rasterio 1.4.0/1.4.1 flips y-axis in compute_availabilitymatrix

[fneum]

This is a bit of a random find, but caused some significant faults in renewable potential calculations...

Versions 1.4.0 and 1.4.1 or rasterio cause a flipped y-axis in cutout.compute_availabilitymatrix(), which neither earlier nor later versions of rasterio do. With no changes but changed rasterio version:

With rasterio==1.4.{0,1}

With rasterio<1.4,>=1.4.2

I think excluding these versions makes more sense than implementing version-specific behavior in atlite.

[lkstrp]

Quite a random find, great!

I think this is also what's causing the pypsa-eur problems, which is why we've pinned it to rasterio==1.4.1 right now. Do you think this could be solved with a fix in atlite or do we need to upload updated cutouts somewhere as well?
At least hopefully it does.

EDIT: Hmm, it can't be the former, I'll check the cached cutouts.

[fneum]

I don't think the cutouts are affected by this.

[lkstrp]

rasterio==1.4.1 pin in pypsa-eur is no longer needed. So either the cutouts have been affected or some other rasterio dependency issue. Anyway