Horizontal momentum flux in a global ocean climate model is formulated as a
n anisotropic viscosity with two spatially varying coefficients. This frict
ion can be made purely dissipative, does not produce unphysical torques, an
d satisfies the symmetry conditions required of the Reynolds stress tensor.
The two primary design criteria are to have viscosity at values appropriat
e for the parameterization of missing mesoscale eddies wherever possible an
d to use other values only where required by the numerics. These other visc
osities control numerical noise from advection and generate western boundar
y currents that are wide enough to be resolved by the coarse grid of the mo
del. Noise on the model gridscale is tolerated provided its amplitude is le
ss than about 0.05 cm s(-1). Parameter tuning is minimized by applying phys
ical and numerical principles. The potential value of this line of model de
velopment is demonstrated by comparison with equatorial ocean observations.
In particular, the goal of producing model equatorial ocean currents compar
able to observations was achieved in the Pacific Ocean. The Equatorial Unde
rcurrent reaches a maximum magnitude of nearly 100 cm s(-1) in the annual m
ean. Also, the spatial distribution of near-surface currents compares favor
ably with observations from the Global Drifter Program. The exceptions are
off the equator; in the model the North Equatorial Countercurrent is improv
ed, but still too weak, and the northward flow along the coast of South Ame
rica may be too shallow. Equatorial Pacific upwelling has a realistic patte
rn and its magnitude is of the same order as diagnostic model estimates. Th
e necessary ingredients to achieve these results are wind forcing based on
satellite scatterometry, a background vertical viscosity no greater than ab
out 1 cm(2) s(-1), and a mesoscale eddy viscosity of order 1000 m(2) s(-1)
acting on meridional shear of zonal momentum. Model resolution is not criti
cal, provided these three elements remain unaltered. Thus, if the scatterom
eter winds are accurate, the model results are consistent with observationa
l estimates of these two coefficients. These winds have larger westward str
ess than NCEP reanalysis winds, produce a 14% stronger EUC, more upwelling,
but a weaker westward surface flow.
In the Indian Ocean the seasonal cycle of equatorial currents does not appe
ar to be overly attenuated by the horizontal viscosity, with differences fr
om observations attributable to interannual variability. However, in the At
lantic, the numerics still require too large a meridional viscosity over to
o much of the basin, and a zonal resolution approaching 1 degrees may be ne
cessary to match observations. Because of this viscosity, increasing the ba
ckground vertical viscosity slowed the westward surface current; opposite t
o the response in the Pacific.