Content uploaded by Eckart Schumann
Author content
All content in this area was uploaded by Eckart Schumann on Jan 12, 2015
Content may be subject to copyright.
Journal of Marine Research, 53,
231-248,1995
Coastal sea surface temperature variability along the south
coast of South Africa and the relationship to regional
and global climate
by E. H. Schumannl, A. L. Cohen2 and M. R. Jury3
ABSTRACT
The southern coastline of South Africa is approximately zonal, with a wide (up to 270 km)
shelf region. Intense thermoclines are known to be established by insolation on the inner shelf
region during summer, upwelling is generated by easterly-component winds, and occasionally
Agulhas Current water can be advected close to the coast, particularly in the east. These
processes induce daily and seasonal fluctuations of coastal sea-surface temperature (SST), but
their influence over longer time scales (interannual) has not yet been tested. Here time series
of SST ranging from 12 to 31 years are examined for inter-relationships with local and regional
winds, and the southern oscillation index (SOI). The emphasis is on the summer period, and it
is found that the correlation between SST and major axis wind anomalies can be improved
substantially by considering the frequency of occurrence of winds above given thresholds.
Moreover, winds and SSTs are also correlated with the SOI, such that fewer easterly-
component winds are experienced at low phases (El NiAo) with consequent increases in
coastal SST, and correspondingly more easterly-component winds at high phases (La Nina)
result in decreased coastal SST; however, these relationships did not hold for a measuring site
within a large open bay area. Long-term trends are also established, with substantial increases
in SST (0.2SWdecade) in association with greater increases in air temperature (0.36”C/
decade).
1. Introduction
Tracking changes in sea-surface temperatures (SST) is particularly important to
the study of global climate change because much of the residual heat which drives
atmospheric anomalies is stored in the upper ocean; at the same time, it is also an
indicator of dynamic processes occurring in the ocean. Fortunately, SST is a
parameter which is relatively easy to measure at the coast, and reasonably long
records are available at some sites. As a consequence, statistics relating to inter-
annual variability can be established, and trends and inter-annual cycles connected
1. Department of Geology, Institute for Coastal Research, University of Port Elizabeth, P. 0. Box 1600,
6000 Port Elizabeth, South Africa.
2. Archaeometry Laboratory, University of Cape Town, Private Bag, 7700 Rondebosch, Cape Town,
South Africa.
3. Department of Oceanography, University of Cape Town, Private Bag, 7700 Rondebosch, Cape
Town, South Africa.
231
232 Journal of Marine Research
ALGOA BAY
PORT
Figure 1. Map of the south coast of South Africa, showing the position of the different sites
where measurements analyzed here were made. The inset map shows the position of
Humewood beach in Port Elizabeth (*) and the airport (0); the dashed region over the
Aguihas Bank covers the VOS shipping reports.
with regional and global climatic events such the El Nifio-Southern Oscillation
(ENSO) can be investigated.
SSTs along the south coast of South Africa are analyzed here in such a broader
context of climate. The coastline is approximately zonal, and is characterized by
several large eastward-facing bays (Fig. 1). The shelf region known as the Agulhas
Bank widens from about 50 km off Port Elizabeth in the east, to about 270 km in the
west.
A marked seasonal variability occurs in the temperature structure of the water
column over the Agulhas Bank. Intense seasonal thermoclines develop in summer
(Schumann and Beekman, 1984), with the warm surface layer maintained by
insolation and occasionally replenished by surface plumes of warm Agulhas Current
water, particularly on the narrower shelf to the east (Schumann and van Heerden,
1988; Lutjeharms et al., 1989; Goschen and Schumann, 1990). In winter a more
homogeneous storm-mixed structure is prevalent, though Swart and Largier (1987)
reported marginally colder bottom temperatures in summer. Recently the oceanog-
raphy of the Agulhas Bank has received renewed attention as a focal point of
fisheries inputs to the productive southern Benguela upwelling region to the west
(Boyd and Shillington, 1994).
In winter frontal perturbations of the circumpolar westerlies sweep eastward over
the coast, while in summer the southward movement of the westerly belt means that
19951
Schumann et al.: South African coastal SST
233
the sub-tropical anticyclones exert a greater influence (Preston-Whyte and Tyson,
1988). The weather systems also initiate the regular eastward passage of small,
shallow low pressure systems, termed coastal lows, which enhance the variability
(Gill, 1977). Schumann and Martin (1991) found that the major axis wind directions
lie approximately parallel to the orientation of the coast; at Port Elizabeth westerly
winds dominated throughout the year, though in summer the percentage of easterly
winds reached more than 40%. To the west at Cape Town the southeasterly winds
blew for more than 70% of the time during November to March, decreasing to 40%
in June, and intermediate conditions are likely to prevail along other sections of the
south coast.
Upwelling-favorable winds are those from the east (Schumann
et al.,
1982; 1988)
associated with the southeastward ridging of the South Atlantic anticyclone; as a
result, abrupt changes in sea temperature of more than 10°C within a few hours have
been registered at coastal sites. A pulsed wind regime is driven by the continual
passage of upper level Rossby wave trains (Jury
et al.,
1992), so upwelling events tend
to be intermittent and short lived. The upwelling is initiated at the prominent capes,
and then moves westwards. Goschen and Schumann (1995) showed that there were
marked differences in the response to wind forcing within the bays compared with
sites along the open south coast, and in fact the two systems acted out of phase to
each other. Beckley (1983; 1988) analyzed sea temperatures around Algoa Bay over
one-year periods, and found considerably spatial variability.
Various studies have investigated interannual variability on the shelf zone of South
Africa, in particular in the west coast upwelling region because of its commercial
importance. Shannon
et al.
(1986) analyzed wind and temperature records, and
concluded that there is a South Atlantic equivalent of the Pacific El Nirio, but that
the events are less pronounced and less frequent. On the other hand, Walker
et al.
(1984) showed that, since 1950, southern Benguela warm events in the southeast
Atlantic did correspond temporally to the Pacific El Nifio: events in both oceans
peaked in the southern summer and the inter-ocean link appeared to be via the
northward migration of the South Atlantic Anticyclone (SAA) and the westerly wind
belt. A subsequent, independent analysis of nearshore SST in the southern Benguela
also revealed a relationship between locally high SST and Pacific warm events
(Taunton-Clark and Kamstra, 1988). The negative phase of the Southern Oscillation
Index (SOI), or El Nifio event, gave symptoms of an extended winter season
characterized by generally higher interior pressures and an absence of anticyclones
ridging south of the continent. Walker (1990) and Jury and Pathack (1993) demon-
strated that SST anomalies in the Agulhas region to the east and southeast of Africa
exert a control over summer rainfall of the adjacent interior. Schumann (1992) found
a transition in wind directions at Port Elizabeth and Cape Town following the
1982/83 El Nifio event, suggesting a climate change induced by that event.
This analysis utilizes data from three coastal sites over periods of up to 30 years, in
234 Journal of Marine Research
[53,2
order to investigate monthly, seasonal and longer-term variability. One site (Port
Elizabeth) lies within Algoa Bay, while the other two sites (Tsitsikamma and Knysna)
are at positions along straight sections of the coast. As a comparison, data from
voluntary observing ships (VOS) are also used to gain an understanding of conditions
over the broader Agulhas Bank (Fig. 1).
2. Data availability and processing
Daily sea temperatures have been measured at Humewood beach in Port Eliza-
beth since 1962 (Fig. 1); these measurements were made at about 09:OO local time,
and were read on a thermometer to the nearest 1°C. Starting in 1972, daily,
mid-morning measurements have been made at a well-exposed site near the Knysna
estuary mouth using a standard thermometer with an accuracy of 0.5”C. On the other
hand, the Tsitsikamma measurements were made in a well-flushed harbor area on
the inshore side of a small rocky promontory with an accuracy of O.l”C; the series
covers the period 1978 to 1989.
The analysis here concentrates on longer-term variability, with months being the
shortest time period considered. There were some gaps in the data, with two months
missing in October, 1980, at Tsitsikamma, almost one month in April, 1973, at
Knysna, and a few other breaks of less than 10 days; these were replaced by linear
interpolations. Coastal temperatures can exhibit significant diurnal variability, how-
ever, none of the sites is particularly protected, and by taking the measurements
consistently at one time in the day such diurnal fluctuations should not affect the
longer-term characteristics.
In order to determine the seasonal variability, monthly means and their standard
deviations were calculated over one calendar year using all the daily data. For the
longer term trends, running yearly means were calculated from filtered data provid-
ing 36 values per year, i.e. a data point about every ten days. Normalized departures
(or anomalies) were determined by subtracting the historical mean from the monthly
values and dividing by the historical standard deviation.
Voluntary observing ship (VOS) data were obtained from ships in the block
indicated in Figure 1 over the years 1960 to 1992; a total of 8.5289 observations were
available, with the smallest number per month being 6533 during April. The area has
been chosen on the inner shelf to minimize the direct influence of the Agulhas
Current, though at times plumes of Agulhas surface water can extend over much of
the eastern shelf region (Lutjeharms
et al.,
1989). There are problems associated with
such data in terms of accuracy and the distribution of stations (Taunton-Clark and
Shannon, 1988) but in this case monthly averages were calculated over the whole
area to give a general picture of long-term variability. Yearly seasonal averages and
standard deviations were also determined.
Hourly means of wind speed and direction were obtained over the same time
period from the Port Elizabeth airport (3 km inland and 62 m above mean sea level).
19951 Schumann et al.: South African coastal SST
235
- Knysna
-- Tsitsikamma
G 24
0,
22
------Port
Elizabeth
I
I
I
I
/
1980
I
1985 1990
Year
Figure 2.
Time series of the average monthly temperatures at the three coastal sites; Knysna is
shown as a solid dark line, Tsitsikamma as a solid light line and Port Elizabeth as a dashed
line.
The measurements were made using a Dines anemograph at a height of between 10
and 15 m. Wind direction accuracy is given as lo”, and speed as 0.1 m/s, while wind
speeds less than 1 m/s are treated as calms. The series were filtered to remove the
diurnal signal, and principal axes were then determined for the wind direction using
the method of Kundu and Allen (1976). Taking into account the transition reported
by Schumann (1992) the major axis direction was finally assumed to lie along
70”True (or T, being the angle clockwise from true north).
Hourly air temperature data measured with mercury thermometers with an
accuracy of O.l”C were also obtained from the Port Elizabeth airport. These data
were subjected to the same filtering process as the wind data to remove the higher
frequencies signal and obtain daily values, and thereafter to obtain 36 values per
year. The same procedure as for SST was then adopted to obtain running yearly
means.
The SOI represents the sea-level air pressure difference between Darwin in
Australia and Tahiti standardized anomalies. The SO1 time series was taken from
the Climate Analysis Centre, NOAA (CAC, 1992). Regional wind anomalies for
selected late summer months were re-drawn from the CAC publications.
3. Seasonal variability
Time series of the mean monthly temperatures at the three coastal sites are shown
in Figure 2. The seasonal fluctuation is the dominant variability on this scale, and the
similarity between the different sites is evident. Nonetheless, in certain summers,
notably 1972/73 and 1988/89, the temperatures at both Knysna and Tsitsikamma did
not reach the normal summer highs; this will be investigated further below.
236
Journal of Marine Research
[53,2
144 : : : : : : : : : : : I
Jan Feb Mar Apr May Jun Jul Aug 5-p Ott NOV Oec
Month
Figure 3. Monthly annual variability at the four sites indicated for all the data available in
each case. The lighter lines on either side give the standard deviation about the mean.
Average monthly sea temperatures and their standard deviations are shown in
Figure 3 from the three coastal sites as well as the VOS data over the total recording
periods. The amplitudes of the seasonal signals agree with that reported by Taunton-
Clark and Shannon (1988) for the Agulhas Bank, but contrast with the Cape Town
region where sustained summer upwelling offsets the effect of increased insolation
(Taunton-Clark and Kamstra, 1988).
Table 1 gives details of the temperature variability. The summer maximum mean
temperatures all occur in January, with Tsitsikamma the lowest at 19.07”C and Port
Elizabeth the highest at 21.03”C. However, the greatest variability also occurs in the
summer months because of thermocline development, and here it is the south coast
sites of Tsitsikamma and Knysna which show the largest standard deviations. This is
to be expected, since colder water is more likely to upwell at the coast over the more
abrupt bathymetry off the south coast than in the shallower water of Algoa Bay.
Absolute maximum and minimum daily temperatures over the recording periods
are also given in Table 1. Of interest is the fact that the minimum temperatures at all
three coastal sites were measured in spring and autumn i.e. during those months
19951
Schumann et al.: South African coastal SST
237
Table 1. Temperature values extracted from the data available at each of the sites. The
maximum and minimum values are those from the mean monthly values shown in Figure 2b,
with the month in which it occurred given below; the standard deviation is given similarly.
The absolute maximum and minimum values refer to the daily temperatures over the whole
time series; note that at Port Elizabeth the temperature was only given to an accuracy of
1°C while absolute maxima and minima were not considered valid for the VOS data.
Knysna Tsitsikamma
Port
Elizabeth
Max Temp(“C)
Month
Min Temp (“C)
Month
Max Std Dev (“C)
Month
Abs Max (“C)
Month
Abs Min (“C)
Month
20.53
January
14.67
July
3.05
February
27.0
January
10.5
March
19.07
January
15.17
August
3.79
February
26.0
December
9.2
November
21.03
January
15.54
July
2.06
March
25.0
February
10.0
October
vos
20.74
January
16.31
August
1.94
April
-
when the thermocline was being established or broken down; this is in agreement
with the finding of Swart and Largier (1987) that the coldest bottom temperatures on
the Agulhas Bank do not occur in winter. However, as expected the maximum
temperatures occurred during summer, with the highest temperature of 27.O”C
recorded at Tsitsikamma in January; it is likely that this very high temperature was a
result of diurnal heating in the habor area during an unusually calm period.
Further offshore the absolute maximum and minimum temperatures have not
been given for the VOS data since it is known that large errors can occur in individual
measurements but that the large number of observations should ensure that such
imperfections will be evened out (Christensen, 1980). The trend in the mean is very
similar to the coastal sites, but the standard deviations are much less, indicating
much less variability.
4. Summer upwelling
It is known that, along the Cape south coast, intense thermoclines are established
over the Agulhas Bank in summer, upwelling is generated by easterly component
winds, and that easterly winds are more prevalent in summer than in winter. This is
reflected in the variability shown in Figure 3, and it is therefore worthwhile
investigating the summer SSTs further; in particular, also the relationship with wind
in order to ascertain some of the processes involved. For this representation, the
combined months of January and February will be taken to be typical of summer
conditions.
Various correlations were made involving wind and SST. In particular wind run
238 Journal of Marine Research
-6 -6
-4 -2 0 2 4 6 6
Wind Exceedance Value (m/s)
Figure 4. Correlations between monthly temperature anomalies over January/February at
Knysna (solid dark line), Tsitsikamma (solid light line) and Port Elizabeth (dashed line) and
the anomalies for the number of days per month (also over January/February) with winds at
Port Elizabeth greater than the given exceedance values.
(effectively the distance travelled considering the strength, direction and duration of
the wind) and percentage occurrence of easterly winds were tried, as well as for
westerly winds and both directions combined. However, none of these methods
provided significantly high correlations between the wind parameters and SST.
A calculation was then made to determine the number of days in each January/
February period where the easterly wind component exceeded 1 m/s, 2 m/s,
3
m/s. . . etc., since this should give a measure of the overall effect of the wind. The
mean number of such exceedance days was determined, and anomalies calculated for
each January/February period. Figure 4 shows the correlations between the anoma-
lies for these wind exceedance values and the anomalies in the SSTs at Knysna and
Tsitsikamma for each of the wind exceedance values. A high negative correlation
shows that an increase in the number of days with easterly component winds
exceeding the particular wind exceedance value led to a
decrease
in the mean SST
over the January/February period. Negative wind exceedance values denote westerly
component winds, but it is important to note that the correlation is between the
positive parameter
number of days,
and therefore the negative correlation with
stronger westerly winds means that these winds also led to a
decrease
in SST.
19951
Schumann et al.: South African coastal SST 239
Several conclusions can be drawn from these curves. Firstly, significant correla-
tions are obtained between average monthly sea temperatures anomalies at Knysna
and Tsitsikamma and the easterly component winds. The shape of the curves can also
give some indication of the dynamics associated with the changes in temperature.
The negative correlation shows that colder mean temperatures over January/
February are associated with increases in the frequency of easterly winds, i.e. more
upwelling occurs. However, peaks in the correlation are obtained with relatively low
wind values, lying between 1 and 2 m/s. The implication of these results is that
upwelling is initiated with relatively low winds, although not too much credence
should be placed on the absolute values, since the airport winds are likely to
underestimate the actual values occurring at sea (Hunter, 1982; Schumann
et al.,
1991).
The fact that the correlation then decreases with increasing wind speeds, means
that no further decreases in temperature were achieved with stronger winds, i.e.
upwelling had already progressed to the stage where the colder bottom water had
reached the coastal measuring sites. Schumann and Martin (1991) found that the
peak wind variability at the Port Elizabeth airport occurred at periods around 6 days,
though with a very flat spectral distribution. The implication is that the upwelling
process proceeded very rapidly after the onset of the easterly-component winds, and
Goschen and Schumann (1995) also found a coastal SST response within a day of a
wind change on the coast south of Algoa Bay.
On the other hand, the correlation with westerly winds is initially very low, showing
that downwelling did not move water of consistently different temperature to the
measuring sites. However, the increasing negative correlation as the westerly winds
increased in strength indicates decreasing temperatures, the implication being that
substantial mixing with water below the thermocline occurred, leading to the
temperature decrease at the coast.
For the Port Elizabeth temperatures a very low correlation occurs over the whole
wind speed range, indicating that the same processes were not operative within
Algoa Bay. Goschen and Schumann (1995) found an out of phase relationship
between winds and temperatures at a depth of 17 m at an offshore site in Algoa Bay,
and it is likely that the situation at Humewood beach was more complex and not
consistent.
5. Summer season interannual variability
The time series over the whole recording period (Fig. 2) reflect the seasonal
variations, but clearly there are also interannual fluctuations. This is particularly
evident in the differences between the situation in Algoa Bay and that along the
south coast. Thus during the summers of 1972/73 and 1988/89 the sea temperatures
at Knysna and Tsitsikamma remained low, while those in Algoa Bay followed the
240 Journal
of
Marine Research
P372
2.0
1.5
1.0 i
li
4 0.5~-
A
p o.o--
2
a
-0.5--
lY
: -l.O--
;=
p -1.5--
k
z
-2.o--
-2.5
t
-3.04 : : : : : : : : : : : : : : : : : : : : I
1975 1980 1905
1990
Year
Figure 5. Normalized anomalies over the January/February period of each year for the
Knysna temperatures (solid dark line), Tsitsikamma temperatures (solid light line), the
frequency of occurrence of Port Elizabeth 70”T easterly component winds greater than
1 m/s (dashed dark line), and the Southern Oscillation Index (dashed light line).
normal summer warming; on the other hand in other years the three sites reacted
very similarly, in particular during 1982/83.
With the importance of easterly winds in the summer upwelling season, the wind
exceedance values for Port Elizabeth giving the maximum correlations with tempera-
ture variations at Knysna and Tsitsikamma were chosen from Figure 4, i.e., 1 m/s.
The normalized monthly departures for the January/February seasons are shown in
Figure 5, together with the normalized temperature anomalies from Knysna and
Tsitsikamma; at this stage the sea temperatures from Port Elizabeth have been
excluded because of the low correlations found in Figure 4.
As expected, a pattern emerges where lower sea temperatures are associated with
more easterly winds, while higher temperatures are associated with less easterly
winds. This does not imply that there were more westerly winds with higher
temperatures, since they do not necessarily follow with less easterly winds.
Also included in Figure 5 are the Southern Oscillation Index (SOI) normalized
anomalies for the January/February period, and again a distinct pattern is estab-
lished between the different parameters. When the SO1 was low (the El Nina
situation) the frequency of easterly winds decreased, and correspondingly the coastal
sea temperatures increased. Quinn et al. (1987) identified the El Niiio of 1982/83 as a
very strong event, that of 1972/73 as strong, and those of 1976 and 1987 as moderate.
Figure 5 only considers the January/February situation, and while the 1982/83 event
reached its maximum during these two months, it peaked in the middle of 1972 and
1976 and had partially recovered by the subsequent January/February; this is
lYY5] Schumann et al.: South African coastal SST
241
therefore the reason for the apparent phase lag between the SO1 and winds in 1972,
and also why it does not appear as strong in the figure.
When the SO1 was high (the La Nina situation) the easterly winds increased, and
the sea temperatures decreased. The most dramatic case here is the summer of 1989,
when very low temperatures were measured at both Knysna and Tsitsikamma.
Overall, flucuations of the SOI, SSTs and the wind exceedances appear better
related after 1983.
Some of these events had counterparts in the southeast Atlantic. Offshore SSTs in
the southern Benguela were cooler than average during the early 1970’s (Shannon
et al.,
1988). During the summer of 1982-83, high coastal SSTs were recorded in the
Cape Peninsula area (Walker
et al.,
1984) as a result of a northward shift of westerly
winds (Gillooly and Walker, 1984). The global 1988-89 cool event has not been
recognized as having a great impact on the southern Benguela or Cape south coast,
although Hutchings (1992) referred to it in relation to fisheries productivity along the
south coast.
Correlation coefficients were calculated, with a value of 0.45 between the SO1 and
the Port Elizabeth wind exceedances, and -0.74 and -0.40 between the SOI and
Tsitsikamma and Knysna temperatures respectively; the high value at Tsitsikamma is
because the record excludes most of the poorer agreement region in the early
seventies. The correlation between wind and temperatures is higher, with values of
-0.79 at Tsitsikamma, and -0.67 at Knysna. As expected the Tsitsikamma and
Knysna temperatures follow each other closely, with a correlation of 0.95. Of course,
these correlations are essentially testing for linear relationships, and it is very
unlikely that such correspondences occur over the whole range of parameter values,
particularly for more extreme departures. Nonetheless, the correlation values ob-
tained confirm the general relationships shown in Figure 5. Of interest is the fact that
a correlation of -0.42 was also found between Knysna SST and the SO1 with a lag of
one year, indicating that changes in SST may anticipate the ENS0 cycle.
The regional wind fields for low and high phases of the SO1 can be investigated by
considering the low level 850 hPa wind anomalies over the recent ENS0 cycle of
1989-92. These circulation anomaly patterns are shown in Figure 6, and identify a
cyclonic (anticyclonic) gyre south of Africa in the 1992 El Nina (1989 La Nina). The
coastal symptoms of the regional circulation patterns can now be seen in context. The
gyres are consistent with the size of transient anticyclones, about 4000 km longitudi-
nally, and 2000 km latitudinally. The ENSO-driven circulation anomalies are con-
fined to the longitudes south of Africa and do not represent a wholesale retreat or
advance of the westerly wind belt. Rather the gyres owe their presence to a peculiar
regional response to the global ENSO, primarily through linkages between the
tropics and mid-latitude circulation systems; a discussion of this is beyond the scope
of this paper.
242
Journal of Marine Research
[53,2
+<\Y
c,F--!LA%
Y/J-
-%%
II
%\\
11
%
I
I t
0’ 20-E
40-E 60PE
Figure 6. Regional low level wind anomaly vectors at 850 hPa for examples of El Nifio (1992,
top) and La Niiia (1989, bottom) summers (from CAC bulletins). Vectors with speeds less
than 3 m/s are omitted, while the largest vector is 6 m/s. The small square on the south coast
shows the study region.
6. Long-term trends
The sea temperature
record at Port Elizabeth extends over a 31 year period, that
at Knysna over 21 years, 11 years-at Tsitsikamma and the VOS record is 33 years
long. These records are used here to investigate long-term trends. Moreover, the air
temperature record at the Port Elizabeth airport is also used for comparison
purposes.
Figure 7 shows the yearly running means for the various records. The interannual
variability discussed earlier is evident, in particular the warming associated with the
1976/77 and 1982/83 ENS0 events occurred at all the sites, even in the air
temperatures. The subsequent cooling was also registered in all the SST measure-
ments, as was the cooling in 1973 and 1989, though with an apparent lag at Port
Elizabeth.
Longer-term variability is also evident,
and Table 2 gives the correlation with time
of the temperature trends. Clearly Knysna and Tsitsikamma display greater variabil-
ity, but the records are also too short for any meaningful correlation to be estab-
lished. For the Port Elizabeth record a temperature increase of over 0.25’Wdecade
is found; the VOS increase is less, though with a lower correlation coefficient.
Schumann et al.: South African coastal SST
243
Year
Figure 7. Yearly running means of the Tsitsikamma, Knysna, Port Elizabeth and VOS sea
temperatures, as well as the Port Elizabeth air temperatures. Interannual variability as well
as long-term trends are evident.
A substantial and significant air temperature increase is also found for the Port
Elizabeth airport data. This amounts to 0.36”C/decade, in rough agreement with the
O.S”C/decade found by Jones (1988) for the southern tip of Africa over the period
1967 to 1986. Miihlenbruch-Tegen (1992) also found an increase of 1.7”C in Port
Elizabeth’s air temperature over a fifty year period, though this was not reflected in
sites in the interior of the country.
7. Discussion
This analysis has sought to establish patterns and trends in SST variability along a
section of the South African south coast. Variability has been analyzed over monthly
and longer periods, although it is the processes taking place at periods of a day and
less that affect these longer-term variations.
Table 2. Correlations with time of the sea temperature (“C) series shown in Figure 7; the last
column denotes air temperatures (“C). Where the correlation is low for the first two sites,
the temperature increase per decade is not given.
Knysna Tsitsikamma
Corr. Coeff. 0.103 -0.130
“C/Decade -
-
Port
Elizabeth
0.49
0.258
vos
0.36
0.113
PE Air
Temp
0.82
0.361
244
Journal
of
Marine Research
[53,2
The establishment of the summer thermocline and the concurrent upwelling
driven by easterly component winds dominates the temperature variability at the
coastal sites. There have been some measurements of the mixed layer depth reported
e.g. Beckley (1983); Schumann and Beekman (1984) with values generally between
10 and 20 m. However, no means have been established, but comparison with results
from elsewhere can give an indication of the expected characteristics. Thus the
analysis of Lentz (1992) shows that the mean mixed layer depths at four upwelling
sites (Oregon, California, Northwest Africa and Peru) were all less than 20 m. Off
Oregon this depth was as little as 1.9 m, though considerable variability occurred
about these mean values.
The classical cross-shelf Ekman transport U, is given by
where
T,
is the alongshore surface wind stress generally taken to be proportional to
the square of the wind speed, p the water density and
f
the Coriolis parameter. For a
wind of 3 m/s and an Ekman depth of 2 m, at a latitude of 34s the offshore velocity is
about 7 cm/s. However, Lentz also established that a significant proportion of the
wind-driven transport occurs below the surface mixed layer, and in practice the
velocity should then be even less than 7 cm/s. With such small velocities it may be
expected that it would take some time for an upwelling front to break the surface,
however, most reported investigations have not had temperature measurements at
the coast itself, and in an idealized situation even with such low velocities a front
could move offshore by a few hundred meters in a few hours. Certainly Goschen and
Schumann (1995) found abrupt temperature changes occurring at the coast within a
few hours of the onset of easterly winds with speeds of 10 m/s and less.
A further aspect to be considered is the time taken to establish the Ekman layer,
according to Brink (1983) about half an inertial period. At 34s this is about 10 or 11
hours, but it appears from the results of Goschen and Schumann (1995) that the
response can be faster in terms of a drop in SST at the coast.
It is not known how representative the winds measured at the airport in Port
Elizabeth are of the winds offshore of Tsitsikamma and Knysna, both because of the
effect of the land-based site, and because of changes in the weather systems in their
propagation eastward along the coast. Thus Schumann and Martin (1991) found that
the easterly component winds in summer were much stronger at Cape Town, and
also blew for longer periods. Knysna and Tsitsikamma can be expected to have
intermediate characteristics, though probably more akin to Port Elizabeth.
Evidence from aircraft surveys indicates that shelf winds are double those mea-
sured over land, and create a situation of cyclonic vorticity of order 10e4 s-l
conducive to upwelling (Jury, 1994). Moreover, the winds given in the figure are
filtered to daily values, i.e. the maximum speeds could have been considerably
19951 Schumann et al.: South Aj?ican coastal SST 245
higher, and then the effect of the square of the wind speed given in (1) could be
significant. Nonetheless, the results would seem to indicate that the layer depths at
the coast were also small, possibly similar to the Oregon situation. Another factor
which could influence the marked response to the low upwelling-favorable winds is
the reasonably abrupt drop-off at the coast to a depth of around 60 to 80 m; Largier et
al. (1992) found that subsurface isotherms remain tilted upward toward the coast
throughout the upwelling season, thus further expediting the upwelling process.
Coastal trapped waves (CTWs) may also play a part in the upwelling response,
since such baroclinic fluctuations can result in substantial movements of the thermo-
cline (Gill and Clarke, 1974). Schumann and Brink (1990) found barotropic CTWs
with amplitudes of up to 0.5 m on the south coast, and Jury and Brundrit (1992)
showed evidence to suggest that CTWs are an important component of upwelling on
the west coast of South Africa. Quite clearly all these factors need to be investigated
further to establish the dominant processes operating in this south coast upwelling.
The poor correlation of coastal temperatures with westerly-component winds
implies that the downwelling process does not bring warmer water in to the coast
from further offshore. In particular, in the western regions the warmer waters of the
Agulhas Current are probably so far offshore that they would not play much of a part
in the coastal temperatures; on the other hand Agulhas waters are known to
penetrate into Algoa Bay on occasion (Schumann, 1987; Goschen and Schumann,
1990). Although the VOS data are not ideal for analyzing shorter-term variability, it
is likely that the sea surface temperatures on the wider Agulhas Bank do not vary as
dramatically as those at the coast since upwelling effects would not be felt as strongly;
in fact it is possible that in places increases in temperature would be more dramatic
on those occasions when warmer Agulhas waters reached the site.
The results have emphasized the fact that the sea temperatures within the large
bays react markedly different to winds than those on the southern side of the bays.
Goschen and Schumann (1995) have demonstrated this variability with direct
measurements of specific events, and the results shown here have investigated the
longer term consequences.
The analysis of interannual variability emphasizes global teleconnections, and
again it is the summer period where this is most apparent. A strong link has been
established with the SOI though it is of interest that Shannon et al. (1986) did not find
such directly-related El Nifio occurrences on the west coast. On the other hand, Jury
et al. (1992) attempted further cross-correlations using a summer rainfall index over
the central interior of South Africa, and found the summer SOI to be well correlated
with summer rainfall (Correlation coefficient = +0.64). Indications are that the
sub-tropical humid air mass which overlies northeastern South Africa and the
Agulhas Current region is swept southwestward in La Nifia summers by anticyclonic
weather conditions. This explains why increased coastal easterly winds at this time
are related to increased summer rainfall over the adjacent plateau. In contrast,
246
Journal of Marine Research
[53,2
during El Nirio periods increased westerly winds lead to reduced rainfall. Cohen and
Tyson (1995) constructed a record of sea temperature fluctuations from oxygen
isotope values in the shells of marine molluscs from a site between Knysna and
Tsitsikamma, and speculate that distinctive periods of warming and cooling over the
last 10000 years can be attributed to the same processes which cause interannual
variability in the region today.
In conclusion, it is apparent that the global ENS0 cycle is an important compo-
nent of climate variability along the South African south coast. It has many economic
spin-offs, from seven-fold fluctuations in maize yield over the interior plateau to
fisheries productivity. Thus the chokka catch appears to vary directly with the
temperature (Sauer
et al.,
1991), and the extended upwelling season in 1989 resulted
in a bumper chokka season, while the 1992 season saw quotas going unfilled.
Acknowledgments. This analysis was made possible by the kind assistance of several
organizations and individuals in making data available from the various sites. Thus the Knysna
temperatures were made by Mr. C. Douglas, the Tsitsikamma data were provided by Dr. Nick
Hanekom, the data at Humewood beach were provided by the Port Elizabeth Beach Office,
the VOS data by Dr. Marten Grtindlingh of the South African Data Centre for Oceanography
(SADCO), and the wind and air temperature data from the Port Elizabeth airport by the
Weather Bureau. Mrs. Judy Martin and Mrs. Elsie Sauti assisted with the initial editing and
processing of the data. Funding for the analysis was provided by the South African Foundation
for Research Development.
REFERENCES
Beckley, L. E. 1983. Sea-surface temperature variability around Cape Recife, South Africa,
S. Afr. J. Sci. 79, 436439.
~ 1988. Spatial and temporal variability in sea temperature in Algoa Bay, South Africa,
S. Afr. J. Sci., 84, 67-70.
Boyd, A. J. and F. A. Shillington. 1994. Physical forcing and circulation patterns on the
Agulhas Bank. S. Afr. J. Sci., 90, 114-120.
Brink, K. H. 1983. The near-surface dynamics of coastal upwelling. Prog. Oceanogr., 12,
223-257.
CAC, Climate Analysis Centre. 1992. Climate diagnostics Bulletin, NOAA, Washington, D.C.
(October).
Christensen, M. S. 1980. Sea-surface temperature charts for southern Africa, south of 26s
S. Afr. J. Sci. 76, 541-546.
Cohen, A. L. and P. D. Tyson. 1995. Sea temperature fluctuations during the Holocene on the
south coast of Africa: implications for terrestrial climate and rainfall. The Holocene (in
press).
Gill, A. E. 1977. Coastally trapped waves in the atmosphere. Quart. J. Met. Sot., 103, 431-440.
Gill, A. E. and A. J. Clarke. 1974. Wind-induced upwelling, coastal currents and sea-level
changes. Deep-Sea Res., 21, 325-345.
Gillooly, J. F. and N. D. Walker. 1984. Spatial and temporal behaviour of SST in the south
Atlantic. S. Afr. J. Sci., 80, 97-100.
Goschen, W. S. and E. H. Schumann. 1990. Agulhas current variability and inshore structures
off the Cape Province, South Africa. J. Geophys. Res., 95, 667-678.
19951
Schumann et al.: South African coastal SST 247
~ 1995. Upwelling off prominent headlands of the eastern Cape province, South Africa.
S. Afr. J. Mar. Sci. (in press).
Hunter, I. T. 1982. The coastal wind field