Hi;
I'm running a configuration of ROMS 2.1 for the Gulf of Alaska that uses both biology and tides - a problem originally developed by Kate. My issue is that adding biology doubles the run-time of the model!
The biology subroutine was developed here and has 11 species in it. It isn't efficient code, but the massive time increase is only partly due to the biology; 80% of the increase is due to dealing with 11 extra tracers in pre_step3d.F and step3d_t.F.
I want to improve performance, and would like to increase the timesteps, but am limited because the model also includes tides. The increased velocity requires a pretty short timestep. I can't mix too much, because I am also trying to retain eddies, which presently don't live long enough. So I'm wondering if I'm using all the newest ROMS options to their fullest potential. Here's a short list of the relevant options I have enabled:
#define UV_ADV /* turn ON or OFF advection terms */
#define UV_VIS2 /* turn ON or OFF Laplacian horizontal mixing */
#define TS_U3HADVECTION /* define if 3rd-order upstream horiz. adv. */
#define TS_SVADVECTION /* define if splines vertical advection */
#define TS_DIF2 /* turn ON or OFF Laplacian horizontal mixing */
#define DJ_GRADPS /* Splines density Jacobian (Shchepetkin, 2000) */
#define SPLINES /* turn ON or OFF parabolic splines reconstruction */
tnu2 = 1
visc2 = something between 50 and 200. Still playing with that.
tracers OBC = radiation with nudging
Does anyone have suggestions for new advection/diffusion/pressure gradient options that emphasize effeciency for the tracer arrays while dealing well with high velocities?
Thanks,
Liz Dobbins