staticmaps.nc to gtiff and back

[jurianbeunk]

Converting staticmaps.nc to gtiff and back to staticmaps.nc as described here:
https://deltares.github.io/hydromt_wflow/latest/examples/examples/convert_staticmaps_to_gtiff.html

Results in bad maps (striped patterns in the maps) even without editing any of the gtiff rasters.

ds_updated = hydromt.open_mfraster(join(root, 'staticmaps', '*.tif')) does create good maps (when exporting ds_updated to .nc file), but writing it back to a model (i.e. mod.set_staticmaps(ds_updated) & mod.write() seems to create the issue somehow.

[jurianbeunk]

Other, but potentially related issue:

Running hydromt update to setup_gauges, based on a .gpkg file providing the gauge locations, results in a staticmap (i.e. wflow_gauges) with missing gauges, even though hydromt finds all gauge locations in domain.

[hboisgon]

From Huite:
I notice that that the values are located in the right place; all the spatial patterns appear to be the same. So I'm quite confident that you're running into one of my least favorite features of xarray: the automatic alignment. Somewhere in the workflow, two different datasets are being combined (e.g. via some arithmetic operation). Xarray will look for the intersection of the coordinates here (i.e. the coordinates that are available in both datasets).

The problem here is being caused by having a dataset with e.g. 500 m resolution versus one with 100 m resolution. In this case, xarray will select the coordinates that are available in both and compute a result with just that.

So my guess is, if you know which dataset has a different cellsize and is being used in this workflow, you can fix the issue by changing that dataset to the appropriate cellsize (at the start)

[hboisgon]

The problem indeed comes from calling the set_staticmaps line 'too soon' (without any further processing of the mapstack).
Additionally Wflow staticmaps has two parameters with a third dimension (layer for c and time for LAI) so anyway more processing is needed than just read and write.

In #42 I renamed the notebook to convert_staticmaps_to_mapstack and recommend to use the built-in read_staticmaps_pcr and write_staticmaps_pcr in order to be able to use hydromt to manually update wflow staticmaps.

@DirkEilander : shall we add some extra lines in set_staticmaps in order to check that the new data has the same grid/dimensions and if not throw an error?

[DirkEilander]

@hboisgon good suggestion! I think we can simply added join='exact' to the xarray.merge method. see https://xarray.pydata.org/en/stable/generated/xarray.merge.html

[hboisgon]

Closed as fixed in setup_grid in hydromt core