(back to up to ScienceWorkRobin)
Timestepping issues
Ron Kahana in Bristol was having trouble with crashes in his paleoruns and noticed that increasing the number of passes through the dynamic routines in each timestep made the model a lot more stable. I used this for a couple of problematic test runs, in which I then noticed large surface temperature differences I wasn’t expecting.
Turns out that FAMOUS is rather sensitive to the timestep length chosen. You can get non-tiny differences in surface temperature by dividing the timestep by 2 or 3 - presumably because various parameterisations have been tuned for 24 timesteps a day. Of particular note, going from 1 to 2 sweeps through the dynamic routines seems to resolve mid-latitude processes much better without perturbing whatever it is that throws things off by halving the whole timestep. These effects seem present in HadCM3 as well, but they result in much smaller perturbations to the climate.
At any rate, 2 sweeps improves mid-latitude eddy activity, improves the representation of the Atlantic stormtrack no end and atmospheric heat transport is increased. The fly in the ointment is that if used with the new, lower sea-ice albedoes of XCPSA and onward (as Ron did - sorry Ron) you end up with an annual average 4 degree warm bias at the poles. Used with the original, higher sea-ice albedoes and you have a climate very similar to XCPSA/XDBUA but with a better stormtrack and without the overly-large seasonal swing in sea-ice that came with the low albedoes of XCPSA.
Not available in any current FAMOUS release. I’m currently planning on tuning the sea-ice albedoes with this to see if I can get the coverage down a bit without causing the too-large die-back in summer now that I don’t need the albedo to fix the whole cold-bias issue.
UPDATE: with longer runs, it becomes apparent that however you use 2 sweeps it somehow makes the Pacific MOC very sensitive to changes in surface forcing, with knock-on compensating changes in AMOC strength. This makes it very unsuitable for general use in FAMOUS. See Current Work Filtering for more.