Explore what we would need to modify/add to support voronoi grids

[dbrakenhoff]

I got some questions in the last weeks about supporting Voronoi type grids in nlmod. Theoretically the disv structure is supported in our model datasets, but some functionality undoubtedly assumes quad-tree style grids. So in this issue I'd like to explore what would be needed to support this more comprehensively.

List of things:

• Resampling rasters currently uses raster methods, this will presumably fail for voronoi grids. What would be needed for more advanced resampling techniques?
  • e.g. structured_da_to_ds()
• Plotting using pcolormesh (though nlmod.plot.flopy methods will work)
• ...

[rubencalje]

Good idea to make a list of possible problems. I was thinking about structured_da_to_ds as well.

I think we need to store the type of grid-refinement somewhere in the attributes, so we can inform the user when methods will not perform as expected. Maybe we can use the 'gridtype' attribute, as we do not use it anymore to determine the type of grid (we now look at the dimensions right?) So ds['attrs']['gridtype'] = 'quadtree'. Or should we add a new property? I am in favor of just using the gridtype-property.

[dbrakenhoff]

What would the different gridtypes be?

• voronoi
• triangle
• quadtree
• regular? --> all cells equal shape
• structured? (nrow by ncol), but not all equal shape

And how do we distinguish between them... ? I think creating a model dataset from a voronoi disv might already work?