drift (Richardson, 1997), and the drifter observations were also made over broad ocean areas
where ships do not go. New current patterns were
discovered; several of these are discussed below.
The mean velocity vector map on Fig. 4.1.5a
(see Plate 4.1.5a, p. 300) includes an ellipse at the
tip of each vector. This is the standard error of the
mean, or uncertainty, ellipse at each location. To
compute this ellipse, let :u9, :v9 be the ensemble mean velocity components within each bin in
the latitude (x), longitude (y) directions, respectively. The symbol :9 is the arithmetic average of
the ensemble of observations within each bin. Let
uЈ, vЈ be the corresponding deviations from these
means. Rotate the x, y coordinate system into the
principal variance axes so that in the rotated coordinate system :uЈvЈ9:0. In the rotated coordinate system the equation for the variance ellipse is
{x
2
:uЈ
2
9
91
;y
2
:vЈ
2
9
91
}:1. If N is the number
of ‘independent’ velocity observations within each
bin, then the equation for the standard error ellipse
is {x
2
:uЈ
2
/N9
91
;y
2
:vЈ
2
/N9
91
}:1. The number of independent observations depends upon the
Lagrangian auto-decorrelation time scale in each
bin. Here it was chosen to be 6 days, or an average
of the many estimates computed at various locations in the Pacific (Poulain and Niiler, 1989;
Paduan and Niiler, 1993; Bi, 1995). Thus every
third observation in a bin was considered independent. For a normal distribution of variability, there
is an 18% probability that a mean vector computed from a different large and random sample of
data within the bin will fall outside the ellipses
drawn on Fig. 4.1.5a.
The most completely sampled distribution of
circulation was in the tropical Pacific where, commencing in 1978, small numbers of FGGE-type
drifters were deployed over a 10-year period. The
basin-scale observations with the WOCE/TOGA
Lagrangian Drifters have been continuous since
1988. From these data, Reverdin et al. (1994)
calculated the seasonal cycle of currents and
Frankignoul et al. (1996) computed the anomalous
currents. During the 1986–87, 1992–93 and
1997–98 El Niños, the surface circulation of the
entire Pacific within 6°S of the equator was to
the east (NOAA, 1997–1998). A comparison of
the observed and modelled circulation was made
with coupled models of tropical Pacific circulation
at the end of TOGA (WCRP, 1995b). The least
sampled basin was the tropical Atlantic, where
observations began on a systematic basis in 1998
and will continue as part of ongoing WCRPsponsored programmes.
The surface geostrophic currents, relative to
1000 m or 3000 m depth, inferred from hydrographic data in the tropical and subtropical Pacific
Ocean basin are toward the equator (e.g. Wytrki,
1975). The observed currents from drifters within
the 25° latitude belt around the equator, however,
were toward the pole. Thus, a most striking difference was found between the observed meridional
component of velocity and that computed from
hydrographic data. This phenomenon had been
documented from a comparison of the ship-borne
Acoustic Doppler Current Profiler (ADCP) observations of currents and those computed from
hydrographic data from an 11°N section across
the Atlantic (Chereskin and Roemmich, 1991) and
10°N section across the Pacific (Wijffels et al.,
1994). These direct observations of circulation
implied that generally in the subtropics, the locally
wind-driven meridional circulation was much
stronger and in the opposite direction from that
derived from the geostrophic balance alone.
Within the winds of the Northeast Trades, both
the meridional and the zonal components of the
Ekman velocity were stronger than the surface
geostrophic velocity (Niiler and Barth, 2001). The
meridional velocity plays an important role in
the transport of heat from the equator toward the
poles. Its strength and its depth distribution are a
function of the vertical turbulent processes, which
are not well known in the upper ocean and thus
heavily parameterized in ocean circulation models.
In Section 4.1.4, the Ekman velocity component is
further discussed with a view toward improving
this parameterization.
On Fig. 4.1.5a (see Plate 4.1.5a, p. 300), two
different scales of velocity vectors were used in
order to display the boundary currents and the
mid-ocean circulation, with their respective standard error ellipses. Mean surface circulation in the
centre of gyres was not known before the large
number of drifter observations became available,
as ship drift was not accurate enough to resolve
the mean surface velocity and hydrographic methods did not include the strong, local wind-driven
component. A discovery in the mid-Pacific was
that of a ‘gyral’ current system north and east of
the Hawaiian Islands, whose existence was predicted by Sverdrup et al. (1942), but had remained
4.1 The World Ocean Surface Circulation
199
Niiler
where ships do not go. New current patterns were
discovered; several of these are discussed below.
The mean velocity vector map on Fig. 4.1.5a
(see Plate 4.1.5a, p. 300) includes an ellipse at the
tip of each vector. This is the standard error of the
mean, or uncertainty, ellipse at each location. To
compute this ellipse, let :u9, :v9 be the ensemble mean velocity components within each bin in
the latitude (x), longitude (y) directions, respectively. The symbol :9 is the arithmetic average of
the ensemble of observations within each bin. Let
uЈ, vЈ be the corresponding deviations from these
means. Rotate the x, y coordinate system into the
principal variance axes so that in the rotated coordinate system :uЈvЈ9:0. In the rotated coordinate system the equation for the variance ellipse is
{x
2
:uЈ
2
9
91
;y
2
:vЈ
2
9
91
}:1. If N is the number
of ‘independent’ velocity observations within each
bin, then the equation for the standard error ellipse
is {x
2
:uЈ
2
/N9
91
;y
2
:vЈ
2
/N9
91
}:1. The number of independent observations depends upon the
Lagrangian auto-decorrelation time scale in each
bin. Here it was chosen to be 6 days, or an average
of the many estimates computed at various locations in the Pacific (Poulain and Niiler, 1989;
Paduan and Niiler, 1993; Bi, 1995). Thus every
third observation in a bin was considered independent. For a normal distribution of variability, there
is an 18% probability that a mean vector computed from a different large and random sample of
data within the bin will fall outside the ellipses
drawn on Fig. 4.1.5a.
The most completely sampled distribution of
circulation was in the tropical Pacific where, commencing in 1978, small numbers of FGGE-type
drifters were deployed over a 10-year period. The
basin-scale observations with the WOCE/TOGA
Lagrangian Drifters have been continuous since
1988. From these data, Reverdin et al. (1994)
calculated the seasonal cycle of currents and
Frankignoul et al. (1996) computed the anomalous
currents. During the 1986–87, 1992–93 and
1997–98 El Niños, the surface circulation of the
entire Pacific within 6°S of the equator was to
the east (NOAA, 1997–1998). A comparison of
the observed and modelled circulation was made
with coupled models of tropical Pacific circulation
at the end of TOGA (WCRP, 1995b). The least
sampled basin was the tropical Atlantic, where
observations began on a systematic basis in 1998
and will continue as part of ongoing WCRPsponsored programmes.
The surface geostrophic currents, relative to
1000 m or 3000 m depth, inferred from hydrographic data in the tropical and subtropical Pacific
Ocean basin are toward the equator (e.g. Wytrki,
1975). The observed currents from drifters within
the 25° latitude belt around the equator, however,
were toward the pole. Thus, a most striking difference was found between the observed meridional
component of velocity and that computed from
hydrographic data. This phenomenon had been
documented from a comparison of the ship-borne
Acoustic Doppler Current Profiler (ADCP) observations of currents and those computed from
hydrographic data from an 11°N section across
the Atlantic (Chereskin and Roemmich, 1991) and
10°N section across the Pacific (Wijffels et al.,
1994). These direct observations of circulation
implied that generally in the subtropics, the locally
wind-driven meridional circulation was much
stronger and in the opposite direction from that
derived from the geostrophic balance alone.
Within the winds of the Northeast Trades, both
the meridional and the zonal components of the
Ekman velocity were stronger than the surface
geostrophic velocity (Niiler and Barth, 2001). The
meridional velocity plays an important role in
the transport of heat from the equator toward the
poles. Its strength and its depth distribution are a
function of the vertical turbulent processes, which
are not well known in the upper ocean and thus
heavily parameterized in ocean circulation models.
In Section 4.1.4, the Ekman velocity component is
further discussed with a view toward improving
this parameterization.
On Fig. 4.1.5a (see Plate 4.1.5a, p. 300), two
different scales of velocity vectors were used in
order to display the boundary currents and the
mid-ocean circulation, with their respective standard error ellipses. Mean surface circulation in the
centre of gyres was not known before the large
number of drifter observations became available,
as ship drift was not accurate enough to resolve
the mean surface velocity and hydrographic methods did not include the strong, local wind-driven
component. A discovery in the mid-Pacific was
that of a ‘gyral’ current system north and east of
the Hawaiian Islands, whose existence was predicted by Sverdrup et al. (1942), but had remained
4.1 The World Ocean Surface Circulation
199
Niiler