for multiple time frequency (day, month, ...)
for multiple domains (EUR-11, EUR-44, ...)
for multiple ensembles (r1i1p1, r2i1p1, ...)
for multiple downscale realizations (v1, v2, ..)
for complete periods of simulations: historical usually goes approx.
from 1950/70 - 2005, and rcp* from 2006 - 2100; evaluation is not checked,
because it has very heterogeneous periods; cordex-adjust has historical and
rcp* combined
that each rcp* has a corresponding historical run (optional, off by
default; otherwise problematic with merged hist and rcp runs, as in cordex-adjust)
that all variables (tas, pr, ...) are available for all models (optional, off by default)