Temperature fluctuations on the western continental slopes of the Atlantic Ocean have been measured at three stations on the ocean bottom at depths from 1400 to 2100 m and from 17° to 35°N. All three stations were in or near the deep western boundary current. Daily readings were taken for four to six years. Maximum-to-minimum fluctuations of the isotherms were 800 m off Cape Hatteras, with average temperature of 3.80°C. Temperature appeared episode-driven, with minima consistently near 3.60°C and maxima near 4.10°C. Large swings in temperature had a duration of 10–30 days. Data from Antigua also appeared episode-driven but showed fewer large shifts in temperature, but the maximum-to-minimum fluctuations of the isotherms were 400 m. Hence interesting motions may exist in the seldom-studied region near the bend in the Antilles Islands chain. These two stations are vastly different from previously reported data from the Blake Plateau in which temperature increases are consistently more rapid than decreases and suggest frequent overflow currents. Annual oscillations were seen at the Antigua and Blake Plateau stations, but no semiannual effect was seen.
Stommel (1958) predicted a deep western boundary current (DWBC) below 1000-m depth in the North Atlantic Ocean. The DWBC transports water southward from the Labrador Sea and from overflow from the Norwegian Sea in the opposite direction to the Gulf Stream. Swallow and Worthington (1961) made the first observation of the DWBC and placed the level of no motion at 1900 m off the Blake Plateau. But several observers found that the Gulf Stream extends to the bottom, even in the deep ocean (Fuglister 1963). The question of how the currents cross was answered by Richardson (1977), saying that the DWBC goes under the Gulf Stream “except in brief current reversals.” However, the shallow part of the DWBC can sometimes cross the Gulf Stream: meanders of the Gulf Stream allow the DWBC to continue south between the continental slope and the Gulf Stream and then to (occasionally) push through the Gulf Stream (Bower and Hunt 2000).
Off the Blake Plateau deep horizontal motions are found (Riser, Freeland, and Rossby 1978): eddies with radius near 40 km and orbital speeds near 40 cm s−1. The fluctuations in the DWBC were soon found to be as great as the current itself (Mills and Rhines 1979; Lai 1984). The DWBC was shown to be continuous from Abaco (26.5°N) to Barbados (13°N) by Fine and Molinari (1988). Lee et al. (1990) give a detailed history of DWBC studies; east of Abaco they found no annual effect at any depth, but an “event-dominated record, with events occurring on the average every 100 days.” Lee et al. (1996) discovered that the DWBC meanders sideways; sometimes it touches the continental slope and sometimes it is over 120 km away from the slope.
The continental slope east of Antigua may be affected by variability of the North Brazil Current (NBC). Johns et al. (1990) found that circa 400-km eddies propagate westward from the NBC. But the DWBC is at 4300-m depth, far below the thermistors in this report (Johns et al. 1993). Time scales of 60–90 days are found below the surface, as well as “large anticyclonic eddies” that propagate northwestward (Johns et al. 1998).
Although many papers address fluctuations in horizontal currents in the deep ocean, rather few reports consider years of temperature fluctuations on the continental slopes. Northeast Pacific Ocean data off Cape Mendocino (Broek 1969a) showed sizable aperiodic declines in 1962, a fortnightly tide, three diurnal tidal components, and one semidiurnal tidal component. A strong periodicity at 49 days was identified with the Madden–Julian oscillation (Madden and Julian 1971, 1972, 1994. On the other hand, temperature fluctuations from 2000 m on the northwest Atlantic slope near the Blake Plateau show no spectral peaks at all (except possibly near 50 days periodicity), but increases in temperature are consistently more sudden than decreases (Broek 1969b). Tides are also found near 3400 m on the continental slope of the northeast Atlantic (Thorpe 1987). Temperature fluctuations on the continental slope off central Chile have been reported (Shaffer et al. 1999), and temperature fluctuations off Bermuda at depths to 2000 m have also been reported (Frankignoul 1981). Data from seven thermistors on the continental slope of the northeast Pacific showed a strong semiannual effect in 1962, various tidal oscillations, the 50-day Madden–Julian oscillation, and a temperature decline in March, when the surface current reverses (Broek 2000). The present paper is only the second report of long-duration temperature fluctuations on the western continental slope of the North Atlantic, the other being Broek (1969b).
Temperature fluctuations on the Hatteras Abyssal Plain are small, about 0.05°C, with long stretches (months) of nearly constant temperature (Brown et al. 1975). Temperature fluctuations on the abyssal plains in the eastern North Atlantic are so small as to be almost undetectable (Saunders and Cherriman 1983). Davis et al. (2003) have the longest temperature series of all, 40 years long, which show large fluctuations in the western North Atlantic (Labrador Sea) but nearly constant temperature at abyssal depths in the central North Atlantic and eastern North Pacific.
In general, temperature fluctuations are often very small at abyssal depths and greater near continental margins.
Temperature measurements were made by use of thermistors placed between pairs of wires in telephone cables that were lowered to the ocean bottom to depths of 1400–2000 m (Table 1). The thermistors were placed at three locations: east of Cape Hatteras, North Carolina, east of Eleuthera, Bahamas, and east of Antigua, West Indies. The Antigua experiment had two thermistors, 2 km apart, which recorded almost the same temperatures. Temperatures were recorded manually, by balancing a galvanometer once a day, usually at the same time of day. The thermistors were in the middle of a long telephone cable. The thermistors were not moored or buoyed, but were not necessarily on the bottom. If the local bottom was pockmarked or rough, the thermistor might be slightly above the bottom. Conversely, if the thermistor hit a soft bottom, it might be buried in sediment, or subsequent overflow currents or slumping might bury it. (Results presented below suggest that the Blake Plateau thermistor may have become buried.) Locations are given in Table 1 and Fig. 1.
Data from near Cape Hatteras were available for the six calendar years 1961–66. Data from the Blake Plateau—that is, Eleuthera, reported by Broek (1969b)—covered 1 January 1960–1 April 1966. Data from Antigua were available for the calendar years 1961–62 and 1965–66. Antigua data were taken on two thermistors separated horizontally by 2 km; these provided almost identical data. Occasional missing data were supplied by interpolation. A few badly inconsistent measurements were replaced with interpolated values. Plots of data reveal gradual temperature changes on top of large excursions in temperature (Figs. 2–4).
c. Spectrum analysis
Techniques of spectrum analysis were about the same as in earlier work (Broek 2000). The data were averaged by the S4S5S6 filter (summing four adjacent data, then five of the result, then six of that result), then decimated by keeping every fourth datum of the averaged data. Spectrum analyses were performed on the decimated data. Spectrum estimates have a 90% chance of being within a factor of 2.23 of the true value.
d. Tides and tidal aliasing
The periodicities and potentials of the tides were computed exhaustively by Doodson (1922) and by Wunsch (1967). The semiannual tide has a potential 6.3 times as great as that of the annual tide, but annual oscillations may be greatly increased by seasonal heating. The strongest semimonthly tides are at 13.660 and 13.633 days, or 0.0732 and 0.0734 cycles per day.
Sampling once per day introduces the possibility of aliases of the diurnal and semidiurnal tides. The semidiurnal and diurnal tides are expected to produce their strongest alias peaks at 14.765 and 14.191 days, or 0.0677 and 0.0705 cycles per day. Diurnal tides can alias to one cycle per year, and the semidiurnal tide can alias to two cycles per year.
a. Temperature data
Daily temperature data from the Cape Hatteras station are shown in Fig. 2. Averaged temperature from the Cape Hatteras station is shown in Fig. 3. Small fluctuations are superimposed on occasional large sudden processes. Excursions to low temperature consistently reach a minimum near 3.60°C while upward swings peak near 4.00°–4.08°C.
The Blake Plateau data have a different appearance: “slow decreases in temperature (over seven days or more) followed by a much more rapid increase” (Broek 1969b).
Antigua data are similar to that that from Cape Hatteras, and both appear different from the Blake Plateau data. Averaged temperature data from Antigua for 1961, 1962, 1965, and 1966 are shown in Fig. 4. As at Cape Hatteras, small temperature fluctuations are overwhelmed by occasional large, sudden changes. Downswings usually reach a minimum at about 3.70°C, while upswings may peak between 3.94° and 4.08°C. February 1966 has a unique pattern: an upward step function followed by a damped oscillation.
The major departures in temperature have a duration of roughly 10–30 days at both stations, but the CapeHatteras station has more jagged curves (more large swings in temperature).
b. Temperature densities
The data from the Cape Hatteras station can be fitted by a Gaussian with mean of 3.80°C and standard deviation of 0.098°C. Depth is about 1400 m.
In the Blake Plateau data, temperature has an average of 3.28°C. Temperatures above average are fitted by a Gaussian with standard deviation of 0.07°C, but at low temperature the density lies above the Gaussian. On 10 occasions temperatures as low as 2.80°C were observed (Broek 1969b).
The data from the Antigua station are fitted by a Gaussian with mean of 3.82°C and standard deviation of 0.082°C. Depth was about 1700 m.
These standard deviations are only slightly larger than those observed on the continental slope of the northeast Pacific (Broek 2000), which were in the interval 0.067–0.097°C. The extent of variation, from lowest observed temperature to highest, was 0.508°C at the Cape Hatteras station and 0.404°C at the Antigua station. These ranges are similar to Pacific data, but temperature gradients versus depth are greater at the Atlantic stations. Temperature gradients derived from the atlas compiled by Fuglister (1960) imply that isotherms at the Cape Hatteras station can move roughly 400 m in either direction and 200 m at the Antigua station.
Data from the Antigua station suggest a possible semiannual effect only for 1962. Higher-than-average temperature prevailed around 10 March and 10 September, and low temperature around 10 January and 10 July.
Temperature densities at the Cape Hatteras and Antigua stations are not far from Gaussian. Scatterplots of temperature change in 8 days versus temperature at the middle of the 8-day interval gave only a random scatter pattern. Density of temperature change in 8 days is symmetric about zero temperature change, but falls off like an exponential, and not like the sum of two or more Gaussians. The median absolute value of temperature change at the Cape Hatteras station is 0.045°C versus only 0.026°C at the Antigua station. Likewise the 90% point: the absolute value of the 8-day temperature change at the Cape Hatteras station is above 0.127°C in only 10% of observations. The corresponding number at the Antigua station is 0.077°C.
c. Temperature spectra
A tidal peak at 14.8-days periodicity in the Antigua spectrum was also prominent in the Pacific data. This peak may be either a fortnightly tide or an alias of the largest tide (the M2 tide, two cycles per lunar day). The Blake Plateau spectra has a large annual peak and a possible broad 50-day peak (Broek 1969b).
Coherency between the two thermistors off Antigua separated by 2 km falls from unity at zero frequency to 0.6 at the Nyquist frequency (0.5 cycles per day), as shown in Fig. 6. The phase difference is usually less than 30°. No coherency is seen between any other pair of stations. In this paper coherency is defined by the root-mean-square formula.
a. Time domain
The DWBC, which is normally offshore from the continental slope, can meander toward the slope or away from it (Lee et al. 1996). When the DWBC is touching the continental slope, below-average temperatures would be expected. This process might explain the occasional appearance of very low temperature (non-Gaussian density) at the Blake Plateau station (Broek 1969b), but the station at the Blake Plateau shows “a tendency for slow decreases in temperature (over seven days or more) followed by a much more rapid increase.” Conversely a rapid increase may be followed by a slow decrease. Such an effect on the steep slopes around the Blake Plateau might be caused by episodes of overflow currents (often called turbidity currents). The duration of the observed uptrends in temperature is consistent with simulations of overflow currents (Jiang and Garwood 1996; Kaempf and Backhaus 1999). Possibly both overflow currents and DWBC meanders are in operation.
At the stations off Cape Hatteras and Antigua temperature increases as fast as it decreases, and no apparent overflow currents are seen. Off Cape Hatteras the fluctuations are larger and more frequent than off Antigua, possibly due to the head-on meeting and crossing of the Gulf Stream and the DWBC near Cape Hatteras. The Gulf Stream splits the DWBC in two (Richardson and Knauss 1971), and the part of the DWBC next to the continental slope slides down the slope to conserve potential vorticity (Hogg and Stommel 1985). This part of the DWBC occasionally pushes the Gulf Stream out of its way (Bower and Hunt 2000).
b. Frequency domain
No spectrum peaks were seen in data from the Cape Hatteras station, but two peaks were seen at the Antigua station. The peak at 14.8 days is likely to be an alias of the M2 tide, since the M2 tide is strong at the nearby Jungfern–Grappler Sill (MacCready et al. 1999).
Spectra from the Blake Plateau station showed a very large peak at one cycle per year, and a possible broad peak at 50-day periodicity (Broek 1969b).
c. Caribbean Deep Water
The Antigua stations at 1700-m depth are 300 km away from the Jungfern–Grappler Sill at 1815 m depth. This sill is a source of Caribbean Deep Water (MacCready et al., 1999). The episodic fluctuations in temperature have a superficial resemblance to episodic current fluctuations at the Jungfern–Grappler Sill, as reported in Figs. 4a and 20 of the paper by MacCready et al. (1999). Both processes appear to have upper and lower limits, like shifts between two quasi-stable states. Hence the temperature fluctuations off Antigua may be part of a large process that results in the episodic overflow of North Atlantic Deep Water into the Caribbean.
d. Sea surface height
Reversing the sign of the temperature fluctuations gives a quantity that is proportional to the height of the isotherms. This quantity can then be compared with sea surface height observed at nearby shore stations. No semiannual effect is found in the temperature data from the Atlantic stations, except that Antigua data have maxima in October 1961 and April and December 1962, close but not quite semiannual.
e. Comparison with northeast Pacific data
In the Pacific data (Fig. 1 of Broek 2000), the temperature had a minimum near 1 May and 1 December 1962. The spectra showed a semiannual peak, and the semiannual tide has a potential 6.3 times as great as that of the annual tide. But the tidal explanation is dubious for three reasons: the temperature minima are sharp and not sinusoidal; the minima are 7 months apart; and the “semiannual” process is not seen in most of the other years.
Sea surface height data from the Pacific (Fig. 2 of Chelton and Davis 1982) have maxima near February, May, and October 1962 from Unalaska to Los Angeles. Hence sea surface height fluctuations can resemble fluctuations in temperature on the continental slope (with reversed sign), but the resemblance may be coincidental. The semiannual fluctuations do not appear to be caused by an astronomical tide.
A tidal peak at 14.8 days periodicity in the Antigua spectrum was also prominent in the Pacific data. This peak may be either a fortnightly tide or an alias of the largest tide (the M2 tide, two cycles per lunar day). Analysis of the Pacific data advanced the fortnightly tide as the more likely explanation (Broek 1969a).
Isotherms at the Cape Hatteras station and at the Blake Plateau station can move roughly 400 m in either direction, and 200 m at the Antigua station. These motions are greater than the 86–160 m observed in the Pacific data (Broek 2000).
The station at 1400 m off Cape Hatteras is often in the DWBC, composed of Labrador Sea Water. The average temperature is 3.80°C with standard deviation of 0.098°C. Data range between 3.55° and 4.10°C. No periodicities are seen.
Data from off the Blake Plateau at 2000-m depth are vastly different (Broek 1969b). Temperature rises are consistently faster than declines. Average current is believed to be small, but occasionally the DWBC may move westward to the continental slope. Motion of the DWBC and overflow currents from the shelf is a possible explanation for the observed temperature fluctuations. Spectra show a large annual peak and a possible broad peak near 50-day periodicity.
The station near Antigua at 1700-m depth had average temperature of 3.82°C and standard deviation of 0.082°C. The data resemble those from the Cape Hatteras station except that the latter have more large fluctuations, possibly from the opposition between the Gulf Stream and the DWBC. Even though data from the Antigua station have less fluctuations, the similarity between data from the two stations (and the fact that their standard deviations are comparable) suggest that possible complex current fluctuations may exist near the “Barbuda right angle” in the Antilles chain.
The Antigua station is at nearly the same depth as the Jungfern–Grappler Sill (through which flows water that becomes Caribbean Deep Water). Temperature fluctuation episodes at the Antigua station show a superficial resemblance to current fluctuations reported by MacCready et al. (1999) at the Jungfern–Grappler Sill.
The only tide observed was a small 14.8-day spectrum peak from the Antigua station. In general, the time domain seems more instructive than the frequency domain. The data and the references point toward the existence of large episodic currents in the deep waters of the northwest Atlantic. These episodic curents may be related to the “periodic surges of cold bottom waters” reported by Johns et al. (1993), or they may be related to the “large anticyclonic eddies” that propagate northwestward from the North Brazil Current (Johns et al. 1998).
Temperature fluctuations on the continental slope often resemble (with reversed sign) fluctuations in sea surface height; however, the resemblances may be coincidental. “Semiannual” temperature fluctuations observed in 1962 are not attributable to astronomical tides.
Gas hydrates are found below the seafloors of the margins of all oceans, and also below permafrost (Mazurenko and Soloviev 2003). These hydrates often emit methane into the ocean in either gradual or violent releases (Judd 2003), but it is not known if such processes have a detectable effect on ocean-bottom temperatures.
I am grateful to Charles F. Wiebusch and Warren A. Tyrrell for suggesting this project. It is a pleasure to recall conversations with John C. Beckerle, R. H. (Nick) Nichols, and R. D. (Bob) Worley concerning ocean data. I thank J. A. Pecon for work on the instrumentation for this project. Thanks are given to Mrs. J. E. Larkin for the computer programming. I thank Christopher John Broek for a fine job of producing the illustrations. I acknowledge fine cooperation from the librarians at the University of Wisconsin, Nova Southeastern University, Northern Illinois University, and Florida Atlantic University.
Corresponding author address: Howard W. Broek, 57 White Oak Circle, St. Charles, IL 60174-4164. Email: firstname.lastname@example.org