Fix crs of buidrain data

[rubencalje]

The original data for pipe drainage (buisdrainage) contains a wrong coordinate system, causing the data to be shifted to the northeast by more than 100 m. This PR fixes this issue by adding the argument fix_crs to nlmod.read.nhi.discretize_buisdrainage. The value of this argument by default is False, so existing models that use this method will not change, but the user will see a warning.

This PR also solves some of the Use "@" matmul instead of "*" warnings, and a bathemetry-error.

The following code:

import nlmod
extent = [115_000, 122_000, 437_000, 442_000]
ds = nlmod.get_ds(extent, delr=25)
pathname = r"buidrain_test\cache"
ds_out = nlmod.read.nhi.discretize_buisdrainage(ds, pathname, fix_crs=False)

f, ax = nlmod.plot.get_map(extent)
da = ds_out['buisdrain_cond']
nlmod.plot.data_array(da.where(da> 0), ds=ds, ax=ax, alpha=0.5)
backgrounds.BrtAchtergrondKaart(ax=ax).plot(alpha=0.5)

produces the image below. The data is placed too far northeast, see the offset near the roads and the pixels in the river.

When we set fix_crs to True in the code above, we get the following image, in which you can see that the data now aligns with the background map.

[OnnoEbbens]

Nice work! I just have some minor remarks.

[OnnoEbbens]

maybe add a reference to this PR. The image in the PR speaks a thousand bookparts.

[rubencalje]

good idea, will add a link to the docstring

[OnnoEbbens]

But you did fix the matmul right? Is this <3.0.1 still necessary?

[rubencalje]

I fixed it in our own code. But it still needs to be fixed in methods within rasterio and rioxarray, that we use.

[OnnoEbbens]

Not sure what is happening with these changes but I assume it is some fix for RWS bathymetry download?

[rubencalje]

Yes, the FeatureServer is not avialable anymore, but the MapServer is. Then in Python 3.12, the stuff with the env-variable GDAL_DISABLE_READDIR_ON_OPEN was needed, otherwise it would look for non-existing files and fail.