buoyancy forcing (with a maximum in March–
April, see Fig. 3.3.10c; Plate 3.3.10c, p. 172),
whereas to the south of the ACC, it is dominated
by the wind forcing.
After the removal of the steric component, the
residual annual cycle is prominent in the tropics,
where the seasonal wind forcing is the main driver
of the annual cycle. The annual cycle in the Pacific
has two bands of high amplitude: one along 5–7°N
and the other along 12–15°N. The former is associated with the annual variation of the North
Equatorial Countercurrent which reaches its maximum around September–November. A westward
phase propagation is observed across the current;
the month of maximum sea level changes from
September in the eastern part of the current to
November in the western part. This phase pattern
is consistent with the explanations of Meyers
(1979) in terms of wind-driven Rossby waves. The
variability along 12–15°N is indicative of the
annual variations of the North Equatorial Current.
Westward phase propagation is observed only in
the eastern part of the current. There is little phase
propagation in the western part, where Meyers’
study suggested the existence of a near in-phase
relationship between sea level and wind stress curl
(also see Vivier et al., 1999). The maximum in the
Atlantic (along 5–7°N) is associated with the
Atlantic North Equatorial Countercurrent with a
phase similar to its Pacific counterpart (Richardson
and Reverdin, 1987). Previous studies of the
Indian Ocean (Woodbury et al., 1989; Perigaud
and Delecluse, 1992) suggested that the annual
cycle in the southern tropical Indian Ocean was
associated with Rossby waves driven by the
annual cycle of trade winds (also see Masumoto
and Meyers, 1998). It is also interesting to note
the 180° phase change across the Arabian Sea, representing the ocean’s response to the annual monsoon wind cycle during its two opposite phases
(Bruce et al., 1994). Significant semiannual signals
are found in the Indian Ocean (Jacobs et al., 1992;
Basu et al., 2000), as well as the central equatorial
Pacific and the southwest Atlantic.
The ability of numerical models to simulate the
oceanic annual cycle is an important test of the usefulness of the model for climate studies, because
the annual cycle is the result of the complicated
interaction between the ocean and the atmosphere.
Comparisons of ocean general circulation models
with the T/P altimeter data have revealed the
strengths and weaknesses of the models (Fu and
Smith, 1996; Jacobs et al., 1996; Stammer et al.,
1996). In general the models have good skills in
reproducing the wind-driven component of the
variability, but not the steric component. The
model-simulated amplitude at mid-latitudes where
the steric component dominates the annual cycle is
generally too weak (by as much as 3 cm), reflecting
possible problems in the model’s mixing mechanism and/or the poor quality of heat flux data
driving the model. Although the error in the
ECMWF heat flux could cause an error of 2 cm in
the annual cycle (Stammer et al., 1996), there is no
evidence for a systematic low bias in the ECMWF
heat flux. The major culprit should be the model’s
lack of a good mixed layer. Li et al. (2000)
demonstrated that the use of a state-of-the-art
mixing scheme significantly improved the simulation of the annual cycle by an ocean model.
3.3.3.6 Interannual variability
It is well known that the oceanic frequency spectrum is ‘red’ with the background spectral level ever
increasing with decreasing frequency, except for
peaks such as the annual cycle. There is substantial
variability at periods longer than the annual period.
The relatively short record of altimetry allows one
to study only the interannual scales (between a
year and a decade). Shown in Fig. 3.3.11 (see Plate
3.3.11, p. 172) are the yearly-averaged sea-level
anomalies computed from the T/P data for
1996–99. The anomalies were relative to a 4-year
mean computed from the data covering 1993–96.
The 4-year period of 1993–96 has a weak warm
event (1994–95) and a weak cold event (1996) and
is thus considered a reasonable period for computing a mean somewhat close to the norm. The
yearly averaging has filtered out the annual cycle
and other high-frequency signals. The resulting
yearly anomalies thus reveal primarily the variability on time scales longer than a year. However,
the mesoscale variability is still visible in many
places after the averaging, especially along the
Indian Ocean sector of the ACC as well as in the
Brazil/Malvinas Confluence region of the South
Atlantic. The large-scale features clearly illustrate
the effects of the dramatic 1997–98 El Niño and
its transition into La Niña in 1998–99. The high
sea levels in the western Pacific and the eastern
Indian Ocean in 1996 set the build-up stage for El
Niño, during which the tropical Pacific Ocean and
3.3 Ocean Circulation and Variability from Satellite Altimetry
159
Fu
April, see Fig. 3.3.10c; Plate 3.3.10c, p. 172),
whereas to the south of the ACC, it is dominated
by the wind forcing.
After the removal of the steric component, the
residual annual cycle is prominent in the tropics,
where the seasonal wind forcing is the main driver
of the annual cycle. The annual cycle in the Pacific
has two bands of high amplitude: one along 5–7°N
and the other along 12–15°N. The former is associated with the annual variation of the North
Equatorial Countercurrent which reaches its maximum around September–November. A westward
phase propagation is observed across the current;
the month of maximum sea level changes from
September in the eastern part of the current to
November in the western part. This phase pattern
is consistent with the explanations of Meyers
(1979) in terms of wind-driven Rossby waves. The
variability along 12–15°N is indicative of the
annual variations of the North Equatorial Current.
Westward phase propagation is observed only in
the eastern part of the current. There is little phase
propagation in the western part, where Meyers’
study suggested the existence of a near in-phase
relationship between sea level and wind stress curl
(also see Vivier et al., 1999). The maximum in the
Atlantic (along 5–7°N) is associated with the
Atlantic North Equatorial Countercurrent with a
phase similar to its Pacific counterpart (Richardson
and Reverdin, 1987). Previous studies of the
Indian Ocean (Woodbury et al., 1989; Perigaud
and Delecluse, 1992) suggested that the annual
cycle in the southern tropical Indian Ocean was
associated with Rossby waves driven by the
annual cycle of trade winds (also see Masumoto
and Meyers, 1998). It is also interesting to note
the 180° phase change across the Arabian Sea, representing the ocean’s response to the annual monsoon wind cycle during its two opposite phases
(Bruce et al., 1994). Significant semiannual signals
are found in the Indian Ocean (Jacobs et al., 1992;
Basu et al., 2000), as well as the central equatorial
Pacific and the southwest Atlantic.
The ability of numerical models to simulate the
oceanic annual cycle is an important test of the usefulness of the model for climate studies, because
the annual cycle is the result of the complicated
interaction between the ocean and the atmosphere.
Comparisons of ocean general circulation models
with the T/P altimeter data have revealed the
strengths and weaknesses of the models (Fu and
Smith, 1996; Jacobs et al., 1996; Stammer et al.,
1996). In general the models have good skills in
reproducing the wind-driven component of the
variability, but not the steric component. The
model-simulated amplitude at mid-latitudes where
the steric component dominates the annual cycle is
generally too weak (by as much as 3 cm), reflecting
possible problems in the model’s mixing mechanism and/or the poor quality of heat flux data
driving the model. Although the error in the
ECMWF heat flux could cause an error of 2 cm in
the annual cycle (Stammer et al., 1996), there is no
evidence for a systematic low bias in the ECMWF
heat flux. The major culprit should be the model’s
lack of a good mixed layer. Li et al. (2000)
demonstrated that the use of a state-of-the-art
mixing scheme significantly improved the simulation of the annual cycle by an ocean model.
3.3.3.6 Interannual variability
It is well known that the oceanic frequency spectrum is ‘red’ with the background spectral level ever
increasing with decreasing frequency, except for
peaks such as the annual cycle. There is substantial
variability at periods longer than the annual period.
The relatively short record of altimetry allows one
to study only the interannual scales (between a
year and a decade). Shown in Fig. 3.3.11 (see Plate
3.3.11, p. 172) are the yearly-averaged sea-level
anomalies computed from the T/P data for
1996–99. The anomalies were relative to a 4-year
mean computed from the data covering 1993–96.
The 4-year period of 1993–96 has a weak warm
event (1994–95) and a weak cold event (1996) and
is thus considered a reasonable period for computing a mean somewhat close to the norm. The
yearly averaging has filtered out the annual cycle
and other high-frequency signals. The resulting
yearly anomalies thus reveal primarily the variability on time scales longer than a year. However,
the mesoscale variability is still visible in many
places after the averaging, especially along the
Indian Ocean sector of the ACC as well as in the
Brazil/Malvinas Confluence region of the South
Atlantic. The large-scale features clearly illustrate
the effects of the dramatic 1997–98 El Niño and
its transition into La Niña in 1998–99. The high
sea levels in the western Pacific and the eastern
Indian Ocean in 1996 set the build-up stage for El
Niño, during which the tropical Pacific Ocean and
3.3 Ocean Circulation and Variability from Satellite Altimetry
159
Fu