the absolute temperature field. Such temperature
estimates were able to account for 53% of the total
variance, with a maximum rms error of 0.7°C at
depths of 80–160 m in the central basin. The error
increases to 1°C in the thermocline at both the
eastern and western ends of the basin. Improvement of such estimates is expected from a longer
data set that will allow increased signal-to-noise
ratio and more accurate regression analysis.
Gilson et al. (1998) also estimated the specific
volume anomaly at depths from the altimetric sealevel anomaly. With the vertical profiles of specific
volume anomaly determined, they then computed
the basin-wide meridional volume and temperature transports of the upper 800 m. They compared the altimetrically determined transports with
those computed from the in-situ data and found
reasonable agreement. The study of Gilson et al.
(1998) has thus demonstrated the utility of combined use of altimeter data with in-situ data for
making estimates of basin-wide subsurface fields
and transports that are of importance to the study
of climate. The potential of this combination has
motivated plans for deploying a network of floats
in the ocean to complement future altimetry missions (The programme is called Argo, see Wilson,
2000.) More discussions on the use of altimeter
data for estimating the transports of ocean currents
will be given in Section 3.3.4.3. The correlation
between sea-level variability and subsurface variability makes satellite altimetry a powerful tool
for a global observing system. A more effective
approach to the estimation of subsurface variability
from altimeter data has been provided by the use of
an ocean general circulation model through the
techniques of data assimilation (Fukumori, 2000).
Stammer et al. (1997b) applied an adjoint method
to a coarse-resolution ocean general circulation
model (2°2°, 20 vertical levels) for assimilation
of T/P data along with other observations in the
determination of the entire state of the ocean.
They showed the impact of altimeter data on
improving the meridional heat transport of the
ocean. Fukumori et al. (1999) applied an approximate Kalman filter to a similar coarse-resolution
model and showed improved estimates of subsurface temperature and velocity.
3.3.3.5 The annual cycle
The annual cycle is a major component of the
large-scale variability over most of the ocean. It is
caused by a combination of many different processes and hence exhibits a very complicated geographic pattern in its amplitude and phase. Owing
to its dense spatial coverage, satellite altimetry
provides the first detailed description of this complicated pattern. Jacobs et al. (1992) presented the
first results of the global ocean annual cycle from
altimetry using the GEOSAT data. The analysis was
complicated by large tidal and orbital errors in the
data. Although ingenious methods were applied to
remove these errors, the results inevitably suffer
from the residual errors because both the signals
and errors have similar large scales. Using the
more accurate data from T/P, Stammer (1997a)
computed the amplitude and phase of an annual
harmonic fit to 3 years of data (1993–95, Fig.
3.3.10, see Plate 3.3.10, p. 172). Also shown are
the steric component and the residual sea level.
As discussed in Section 3.3.3.1, p. 148, a major
portion of the annual cycle at mid-latitudes is
due to steric effect with opposite phase between
the two hemispheres. The larger amplitude in the
northern hemisphere is clearly revealed. However,
this hemispheric asymmetry was not shown in the
GEOSAT result, probably due to the effect of the
orbit-error-removal procedure applied to the data.
In fact, the annual cycle at mid-latitudes away
from the western boundary currents was mostly
absent in Jacobs et al. (1992).
At high latitudes, where air–sea heat flux has
the largest annual variability, the sea-level annual
cycle is actually less than that at mid-latitudes.
This is because the coefficient of heat expansion
becomes smaller when the water gets colder. The
coefficient at high latitudes is about only 1/3 of its
value at low latitudes. In the high-latitude Southern Ocean, there is a sharp change in the phase of
the annual cycle, roughly within 50–60°S where
the Antarctic Circumpolar Current (ACC) flows.
The annual sea-level maximum occurs in March–
April to the north and in August–September to the
south. This front of phase change is consistent
with the GEOSAT study of Chelton et al. (1990)
in which the annual cycle accounts for the first
empirical orthogonal mode. After the steric component is removed, the phase of the wind-forced
residual variability in the Southern Ocean
(Fig. 3.3.10f, see Plate 3.3.10f, p. 172) has a maximum in August–October when the wind is the
strongest. It is therefore apparent that the annual
cycle to the north of the ACC is dominated by the
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
158
estimates were able to account for 53% of the total
variance, with a maximum rms error of 0.7°C at
depths of 80–160 m in the central basin. The error
increases to 1°C in the thermocline at both the
eastern and western ends of the basin. Improvement of such estimates is expected from a longer
data set that will allow increased signal-to-noise
ratio and more accurate regression analysis.
Gilson et al. (1998) also estimated the specific
volume anomaly at depths from the altimetric sealevel anomaly. With the vertical profiles of specific
volume anomaly determined, they then computed
the basin-wide meridional volume and temperature transports of the upper 800 m. They compared the altimetrically determined transports with
those computed from the in-situ data and found
reasonable agreement. The study of Gilson et al.
(1998) has thus demonstrated the utility of combined use of altimeter data with in-situ data for
making estimates of basin-wide subsurface fields
and transports that are of importance to the study
of climate. The potential of this combination has
motivated plans for deploying a network of floats
in the ocean to complement future altimetry missions (The programme is called Argo, see Wilson,
2000.) More discussions on the use of altimeter
data for estimating the transports of ocean currents
will be given in Section 3.3.4.3. The correlation
between sea-level variability and subsurface variability makes satellite altimetry a powerful tool
for a global observing system. A more effective
approach to the estimation of subsurface variability
from altimeter data has been provided by the use of
an ocean general circulation model through the
techniques of data assimilation (Fukumori, 2000).
Stammer et al. (1997b) applied an adjoint method
to a coarse-resolution ocean general circulation
model (2°2°, 20 vertical levels) for assimilation
of T/P data along with other observations in the
determination of the entire state of the ocean.
They showed the impact of altimeter data on
improving the meridional heat transport of the
ocean. Fukumori et al. (1999) applied an approximate Kalman filter to a similar coarse-resolution
model and showed improved estimates of subsurface temperature and velocity.
3.3.3.5 The annual cycle
The annual cycle is a major component of the
large-scale variability over most of the ocean. It is
caused by a combination of many different processes and hence exhibits a very complicated geographic pattern in its amplitude and phase. Owing
to its dense spatial coverage, satellite altimetry
provides the first detailed description of this complicated pattern. Jacobs et al. (1992) presented the
first results of the global ocean annual cycle from
altimetry using the GEOSAT data. The analysis was
complicated by large tidal and orbital errors in the
data. Although ingenious methods were applied to
remove these errors, the results inevitably suffer
from the residual errors because both the signals
and errors have similar large scales. Using the
more accurate data from T/P, Stammer (1997a)
computed the amplitude and phase of an annual
harmonic fit to 3 years of data (1993–95, Fig.
3.3.10, see Plate 3.3.10, p. 172). Also shown are
the steric component and the residual sea level.
As discussed in Section 3.3.3.1, p. 148, a major
portion of the annual cycle at mid-latitudes is
due to steric effect with opposite phase between
the two hemispheres. The larger amplitude in the
northern hemisphere is clearly revealed. However,
this hemispheric asymmetry was not shown in the
GEOSAT result, probably due to the effect of the
orbit-error-removal procedure applied to the data.
In fact, the annual cycle at mid-latitudes away
from the western boundary currents was mostly
absent in Jacobs et al. (1992).
At high latitudes, where air–sea heat flux has
the largest annual variability, the sea-level annual
cycle is actually less than that at mid-latitudes.
This is because the coefficient of heat expansion
becomes smaller when the water gets colder. The
coefficient at high latitudes is about only 1/3 of its
value at low latitudes. In the high-latitude Southern Ocean, there is a sharp change in the phase of
the annual cycle, roughly within 50–60°S where
the Antarctic Circumpolar Current (ACC) flows.
The annual sea-level maximum occurs in March–
April to the north and in August–September to the
south. This front of phase change is consistent
with the GEOSAT study of Chelton et al. (1990)
in which the annual cycle accounts for the first
empirical orthogonal mode. After the steric component is removed, the phase of the wind-forced
residual variability in the Southern Ocean
(Fig. 3.3.10f, see Plate 3.3.10f, p. 172) has a maximum in August–October when the wind is the
strongest. It is therefore apparent that the annual
cycle to the north of the ACC is dominated by the
SECTION 3 NEW WAYS OF OBSERVING THE OCEAN
158