Generalize EC cell widths function #526

xylar · 2020-04-19T08:30:06Z

Many supposed parameters are actually hard-coded. Documentation has been updated to adhere to numpydoc formatting. Unnecessary looping has been removed.

This merge also improves plotting of cell widths in define_base_mesh. Imports are now at the top of the file. Colormap is now perceptually uniform (not "jet"!) Plot layout is now tighter (though there's still considerable whitespace).

Many supposed parameters are actually hard-coded. Documentation has been updated to adhere to numpydoc formatting. Unnecessary looping has been removed.

Imports should be at the top of the file. Colormap is now perceputally uniform (not "jet"!) Plot layout is now tighter (though there's still considerable whitespace.

xylar · 2020-04-19T08:31:42Z

@sbrus89 and @pwolfram, these changes may affect you depending on what you use as your background mesh. Either way, these changes would be a good template for future changes to mesh_definition_tools.py

xylar · 2020-04-20T14:39:05Z

testing_and_setup/compass/ocean/jigsaw_to_MPAS/build_mesh.py

-            ), extent=[-180, 180, -90, 90], cmap='jet')
+            im = ax.imshow(cellWidth, origin='lower',
+                           transform=ccrs.PlateCarree(),
+                           extent=[-180, 180, -90, 90], cmap='viridis_r')


Make this rainbow.

Please don't-- this doesn't add value and rainbow is a notoriously bad colormap.

(unless rainbow is the perceptually uniform one, not "jet") In that case this is ok.

@pwolfram, thanks for the feedback. I agree, rainbow isn't a big improvement over jet. How about another solution? I have been using a colormap from paraview for resolution that is made up of 3 perceptually uniform segments ("Blue - Green - Orange"), the second colormap here: https://sciviscolor.org/3-wave-colormaps/. It will be more complicated to add this to COMPASS, since it's not standard in matplotlib, but I'll see what I can do. Would that work for you? @mark-petersen, what do you think?

The problem with viridis is that it's too dark and the black outlines from cartopy don't show up well, but this could be fixed by drawing in the continents, not just the outlines.

I'm happy with perceptually uniform solutions, but don't like it when you can see "lines" like yellow, in the colormap. But, I'm ok with non-jet solutions like proposed below-- thanks!

xylar · 2020-04-20T14:39:28Z

testing_and_setup/compass/ocean/jigsaw_to_MPAS/build_mesh.py

+            fig.canvas.draw()
+            plt.tight_layout()


This cleans up some whitespace.

xylar · 2020-04-20T16:32:40Z

@mark-petersen, I'm still tinkering with this one. Please don't merge yet.

xylar · 2020-04-20T16:59:25Z

@pwolfram, @mark-petersen and anyone else who wants to weigh in, which is better?
1.

2.

I personally prefer 2 (viridis_r) because it's perceptually uniform. I can see how you could get more granularity from the blue-green-orange version. But it's much harder to tell which is higher and which is lower res.

xylar · 2020-04-20T17:32:49Z

Pending @pwolfram and @mark-petersen's approval of using viridis now that I'm drawing the continents and not just their outlines, I think this is ready to merge.

xylar · 2020-04-20T18:53:06Z

testing_and_setup/compass/ocean/jigsaw_to_MPAS/build_mesh.py

bbox_inches='tight' eliminates some of the remaining white space around the image so it looks right.

Thanks, good to know.

mark-petersen · 2020-04-20T19:01:24Z

@xylar I strongly prefer your first one above (de-saturated rainbow?) because I can perceive six colors, and they are in an order I expect. In viridis_r I only see three colors, and the top half of the colorbar looks the same to me, all dark blue. I made a similar one at the same time, but with slightly muted colors compared to above. I didn't realize we were both working on it:

Here is the colorbar code.
I really like your continents filled in. Thanks for changing it.
Despite my work on this one, I would go with your de-saturated rainbow above, out of these three.

mark-petersen · 2020-04-20T19:05:00Z

testing_and_setup/compass/ocean/jigsaw_to_MPAS/3Wbgy5.xml

It looks like you downloaded this directly from Paraview? That was smart.

No, copied from MPAS-Analysis, which gives proper credit and got it from the original source (LANL's own SciVisColor web page, I believe).

xylar · 2020-04-20T19:31:20Z

@mark-petersen, so the problem is a really hard one. More colors does give more granularity in some sense but it really leaves you with no intuition about where is high res and where is low. As Leif pointed out, he couldn't even tell how many regions of different resolution there were in my plot 1. above -- he thought there were 9, and that the yellow bands were really distracting as if they had extra meaning. This is the general complaint about color maps that don't transition uniformly from dark to light or visa versa, and the reason for the push for perceptual uniformity.

xylar · 2020-04-20T19:46:04Z

I played around with a bunch more color maps and I do think 3Wbgy5, my option 1. above, might be the best option. As @mark-petersen points out, anything that's perceptually uniform tends to be made up of 3 distinct colors or so (the ones that aren't are gawd-awfully ugly). So I think we sacrifice the intuitive transition from light meaning high res to dark meaning low res in order to get more granularity in what isocontours correspond to what resolutions.

Add support for a colormap form SciVisColor (3Wbgy5).

mark-petersen · 2020-04-20T20:37:23Z

Added a min/max to title because it is easier to read there:

mark-petersen

@xylar thanks for your work on this. It all looks great!

mark-petersen · 2020-04-20T20:46:36Z

@pwolfram and @sbrus89 I think this is a nice improvement. I'd like to merge today, so we can rebase the mesh PRs on it.

pwolfram

Approved by inspection

pwolfram · 2020-04-20T21:21:10Z

@sbrus89, can you please review the above?

sbrus89

I agree this is a nice improvement. I approve by inspection.

mark-petersen · 2020-04-20T23:20:02Z

Thanks. Tested one last time with QU240, QU240wISC, EC60to30wISC, EC60to30, SO60to10wISC. Everything appears to be working and making the plots correctly.

Generalize EC cell widths function MPAS-Dev#526 Many supposed parameters are actually hard-coded. Documentation has been updated to adhere to numpydoc formatting. Unnecessary looping has been removed. This merge also improves plotting of cell widths in define_base_mesh. Imports are now at the top of the file. Colormap is now perceptually uniform (not "jet"!) Plot layout is now tighter (though there's still considerable whitespace).

xylar added 2 commits April 19, 2020 10:25

Generalize EC cell widths function

a95a22d

Many supposed parameters are actually hard-coded. Documentation has been updated to adhere to numpydoc formatting. Unnecessary looping has been removed.

Improve plotting of cell widths in define_base_mesh

b3cc8a4

Imports should be at the top of the file. Colormap is now perceputally uniform (not "jet"!) Plot layout is now tighter (though there's still considerable whitespace.

xylar added Ocean COMPASS clean-up labels Apr 19, 2020

xylar requested a review from mark-petersen April 19, 2020 08:30

xylar assigned mark-petersen Apr 19, 2020

xylar requested review from pwolfram and sbrus89 April 19, 2020 08:30

xylar mentioned this pull request Apr 19, 2020

Convert CUSP to signed distance #523

Merged

xylar commented Apr 20, 2020

View reviewed changes

xylar added the in progress label Apr 20, 2020

xylar removed the in progress label Apr 20, 2020

xylar commented Apr 20, 2020

View reviewed changes

mark-petersen reviewed Apr 20, 2020

View reviewed changes

Tighten plotting and add more colormap support

964e31a

Add support for a colormap form SciVisColor (3Wbgy5).

xylar force-pushed the ocean/generalize_ec branch from 67f6b51 to 964e31a Compare April 20, 2020 19:47

Add min and max to plot title

0307149

mark-petersen self-requested a review April 20, 2020 20:37

mark-petersen approved these changes Apr 20, 2020

View reviewed changes

pwolfram approved these changes Apr 20, 2020

View reviewed changes

sbrus89 approved these changes Apr 20, 2020

View reviewed changes

mark-petersen merged commit 5c7fc77 into MPAS-Dev:ocean/develop Apr 20, 2020

xylar deleted the ocean/generalize_ec branch April 21, 2020 09:27

matthewhoffman added clean-up COMPASS Ocean labels Mar 17, 2021

Generalize EC cell widths function #526

Generalize EC cell widths function #526

Uh oh!

Conversation

xylar commented Apr 19, 2020

Uh oh!

xylar commented Apr 19, 2020

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xylar Apr 20, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xylar commented Apr 20, 2020

Uh oh!

xylar commented Apr 20, 2020

Uh oh!

xylar commented Apr 20, 2020

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mark-petersen commented Apr 20, 2020

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

xylar commented Apr 20, 2020

Uh oh!

xylar commented Apr 20, 2020

Uh oh!

mark-petersen commented Apr 20, 2020

Uh oh!

mark-petersen left a comment

Choose a reason for hiding this comment

Uh oh!

mark-petersen commented Apr 20, 2020

Uh oh!

pwolfram left a comment

Choose a reason for hiding this comment

Uh oh!

pwolfram commented Apr 20, 2020

Uh oh!

sbrus89 left a comment

Choose a reason for hiding this comment

Uh oh!

mark-petersen commented Apr 20, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

xylar Apr 20, 2020 •

edited

Loading