gent = xr.open_dataarray("./data/gent/raster/2020-09-17_Sentinel_2_L1C_B0408.tiff", engine="rasterio", mask_and_scale=False)
gent = xr_array.assign_coords(band=("band", ["b4", "b8"]))

should be

gent = xr.open_dataarray("./data/gent/raster/2020-09-17_Sentinel_2_L1C_B0408.tiff", engine="rasterio", mask_and_scale=False)
gent = gent.assign_coords(band=("band", ["b4", "b8"]))

• Better explain DataArray vs DataSet:

  • Probably we can just start with explaining DataArray first (and do exercises on it), and only later Dataset
  • The raster satellite image might be less suited to explain the Dataset, we can better take an example where its clearer that there are multile variables (like temperature and pressure, ..)

• Only explain "there can be more coordinates as dimensions" later on (eg when it actually occurs in the data of the exercises)

• Be consistent with .values vs .data

• (for pandas -> also use the .plot.<type>() pattern instead of .plot(kind=<type>) for consistency)

• mistake in exercise: use salinity instead of pressure to calculate the buoyancy

• only explain conversion of dataarray <-> dataset when both concepts are already known (and then can show benefit of either way)

Both eg writing a Shapefile/GeoJSON/.. with geopandas as writing eg a netcdf file with xarray (with rasterio I think we showed how to do this)

• use drop() method instead of del for removing columns

In the first set of exercises in https://github.com/jorisvandenbossche/DS-python-geospatial/blob/master/notebooks/04-spatial-relationships-joins.ipynb we work with bike stations and the Eiffel tower yet a comment mention restaurants which is certainly a typo:

"# Make a plot of the close-by restaurants"

See https://gist.github.com/stijnvanhoey/5eb080605d569e4a06b1a8fb0a5ab4aa for the examples

See https://github.com/makepath/xarray-spatial

If these projects like rioxarray and this are mature enough, we can reformat the raster part of the course and build it entirely around xarray instead of the mix rasterio-xarray-numpy

• Ex. 3 of "Combining with municipilaties": "TODO HINTS" can be removed.
• "Combining with land use data": the code for downloading/cropping LC is not yet included ([INCLUDE LINK])
• "The goal is now to to query" -> "The goal is now to do query"

In notebook 02, the areas you compute with countries.geometry.area are not areas; they're computed as if geographic coordinates are Cartesian - their unit is squared degrees, which is meaningless (because dependent on latitude).

The instructions still indicate the (Raw) version

The link to the matplotlib gallery contains an extra ' " ', which disables the link.

We should remove these in the main/master branch as these are not in line with the new version of the course

Is currently confusing: we apply this on the bands dim and the variable is called x, which is another dimension.

@jorisvandenbossche , I think we should do some rework on the current state of the raster info in the repo. My proposal:

• Remove 11-numpy.ipynb. Could be an option to integrate the convolution example at the end in an advanced exercise or use-case as it is a useful GIS element
• Remove 12-rasterio.ipynb
• Combine 10-introduction-raster.ipynb, 13-xarray.ipynb & 14-xarray-intro.ipynb to have a single introduction on rasters, but using xarray and rio-xarray directly. Focus on xarray based on analogy with geopandas instead of analogy/extension of numpy ('adds context to NumPy'). Start with single data source, i.e. DataArray; add write to file example/exercise

For clarity/consistency during the course: we handle bands/channels of a single data source as a DataArray dimension (RGB,..), we handle different kinds of data (temperature, salinity,...) as DataSets. We might make a remark in the end that switching is fine.

• 15-xarray-datasets.ipynb -> rework towards DataSets containing different variables instead of the b4/b8 example; add write to file example/exercise

• Combine 16-raster-processing.ipynb and 20-raster-vector-tools.ipynb into a single notebook on raster/vector tooling. Proposal of topics to cover: clip region (with/without buffer), conversion raster/vector, proximity (xarray-spatial ) and rasterstats; move "Cloud: only download what you need" to the 'big-data' notebook. I would add an addendum on 'calling external tools' like https://www.whiteboxgeo.com/, http://www.saga-gis.org/saga_tool_doc/2.2.7/a2z.html,...

• 21-xarray-dask-big-data.ipynb: extend with the COG-info, mention also https://geemap.org/ (requires account of google earth engine)?

• What to use for the zonal statistics -> what about https://github.com/corteva/geocube vs https://xarray-spatial.org/user_guide/zonal.html vs rasterstats package?

Some sections are now a bit redundant to show clip and ``rasterize; we should improve the structure to make clear what the repitition is and what is new.

• mask_and_scale has already been tackled before -> shorten in this notebook (just a recap)
• it is a lot of clipping no (both on initial section as on dem example); might be better to stick to 1 of those or just start from the clipped dem directly to have the focus on the rasterize only.

When someone uses latlon, not projected, use Plattecarree. Other situations the link in between epsg and projecten in cartopy need to be clarified. Specific, add the function ccrs.epsg(....).

e.g. case of Lambert72 (EPSG:31370), the ccrs.epsg(31370) works and zooms the the proper extent as well. Cfr also ccrs.epsg(3035) as well.

in cartopy vis notebook , download not functional for different participants.

There is a typo in the table "An overview of the possible comparison operations"
It should be:

An overview of the possible comparison operations:

and to combine multiple conditions:

What is the average popolation number of the districts? -> What is the average population number of the districts?

https://aballatore.space/geospatial-datasets/?s=03 or https://github.com/andrea-ballatore/open-geo-data-education

in sjoin the op is going to be deprecated and predicate will be used

https://scitools.org.uk/cartopy/docs/latest/reference/projections.html#cartopy-projections instead of https://scitools.org.uk/cartopy/docs/latest/crs/projections.html

• Do we need squeeze() ?
  • Or can we use indexing as well?
  • If we use it for numpy, should maybe use it for geopandas as well? (instead of .item())
• Leave where for the "extra" section at the bottom
• Add some more visual illustrations (see new numpy docs)

Other metadata, such as the spatial information) provided by the tiff are stored in the Attributes is not fully correct as we now read the metadata in the crs using DataArray

• north arrow, scalebar
• inset (to locate study area) (see https://matplotlib.org/3.3.1/gallery/subplots_axes_and_figures/zoom_inset_axes.html#sphx-glr-gallery-subplots-axes-and-figures-zoom-inset-axes-py, https://basemaptutorial.readthedocs.io/en/latest/locator.html#inset-locator)
• customize title, legend, ...

In exercise 7 of this notebook (part I), the question is to select all districts in the 1st arrondissement, while the solution is showing how to select all districts in the 3rd arrondissement.

As the integration works nicely, we do not need to provide the full section with the transpose and extent example. Move this to the end as an optional 'manual' section as what xarray is doing under the hood.

"the code to do this cropping can be see at [INCLUDE LINK]"

should be - instad of _

Small typo in the section 'Importing raster data', line 1: rastyer where it should be raster.