the general circulation model simulations provide
convincing evidence that the large-scale, intraseasonal sea-level variability is largely due to windforced barotropic motion of the ocean.
Sea-level variability has significant amount of
energy at periods shorter than the repeat period
of altimetry (e.g. 10 days for T/P, 17 days for
GEOSAT, 35 days for ERS). Fukumori et al.
(1998) demonstrated that model simulations driven by wind could partially account for these highfrequency barotropic variabilities in the T/P data.
Although the 10-day repeat period of T/P was not
adequate for resolving the variabilities, Fukumori
showed that 12-h sampled model simulations were
able to explain more variance of T/P data than
3-day sampled simulations. These high-frequency
variabilites, if not removed, would create aliasing problems for studying low-frequency variability using altimetry data. Recent studies of
Stammer et al. (2000a) and Tierney et al. (2000)
have suggested that de-aliasing of this kind can
be performed using ocean general circulation
models forced by accurate wind and atmospheric
pressure.
As time scale increases, the ocean’s response to
wind becomes increasingly baroclinic. This is
because the wavelength of barotropic Rossby
waves decreases with increasing period. For example, the wavelength is about 100 km for a wave
period of 300 days at 45° latitude. The wind forcing has relatively little energy in this frequency/
wavenumber band to generate barotropic waves.
The minimum period of baroclinic Rossby waves
is basically a function of latitude (Gill, 1982). Such
period is on the order of 500 days and longer poleward of the 50° latitude, where there is not sufficient wind energy at such long periods (Fukumori
et al., 1998, their Fig. 13) to force the baroclinic
waves. However, due to the limited duration of
available observations, we are not able to address
the baroclinic energy at decadal scales when the
basin-wide baroclinic adjustment processes become
important. On the other hand, the minimum wave
period becomes shorter at low latitudes (e.g. 100
days equatorward of 20°), where the wind has
substantial energy available in the period band to
force the baroclinic waves. Therefore the ocean’s
baroclinic response to wind forcing is primarily
originated at mid- and low latitudes. This response
is discussed below in terms of baroclinic waves in
the extratropical and tropical regions separately.
3.3.3.2 Extratropical Rossby waves
Westward propagation is a ubiquitous characteristic in a display of sea-level anomalies with a time–
longitude section except within the vicinity of the
equator (Fig. 3.3.7, see Plate 3.3.7, p. 172). Its
interpretation in terms of Rossby waves has been
documented in a large body of literature (Fu and
Chelton, 2000). Identification of the source of these
waves is not easy, but they are to a large extent
forced by wind remotely. Because of the variability
of the sea surface temperature associated with the
waves, local wind could be affected by the waves
and becomes coupled to them, leading to an ocean–
atmosphere-coupled system (White et al., 1998b).
Many studies were focused on regional analyses
in which evidence of Rossby waves was documented
and standard theories were invoked to interpret the
observations (White et al., 1990a; Kelly et al., 1993;
Le Traon and Minster, 1993; van Woert and Price,
1993). A systematic search for Rossby waves was
conducted by Jacobs et al. (1993) in the Pacific
Ocean. They fitted the GEOSAT sea surface height
anomalies to a set of waves with wavenumbers
and frequencies obeying a quasigeostrophic Rossby
wave dispersion relation. Such fitting was performed using a least-squares technique in 10°10°
boxes. They found evidence for the generation of
baroclinic waves along the eastern boundaries and
the so-called refraction along the path of westward propagation into the ocean interior. The
refraction is caused by the fact that the wave’s zonal
phase speed is faster at low latitudes than at high
latitudes due to the latitudinal dependence of
(ϵdf/dy). The propagation direction of the waves
is thus changing with latitude.
Polito and Cornillon (1997) conducted a basinwide survey in the North Atlantic using T/P data.
They estimated the phase speed of Rossby wave
propagation in two period bands centred on the
annual and semiannual periods. The westward
phase speeds were found to increase equatorward
as predicted by theory (also see Nerem et al.,
1994). The wavelengths range from 400 to 4600 km
for the annual band and 270 to 2500 km for the
semiannual band. The wave amplitudes in the
annual band (2–12 cm) are larger than those in
the semiannual band (1–4 cm). They also reported
evidence for remote generation of the waves at the
eastern boundary by fluctuating wind stress curl
and the interaction of the waves with the Mid
Atlantic Ridge.
3.3 Ocean Circulation and Variability from Satellite Altimetry
153
Fu
convincing evidence that the large-scale, intraseasonal sea-level variability is largely due to windforced barotropic motion of the ocean.
Sea-level variability has significant amount of
energy at periods shorter than the repeat period
of altimetry (e.g. 10 days for T/P, 17 days for
GEOSAT, 35 days for ERS). Fukumori et al.
(1998) demonstrated that model simulations driven by wind could partially account for these highfrequency barotropic variabilities in the T/P data.
Although the 10-day repeat period of T/P was not
adequate for resolving the variabilities, Fukumori
showed that 12-h sampled model simulations were
able to explain more variance of T/P data than
3-day sampled simulations. These high-frequency
variabilites, if not removed, would create aliasing problems for studying low-frequency variability using altimetry data. Recent studies of
Stammer et al. (2000a) and Tierney et al. (2000)
have suggested that de-aliasing of this kind can
be performed using ocean general circulation
models forced by accurate wind and atmospheric
pressure.
As time scale increases, the ocean’s response to
wind becomes increasingly baroclinic. This is
because the wavelength of barotropic Rossby
waves decreases with increasing period. For example, the wavelength is about 100 km for a wave
period of 300 days at 45° latitude. The wind forcing has relatively little energy in this frequency/
wavenumber band to generate barotropic waves.
The minimum period of baroclinic Rossby waves
is basically a function of latitude (Gill, 1982). Such
period is on the order of 500 days and longer poleward of the 50° latitude, where there is not sufficient wind energy at such long periods (Fukumori
et al., 1998, their Fig. 13) to force the baroclinic
waves. However, due to the limited duration of
available observations, we are not able to address
the baroclinic energy at decadal scales when the
basin-wide baroclinic adjustment processes become
important. On the other hand, the minimum wave
period becomes shorter at low latitudes (e.g. 100
days equatorward of 20°), where the wind has
substantial energy available in the period band to
force the baroclinic waves. Therefore the ocean’s
baroclinic response to wind forcing is primarily
originated at mid- and low latitudes. This response
is discussed below in terms of baroclinic waves in
the extratropical and tropical regions separately.
3.3.3.2 Extratropical Rossby waves
Westward propagation is a ubiquitous characteristic in a display of sea-level anomalies with a time–
longitude section except within the vicinity of the
equator (Fig. 3.3.7, see Plate 3.3.7, p. 172). Its
interpretation in terms of Rossby waves has been
documented in a large body of literature (Fu and
Chelton, 2000). Identification of the source of these
waves is not easy, but they are to a large extent
forced by wind remotely. Because of the variability
of the sea surface temperature associated with the
waves, local wind could be affected by the waves
and becomes coupled to them, leading to an ocean–
atmosphere-coupled system (White et al., 1998b).
Many studies were focused on regional analyses
in which evidence of Rossby waves was documented
and standard theories were invoked to interpret the
observations (White et al., 1990a; Kelly et al., 1993;
Le Traon and Minster, 1993; van Woert and Price,
1993). A systematic search for Rossby waves was
conducted by Jacobs et al. (1993) in the Pacific
Ocean. They fitted the GEOSAT sea surface height
anomalies to a set of waves with wavenumbers
and frequencies obeying a quasigeostrophic Rossby
wave dispersion relation. Such fitting was performed using a least-squares technique in 10°10°
boxes. They found evidence for the generation of
baroclinic waves along the eastern boundaries and
the so-called refraction along the path of westward propagation into the ocean interior. The
refraction is caused by the fact that the wave’s zonal
phase speed is faster at low latitudes than at high
latitudes due to the latitudinal dependence of
(ϵdf/dy). The propagation direction of the waves
is thus changing with latitude.
Polito and Cornillon (1997) conducted a basinwide survey in the North Atlantic using T/P data.
They estimated the phase speed of Rossby wave
propagation in two period bands centred on the
annual and semiannual periods. The westward
phase speeds were found to increase equatorward
as predicted by theory (also see Nerem et al.,
1994). The wavelengths range from 400 to 4600 km
for the annual band and 270 to 2500 km for the
semiannual band. The wave amplitudes in the
annual band (2–12 cm) are larger than those in
the semiannual band (1–4 cm). They also reported
evidence for remote generation of the waves at the
eastern boundary by fluctuating wind stress curl
and the interaction of the waves with the Mid
Atlantic Ridge.
3.3 Ocean Circulation and Variability from Satellite Altimetry
153
Fu