well-sampled, time-averaged general circulation
should emerge. As discussed earlier, the errors
in the ocean models are generally larger than the
altimetric ocean topography at large scales. The
combined solution through data assimilation
should produce an optimal estimate that is better
than the one based on either data or model alone.
Such an approach was taken by Stammer et al.
(1997). Some preliminary results from the study
can be found in Wunsch and Stammer (1998).
The effects of altimetry on the model simulation
are clearly identified. However, it is not yet
clear whether such estimates are consistent with
the hydrographic estimates as well as other types
of observations. The difficulty lies in the determination of the error covariance of the estimates
resulting from complex calculations involving
highly non-linear general circulation models.
3.3.3 Large-scale sea-level variability
After the tidal signals are removed (including the
effects of the solid earth tides), the sea-level variability has a total amplitude of 12 cm (global rms).
After making the inverted barometer correction
(see discussions below), the amplitude is reduced
to about 10 cm (Wunsch and Stammer, 1997).
The large scales (wavelength larger than about
500 km) account for about 6–7 cm (Wunsch and
Stammer, 1998), while the mesoscales account for
about 7–8 cm. As noted in 3.3.1, the large-scale
variability was often partly removed in the analysis of altimetry data in order to reduce the large
orbit errors present in the data collected by missions flown before T/P. Therefore, the large-scale
sea-level variability has not been fully accessible
to satellite altimetry until the advent of T/P.
The sampling of T/P is particularly suited for
resolving large-scale variability. Its 10-day repeat
period is able to sample fast-moving equatorial
waves as well as barotropic variabilities. The time
separation between adjacent tracks, whose longitudinal separation is 315 km at the equator and
220 km at 45° latitude, is only 3 days. Therefore,
T/P is able to sample partially large-scale signals
even at periods shorter than its 20-day Nyquist
period.
Kuragano and Kamachi (2000) investigated the
characteristics of spatial and temporal scales of
sea-level variability using T/P data. They fitted a
three-dimensional Gaussian model to the observed
covariance as a joint function (as opposed to
separate functions) in latitudinal, longitudinal and
temporal lags. They found that in areas where the
mesoscale energy is low, the dominant scales
are generally larger than 500 km. At mid- and low
latitudes, the large-scale variability is generally
anisotropic with larger zonal scales and westward
phase propagation, exhibiting the presence of
Rossby waves (see Sections 3.3.3.2 and 3.3.3.3).
At high latitudes, the large-scale variability becomes
more isotropic, perhaps reflecting locally forced
response. In areas where the mesoscale energy is
high, the large-scale variability becomes obscured
without some space-time filtering.
The large-scale variability of sea level is basically caused by the ocean’s response to the forcing
by the atmosphere. The primary forcing mechanisms include wind stress, pressure, and air–sea
exchange of heat and fresh water (Section 3.3.3.1).
These are part of the air–sea interaction processes.
Large-scale sea-level variability is in fact a key
indicator of climate variability. Mostly driven by
wind, the large-scale, baroclinic waves are
observed extensively by altimetry. They are discussed in terms of the extratropical and tropical
regions in Sections 3.3.3.2 and 3.3.3.3, respectively. The relation between large-scale sea-level
variability and subsurface fields and motions is
discussed in Section 3.3.3.4. The global annual
cycle, interannual variability, and global mean-sealevel changes are discussed in Sections 3.3.3.5,
3.3.3.6 and 3.3.3.7, respectively.
3.3.3.1 The ocean’s response to the
atmosphere
Atmospheric pressure forcing
The response of the ocean to atmospheric pressure
forcing has not received the same attention as the
response to wind and buoyancy forcing. This is
because pressure forcing is a much less effective
mechanism for generating oceanic motions. The
conventional wisdom is that, to first order, the
ocean responds to pressure forcing in a static manner. An increase (decrease) in atmospheric pressure
by 1 mbar would depress (raise) sea level by nearly
1 cm. The ocean acts like an ‘inverted barometer’.
The horizontal gradient of the atmospheric pressure
is completely compensated by the adjustment of sea
level, leaving no pressure gradient just below the
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
148
should emerge. As discussed earlier, the errors
in the ocean models are generally larger than the
altimetric ocean topography at large scales. The
combined solution through data assimilation
should produce an optimal estimate that is better
than the one based on either data or model alone.
Such an approach was taken by Stammer et al.
(1997). Some preliminary results from the study
can be found in Wunsch and Stammer (1998).
The effects of altimetry on the model simulation
are clearly identified. However, it is not yet
clear whether such estimates are consistent with
the hydrographic estimates as well as other types
of observations. The difficulty lies in the determination of the error covariance of the estimates
resulting from complex calculations involving
highly non-linear general circulation models.
3.3.3 Large-scale sea-level variability
After the tidal signals are removed (including the
effects of the solid earth tides), the sea-level variability has a total amplitude of 12 cm (global rms).
After making the inverted barometer correction
(see discussions below), the amplitude is reduced
to about 10 cm (Wunsch and Stammer, 1997).
The large scales (wavelength larger than about
500 km) account for about 6–7 cm (Wunsch and
Stammer, 1998), while the mesoscales account for
about 7–8 cm. As noted in 3.3.1, the large-scale
variability was often partly removed in the analysis of altimetry data in order to reduce the large
orbit errors present in the data collected by missions flown before T/P. Therefore, the large-scale
sea-level variability has not been fully accessible
to satellite altimetry until the advent of T/P.
The sampling of T/P is particularly suited for
resolving large-scale variability. Its 10-day repeat
period is able to sample fast-moving equatorial
waves as well as barotropic variabilities. The time
separation between adjacent tracks, whose longitudinal separation is 315 km at the equator and
220 km at 45° latitude, is only 3 days. Therefore,
T/P is able to sample partially large-scale signals
even at periods shorter than its 20-day Nyquist
period.
Kuragano and Kamachi (2000) investigated the
characteristics of spatial and temporal scales of
sea-level variability using T/P data. They fitted a
three-dimensional Gaussian model to the observed
covariance as a joint function (as opposed to
separate functions) in latitudinal, longitudinal and
temporal lags. They found that in areas where the
mesoscale energy is low, the dominant scales
are generally larger than 500 km. At mid- and low
latitudes, the large-scale variability is generally
anisotropic with larger zonal scales and westward
phase propagation, exhibiting the presence of
Rossby waves (see Sections 3.3.3.2 and 3.3.3.3).
At high latitudes, the large-scale variability becomes
more isotropic, perhaps reflecting locally forced
response. In areas where the mesoscale energy is
high, the large-scale variability becomes obscured
without some space-time filtering.
The large-scale variability of sea level is basically caused by the ocean’s response to the forcing
by the atmosphere. The primary forcing mechanisms include wind stress, pressure, and air–sea
exchange of heat and fresh water (Section 3.3.3.1).
These are part of the air–sea interaction processes.
Large-scale sea-level variability is in fact a key
indicator of climate variability. Mostly driven by
wind, the large-scale, baroclinic waves are
observed extensively by altimetry. They are discussed in terms of the extratropical and tropical
regions in Sections 3.3.3.2 and 3.3.3.3, respectively. The relation between large-scale sea-level
variability and subsurface fields and motions is
discussed in Section 3.3.3.4. The global annual
cycle, interannual variability, and global mean-sealevel changes are discussed in Sections 3.3.3.5,
3.3.3.6 and 3.3.3.7, respectively.
3.3.3.1 The ocean’s response to the
atmosphere
Atmospheric pressure forcing
The response of the ocean to atmospheric pressure
forcing has not received the same attention as the
response to wind and buoyancy forcing. This is
because pressure forcing is a much less effective
mechanism for generating oceanic motions. The
conventional wisdom is that, to first order, the
ocean responds to pressure forcing in a static manner. An increase (decrease) in atmospheric pressure
by 1 mbar would depress (raise) sea level by nearly
1 cm. The ocean acts like an ‘inverted barometer’.
The horizontal gradient of the atmospheric pressure
is completely compensated by the adjustment of sea
level, leaving no pressure gradient just below the
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
148