CLI improvement

[marinebcht]

Resolves #841 #922

ACCURACY

Default method in the doc vs method in the accuracy_config.yaml

Changed NuthKaab to LZD

Preprocessed datasat presentation

Save the projected/cropped DEM after data preparation if activated :

• preprocessed_reference_elev_map.png if sampling_grid = to_be_aligned_elev
• preprocessed_to_be_aligned_elev_map.png if sampling_grid = reference_elev

Show them in the report, in the section after inputs metadata information.

Turned statistics table

Before in HTML and PDF:

After in HTML and PDF:

Elevation difference plot with the same scale

Mask management

Masked data were not plot and correctely handle

• in the stats now:

TOPO

Remove empty terrain attributes case : before/After

BOTH

Dynamic unit

(m) is now taken from the CRS info (here: metre)

Same font size for each plot

Stats :

Format data in the stats :

• if "count": int
• if "percentage": '{:.2f}'.format(val) + "%"
• if abs(val) > 10E4 or abs(val) < 10E-4: np.format_float_scientific(val, precision=3)
• else '{:.3f}'.format(val)

• Remove std/standarddeviation duplicate in each template and alias std gives "Standart deviation" too

See before for accuracy and topo :

Reduce default stats

[marinebcht]

@rhugonnet I see two additonnials places we could use the method you will choose for the Limit the number of decimal places when displaying floats with NumPy issue in your PR #759 :)

• when we print the statistics in the accuracy especially and topo workflow

ex: would be more readable

• when we print the stats in the difference elevation plot here and here

ex: here median after = 0.00064086914, maybe we can plot 0.00064 (display.precision 3) or with scientific format... or let it like this

[marinebcht]

@rhugonnet @belletva @adebardo ready for review :)

[marinebcht]

News !

• added the missing section

• change the generate_plot and others plot to have always the same font size (6) and choose to have all dem = 50% page for the visibility
• reference_elev + to_be_aligned_elev (identical) documentation in the same tab

[rhugonnet]

Great @marinebcht! 😄

For the mask, two remarks:
1/ As it is ultimately a boolean raster (but can be provided as a vector that gets rasterize inside or outside polygons), plotting the vector is ambiguous (are we masking inside or outside polygons?).
To remedy this, we could plot either the actual raster mask (0 where terrain is masked, 1 where it is not), or the DEM with gaps (NaNs) where the mask is equal to 1. Or we could keep the vector (when it is one), and print one the figure (masked = "inside" or "outside" geometries).
2/ A mask always refers only to the "reference" raster for coregistration (the other being misaligned, the masked terrain will not cover the same areas). This is something we should clarify also in the documentation/Coreg function descriptions, as I think it is implicit right now...

[rhugonnet]

@marinebcht For displaying the masks: Perfect! It's a shame that the "terrain" colormap has white in it, because we can confuse nodata or the edge of the colormap. We could maybe cut the colorbar, or choose another one...
What do you think?

[marinebcht]

@rhugonnet @adehecq
Here another prop of cmap : gist_earth

version with mask = black

What do you think ?

[rhugonnet]

As you prefer, both work well I think! 🙂
(gist_earth + white, or terrain + black)

[belletva]

As you prefer, both work well I think! 🙂 (gist_earth + white, or terrain + black)

terrain + black :)

[marinebcht]

@rhugonnet @adehecq @belletva ready for the reviews :)

[rhugonnet]

Perfect, thanks! 😄

I don't have much to re-review.
My only two comments:

• It would be nice to get a RTD link (following the Wiki) to see the new doc rendered, easier to read + get a sense of the changes + potentially help spot typos/issues.
• Should we define default statistics? The full list is a bit long, we already discussed this... Maybe simply Min/Max/Mean/Median/STD/NMAD and Valid/Total/Percent counts is already plenty.

[adehecq]

As you prefer, both work well I think! 🙂 (gist_earth + white, or terrain + black)

I'm happy with both, although I find terrain a little bit flashy 😉
Otherwise, happy with the current reports (once the rounding of statistics is fixed).

[belletva]

Good for me!