Raster processing (notebook 13) - open methods and nodata handling exercise

[jorisvandenbossche]

In this notebook we use rioxarray.open_rasterio in the explanation, but then in the exercises, we use the standard xr.open_dataarray, and their behaviour regarding nodata is different (xr.open_dataarray uses mask_and scale=True by default).

The second exercise is about this topic, but should in the instructions also mention mask_and_scale.

And we might want to consolidate the way the dataset is opened (eg always xr.open_dataarray)

[jorisvandenbossche]

More in general, we might also want to consolidate our explanation and exercises on the "mask_and_scale" / nodata topic. Currently:

• Notebook 11 (xarray intro):
  • First exercise asks to use this keyword (without prior introduction)
  • After this exercise there is a "remember" box, but so this is the first time it is explained
  • The 4th exercise explains how the biggest integer value is "nodata", and converts the data manually to float and replaces that value with NaN -> essentially manually doing what mask_and_scale does
    • We should also point to the _FillValue attribute that holds this information (or check .rio.nodata)
  • The second exercise after reductions continues with this dataset and rescales the uint16 numbers to [0-1] floats to plot as true color, but actually forgets to convert the largest values to NaN to be fully correct
• Notebook 13 (raster processing):
  • In the first exercises reading the DEM, we repeat this with reading it with mask_and_scale=False to inspect and handle nodata value of -9999 -> this is probably a good repetition

[stijnvanhoey]

see #65

In this notebook we use rioxarray.open_rasterio in the explanation, but then in the exercises, we use the standard xr.open_dataarray, and their behaviour regarding nodata is different (xr.open_dataarray uses mask_and scale=True by default).
And we might want to consolidate the way the dataset is opened (eg always xr.open_dataarray)

Tackled and using by default (except the example in the end which uses an argument that can not be passed to open_dataarray)

After this exercise there is a "remember" box, but so this is the first time it is explained

First exercise converted to additional 'theoretical part' and a small recap exercise with focus on the .rio.nodata

The 4th exercise ... We should also point to the _FillValue attribute

Unfortunately, the data encoding is actually not correct in the geotiff file, hence the fillvalue is also not available after reading in the data. I made a note/warning about this.

The second exercise after reductions continues

The conversion into NaNs is now explicitly in the exercise assignment

In the first exercises reading the DEM, we repeat this with reading it with mask_and_scale=False to inspect and handle nodata value of -9999 -> this is probably a good repetition

Left the exercise in and adjusted the text a bit