is predominantly caused by adiabatic, mechanical
forcing and thus has a tight relation with subsurface variability as part of an organized motion
field of the water column. For the barotropic
mode, the surface geostrophic velocity determined
from sea-level variability through equation (3.3.3)
represents the uniform horizontal velocity of the
entire water column. For the baroclinic mode, sealevel change is related to the change in the height
of the subsurface constant-density surfaces, the
isopycnals. Therefore, sea level has been used to
estimate subsurface density field, from which horizontal geostrophic velocity at depths has also been
estimated.
Carnes et al. (1990) used GEOSAT altimeter
data to estimate the subsurface temperature field
in the Gulf Stream area. They applied statistical
regression analysis to a set of AXBT (Air-dropped
expendable bathythermograph) data deployed along
the GEOSAT ground tracks and derived an empirical relation between subsurface temperature and
surface dynamic height. They then applied the
empirical relation to the GEOSAT altimeter data
for estimating the subsurface temperature field.
They used the ocean topography derived from the
altimeter data relative to a geoid model as a surrogate for the dynamic height. The rms difference
between the GEOSAT-derived ocean topography
and the dynamic height is 15–19 cm, mostly reflecting the geoid errors and the GEOSAT altimeter
measurement errors. The resulting temperature
estimates have an rms error of about 1°C below
200 m. The error increases to 2°C near the surface,
where there is significant change in temperature
due to the annual cycle of heat exchange with the
atmosphere. The thermally driven variability of
temperature as a function of depth does not have a
tight relation with the sea-level variability.
Using T/P data with a series of repeating transects (repeat every 3 months) of XBT and XCTD
(Expendable Conductivity and Temperature profiler) across the North Pacific Ocean (from Taiwan
through Guam to San Francisco), Gilson et al.
(1998) studied the relationship between altimetric
sea-level measurements and the subsurface temperature and currents. A total of 5 years of simultaneous in-situ and satellite data were analysed,
allowing the relation between the two data sets to
be examined over a wide range of spatial and temporal scales. The anomalies of the altimetric sea
level relative to a 5-year mean were compared
with the anomalies of the dynamic height computed from the in-situ data relative to the same
5-year mean. The altimeter data were interpolated
to the times and locations of the in-situ observations using an objective analysis scheme. The overall rms difference between the two anomaly fields
is 5.2 cm. At wavelengths longer than 500 km,
where lies 65% of the variance of the dynamic
height, the two are highly coherent (0.89) with an
rms difference of 3.5 cm. This difference is consistent with the measurement errors of T/P (Chelton
et al., 2000) and the errors in the dynamic height
estimates, plus the residual dynamic height variability below 800 m, which is the deepest level of
the XBT observations. At wavelengths shorter
than 500 km, the coarse spacing of the T/P ground
tracks led to the underestimation of the mesoscale
variability, which, however, was well sampled by
the in-situ data. The correlation between the two
decreased to 0.56 at these scales.
From analysing the XBT/XCTD data, Gilson
et al. (1998) found that, except for the annual
variations, the dynamic height anomalies were
largely caused by the vertical motion of the thermocline. The entire thermocline moved coherently in
the vertical, causing the temperature at the thermocline depths to be highly correlated with the
surface dynamic height but with an opposite sign.
The high correlation between the altimetric sealevel anomaly and the surface dynamic height
anomaly discussed above thus allows the use of
the former to make estimates of the subsurface
temperature anomaly. The correlation between the
dynamic height and the subsurface temperature
was derived after the annual cycle was removed
from both fields. This is because the annual temperature change in the extratropical regions is
primarily caused by the heat exchange with the
atmosphere instead of the vertical motion of the
isopycnals. The vertical distribution of the temperature change at the annual period is thus not
highly correlated with the surface dynamic height,
except for the tropics where the annual cycle is
wind-driven (see Section 3.3.3.5) and hence the
correlation between dynamic height and temperature is high. Therefore, Gilson et al. (1998) used
the altimeter data only to estimate the subsurface
temperature anomaly at non-annual time scales.
The temperature anomalies derived from the altimeter data were added to the mean and the annual
cycle determined from the in-situ data to obtain
3.3 Ocean Circulation and Variability from Satellite Altimetry
157
Fu
forcing and thus has a tight relation with subsurface variability as part of an organized motion
field of the water column. For the barotropic
mode, the surface geostrophic velocity determined
from sea-level variability through equation (3.3.3)
represents the uniform horizontal velocity of the
entire water column. For the baroclinic mode, sealevel change is related to the change in the height
of the subsurface constant-density surfaces, the
isopycnals. Therefore, sea level has been used to
estimate subsurface density field, from which horizontal geostrophic velocity at depths has also been
estimated.
Carnes et al. (1990) used GEOSAT altimeter
data to estimate the subsurface temperature field
in the Gulf Stream area. They applied statistical
regression analysis to a set of AXBT (Air-dropped
expendable bathythermograph) data deployed along
the GEOSAT ground tracks and derived an empirical relation between subsurface temperature and
surface dynamic height. They then applied the
empirical relation to the GEOSAT altimeter data
for estimating the subsurface temperature field.
They used the ocean topography derived from the
altimeter data relative to a geoid model as a surrogate for the dynamic height. The rms difference
between the GEOSAT-derived ocean topography
and the dynamic height is 15–19 cm, mostly reflecting the geoid errors and the GEOSAT altimeter
measurement errors. The resulting temperature
estimates have an rms error of about 1°C below
200 m. The error increases to 2°C near the surface,
where there is significant change in temperature
due to the annual cycle of heat exchange with the
atmosphere. The thermally driven variability of
temperature as a function of depth does not have a
tight relation with the sea-level variability.
Using T/P data with a series of repeating transects (repeat every 3 months) of XBT and XCTD
(Expendable Conductivity and Temperature profiler) across the North Pacific Ocean (from Taiwan
through Guam to San Francisco), Gilson et al.
(1998) studied the relationship between altimetric
sea-level measurements and the subsurface temperature and currents. A total of 5 years of simultaneous in-situ and satellite data were analysed,
allowing the relation between the two data sets to
be examined over a wide range of spatial and temporal scales. The anomalies of the altimetric sea
level relative to a 5-year mean were compared
with the anomalies of the dynamic height computed from the in-situ data relative to the same
5-year mean. The altimeter data were interpolated
to the times and locations of the in-situ observations using an objective analysis scheme. The overall rms difference between the two anomaly fields
is 5.2 cm. At wavelengths longer than 500 km,
where lies 65% of the variance of the dynamic
height, the two are highly coherent (0.89) with an
rms difference of 3.5 cm. This difference is consistent with the measurement errors of T/P (Chelton
et al., 2000) and the errors in the dynamic height
estimates, plus the residual dynamic height variability below 800 m, which is the deepest level of
the XBT observations. At wavelengths shorter
than 500 km, the coarse spacing of the T/P ground
tracks led to the underestimation of the mesoscale
variability, which, however, was well sampled by
the in-situ data. The correlation between the two
decreased to 0.56 at these scales.
From analysing the XBT/XCTD data, Gilson
et al. (1998) found that, except for the annual
variations, the dynamic height anomalies were
largely caused by the vertical motion of the thermocline. The entire thermocline moved coherently in
the vertical, causing the temperature at the thermocline depths to be highly correlated with the
surface dynamic height but with an opposite sign.
The high correlation between the altimetric sealevel anomaly and the surface dynamic height
anomaly discussed above thus allows the use of
the former to make estimates of the subsurface
temperature anomaly. The correlation between the
dynamic height and the subsurface temperature
was derived after the annual cycle was removed
from both fields. This is because the annual temperature change in the extratropical regions is
primarily caused by the heat exchange with the
atmosphere instead of the vertical motion of the
isopycnals. The vertical distribution of the temperature change at the annual period is thus not
highly correlated with the surface dynamic height,
except for the tropics where the annual cycle is
wind-driven (see Section 3.3.3.5) and hence the
correlation between dynamic height and temperature is high. Therefore, Gilson et al. (1998) used
the altimeter data only to estimate the subsurface
temperature anomaly at non-annual time scales.
The temperature anomalies derived from the altimeter data were added to the mean and the annual
cycle determined from the in-situ data to obtain
3.3 Ocean Circulation and Variability from Satellite Altimetry
157
Fu