were mapped onto standard grids, regularly spaced
in the vertical and horizontal, before differencing.
Pressure is the vertical coordinate used in the grid.
Figure 7.3.4 (see Plate 7.3.4, p. 588) summarizes
the calculated changes, averaged over the 1000–
2000 dbar depth interval, except at 48°N where the
maximum cooling signal was between 500 and
1500 dbar. The changes are partitioned into three
components: (a) the isobaric change, (b) the isopycnal change, and (c) changes in due to isopycnal
displacements. The components of the 1950s-minus1920s changes are calculated on the assumption that
the titrated Meteor salinities, although containing
large random errors, do not contain significant systematic errors (Mantyla, 1994).
The measurement errors of Meteor and IGY
temperatures (Fuglister, 1960), as well as those of
the more recent Conductivity-Temperature-Depth
(CTD) temperatures, are much smaller than the
changes reported here. Errors caused by sparser
sampling of the Meteor and IGY sections, and by
the non-coincidence of cruise tracks from different
decades, were investigated and found to be small.
For all of our sections, the largest source of error
in the changes between 1000 and 2000 dbar is
that due to mesoscale eddies. Following previous
work (Parrilla et al., 1994), we estimate eddy
errors by taking the standard deviation of the section-averaged signals and dividing by the square
root of the number of eddy diameters (taken to be
100 km) across the section. The eddy errors of the
vertically averaged isobaric changes are shown in
Figure 7.3.4a (see Plate 7.3.4a, p. 588).
Except at 48°N, the section changes show
an isobaric warming, of 0.5°C century
91 , between
1000 and 2000 dbar. This warming rate is consistent with the trend seen at the Bermuda time series
station (Roemmich and Wunsch, 1984; Joyce and
Robbins, 1996). Our map, with its comprehensive
geographical coverage, is a good complement to
the Bermuda time series, which samples only one
point in the ocean, but has excellent temporal coverage. Together, they argue convincingly that there
have been widespread changes in temperature at
intermediate depths over the entire Atlantic Ocean
since at least the 1920s. In the North Atlantic
between the 1950s and 1980s, the warmings are
mostly due to downward displacements of isopcynal (neutral) surfaces, while isopycnal change
dominates all but the most recent of the South
Atlantic signals.
Though depressions of isopycnal surfaces in the
North Atlantic may arise from a range of causes,
such coherent and large-scale changes as we see
here suggest that the cause may involve variations
in the formation rate of intermediate water masses
at source. Correlations between time series of the
source-thickness of Labrador Sea Water (LSW) –
the major water mass in the North Atlantic
between 1000 and 2000 dbar – and the temperature anomalies at Bermuda have led to the proposal (Curry et al., 1998) that the thickness of the
LSW layer is a major controlling factor in intermediate-depth temperature anomalies of the subtropical North Atlantic. We suggest that, additionally,
there may have been changes in the thickness of
the layers above or below the LSW intermediatelayer. If the intermediate-layer temperature changes
were due only to thickening and thinning of the
LSW, one might expect to see isopycnal displacements of opposite sign at the top and bottom of
the LSW layer – roughly at 1000 and 2000 dbar,
respectively. However, the 1980s-minus-1950s
differences on all of our North Atlantic sections
from 8 to 36°N show a consistent downward displacement of isopycnals of ϳ30–100 dbar (not
shown) throughout the entire extent of the 1000–
2000-dbar layer. This is consistent with either a
deepening thermocline or a shallowing of the
abyssal layer (Bacon, 1998). Our calculations indicate that the thermocline motions concurrent with
the wind- driven sea-level changes calculated by
Sturges et al. (1998) are not consistent with the
isopycnal displacements we observe. Thus, our
analysis does not support the idea that these changes
are simply due the effect of wind forcing on thermocline depth. Another possibility is that the thermocline has deepened due to changes in the amount of
fluid being pumped downwards from its surface
outcrops (Huang and Pedlosky, 1999), but we do
not yet have a quantitative measure of this effect.
In contrast to the North Atlantic warmings,
those in the South Atlantic are generally dominated
by isopycnal change, the 1990s-minus-1980s signal
at 11°S being an exception. This isopycnal warming extends north into the southern North Atlantic.
There are several sources for waters between 1000
and 2000 dbar: deep convection in the Labrador
Sea, overflows from the Mediterranean Sea, and
Antarctic Intermediate Water (AAIW). Although
the core of the AAIW layer is shallower, salinity
data (Bainbridge, 1981) suggest that AAIW
SECTION 7 INSIGHTS FOR THE FUTURE
566
in the vertical and horizontal, before differencing.
Pressure is the vertical coordinate used in the grid.
Figure 7.3.4 (see Plate 7.3.4, p. 588) summarizes
the calculated changes, averaged over the 1000–
2000 dbar depth interval, except at 48°N where the
maximum cooling signal was between 500 and
1500 dbar. The changes are partitioned into three
components: (a) the isobaric change, (b) the isopycnal change, and (c) changes in due to isopycnal
displacements. The components of the 1950s-minus1920s changes are calculated on the assumption that
the titrated Meteor salinities, although containing
large random errors, do not contain significant systematic errors (Mantyla, 1994).
The measurement errors of Meteor and IGY
temperatures (Fuglister, 1960), as well as those of
the more recent Conductivity-Temperature-Depth
(CTD) temperatures, are much smaller than the
changes reported here. Errors caused by sparser
sampling of the Meteor and IGY sections, and by
the non-coincidence of cruise tracks from different
decades, were investigated and found to be small.
For all of our sections, the largest source of error
in the changes between 1000 and 2000 dbar is
that due to mesoscale eddies. Following previous
work (Parrilla et al., 1994), we estimate eddy
errors by taking the standard deviation of the section-averaged signals and dividing by the square
root of the number of eddy diameters (taken to be
100 km) across the section. The eddy errors of the
vertically averaged isobaric changes are shown in
Figure 7.3.4a (see Plate 7.3.4a, p. 588).
Except at 48°N, the section changes show
an isobaric warming, of 0.5°C century
91 , between
1000 and 2000 dbar. This warming rate is consistent with the trend seen at the Bermuda time series
station (Roemmich and Wunsch, 1984; Joyce and
Robbins, 1996). Our map, with its comprehensive
geographical coverage, is a good complement to
the Bermuda time series, which samples only one
point in the ocean, but has excellent temporal coverage. Together, they argue convincingly that there
have been widespread changes in temperature at
intermediate depths over the entire Atlantic Ocean
since at least the 1920s. In the North Atlantic
between the 1950s and 1980s, the warmings are
mostly due to downward displacements of isopcynal (neutral) surfaces, while isopycnal change
dominates all but the most recent of the South
Atlantic signals.
Though depressions of isopycnal surfaces in the
North Atlantic may arise from a range of causes,
such coherent and large-scale changes as we see
here suggest that the cause may involve variations
in the formation rate of intermediate water masses
at source. Correlations between time series of the
source-thickness of Labrador Sea Water (LSW) –
the major water mass in the North Atlantic
between 1000 and 2000 dbar – and the temperature anomalies at Bermuda have led to the proposal (Curry et al., 1998) that the thickness of the
LSW layer is a major controlling factor in intermediate-depth temperature anomalies of the subtropical North Atlantic. We suggest that, additionally,
there may have been changes in the thickness of
the layers above or below the LSW intermediatelayer. If the intermediate-layer temperature changes
were due only to thickening and thinning of the
LSW, one might expect to see isopycnal displacements of opposite sign at the top and bottom of
the LSW layer – roughly at 1000 and 2000 dbar,
respectively. However, the 1980s-minus-1950s
differences on all of our North Atlantic sections
from 8 to 36°N show a consistent downward displacement of isopycnals of ϳ30–100 dbar (not
shown) throughout the entire extent of the 1000–
2000-dbar layer. This is consistent with either a
deepening thermocline or a shallowing of the
abyssal layer (Bacon, 1998). Our calculations indicate that the thermocline motions concurrent with
the wind- driven sea-level changes calculated by
Sturges et al. (1998) are not consistent with the
isopycnal displacements we observe. Thus, our
analysis does not support the idea that these changes
are simply due the effect of wind forcing on thermocline depth. Another possibility is that the thermocline has deepened due to changes in the amount of
fluid being pumped downwards from its surface
outcrops (Huang and Pedlosky, 1999), but we do
not yet have a quantitative measure of this effect.
In contrast to the North Atlantic warmings,
those in the South Atlantic are generally dominated
by isopycnal change, the 1990s-minus-1980s signal
at 11°S being an exception. This isopycnal warming extends north into the southern North Atlantic.
There are several sources for waters between 1000
and 2000 dbar: deep convection in the Labrador
Sea, overflows from the Mediterranean Sea, and
Antarctic Intermediate Water (AAIW). Although
the core of the AAIW layer is shallower, salinity
data (Bainbridge, 1981) suggest that AAIW
SECTION 7 INSIGHTS FOR THE FUTURE
566