NEXRAD examples are working as expected, could this be something that changed with the NEXRAD level2 files? It seems unlikely to me, but I can't see anything that changed in the gridding algorithm. Any insight would be appreciated!

[zssherman]

Hi @lauratomkins , at the moment there was no changes to the gridding code. Is there any luck when changing the ROI in the gridding? I wonder if its picking up so missing values. I added in @rcjackson as well as he was a apart of this discussion in the past.

[lauratomkins]

@zssherman , it looks like it is an issue with the ROI. I have been using the default values but there were some changes to these default ROI values 7 months ago #1552.
Specific line here

When I change nb back to the original value of 1.5, I get much a much more reasonable interpolation and no fuzzy/artifacts at the edges.

Interpolation with (new) default nb of 1.0

Interpolation with previous default nb of 1.5

Should I update the function calls in my example page to explicitly set the nb values to 1.5? Would it be a possibility to have a dynamic nb value based on the source of the input data? Or at the least perhaps an example page or a blog post that explains the ROI process/variables and lists some best practices for interpolating data from different radars? I can imagine as a first time Py-ART user working with NEXRAD data and the default parameters, this output would be pretty confusing.

[zssherman]

@lauratomkins Good catch! I forgot about the ROI changing in the PR with the new gridding functions. If you don't mind updating the example, be much appreciated! I also agree on a blog post/example, I think short term I can work on that. Long term I would have to think on an approach for a dynamic value based on input data (users might want a different values for specific reasons possibly) but not 100% sure, but I'll think on that part of it.