Missing areas when using ISEA projection when +orient is "isea"

[mikima]

Example of problem

When using the ISEA projection, parts of the image are missing (see image with missing areas in the red circles:

the error doesn't happen when using +orient=pole:

Images were generated on MacOS 15 using gdalwarp. GDAL version 3.10.1, PROJ version 9.5.1

gdalwarp input.tif output.tif -s_srs "EPSG:4326" -t_srs "+proj=isea +R=6371007.18091875" -te -25000000 -25000000 25000000 25000000

The error doesn't happen when using the orient=pole

gdalwarp input.tif output.tif -s_srs "EPSG:4326" -t_srs "+proj=isea +R=6371007.18091875 +orient=pole" -te -25000000 -25000000 25000000 25000000

Problem description

When using the ISEA projection, parts of the image are missing

Expected Output

The projected image without missing parts

Environment Information

• PROJ version (proj) 9.5.1
• GDAL version 3.10.1
• Operation System Information MacOS 15.2 (24C101)

Installation method

• conda

[rouault]

CC @jerstlouis

[jerstlouis]

Thanks @mikima @rouault

I tried to reproduce this using https://maps.gnosis.earth/ogcapi/collections/blueMarble/map.tiff?width=2048 as input and GDAL 3.11.0dev-51337551c3

The Alaska vertex seems fine:

I also did similar tests last year when adding support for the inverse projection and don't remember seeing something like this.

Not sure why you get these results. Those corners are the wrap-around points / edge of the interruptions which can be quite tricky.

@mikima Would you able to share the input file, and potentially that orange vector outline to compare against (or is that just extending lines in an image editor)?

I hope to have time to spend improving the ISEA projection module in the coming month.

This does not seem directly related to the planned improvements, but I could take a look at this at the same time.

If a bug fix is urgently needed, please let me know and I'll see what I can do. Also happy to help if anyone else is investigating this sooner.

[mikima]

Hello

To reproduce, here is the source tif. Since I have georeferenced it (and maybe georeferenced it wrongly) i reproduced it using Natural earth 2 relief.

Again, there are missing areas with +orient=isea but this time in the bottom right part:

When using the map you provided there are no errors:

So now i'm bit puzzled.

Here you can find a PDF with the red lines, i added them with a vector editor software:

lines.pdf.zip

[jerstlouis]

Thank you @mikima .

I do notice one ERROR 1: Point outside of projection domain when running that gdalwarp command.

I get even worst results and a lot more transformation errors with a similar GEBCO tiff generated from our server:

http://maps.gnosis.earth/ogcapi/collections/gebco/coverage.tiff?width=5400

We will need to investigate this more.

[mikima]

@jerstlouis today i tried to scale down my geotiff to 1024x2048 and i got the result without missing areas:

I then tried to scale it up to 2048x4096 hoping it was somehow related to source file, but i got again the errors:

Here the files to reproduce:

2048-4096 geotiff.zip