Professional Documents
Culture Documents
Insensitivity of Subtropical Mode Water Characteristics To Meteorological Fluctuations
Insensitivity of Subtropical Mode Water Characteristics To Meteorological Fluctuations
BRUCE A. WARREN~
Abstraet--A striking feature of subtropical mode water ('18" water' in the North Atlantic) is the lack
of change in its temperature and salinity despite exposure to the atmosphere under varying winter
conditions. A local, one-dimensional model is developed for the determination of mode water
characteristics, which agrees well with observations for mean conditions, in both the North Atlantic
and the North Pacific. The model is then used to estimate effects of reasonable weather fluctuations,
and it is found that the expected temperature and salinity changes lie within their small observed
ranges of variation. Because of the great thickness of the mode water layer, and the brief time actually
available in any one winter to modify it, winters would have to be very much more severe than really
occur to make any great change in its characteristics.
1. INTRODUCTION
A PROMINENTfeature of the western subtropical North Atlantic and Pacific is a lens
of nearly homogeneous water lying above the main thermocline. Generally it lies
beneath a secondary near-surface thermocline (non-seasonal), but in late winter it
outcrops at the sea surface near the northern b o u n d ~ e s of the subtropical gyres
(Fig. 1). This layer has been called '18 ° water' in the North Atlantic (WORTHINGTON,
1959; SCHROEDER,STOMMEL,MENZBL and SUTCLIFFE, 1959), although its temperature
actually decreases eastward to less than 17-5°C in the longitudes of the Grand Banks;
and 'subtropical mode water' in the North Pacific, where its temperature is also about
18°C due south of Honshu, but is more nearly 16.5°C to the eastward, over the
greater part of its areal extent (MASUZAWA, 1969). Because o f the temperature
variation, and despite the sanction which usage has given to the overly precise term
'18 ° water', we shall use 'subtropical mode water' as a convenient general term for
both distinctive layers.
Definite estimates for the variability in time o f subtropical mode water character-
istics are not available for the North Pacific, but, for the North Atlantic, WORTHmOTON
(1959) indicates 4-0.3°C in temperature and 4-0-10~oo in salinity. It has seemed
remarkable that these characteristics should change so little despite the noticeable
variations in wintertime surface conditions. Indeed the most common interpretation
o f subtropical mode water (WORTHINGTON, 1959) has been that winter cooling effects
massive sinking of surface water, which subsequently spreads southward beneath
warmer surface water drawn northward in the sinking process. If so, then changes
in the driving conditions should impose changes in the temperature and salinity of
the water in the convection cell.
*Contribution No. 2699 from the Woods Hole Oceanographic Institution.
tWeeds Hole Oceanographic Institution, Woods Hole, Massachusetts.
1
2 BRUCE A. WARREN
200 --~ - 2 O0
400- ~ ~ 4OO
600- - 6OO
800 • ~ 1 ~ -800
O m
i' I i I I I I I I I I I I I
v°.o
I l I I I I I I I I t I
om
200,
400"
600-
800 - -800
Fig. 1. Composite meridional profiles of temperature (upper, *C) and salinity (lower, %°) illustrating
late winter conditions in the upper kilometer of the western North Atlantic. Atlantis Stas. 5874--5889,
9-14 April 1960; 40"00'N, 68"30'W to 33"00'N, 68"30'W. CrawfordStas. 312-325, 11-16 February
1958; 32"15'N, 64"40'W to 19"36'N, 68"04'W.
In this paper are presented some calculations concerning the effect of meteoro-
logical anomalies on mode water characteristics. The viewpoint adopted is that
subtropical mode water represents a local response of the ocean to surface heat
fluxes and thermocline geometry, and that its occurrence away from its 'formation'
region results from the general southward motion in the upper several hundred
meters of the subtropical North Atlantic and Pacific. This interpretation contrasts
with the large-scale convection cell proposed earlier, but the latter seems unsatis-
factory in that it requires northward surface flow in the subtropics during winter,
whereas current charts and dynamic topography seem to indicate a prevailing south-
ward motion. When we consider reasonable variations in winter atmospheric con-
ditions, we shall find that they can have only a negligible effect on the mode water
characteristics: in part, because the mode water layer is simply too thick to be affected
very much by the range of conditions which actually occur; and, in part, because most
of a winter's cooling must be spent on erasing the previous seasonal thermoeline,
leaving little time for modification of the mode water. This conclusion is not entirely
unexpected, but it seems worthwhile nonetheless to try to reconcile quantitatively the
observed steadiness of mode-water characteristics and the known variability of
weather conditions•
Before one can calculate the effects of anomalies, one requires a quantitative
model for the mean situation. Since none has been given before, and since the present
interpretation of the mode water differs from some earlier ideas, most of this paper
will be devoted to developing in detail those parts of the local one-dimensional model
mentioned above which are necessary for the anomaly calculations, and to testing
the model against mean conditions for the North Atlantic and North Pacific. At
Insensitivityof subtropical mode water characteristicsto meteorologicalfluctuations 3
the very end we shall tum to the question of 'severe' winters and their effect on the
mode water.
where dU/dt is the time-rate of change of the internal energy of the column, QR
is the short-wave solar radiation absorbed across the sea surface, Q8 is the net long-
wave radiation emitted from the sea surface, and Qs and Qe are the sensible and
evaporative heat flaxes from the ocean into the atmosphere. We next give empirical
formulae for the Q's, make some simplified representations of dU/dt in terms of
surface temperature, and then do initial-value problems to obtain the annual temper-
ature cycles and winter minimum temperatures.
Following Bt~RLIAND(1960) we express QR (cal cm -2 sec -1) as
QR = Q0[1 - (a + 0.38n)n](1 - ~) (2)
where Qo is the short-wave radiation incident through a clear atmosphere on the
Earth's surface, n is the fractional cloud cover over the sky, a is a coefficient which
ranges from 0.35--0.38 for the latitudes with which we shall be concerned, and
is the albedo of the water surface, taken as 0.07 from BUDYKO'S (1956) Table 6.
Mean monthly values of Qo for five-degree intervals of latitude are tabulated by
BERLIAND (1960), and quarterly maps of fractional cloud cover are included in the
Atlas of Climatic Charts of the Ocean (McDONALD, 1939).*
For the long-wave radiation under clear skies QB', we follow BUDYKO (1956)
and use an expression derived theoretically by BERLIAND and BERLIAND (1952):
QB' = s~04(0"39 -- 0.05e~) (3)
where s is the infra-red emissivity of the sea surface, for which we use an average
0.98 of rough-sea values given by SAUND~ (1968)¢, ~ is the Boltzmann constant
(1.356 x 10 -12 cal sec -1 cm -2 deg-4), 0 is the absolute temperature (°K) of the
sea surface, and e is the vapor pressure in mb at ship's level. It is worth noting that,
although various investigations are consistent with the above form for QB', they differ
concerning the actual numbers: mean values cited by BRUNT (1944) of several empirical
determinations of the coefficients in parentheses in (3) were half again as large as
given here. On the other hand, this version is in good agreement with more recent
*Surprisingly, in view of its age, this Atlas provides much readier access to usefd mean climato-
logical data than any more modem sources with which I am familiar.
1"The figure 0.98 applies to the infra-red window 800--1200em-x. For the integrated radiation
considered here, Saunders (personal communication) recommends his calculations (1968) of 0.990-
0.996, pertaining to the infra-red region 200--4000cm -1, as more appropriate. Since the difference
is slight, the present results have not been corrected for the discrepancy.
Insensitivityof subtropical mode water characteristics to meteorologicalfluctuations 5
Values of qa are less accessible than the other atmospheric parameters introduced,
but McDONALD 0938) includes quarterly-mean maps of wet-bulb depression.
These show only small seasonal variation, and can be combined with the monthly-
averaged air temperatures to yield monthly values of specific humidity.
With Q0, Ta, qa, ua, n regarded as given quantities, dU/dt in (1) is specified as a
function of time and surface temperature after substitution of formulas (2), (4), (5)
and (7), and (1) reduces to a form of Newtonian cooling:
dU
-- F(t) -- G(t)r. (8)
dt
It works out, with the climatological data to be considered below, that the time
dependence can be approximated closely by trigonometric functions:
dU
-- A -- B cos (tot + ~b) -- [C + D cos (tot + 0Q]T (9)
dt
where to = 1.99 x 10 -7 see- 1 is the angular frequency of the Earth's motion about
the sun, and 4, ~ are phases to be estimated from the data. The determination of
surface temperature by radiation and atmospheric conditions is now explicit.
We next require a representation of dU[dt in terms of T. Since the surface
temperature does not respond instantaneously to changes in atmospheric conditions,
the formation and destruction of the seasonal thermoeline cannot be neglected in
determining the late-winter minimum temperature. It is not necessary for present
purposes to model the seasonal thermocline in exact detail; all that is needed is some
sort of seasonal heat storage of about the right magnitude, specified by the right
surface temperature. Accordingly, we represent it during the heating period as a
layer having linear temperature decrease with depth, of thickness y ( T - Tw), where
Tw is the temperature of the isothermal layer at the end of the previous winter, and
y is the reciprocal of the average seasonal thermocline temperature gradient, as taken
from observations. Then during summer heating, until T reaches a maximum,
dU dT
dt -- pc~y(T -- Tw) d-'~ (10a)
where p is a mean water density and c~ a mean specific heat at constant pressure.
We model the destruction of the seasonal thermocline by the development of a
progressively thicker and cooler isothermal layer lying above the water with the
constant vertical temperature gradient, 1/y. (We ignore effects of salinity on hydro-
static stability). Therefore during the time when dT/dt < 0, until T falls below Tw,
dU dT
dt -- pc~y(Ts -- T) d-'7 (10b)
thermocline, and we choose the surface temperature TO so that the deeper values
match actual thermoeline temperatures. As implied earlier, values of To work out
to be much too high to be compatible with the annual heat balance; consequently, as
time progresses, the extrapolated main thermoeline gradually erodes away, with the
periodic appearance and disappearance of a seasonal thcrmocline (whose amplitude
increases with time), until the surface temperature cycle has fallen to a level where the
annual heat balance can be maintained without further secular change. We model
the erosion of the extrapolated main thermocline in the same way as the destruction
of the seasonal thermocline, and represent dU[dt during continued cooling after the
disappearance of a seasonal thermocline (i.e. when dT[dt < 0, but T is less than the
previous winter minimum temperature) as:
dU dT
d"-i = Pc~fl(To -- I") d"t " (10c)
3. NORTH A T L A N T I C
The formation region in the North Atlantic covers an extensive area, but appears
to center around 35°N, 60°W. Monthly values of the atmospheric and radiation
parameters, obtained from the sources already cited, are listed for this position in
8 BRUCE A. WARREN
Table 1. Mean monthly climatological data for the position 35°N, 60°W.
Qo (cal cm -~ see-1) n u,(kn) Ta('C) qa(g/kg) T,,('C)
J 4"07 X 10-3 0.65 19 16-4 9'6 19.1
F 5-32 X 10-3 0"65 19 15'7 9.1 18"3
M 6"71 X 10-3 0-60 16 16.1 9.4 18.1
A 7"97 X 10-a 0"60 16 17.4 10.3 18-7
M 8.84 X 10-3 0-60 16 20.1 12"3 20.4
J 9'12 X 10-a 0"50 12 23.0 14.7 22.9
J 8.97 X 10-3 0"50 12 25"4 17.1 25.2
A 8.17 X 10-3 0.50 12 26.2 18.0 26.2
S 6.90 x 10-3 0-55 14 25-2 16.8 25.7
O 5"59 x 10-3 0-55 14 22"8 14"4 24.1
N 4"34 x 10-3 0.55 14 20"1 12"1 22"1
D 3"70 X 10-3 0.65 19 17.7 10-5 20-3
dU
-- 0.66 x 10 -3 - - 7.15 x 10 -3 cos tot - - (1.071 + 0.271 cos tat) x 10-aT, (12)
dt
where it is u n d e r s t o o d that tot -----0 or an integral multiple o f 27r a t the end o f January.
As the initial state we a d o p t a thermocline with 15°C at a d e p t h o f 600 m, a n d
fl = 50 m / d e g (see Fig. 1); the initial surface t e m p e r a t u r e is then 27°C, o r T O = 5 °.
W e t a k e ~ = 15 m / d e g to represent the seasonal thermocline, as i n d i c a t e d by
SCI-IROED•R, STOMMEL, MENZEL a n d SUTCLIFFE (1959, Fig. 1); a n d c~ = 0"93 cal
g - 1 deg-1, p = 1-026 g/cm 8.
Starting at t = 0 (mid-winter), (12) was i n t e g r a t e d numerically until subsequent
winter m i n i m u m t e m p e r a t u r e s differed b y <0"01°C. * Fifteen years were required to
achieve this equilibrium state, suggesting a very sluggish response o f the system to
changes. I n the final limit cycle, the a n n u a l m e a n t e m p e r a t u r e was 21-8°C, the s u m m e r
m a x i m u m 25.9°C, a n d the winter m i n i m u m 18.1°C. As m a y be seen in Table 1,
*For computational stability, the integration was actually carried out for U = U(t); values of
T ~ T(U) were then calculated at each time-step.
Insensitivity of subtropical mode water characteristics to meteorological fluctuations 9
these figures compare very favorably with the observed annual mean of 21-8°C, summer
maximum of 26.2°C, and winter minimum of 18-1°C. The final depth of the isothermal
layer at the end of winter was 445 m, which also compares well with observation
(Fig. 1), but this follows automatically from getting the minimum temperature fight,
with the given initial temperature distribution.
Figure 2 is included to illustrate the slow development of the isothermal layer.
Even after more than seven years the minimum temperature is 0.2°C above its asympto-
tic value. One reason, of course, is that during most of the year the 'main thermocline'
water is shielded from the surface heat flux by the seasonal thermoclin¢. In the
eighth winter the main thermocline is exposed to atmospheric cooling for about one
month, but this has little effect because of the great thickness of the isothermal layer
above, which must also be cooled. The suggestion, obviously, is that short-term
meteorological fluctuations will not have much effect on the mode water, but we
reserve detailed calculations until Section 7.
In Fig. 3 the asymptotic annual surface temperature cycle is compared with the
observed temperature variation in the vicinity of 35°N, 60°W (plotted from the values
in Table 1). The amplitude agreement is good, as noted above, and the phase is also
24
22
20
~e - I | >2s | >2e
OM
_
I00
-- 2 0 0
-- 300
20'
-- 400
18 18
-- 500
26"(
24
22
20
t8 l - 26 ! !
--OM
100
-- 200
f 300
4O0
18 t8
[ 1 T [ T T r 1 T I- T - - 500
12.0 16.0 20.0 ;)4,0 x t 0 7 $e¢
Fig. 2. Plots of surface temperature (upper, "C) and temperature of the upper 500 m against time
(sec), calculated from climatological data for 35"N, 60"W, illustrating the gradual development of a
homogeneous layer from an initial linear distribution of temperature in the vertical. Arrows indicate
time at which seasonal thormocline vanishes.
10 BRUCE A. WARREN
30°C -
. . ."*'"""""..
28-
26-
f.... I///I "~.
."."g/ ~
24-
22-
\
20-
18
16
t4
I, 1 i i i i
Fig. 3. Calculated asymptotic annual surface temperature cycle at 35"N, 60*W (solid line); observed
surface temperature variation at 35"N, 60"W (dashed line); surface temperature variation imposed
by meteorological forcing at 35"N, 60*W if the ocean had zero heat capacity (dotted line).
fairly good at the end of winter; at the end of summer, however, the observed cycle
leads the calculated variation by about one month. Probably this is due to the very
simplified model of the seasonal thermocline: in nature, the seasonal thermocline
continues to deepen after the surface temperature has begun to fall (ScnROEDER,
STOMMEL, MENZELand SUTCLIFFE, 1959), whereas its depth has been held constant
here. Thus the internal energy in the water column (which the theoretical temperature
calculations reflect directly) actually continues to increase for a time in the ocean
after the surface temperature passes its maximum.
Also included in Fig. 3 is a plot of the surface temperature which would obtain
if the ocean had zero heat capacity and responded instantaneously to meteorological
variations; it is the solution to (9) with dU/dt =-- O. The amplitude is nearly twice
that of the other curves, indicating again how the formation and destruction of a
seasonal thermocline damp the annual temperature variation imposed on the ocean.
The mean annual temperature here is about 0.4°C higher than the values cited pre-
viously, resulting from the correlation between T and G(t) in (8).
Because of the evident role of seasonal heat storage in determining the temperature
of the mode water in the model, it is well to verify that error in our representation of
the seasonal thermocline would not invalidate the results. The numerical integration
described above was therefore re-run for y = 20 m/deg (deeper thermoeline) and
~, = 10 m/deg (shallower thermocline). For y = 20 m/deg the asymptotic winter
Insensitivityof subtropical mode water characteristicsto meteorologicalfluctuations 11
minimum temperature was 18.4°C and the summer maximum was 25.4°C. For
), = 10 m/deg, the corresponding values were 17"7°C and 26.6°C. Thus misrepre-
sentation of the seasonal thermocline in the model would probably not affect the
calculation of the mode water temperature by more than a few tenths of a degree.
It is only in a narrow band of latitudes that a deep isothermal layer of water
extends from the sea surface to the main thermocline in late winter (Fig. 1). Therefore
it is also well to verify that the model does not permit such a layer elsewhere too.
We can do so by estimating winter minimum temperatures, from which the depth
of the isothermal layer can be inferred directly, at other latitudes along Long. 60°W.
Since great accuracy is not required here, we can dispense with the detailed numerical
integration, and use the asymptotic winter minimum temperatures Twz given by
the linearized, constant-coefficient approximation to (9). It follows from (11) that
TWL is given as:
Similarly, we need not trouble to calculate F(t) and G(t) in (8) for every month of
the year to obtain values for A, B, and C in (9). Instead we calculate expressions of
the form (8) for the extreme months of the year, February and August, from the data
sources previously cited, and estimate A as the average of the two F values, B as
half the difference between the F values, and C as the average of the two G values.
(D, of course, is neglected in this approximation). The February and August climato-
logical data for Long. 60°W, Lats. 20, 25, 30, 35, and 40°N are listed in Table 2
(sources as above), together with the calculated values of A, B, C, T*, and TWL.
Also included is the parameter a which appears in formula (2), as listed by B~RLIAND
(1960), and the parameter c of formula (4) (BuoYKO, 1956, Table 8).
At Lat. 35°N the approximate value Twn is lower than the exact winter minimum
previously calculated, but only by 0.9°C. About half of this difference comes from
neglecting D in the expression for G(t) [actually T and G(t) are highly correlated],
the rest from the linearization. It appears that at 20°N and 25°N the winter minima
are too high to allow any isothermal layer at all, if referred to the extrapolated thermo-
cline which was used as the initial state for the numerical integrations, while at
30°N, the minimum temperature would require a layer only about 200 m thick. At
40°N the winter minimum is low enough to demand a very deep isothermal layer if
referred to the initial state above, but 40°N is north of the Gulf Stream, where the
main thermocline is some 600 m shallower than in the Sargasso Sea, and the previous
initial state is therefore irrelevant here. (The value of TwL at 40°N is actually in fair
agreement with observed winter surface temperatures in the slope water). Thus it is
only between 30°N, say, and the Gulf Stream that one should expect to find the deep
homogeneous layer in late winter. (As will be discussed in Section 6, a heat balance of
the form of (1) is probably not so applicable south of 30°N as at 35°N, but this broad
conclusion should remain valid).
Use of (13) also provides a convenient way of estimating the uncertainty in the
minimum-temperature calculation which is associated with uncertainty in the heat
flux formulas. The major ambiguities, as noted in Section 2, concern the coefficient K
in (6), for which BUDYKO (1956) cites a value 2"4 X 10-6 g/cma (50 ~o bigger than
Table 2. Mean February and August climatological data for Lats. 20--40°N along Long. 6 0 ° W , including derived quantities
pertaining to sea-surface temperature (see text). Units of.4 and B are cal c m - 2 s e c - 1, and of C, cal c m - 2 s e c - 1 d e g - 1.
Qo n u, Ta q,~ a c A B C T* TWL
(cal c m -2 see-1) (kt) (*C) (g/kg) ( × 10a) ( x 10a) ( × 10a) ('C) ('C)
that used here), and the numerical terms in (3), which BRUNT(1944) gives alternatively
as;
Q ' B = stT04(0 " 5 6 - - 0 " 0 S e t ) •
We have therefore recalculated the February and August values of F(t) and G(t)
at 35°N, 60°W for these different expressions for Qs and Qs. The constants A, B,
C were then A = --2.17 x I0 -a, B = 9.18 x 10-a, and C = 1.460 x 10-a; from
(13), the temperature amplitude comes out to be 4.6°C, and the winter minimum
temperature, 15.9°C, or 1.3°C lower than calculated previously (Table 2). About
0.2°C derives from change in the formula for QB, the remaining 1.1 ° from change in
Qs. Thus uncertainty in representation of the surface heat flux probably makes the
detailed calculation of 18-I°C for the winter minimum temperature questionable to
within 1°, although the good agreement with observation obtained from (12), not
only for the mean temperature but also for the annual range of temperature, suggests
both that the model is valid to first order and that the heat fluxes were fairly well
represented.
4. NORTH PACIFIC
We next apply the model to the temperature of subtropical mode water in the
North Pacific. The formation region has been less clearly identified here than in the
North Atlantic but from maps prepared by MASUZAWA0969, Fig. 6) and REID
(1969), it appears to lie between Longs. 140°E and 160°E, and between Lats. 30°N
and 35°N. Monthly climatological data for 32.5°N, 150°E are listed in Table 3.
They were obtained from the same sources, and in the same way, as those for the
North Atlantic, except that the air and sea-surface temperatures are averages of the
monthly means for the two five-degree squares enclosing this point (i.e. they are
average values for the rectangle 30-35°N, 145-155°E), and the Q0's are averages of
values at Lats. 30°N and 35°N.
The calculated monthly values of F(t) and G(t) in (8) showed the same phase
as those for the North Atlantic, and the least-squares fit for amplitudes gave the
following numerical expression for (9):
dU
-- --1.38 x l0 -a -- 7.06 x 10-a cos cot - (0.968 + 0.272cos o~t) x 10-aT, 04)
dt
where, again, ~ot = 0 or an integral multiple of 2~ at the end of January.
Table 3. Mean monthly climatological data for the position 32.5°N, 150°E.
Oo (cal c m - 2 s e c -1) n u.(kt) T~('C) q~(g/kg) T,o('C)
The main thermocline is sharper and shallower in the subtropical North Pacific
than in the North Atlantic• The reciprocal of the thermocline temperature gradient
/~ is about 40 m/deg just south of the Kuroshio, and the 15° isotherm is at about
400 m depth (MASUZAWA, 1969, Fig. 1); To accordingly was taken to be 3 ° (25°C).
With ~, = 15 m/deg again, successive winter minimum temperatures in the numerical
integration of (14) differed by < 0.01°C after ten years. At this point the winter
minimum was 17.3°C, the summer maximum 25.1°C, and the annual average tempera-
ture 21.0°C. The average agrees well with the observed value of 20.7°C (Table 3),
but the amplitude of the annual variation is 1o less than observed (Table 3). Part of
the discrepancy seems due to error in the seasonal thermocline: MASUZAWA'S(1969)
Fig. 1 suggests ~ = 10 m/deg rather than 15 m/deg. It is quite unclear why the
seasonal thermocline should have a different character in the North Pacific from that
in the North Atlantic, but the numerical integration was re-run anyway for the smaller
value of 7. Convergence was achieved after fifteen years, with a winter minimum
temperature of 16.8°C, summer maximum of 25.9°C, and annual average again of
21.0°C. The amplitude agreement is much better, though not quite so good as in the
North Atlantic calculations.
Curves like those in Fig. 3 are plotted for the North Pacific in Fig. 4. The
calculated temperature is based on 7 = 10 m/deg. Although the agreement is not
so good in either amplitude or phase as for the North Atlantic, probably it is still
good enough to take the local model seriously here too. With regard to the model,
30oc - ''',.
20-
26=
24 i
22 =
20-
.
i6 =
~4
Fig. 4. Calculated asymptotic annual surface temperature cycle at 32.5"N, 150"E (solid line);
observed surface temperature variation at 32.5"N, I50"E (dashed line); surface temperature variation
imposed by meteorological forcing at 32.5"N, 150"E if the ocean had zero heat capacity (dotted line).
Insensitivity of subtropical mode water characteristics to meteorological fluctuations 15
for the North Atlantic, and yield a similar result. With so much question about
precipitation and diffusivity values, there is little basis at this time for speculating as
to why mode water there has a lower salinity than mode water in the North Atlantic.
6. SOUTHWARD EXTENSION OF M O D E W A T E R
The areal distribution of subtropical mode water is not properly our subject
here, since we are mainly concerned with variability in its temperature and salinity.
Nevertheless, some comment seems warranted because the extension of mode water
southward from its formation region between the main thermocline and the secondary
near-surface thermocline (Fig. 1 ; MASUZAWA,1969, Fig. 1) has been taken to indicate
a very different process for forming mode water from the model advanced here.
It has been suggested (see Section 1) that this long tongue represents the lower limb
of a large-scale convection cell operating in late winter, in which intense cooling just
south of the Gulf Stream draws warm surface water northward from low latitudes,
to sink, and then spread southward between the main thermocline and the surface
water. This does not seem a necessary interpretation of Fig. l, and, in fact, inde-
pendent evidence for the large-scale northward surface flow in the Sargasso Sea does
not exist. Indeed, current charts and dynamic topographies suggest a prevailing
southward component of motion in the upper kilometer (say) of the Sargasso Sea,
presumably deriving from the Gulf Stream to the north.
It appears quite possible to interpret the temperature profile (Fig. 1) in terms of a
balance between vertical diffusion and general southward advection (with zonal
components) in the upper few hundred meters, with the upper boundary condition
given as the surface heat flux prescribed as the right-hand side of(8). The mean annual
surface heat input increases progressively to the south (Table 2) and as an initially
isothermal column of water standing above the main thermocline moves southward,
its surface temperature will rise and heat will be diffused downward, leading to
development of a secondary thermocline. In such a scheme, the depth of an isotherm
z would be given very roughly by (Ky/v)J, where K is the vertical diffusion coefficient, v
the southward velocity component, and y the distance south of the latitude where the
isotherm intersects the sea surface. For K = 1 cm2/sec and v = 1 cm/sec, say, and
y = 1000 km, then z = 100 m, which is approximately the depth of the secondary
thermocline illustrated in Fig. 1. We cannot pursue this interpretation any further:
it is mentioned here only to show that the southward-extending tongue of subtropical
mode water does not necessarily argue against the one-dimensional model for mode
water formation.
7. METEOROLOGICAL FLUCTUATIONS
We have described a simple scheme for the existence of subtropical mode water,
which appears to be roughly consistent with the general distribution of properties and
currents in the subtropical gyres of the North Atlantic and Pacific. We have also
developed numerical formulae which permit calculations of the characteristics of
mode water. The results for monthly-mean conditions in both oceans agree well
enough with observation to encourage the belief that these formulae will also give
realistic estimates for the effects of fluctuations in weather on the water-mass charac-
Insensitivityof subtropical mode water characteristicsto meteorologicalfluctuations 17
teristics. Speeitically, how much would a severe winter change the temperature and
salinity ?
The first problem here, of course, is to specify just what a 'severe' winter is.
If one considers the December-February average air temperatures at the ground for
all years on record in southeastern New England, orders them by average temperature,
and cuts the sequence into three groups each containing the same number of years,
one finds that the division between the 'cold' third and the median third lies at a
temperature which is only about 0.5°C lower than the average air temperature for
the period for all years (J. Chase, personal communication). Perhaps, then, one
might consider as 'severe' a winter at 35°N, 60°W where the average air temperature
for the six-month period November-April was 1.0°C below normal. Very little in-
formation is readily available concerning humidity variations, but for these rough
purposes probably one can regard the relative humidity as unchanging from year to
year, and estimate variations in specific humidityfrom the variations in air temperature
(i.e. with the anomaly AT,---- --1 ° we associate an anomaly Aq, of --0"5 g/kg).
For the wind speed, we shall arbitrarily consider an increase of 5 kt in the averages
as 'severe.' We may then use (9) to calculate the effects of these anomalies, persisting
for six months, November-April, on the winter surface temperature minimum.
We consider first, and in conjunction, the anomalies in air temperature and specific
humidity. These, being uniform in time, affect mainly the constant A in (9), which
reflects annual averages; they affect B very little, and C and D not at all. For our
purposes it will be sufficient only to estimate 'corrections' to A in (12). The annual
average wind speed ~,----15¼ kt----785 cm/sec at 35°N, 60°W (Table 1). Then
from (5), an anomaly AT, leads to a contribution to A of 4.16 × 10-Tfi,AT, =
--0.33 × 10 -3 calcm 2 secX; and from (6), an anomaly Aq, gives a contribution
KLa, Aq, = --0-37 × 10-8 cal cm -2 sec-L
Thus we replace A = --0.66 x 10-3 in (12) by A ---- --1.36 × 10-3, andre-run
the numerical integration, starting at the beginning of November in the asymptotic
temperature cycle (Fig. 3), and continuing until the temperature reaches its new
winter minimum. With the more intense cooling the seasonal thermocline is erased
3.8 months after starting (late February); the subtropical mode water is then exposed
to the atmosphere for 1.4 months (into early April), after which seasonal warming
sets in. The cooling of the mode water lowers its temperature by only 0"1°C.
For a 'mild' winter, the integration was run again with temperature and humidity
anomalies of opposite sign (but same magnitude). In this case, of course, the seasonal
thermocline was never entirely erased, and the mode water was never exposed to the
atmosphere. The surface temperature fell to 18.4°C after 4.8 months (late March),
after which it began rising.
It is less easy to tinker with (12) to 'correct' it for increases in u,, because both
Qs and Q~ vary directly with u,, and hence A, B, C, D, in (9) all depend on u,. Since,
to good accuracy, C and D vary directly with u,, and since Au,[ft, ~ 0.3, we adjust
C and D by increasing their values 30 % in (12). To adjust A and B we need to extract
the mean values and amplitudes of the T-independent parts of Qs and Qs. These
assume their maximum and minimum values in February and August, from which
values means and amplitudes may be approximated as: for Q~, mean of 2.80 × 10-3,
amplitude of 3"78 × 10-8 cal cm-2 sec-x; for Qs, mean of 0.74 × 10 -3, amplitude
of 1.82 × 10-3 cal era-2 see -x. To modify A and B for the 30y. increase in wind
1~ BRUCEA. WARREN
speed, we subtract 30 ~ of these means and amplitudes from the values given for
A and B in (12). The adjusted version of (12) is then:
dU
-- 1.72 x 10-3 -- 8.83 x 10-a cos cot -- (1.392 -I- 0-352 cos oJt) x 10-aT. (15)
dt
This expression was integrated numerically under the same conditions as the
integration for the anomalies in air temperature and specific humidity. The seasonal
thermocline was erased after 3.4 months (mid-February); during the next 1.7 months
(to early April), the temperature of the mode water decreased to 17.9°C, after which
seasonal warming began.
For a 30 ~o decrease in wind speed (a 'mild' winter), the seasonal thermocline
was again not entirely removed: the surface temperature decreased to 19.0°C after
5"0 months, and then rose again. The temperature of the mode water was unaffected
of course.
Thus the combined effect of the severe-winter anomalies is to reduce the mode-
water temperature by only 0.3°C, which lies within the observed variation cited by
WORTHINGTON (1959) of 4-0"3°C. In mild winters we expect no change in mode-
water temperature, except as resulting from vertical diffusion and advection, which
we have neglected in this model.
Increased wind speed and reduced specific humidity mean an enhanced evapora-
tion rate, which raises the salinity of the mode water. We make an overestimate of
the salinity change by considering only the effects of wind speed and humidity
anomalies on evaporation rate, and ignore the decrease in saturation specific humidity
at the sea surface, brought about by the decline in surface temperature. The annual-
mean difference qs(T) -- q,~ is about 3.2 g/kg. Decreasing q~ by 0.5 g/kg raises this
difference, and thereby the annual-mean evaporation rate, by 16~. The wind speed
anomaly introduces another increase to E, of 30 ~o, for a total increase in evaporation
rate of 51 ~o. Since in Section 5 we calculated the mean annual value of E to be
4.02x 10-rgcm-2sec -1, the increase A E = 2 . 0 5 X 10-rgcm-2sec -1. If the
precipitation rate stays constant, the increased salt flux down from the sea surface
is SoAE. At this rate, the additional mass of salt accumulated in the water column
during the six-month severe winter is 1-18 g/cm2; and if this amount of salt is mixed
uniformly through the upper 450 m while the mode water is exposed to the atmosphere
and is turning over under the enhanced cooling, then the salinity of the mode water
increases by < 0"03~oo,an increment which is well within the observed salinity varia-
ion of -I-0.10%o (WoRTrnNGTON, 1959).
During mild winters the sea-surface salt flux will be reduced, but only water
within the seasonal thermocline will be affected, since the convective overturn cannot
penetrate to the homogeneous layer.
Meteorological variability over the western North Pacific is probably comparable
to that over the North Atlantic, leading to similar effects on mode-water character-
istics there.
Thus we come finally to the conclusion that, within the terms of our model,
real fluctuations in weather are just not big enough to produce very noticeable varia-
tions in mode-water characteristics; and indeed, that the variations which we might
reasonably expect are actually within the small ranges observed. As we have seen,
even with prolonged intervals of increased cooling, most of this time must be consumed
Insensitivity of subtropical mode water characteristics to meteorologicalfluctuations 19