1. use the DIVA dycore for a temp. spin-up, starting with an initially linear temperature profile,

2. hold the geometry fixed during that spin-up

3. subcycle the temperature calculation relative to the velocity calculation

To avoid having the vel. field immediately get weird because of the linear temp. profile, I suggest: first spinning up for 10 ka (?) without coupling between the temperature and the rate factor. This is a tiny bit tricky; it can be done by setting "flow_factor" to 0 (const. rate factor), but won't work right currently because "flwa" is not in the input .nc file (it would then default to a value for isothermal ice, which we don't want). I've set the .config up now so that you can first do a single diagnostic solve (no evolution), and spit-out the necessary "flwa" field as part of that output. Then, you can use that file as a restart for this same config file, but switching "flow_law" from 2 to 0. Now it will use a const. rate factor but one that is realistic, based on the old temps, and thus the vel field won't change as temps evolve. Once the temp evol. has used the current vel field for a few thousand years, we could switch back to "flow_factor" = 2, to allow the vel. and temp. field to evolve together.

1. Single diagnostic solve using GIS.1km.InitCond.4Glissade.config and GIS.1km.InitCond.4Glissade.nc.

2. Switch flow_law from 2 to 0 in GIS.1km.InitCond.4Glissade.config and use (1.) output *.nc file for "a few thousand years" [lets go with 3].

3. Switch flow_law from 0 back to 2 in GIS.1km.InitCond.4Glissade.config and use (2.) output *.nc file until ten thousand years have elapsed.