sandbox/popinet/gulf-stream.md

The Gulf Stream

The setup is as close as possible to that used by Hurlburt & Hogan, 2000 but uses the layered solver described in Popinet, 2020.

See Hurlburt & Hogan, 2000 for details but the main characteristics of the setup are:

• 5 isopycnal layers (see Table 2 of H&H, 2000 for the reference densities, thicknesses etc.)
• Wind stress in the top layer given by the monthly climatology of Hellerman & Rosenstein, 1983
• “Compressed bathymetry” as in H&H, 2000
• Quadratic bottom friction (Cb = 2 x 10^-3), Laplacian horizontal viscosity (10 m²/s)
• Atlantic Meridional Overturning Circulation (AMOC) driven by fluxes at the northern and southern boundaries (see Table 2 of H&H, 2000 for the values of fluxes)

Some of the differences below can be due to the longer averages taken in Basilisk (> 10 years versus a few years for H&H, 2000).

Movie and snapshots

Snapshot of SSH from Hurlburt & Hogan, 2000, fig. 4, 1/32 degree	Snapshot of SSH, Basilisk, 1/24 degree

Mean Sea Surface Height (SSH)

Mean SSH from Hurlburt & Hogan, 2000, fig. 3, 1/32 degree	Mean SSH, Basilisk, 1/24 degree

Mean SSH from Hurlburt & Hogan, 2008, fig. 2, 1/32 degree	Mean SSH, Basilisk, 1/24 degree

SSH standard deviation

SSH standard deviationfrom Hurlburt & Hogan, 2000, fig. 9, 1/32 degree	SSH standard deviation, Basilisk, 1/24 degree

SSH standard deviation from Chassignet & Xu, 2017, fig. 7

SSH standard deviation, Basilisk, 1/24 degree

Surface kinetic energies

Mean surface kinetic energy from Chassignet & Xu, 2017, fig. 4

Mean surface kinetic energy, Basilisk, 1/24 degree

Surface eddy kinetic energy from Chassignet & Xu, 2017, fig. 14

Surface eddy kinetic energy, Basilisk, 1/24 degree

EKE difference from Chassignet & Xu, 2017, fig. 10, 1/50 degree	EKE difference, Basilisk, 1/24 degree

The high values in C&X, 2017 may be due to non-converged statistics (i.e. shorter averages in Basilisk show similar features/artefacts).

Field transects

The “TOPEX” transect from Figure 5 of C&X, 2017

The “Oleander” transect from Figure 6 of C&X, 2017

The “Oleander” transect, Basilisk, 1/24 degree

Cross-section at 55W from Figure 15 of C&X, 2017, 32 layers

Zonal velocity at 55W, Basilisk, 1/24 degree, 5 layers	EKE at 55W, Basilisk, 1/24 degree, 5 layers

Abyssal currents

Mean abyssal kinetic energy from H&H, 2000, fig. 11a, 1/32 degree	Mean abyssal kinetic energy, Basilisk, 1/24 degree

Mean abyssal layer pressure deviation from H&H, 2000, fig. 10, 1/32 degree	Mean abyssal layer pressure deviation, Basilisk, 1/24 degree

Note that there is a factor \rho_0 = 1000 missing in H&H, 2000.

Abyssal layer EKE from H&H, 2000, fig. 12, 1/32 degree	Abyssal layer EKE, Basilisk, 1/24 degree

Abyssal layer mean current from Figure 3 of H&H, 2008, 1/32 degree

Abyssal layer mean current, Basilisk, 1/24 degree

Abyssal layer mean current from H&H, 2008, fig. 4.a, 1/32 degree	Abyssal layer mean current, Basilisk, 1/24 degree

Abyssal layer interface depth from H&H, 2008, fig. 5.d, 1/32 degree	Abyssal layer interface depth, Basilisk, 1/24 degree

EKE at 700 m from Chassignet & Xu, 2017, fig. 17

EKE in layer 2 (750–500 m), Basilisk, 1/24 degree

EKE at 1000 m from Chassignet & Xu, 2017, fig. 18

EKE in layer 1 (1000–750 m), Basilisk, 1/24 degree

Top layer

Mean top layer thickness (m)

Computational cost etc.

The 1/24 degree domain spans 98W to 14W and 9N to 51N using 2048 \times 1024 grid points. Only five (isopycnal) layers are used in the vertical as in H&H, 2000. The timestep is 150 seconds and the simulation ran for approx 20 years with averages taken after a spinup of 5 years.

The simulation ran at approximately 23 simulated years per day on 2048 cores of the Irene supercomputer at TGCC (i.e. a computational speed close to 10⁹ cells \times timestep / second).

References