perform the unstacking of the paleo PISM ensemble while keeping the attributes of the parameters

unstacking_save_attributes

def unstack_paleo_pism_xarrays(PPds):
    # save the values of id in a data variable before "unstacking", because unstacking will delete id. 
    PPds = PPds.assign(index=PPds["id"])
    # also save the attributes of the ensemble parameters
    A_par_esia = snapshots.par_esia.attrs
    A_par_ppq = snapshots.par_ppq.attrs
    A_par_prec = snapshots.par_prec.attrs 
    A_par_visc = snapshots.par_visc.attrs    
    # make MultiIndex object (this assumes that these do not change over time and that time is the first dimension)
    arrays = [PPds.par_esia[0].values, PPds.par_ppq[0].values, PPds.par_prec[0].values, PPds.par_visc[0].values]
    MI = pd.MultiIndex.from_arrays(arrays, names=('par_esia', 'par_ppq','par_prec','par_visc'))
    # replace the id coordinate with this MultiIndex Object
    PPds.coords['id'] = MI
    # unstack
    out = PPds.unstack(dim='id')
    out.par_esia.attrs = A_par_esia
    out.par_ppq.attrs = A_par_ppq
    out.par_prec.attrs = A_par_prec
    out.par_visc.attrs = A_par_visc
    return out