Handle pressure level data in data preprocessing

[mlangguth89]

The study of [Höhlein et al., 2020] suggests the 500 hPa geopotential as predictor variable (to identify large-scale weather patterns). While pressure-level based variables from ERA5 can be retrieved [here](), the pre-processing does not support handling of these variables yet. This will be fixed in the related issue-branch.

Furthermore, fixes/adaptions in handling the target COSMO REA6 data will be documented in this issue (see below).

[mlangguth89]

The windspeed data downloaded from here cannot be remapped with CDO since the coordinate data cannot be correctly infered.
To enable this, several changes to the netCDF-file have to be undertaken as listed below:

• remove the variables RLAT and RLON: ncks -C -O -x -v <file> <file>
• rename the standard and long-name attributes of rlat and rlon (the actual coordinate data): ncatted -O -a long_name,rlon,o,c,"longitude in rotated pole grid" <file>, ncatted -O -a long_name,rlat,o,c,"latitude in rotated pole grid" <file>, ncatted -O -a standard_name,rlon,o,c,"grid_longitude" <file> and ncatted -O -a standard_name,rlat,o,c,"grid_latitude" <file>
• Add the rotated pole information from some auxiliary file: ncks -A -v rotated_pole ../T_2M/T_2M.2D.201812.nc <file>
• Add grid_mapping-attribute to data variable: ncatted -O -a grid_mapping,wind_speed,o,c,rotated_pole <file>

This ensures that conservative reampping is handling the underlying curvilinear grid in a correct way. The downloaded data has been processed with a Bash-script accordingly.

[mlangguth89]

For consistency, the variable is also renamed into WS_100M (instead of wind_speed).

[mlangguth89]

Final 100m wind speed dataset is now available under /p/scratch/deepacf/maelstrom/maelstrom_data/ap5/downscaling_benchmark_dataset/benchmark_wind/dataset. Support of pressure-based vaiables is implemented and thus, the issue can be closed.