This is an attempt at a very practical explaination of how to set up a PGW simulation using global climate model data in the CMOR-Format as input (for example CMIP5 or CMIP6 data).
What data to get?
You will need data for the following variables: hur, ta, ua, va, zg, pa, hurs, tas, ts, tos
What time resolution should one choose?
Monthly mean data is the easiest. This is called e.g. Amon in CMOR. tos is part of Omon in CMOR.
How to preprocess the data?
For all variables we need to know how they will change under climate change. This needs to be expressed as a mean annual cycle of changes. In practice we can get a time slice of the "historical" period (HIST) and from a future period under a certain emission scenario (SCEN) such as "rcp85". A typical example: For the historical period, get data from 1971-2000. Then construct the mean annual cycle for 1971-2000, for example using the cdo-command "ymonmean". Repeat for 2070-2099 and the rcp85 data. Lastly, subtract the historical monthly-mean annual cycle from the future monthly-mean annual cycle. Save the result from the subtraction, or the difference between the two periods, as a single netcdf-file per variable (e.g. delta_ta.nc, delta_hurs.nc, ....). These netcdf files are needed as input for setp_02_preproc_deltas.py, and the naming convention can be specified in settings.py (look for the dictionary "file_name_bases").