Adding figure to paragraph "Bounds for 2-D coordinate variables with 4-sided cells" in Section 7.1 on bounds

[neumannd]

Title: Adding figure to paragraph "Bounds for 2-D coordinate variables with 4-sided cells" in Section 7.1 on bounds
Moderator: @erget
Requirement Summary: It would be nice to have a visual representation of the paragraph following "Bounds for 2-D coordinate variables with 4-sided cells" (in Section 7.1), which helps users to quickly grasp what is meant. Particularly a user might look for the order in which the four coordinates of the four vertices of a grid cell should be written into the bounds variable.
Technical Proposal Summary: Add a figure that visualizes the textual description and particulary shows the order of the vertices in the bounds variable.
Benefits: A user who looks for the particular information described in the Requirement Summary get this information considerably faster.
Status Quo: Currently, a textual description is provided. It is well written and the reader (who knows for what he/she is looking) understands what is meant. However, it takes some time. Additionally, some readers, who have difficulties imagining the spatial structure, will have issues understanding what is meant.
Detailed Proposal: Add a figure as described above. I will be happy to provide a draft figure next week if there is positive feedback.

[davidhassell]

I am happy, in general, for figures to be used in the conventions document as a supplement to the text, rather than a replacement for text.

So I support this.

[neumannd]

Attached two suggestions for the described figures. The grid lines are drawn in red and the grid cell center in green. Maybe I should add a legend?

I generated the figures with R.

[JimBiardCics]

The figures look great. One thing though. The color choice is problematic for some color-blind people. If they suffer from red-blind protanopia, the image looks like

A different color scheme might be advisable.

[davidhassell]

These are very good diagrams - thanks.

Some thoughts:

The axis labels in the second figure are not quite right - the axes are not longitude and latitude. I think you can be completely general and use discrete axes, i.e. just increasing in index (i, j) space; or else give an example of associated strictly monotonic 1-d coordinates: grid_longitude, grid_latitude or projection_x_coordinate, projection_y_coordinate. I might favour the former, more general, approach, as it more matches the text in 7.1 better, but I don't have a strong opinion.

The conventions numbers vertices 0, 1, 2, 3, so the diagrams should match this.

In the first diagram (1-d case), a legend mentioning that the trailing index numbers are reversed when the coordinates are decreasing would be useful, perhaps.

Thanks,
David

[neumannd]

@JimBiardCics Thanks for the remark. I will update the colors.

@davidhassell Thanks for the remarks.

The conventions numbers vertices 0, 1, 2, 3, so the diagrams should match this.

I will update it.

In the first diagram (1-d case), a legend mentioning that the trailing index numbers are reversed when the coordinates are decreasing would be useful, perhaps.

Or should I just duplicate the first diagram and create one with decreasing lon/x-axis to the right?

The axis labels in the second figure are not quite right - the axes are not longitude and latitude. I think you can be completely general and use discrete axes, i.e. just increasing in index (i, j) space; or [...]

I don't under that remark correctly. The i and j indices are not increasing along the axes. At least it is not meant to be this way. Therefore, it would not be correct to replace the longitude by i index and latitude by j index. I will create another figure and show what I mean.

[neumannd]

updates:

• updated colors (took colors from the beginning and end of the viridis color scale).
• updated indices of bnds (0-based instead of 1-based)
• legend added to 1D case
• added to more grid cell centers to 2D plot to indicate index order

[neumannd]

The source for the pictures is available via a GitHub repository:

https://github.com/neumannd/cell_bounds_figures_for_cf_conventions

[neumannd]

Are the figures OK how they are? Is it feasible to include them into CF-1.8?

push

[neumannd]

Unfortunately, we did not assign a moderator here so that this issue was forgotten. Are there any caveats including these figures into the CF conventions?

[JonathanGregory]

I see no reason in principle against diagrams in the document. On the contrary, they would be an enhancement. I think the question is technical: can they be included in AsciiDoc, and will the software which builds the HTML and the PDF from the AsciiDoc be able to process them properly?

[taylor13]

I'm either confused, or I agree with @davidhassell (#193 (comment)) that in the lower figure the abscissa and ordinate axes are mislabeled. Don't the dashed lines in the figure represent lines of latitude and longitude, not the horizontal and vertical axes? I would think that the horizontal axis would simply be i (or some function of i alone).

[neumannd]

@taylor13

Maybe this figures shows clearer what I mean?

lon1, lon2, ... actually mean lon_bnd1, lon_bnds2, ... .

For my example I assume that we have some grid, which is not orientated along the lon and lat axis. The grid is somehow curved. I tried to include an i-axis into the plot from above (i_{n-1}, i_{n}, i_{n+1} are the i-values):

I could draw the i and j indices into a plot with regular i and j axes. However, we could not reasonably include the lon-lat-information there. It would look like this:

[taylor13]

I was indeed confused, but I probably shouldn't have been. Thanks for going to the trouble of explaining. I think your original diagram should be clear enough to those thinking clearly.

[neumannd]

@taylor13 : Thanks for the feedback. If it is clearer on the first view, I could adapt the original figure to be like the first figure in my last comment -- printing (lon_bnds_n, lat_bnds_n)-tuples at the vertices and writing lon_bnds_n = lon_bnds(i, j, n), ... at the top.