GOSI10: investigating time-step dependency of RK3

[oceanaguiar]

ORCA12 (+ trends diagnostics + fixes from the GOSI10.beta.4 branch)

ID	Time-stepping	Time-step	Advection scheme	Vertical mixing scheme	nn_etau	Shuman averaging	Finished?	Validation note
u-dt293	RK3	7.5 min	FCT 4th order, Vector 2nd order	TKE	1	no	✅
u-dt296	RK3	12 min	FCT 4th order, Vector 2nd order	TKE	1	no	✅	vs u-dt293
u-dt297	RK3	15 min	FCT 4th order, Vector 2nd order	TKE	1	no	✅	vs u-dt293

ORCA025 (reference)

ID	Time-stepping	Time-step	Advection scheme	Vertical mixing scheme	nn_etau	Shuman averaging
u-ds639	RK3	30 min	FCT 4th order, Vector 2nd order	TKE	0	no
u-ds643	RK3	60 min	FCT 4th order, Vector 2nd order	TKE	0	no
u-ds069	RK3	60 min	FCT 4th order, Vector 2nd order	TKE	1	no
u-ds068	RK3	30 min	FCT 4th order, Vector 2nd order	TKE	1	no

ORCA025 (+ trends diagnostics + fixes from the GOSI10.beta.4 branch)

ID	Time-stepping	Time-step	Advection scheme	Lateral viscosity	Vertical mixing scheme	nn_etau	Shuman averaging	Finished?	Validation note
u-ds864	RK3	15 min	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE	1	no	✅
u-ds825	RK3	30 min	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE	1	no	✅
u-ds865	RK3	45 min	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE	1	no	✅	vs u-ds825
u-ds826	RK3	60 min	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE	1	no	✅	vs u-ds825 vs u-ds864
u-ds866	RK3	30 min	FCT 4th order, Vector 2nd order	f(e1, e2)	Richardson number		no	✅
u-ds867	RK3	60 min	FCT 4th order, Vector 2nd order	f(e1, e2)	Richardson number		no	✅	vs u-ds8661
u-ds894	RK3	30 min	MUSCL2, Vector 2nd order	f(e1, e2)	TKE	1	no	✅
u-ds895	RK3	60 min	MUSCL2, Vector 2nd order	f(e1, e2)	TKE	1	no	✅	vs u-ds894
u-ds873	RK3	30 min	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE	1	yes	✅
u-ds874	RK3	45 min3	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE	1	yes	✅	vs u-ds873
u-dt139	RK3	30 min	FCT 4th order + JC fix, Vector 2nd order	f(e1, e2)	TKE	1	no	✅
u-dt140	RK3	45 min3	FCT 4th order + JC fix, Vector 2nd order	f(e1, e2)	TKE	1	no	✅	vs u-dt139
u-dt254	RK3	30 min	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE, no EVD	1	no	✅
u-dt255	RK3	45 min3	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE, no EVD	1	no	✅	vs u-dt254
u-dt256	RK3	30 min	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE, Mass Flux Convection	1	no	✅
u-dt257	RK3	45 min3	FCT 4th order, Vector 2nd order	f(e1, e2)	TKE, Mass Flux Convection	1	no	✅	vs u-dt256
u-dt886	RK3	30 min	FCT 2nd order, Vector 2nd order	f(e1, e2)	TKE	1	no	✅
u-dt887	RK3	60 min	FCT 2nd order, Vector 2nd order	f(e1, e2)	TKE	1	no	✅	vs u-dt886
u-du674	RK3	30 min	FCT 2nd order + JC fix, Vector 2nd order	f(e1, e2)	TKE	1	no	✅
u-du675	RK3	60 min	FCT 2nd order + JC fix, Vector 2nd order	f(e1, e2)	TKE	1	no	✅	vs u-du674
u-dt888	RK3	30 min	FCT 4th order, UP3 3rd order	f(e1, e2)	TKE	1	no	✅	vs u-ds825
u-dt889	RK3	60 min	FCT 4th order, UP3 3rd order	f(e1, e2)	TKE	1	no	✅	vs u-dt888 vs u-ds826
u-du726	RK3	30 min	FCT 4th order, Vector 2nd order	f(u, e12)	TKE	1	no	✅
u-du727	RK3	60 min	FCT 4th order, Vector 2nd order	f(u, e12)	TKE	1	no	❌3
u-du729	RK3	30 min	FCT 4th order, Vector 2nd order	Smagorinsky	TKE	1	no	✅
u-du730	RK3	60 min	FCT 4th order, Vector 2nd order	Smagorinsky	TKE	1	no	✅	vs u-du729

ORCA1 (reference)

ID	Time-stepping	Time-step	Advection scheme	Vertical mixing scheme	nn_etau	Shuman averaging
u-ds644	MLF	60 min	FCT 4th order, Vector 2nd order	TKE	0	no
u-ds645	MLF	30 min	FCT 4th order, Vector 2nd order	TKE	0	no
u-ds700	RK3	30 min	FCT 4th order, Vector 2nd order	TKE	0	no
u-ds699	RK3	60 min	FCT 4th order, Vector 2nd order	TKE	0	no

ORCA1 (+ trends diagnostics + fixes from the GOSI10.beta.4 branch)

ID	Time-stepping	Time-step	Advection scheme	Vertical mixing scheme	nn_etau	Shuman averaging	Finished?	Validation note
u-ds828	RK3	30 min	FCT 4th order, Vector 2nd order	TKE	1	no	✅
u-ds827	RK3	60 min	FCT 4th order, Vector 2nd order	TKE	1	no	✅	vs u-ds828
u-ds870	RK3	120 min	FCT 4th order, Vector 2nd order	TKE	1	no	✅	vs u-ds828
u-ds872	RK3	30 min	FCT 4th order, Vector 2nd order	Richardson number		no	✅
u-ds871	RK3	60 min	FCT 4th order, Vector 2nd order	Richardson number		no	✅	vs u-ds872

Footnotes

1. Meaning period is years 6-10 rather than years 6-20 as for the other validation notes, as u-ds867 crashed with a salinity spike in 1968-08 ↩

2. MUSCL was chosen because the QUICKEST scheme crashed with a salinity spike when using a 30m timestep (so 60m is probably unstable too) and UBS can't be used with implicit vertical advection (ctl_stop error saying it isn't coded). Note that MUSCL doesn't output tracer advection trends (the trd_tra calls are commented out in tra_adv_cen and tra_adv_mus) ↩ ↩²

3. Using a 60 min timestep crashed ↩ ↩² ↩³ ↩⁴ ↩⁵

[dcalve]

Update given to JMMP GO dev meeting and NST meeting on 20th November 2025