Long time series of bottom temperatures in the Southern Ocean are rare. The cDrake array with over 40 current- and pressure-recording inverted echo sounders, moored across Drake Passage to monitor the Antarctic Circumpolar Current (ACC) variability and transport, measured temperature at 1 and 50 m above the seafloor at depths ≥ 3500 m and at the southern continental margin. The 4-yr dataset provided an opportunity to examine the temporal and spatial scales of bottom temperature variability. High variability was observed; ranges were 0.5°–0.9°C in the northern passage and 0.3°–0.6°C in the southern passage. Standard deviations in the two regions were 0.1°–0.15°C and <0.05°C, respectively. Meandering of the ACC with its deep-reaching thermocline accounted for up to 50% of the observed bottom temperature variance. Northern passage temperatures, spaced less than 40 km apart, were correlated with each other, while those in the southern portion, separated by 60–70 km, were not. A gap in the West Scotia Ridge provided a deep passageway for cold water to reach the northern passage from the southern basin; an extreme event during February 2008 brought bottom waters with in situ temperatures below 0.38°C as far north as 57°S. Strong vertical temperature gradients between 1 and 50 m above the bottom occurred intermittently due to intrusions associated with deep eddy circulations arising beneath the meandering jet and to flow over steep topography, permitting the generation of internal waves. High variability in temperature on interannual time scales requires record lengths of 13–17 yr to estimate long-term trends reliably.
Our knowledge about the spatial distribution of seafloor and near-bottom temperature in Drake Passage comes primarily from hydrographic casts (e.g., Orsi et al. 1999; Martinson et al. 2008; Clarke et al. 2009; Meredith et al. 2011; Provost et al. 2011). In the deep sea, bottom temperatures progressively warm eastward around the continent from the Weddell Sea, with the warmest temperatures occurring on the western side of the Antarctic Peninsula and bottom temperatures also warm equatorward from the base of the continental slope (e.g., Orsi et al. 1999; Clarke et al. 2009). Topography influences bottom temperature. Bathymetric ridges provide partial blockages of the deep circulation, and deep sills set preferred pathways (Tarakanov 2012). Less well known is the temporal variability of near-bottom temperature, owing to the lack of sustained observations. Near the Antarctic Peninsula, bottom temperatures have been measured with moored instrumentation on the shallower shelf and slope portions of the continental margin (Meredith et al. 2003; Martinson and McKee 2012) or in deeper regions of the Weddell Sea (Gordon et al. 2010). More typically, however, deep-water mooring configurations place the deepest current meter and temperature measurements 100 m or more above the seafloor, outside the bottom boundary layer, where the movements of the deep thermocline contribute substantially to the variability (Nowlin et al. 1977; Pillsbury et al. 1979; Bryden 1979; Ferrari et al. 2012, 2013).
Southern Ocean deep and near-bottom temperatures vary on multiple time scales from decadal to seasonal to synoptic. Recent analysis of repeat hydrography primarily from the World Ocean Circulation Experiment (WOCE) hydrographic program indicates that deep waters (>3000 m) in the Southern Ocean have warmed between the 1990s and 2000s (e.g., Purkey and Johnson 2010; Kouketsu et al. 2011). Along the western Antarctic Peninsula, a long time series of gridded CTD measurements concentrated mainly on the shallow (<1000 m) shelf region revealed that heat content was increasing due to upwelling of warm Upper Circumpolar Deep Water (Martinson et al. 2008). These decadal trends are superimposed on shorter time-scale variability. Within Drake Passage, using multiple years of data from the World Ocean Circulation Experiment SR1b repeat section, Meredith et al. (2011) observed Antarctic Bottom Water temperature fluctuations on the order of a couple of tenths of a degree over a period of a few years. Shorter-term variability of the same order was also reported by Provost et al. (2011), who conducted two hydrographic sections across the passage three weeks apart. Rubython et al. (2001) reported on a 4-yr bottom temperature record measured by a bottom pressure recorder [Multi-Year Return Tide Level Equipment (MYRTLE)] moored at a depth of 3690 m at a site near the southern end of the WOCE SR1b repeat line. The record exhibited temperature changes of 0.02°–0.06°C at monthly periods and an overall range of 0.15°C. Notable in the time series was a 0.1°C drop in temperature that persisted for about 1 yr. They linked this decrease to changes in the formation of deep and bottom waters in the Weddell Sea.
Time series of near-bottom (less than 500 m above the bottom) temperatures from instruments moored in Drake Passage have been measured at the southern margin (e.g., Nowlin and Zenk 1988; Meredith et al. 2003; Heywood et al. 2007). In 1979, Nowlin and Zenk (1988) placed current meters closer to the bottom on the Antarctic shelf and slope than had been done on previous deployments, where they directly measured persistent westward flow on the continental slope. The 1-yr-long temperatures measured at 2700 m, 300 m above the bottom, were correlated with those measured higher up on the mooring but were uncorrelated with neighboring sites, indicating short horizontal scales. A 7-yr-long time series of bottom temperature measured at 1040 m on the continental slope north of Elephant Island (Meredith et al. 2003) exhibited a strong annual signal with coldest values occurring during or after the austral winter. Local convective events were manifested as large-amplitude cold spikes.
The cDrake experiment (Chereskin et al. 2009, 2012) conducted between 2007 and 2011 provided an opportunity to examine the spatial and time-dependent nature of bottom temperature variability in the deep portions of Drake Passage. Over 40 current- and pressure-recording inverted echo sounders (CPIES) in water depths of 3500–4500 m as well as at two sites on the Antarctic continental slope measured temperatures at 1 and 50 m above the seafloor (D-1 and D-50). The goals of cDrake were to quantify the transport and understand the dynamical balances of the Antarctic Circumpolar Current (ACC) in Drake Passage, and the temperatures were ancillary measurements used to process bottom pressure. Fortunately, deep hydrographic casts taken at each site provided the means to determine temperature offsets and thereby to calibrate the deep temperature measurements; this allowed observations from successive instrument deployments to be patched together to form continuous 4-yr-long time series at numerous sites spanning the deep passage. The observed temperatures exhibit a rich variety of phenomena over a full suite of temporal and spatial scales. Section 2 describes the temperature measurements and calibration procedures. The temperature records are presented and discussed in sections 3 and 4, respectively, with the main focus on the variability. The results are summarized in section 5.
2. Datasets and temperature calibration
From November 2007 to November 2011, over 40 CPIES were moored in Drake Passage as part of the cDrake experiment (Fig. 1). A CPIES is an inverted echo sounder and pressure sensor anchored 1 m above the seafloor with a current meter tethered 50 m overhead. To meet the cDrake goals to quantify transports, over 20 CPIES were deployed in a line spanning the passage between South America and Antarctica (C line), and to understand dynamical balances, additional CPIES were deployed in a 7 by 3 local dynamics array (LDA) in the energetic region between the Subantarctic Front (SAF) and Polar Front (PF). During the final year, five closely spaced CPIES (H array) straddled the base of the Shackleton Fracture Zone (SFZ) to resolve the deep flow near the steep topography. The CPIES were to remain undisturbed on the seafloor for the full observational period; however, instruments that developed problems were replaced as needed on annual cruises. As a result, 4-yr-long data records were obtained at nearly all of the original sites. Of these, 16 sites (solid triangles in Fig. 1) were occupied by a single instrument for the duration of the field program. Water depths at CPIES locations in the LDA and along the C line in the central deep passage (C03–C15) ranged between 3500 and 4500 m. Sites C16 and C17 at the southern continental slope were at depths of 2550 and 1280 m, respectively. The temperature data from C01 and C02 on the northern continental slope (depths of 500 and 1800 m) and C10 (a single-instrument site at 2540 m) on the crest of the SFZ will not be presented here.
Each CPIES records two independent temperature measurements. Both the Paroscientific Digiquartz pressure sensor (Paros) and the Aanderaa acoustic Doppler current sensor (DCS) provide temperature as ancillary measurements. The Paros pressure sensor is mounted inside the CPIES housing, so its temperature measurements lag the surrounding waters. It takes nearly 12 h after launch for the temperature inside the glass sphere to equilibrate within 0.001°C, and these initial measurements are discarded. Once equilibrated, the lag time is about 1 h. The DCS temperature sensor is located inside the current meter; because its housing is small and metal, temperatures rapidly equilibrate with the surroundings. While the absolute accuracy of the uncalibrated sensors is poor, their resolutions are good. The Paros resolution is 0.0002°C. The two DCS models used during cDrake have different resolutions; the resolution of model 3820, used almost exclusively in cDrake, is only 0.012°C, whereas the resolution of model 4390 is 2 × 10−5°C. The Paros and DCS temperature measurements were highly correlated; for the single-instrument sites, correlations between the hourly records exceeded 0.92, except for one site (C13) where the correlation was slightly lower (0.87).
Calibrating the temperature records allows the measurements made by consecutive instruments at one site to be joined together to form a nearly continuous time series. CTDs were taken at each CPIES, and fortuitously, most of the casts reached to within a few tens of meters of the seafloor, permitting the temperature measurements to be calibrated. In general, one CTD was taken at each site on each of five cruises; thus, nearly every temperature record could be calibrated. The hourly Paros and DCS temperature records from each instrument were individually calibrated to agree with the deepest CTD measurements, and the calibrated records agree to within 0.02°C. Calibrating in this manner allowed the two records to be used interchangeably, if necessary, to produce a single time series at each site. The calibrated temperatures were low-pass filtered using a fourth-order Butterworth filter with a 3-day cutoff period, passed forward and backward, and subsequently subsampled to daily intervals. Additionally, the calibrated records from the LDA were optimally interpolated (Bretherton et al. 1976) to produce daily bottom temperature maps. While the focus of this paper is on these daily records, in section 3f, we examine observed differences between the Paros temperatures (D-1) measured 1 m off the bottom and those measured 50 m above by the DCS (D-50) and highlight high-frequency signals apparent in hourly records.
Contemporaneous measurements made by CPIES provide context for the observed temperature variations. Measured acoustic travel times τ are used together with hydrographic lookup tables [gravest empirical mode (GEM) tables; Meinen and Watts 2000; Watts et al. 2001] to produce full-water-column profiles of temperature and salinity from which specific volume anomaly profiles can be determined (Firing et al. 2014; Chidichimo et al. 2014; Donohue et al. 2010). We use the τ records for two purposes: First, we obtain time series of temperature at 3500 dbar at each location to compare with the measurements. Firing et al. (2014) showed that temperatures at 2500 dbar estimated in this manner were well correlated (r > 0.84) with those measured at two nearby French (Ferrari et al. 2012) current meter moorings (squares near C03 and E02 in Fig. 1a). Similarly, temperatures at 2500 m estimated from τ and directly measured by the Paros sensor (the DCS temperature sensor failed) at the cDrake midpassage site C10 were also well correlated (r = 0.90). Second, we integrate the specific volume anomaly profiles divided by gravity to obtain sea surface height (SSH) estimates to determine the approximate path of the ACC. Here, we do not consider the contributions of ocean mass to SSH; since these contributions are generally less than 30% in Drake Passage (Donohue et al. 2016), their exclusion has minimal effect upon the nominal path estimation. In this study we show plan view maps of bottom pressures, currents, and SSH in the LDA produced by Firing et al. (2014) to elucidate the deep and upper flow fields in conjunction with the bottom temperatures.
Satellite-derived SSH maps provide wide areal views of the ACC fronts in Drake Passage. We use the daily, time-delayed, two-satellite anomaly fields and reference them with the 20-yr mean dynamic topography MDT_CNES-CLS13. The altimeter products were produced by Copernicus Marine and Environment Monitoring Service and by CLS Space Oceanography Division and distributed by AVISO with support from CNES (http://www.aviso.altimetry.fr/duacs/).
In section 4, we briefly compare the variability of the cDrake bottom temperatures to the variability measured under the Gulf Stream during the Synoptic Ocean Prediction Experiment (SYNOP; Shay et al. 1995; Watts et al. 1995) using pressure-sensor-equipped inverted echo sounder (PIES) and under the Kuroshio Extension during Kuroshio Extension System Study (KESS; Donohue et al. 2010; Tracey et al. 2012) using CPIES. The data from these two experiments are documented in several reports (Fields and Watts 1990, 1991; Qian et al. 1990; Kennelly et al. 2008). Data from a short mooring M03 (star in Fig. 1a) deployed 8 km inshore of C16 during cDrake from November 2007 to November 2009 are also discussed in section 3. M03 was equipped with two current meters positioned 100 and 300 m above the bottom (1900 m). The cDrake, KESS, and SYNOP datasets are available at the National Centers for Environmental Information.
a. Record-long means
Mean bottom temperatures vary both meridionally and zonally within Drake Passage. The record-long are sorted by latitude and color-coded by location in Fig. 1b, and the vertical bars indicate variability of one standard deviation. The majority of these averages were calculated from 4-yr-long records; the exceptions are 1-yr averages for the H array sites (green triangles) and six other locations with data gaps (1–3-yr averages at C05, C12, C18, C21, C23, and D02).
There are pronounced spatial differences in across the passage, most notably at the SFZ. North of the SFZ, mean temperatures exceed 0.8°C, while south of the SFZ the means are below 0.55°C. Mean temperatures progressively decrease with latitude in the southern passage. The mean temperatures at C13 and C12 are essentially equal, yet their near-bottom currents flowed in nearly opposite directions. Figure 1e suggests a localized clockwise deep circulation occurred around a weak topographic depression in that region, where additional stirring could homogenize the temperatures. The coldest bottom waters ( = 0.35°C) are found at C15, the southernmost site in the deep passage. Mean near-bottom currents there and at C14 (Fig. 1e) persistently flowed to the north-northeast under the Southern ACC Front (SACCF), which separates from the southern margin just upstream of the cDrake line. At C16 and C17, which were located at shallower depths on the continental slope, mean temperatures are 0.43° and 0.82°C, respectively, and the mean near-bottom currents flowed to the southwest parallel to the topography.
In addition, values differ zonally within the LDA. The eastern sites occupying lines D–G exhibit means that are about 0.2°C colder than those at western lines A–C. This is primarily due to changes in the orientation of the ACC fronts as they transit the cDrake region since the baroclinic structure of the flow reaches all the way to the bottom. After crossing the SFZ, the SAF turns abruptly northward, flowing along the northern part of the C line (Fig. 1a), splitting from the PF, which continues flowing eastward along the southern edge of the LDA. Figure 1d shows the record-length-mean, near-bottom currents color-coded by for the sites north of the SFZ. The warmest temperatures were found at sites along the SAF as it enters the western portion of the LDA and turns northward. In the eastern LDA, cooler temperatures were associated with a persistent cyclonic recirculation (Chereskin et al. 2009; Ferrari et al. 2012) in the region between the SAF and PF.
Bottom water masses are cautiously identified using in situ temperature. Naveira Garabato et al. (2002) delineate the deep-water mass boundaries by neutral density γn (Jackett and McDougall 1997). They characterize three water masses found near the bottom in Drake Passage as Lower Circumpolar Deep Water (LCDW) with 28.0 < γn < 28.26 kg m−3, a subrange called Southeast Pacific Deep Water (SPDW) with 28.20 < γn < 28.26 kg m−3, and Weddell Sea Deep Water (WSDW) with 28.26 < γn < 28.4 kg m−3. Since the CPIES does not measure salinity, neither potential temperature θ nor γn can be determined directly. Instead CTDs taken during the cDrake cruises were used to associate in situ temperature T values with θ and γn values (Fig. 2). Waters with T < 0.38°C correspond to γn > 28.26 kg m−3 (Fig. 2c) and, following Naveira Garabato et al. (2002), can be identified as WSDW. Bottom waters with 0.38°C < T < 0.9°C correspond to SPDW, and those with T > 0.9°C correspond to LCDW. Based on the mean temperatures in Fig. 1b, WSDW was confined to site C15 (southernmost site in the deep passage near 61°S), where = 0.35°C, and possibly to C14, where = 0.40°C (given the calibration uncertainty of 0.02°C). SPDW filled the region south of the SFZ and spread as far north as 56.5°S in the eastern portion of the LDA (lines D–G). LCDW was found at the bottom in the western LDA and sites farther north, where mean in situ temperatures exceed 0.9°C.
b. Temporal variability
Generally speaking cold temperatures at the seafloor are perceived to be steady, yet this perception is wrong for the Drake Passage. Time series of daily measurements are shown in Fig. 3 for 15 sites that were occupied by single instruments during the 4-yr field program. The temperature records exhibit both long- and short-period fluctuations.
In the LDA, temperatures at any given site exhibit large ranges of 0.5°–0.9°C. Variability about the mean of one standard deviation σ is typically 0.1°–0.15°C for these sites (Fig. 1b). Primarily in the eastern LDA (e.g., D03, E03, and F02 in Fig. 3), temperatures are punctuated abruptly by pulses of cold water bringing temperature changes of many tenths of a degree over periods as short as 5–10 days, followed by gradual warming over many months. As will be shown in section 3c, these cold pulses occurred at times when the PF meandered northward into the LDA. Noteworthy at these sites is a long period of gradual warming with weak temperature fluctuations between 2008 and 2009; temperatures warmed by about 0.3°C over 260 days. This warming was observed throughout the LDA but was less pronounced in the western portion (e.g., A03, B02, and C06).
South of the SFZ, at southern sites C11–C17, the temperature ranges (0.2°–0.3°C) are smaller and the fluctuations (σ < 0.05°C) are weaker than at sites in the LDA, yet they are still sizable. The smallest temperature range (0.19°C) occurred at C13, which was situated in the relatively quiet region between the PF and SACCF. This range is similar to that observed farther downstream by MYRTLE (0.15°C; Rubython et al. 2001). Plotted as they are in Fig. 3 with the same temperature scale as the northern sites, the sizes of the fluctuations at C13, C14, and C15 are difficult to gauge; changes of 0.05°–0.1°C occurred at monthly periods and twice that amount at periods of 3–6 months. At C17 on the upper continental slope, temperature changes of 0.1°–0.2°C occurred at fortnightly periods.
Long-term (multiyear) temperature trends are ambiguous and highly dependent on the time period over which they are calculated. All 15 sites in Fig. 3 exhibit positive trends toward warmer temperatures ranging from 0.0039° to 0.039°C yr−1 (equivalent to temperature increases of 0.016°–0.16°C over 4 yr) when calculated for the full 4-yr period. Yet, these trends are strongly influenced by very cold temperatures that were observed during the first year. The average temperatures in the LDA from November 2007 to November 2008 are 0.1°–0.15°C colder than the annual averages for each of the remaining 3 yr. In the LDA, warming trends calculated for the first year are large and significant (0.05°–0.26°C yr−1), whereas those calculated for the remaining 3 yr (early 2009 to late 2011) are either negative (cooling) or insignificant. South of the SFZ, warming trends of 0.007°–0.017°C yr−1 over the full 4-yr period occurred, but the trends are inconsistent for shorter 2- or 3-yr periods.
Bottom temperature variations can be attributed, in part, to the meandering ACC fronts because the thermocline is deep reaching. In addition, because frontal movements are tracked by CPIES τ variations, temperature variations at the seafloor associated with lateral shifts of the thermocline can be estimated using τ measurements and the GEM lookup table. Temperatures at 3500 dbar estimated from τ at four sites are shown in Fig. 4 (colored lines). Correlations r between these estimated temperatures and the direct measurements indicate whether the observed variance can be attributed to ACC meandering or to other near-bottom processes. The correlations varied both spatially and temporally; the values plotted (Figs. 4a,b) are statistically different from zero at the 95% significance level. For example, consider three sites in the northern portion of the LDA (Figs. 4c–e). At C19, the records are well correlated (r > 0.65) with about 42% of the observed bottom temperature variance due to the meandering SAF. In the western LDA (e.g., A01), correlations are poor (r < 0.21); prior to 2009, strong meandering of the SAF resulted in large fluctuations in the estimated temperatures (periodicities of 60–65 days), and yet a warming trend with weaker oscillations were measured at the bottom. For the western LDA sites, correlations improve (r > 0.4) for the time period after March 2009 when longer-period fluctuations were prevalent. In the eastern LDA (e.g., F01), measured temperatures exhibit about four cycles of abrupt cooling, which are also evident in the τ-derived records, followed by gradual warming periods, which are largely absent from the τ-derived records. The cooling periods correspond to meanders of the PF protruding northward into the LDA, whereas the warming periods correspond to bottom waters stirring and mixing as they recirculated within a cyclonic gyre found in the Yaghan Basin (Chereskin et al. 2009; Ferrari et al. 2012). For most of the northern sites, the correlations improve slightly (increasing by 0.05–0.1) when they are restricted to times when the near-bottom current speeds exceeded 0.15 m s−1. South of the PF at C13, where the temperature variability is weak, the records are uncorrelated (r < 0.2), yet they exhibit about the same amount of variance. At southern sites C14 and C15 near the SACCF, correlations between the estimated and measured temperatures are higher (r ~ 0.5). The method of combining the τ measurements with the GEM works quite well throughout the water column (Chidichimo et al. 2014; Firing et al. 2014). Below 3500 m, however, the overall temperature signal is weak (range of less than 1°C) compared to the upper water column. The disagreement between the GEM methodology and bottom temperature observations indicates that there are additional near-bottom processes (such as deep circulation around bottom pressure anomalies or localized topographically controlled recirculations) occurring whose temperature–advection signals are comparable in size (σ ~ 0.05°–0.15°C) to meandering–thermocline signals.
Spatial patterns are examined by calculating lateral correlations of the temperature anomalies (time means removed) between sites within three subregions (Fig. 5). These include sites north and south of the SFZ and in the H array at the base of the SFZ. The northern sites (spaced 35–40 km apart) are well correlated at distances less than about 75 km; the correlations are independent of their east/west positioning within the LDA. They smoothly decay with distance before reaching a plateau near 0.4 for distances greater than 100 km. South of the SFZ where the sites were separated by 60–70 km, bottom temperature fluctuations are either weakly correlated (r ~ 0.4) between sites near the SACCF or uncorrelated (r < 0.2). One closely spaced southern pair (10 km apart) is better correlated. Fluctuations at the base of the SFZ exhibit either very high (r > 0.8) or low correlation (r < 0.4). High correlations are associated with pairs of sites oriented along the topographic ridge, whereas the low correlations are associated with pairs straddling it.
The temperature variability can also be characterized through variance-preserving spectra and autocorrelation functions. Records with similar temporal characteristics are grouped together in Fig. 6, and, in general, neighboring sites cluster into similar spectral groups. Spectra were computed using the Welch method; the records were broken into 365-day segments, overlapped by 50%, and scaled with Hanning windows. Autocorrelation functions are truncated at time lags of 296 days (20% of the total 4-yr record length). The spectra for sites at the northern end of the C line (Fig. 6b) exhibit peaks at periods of 30–60 days, and the first zero crossings t0 of autocorrelation functions occur in under 100 days (Figs. 6b,c). For LDA sites (Figs. 6c–i), t0 lengthens from west to east (with increasing distance from the SFZ) from less than 50 to over 200 days. Spectra for western LDA sites (Figs. 6d,e) exhibit peaks near 50–100 days that are associated with the meandering SAF. At the easternmost end of the LDA (Figs. 6h,i), the spectra are red, dominated by longer periods associated with bursts of cold water carried northward by the meandering PF followed by gradual warming afterward as the waters recirculated. The largest spectral peaks occur in the central portion of the LDA (Fig. 6g; note the larger spectral range) at sites influenced equally by the meandering of the SAF and PF. For sites in the southern portion of the passage (open stars), the temperature spectra (Fig. 6j; note the smaller spectral range) are relatively white, and t0 values are short (less than 100 days). At the southern margin, the temperature spectrum of C17 (Fig. 6k) is notable for its prominent peak at the fortnightly period.
c. Extreme cooling event in the LDA
A prominent feature of the temperature records from the LDA is a drastic cooling event in early 2008 followed by a period of relatively steady warming until early 2009. This is particularly conspicuous in the eastern portion of the LDA (D03, E03, F02, and G01 in Fig. 3 and F01 in Fig. 4). Upon close inspection this cooling and subsequent warming can be seen in the western portion (A03, B02, C06, and C07 in Fig. 3 and A01 in Fig. 4) as well. During mid-February 2008, bottom temperatures throughout the eastern LDA plummeted, dropping by more than 0.5°C over a 15-day period. Waters colder than 0.3°C were observed at 6 of the 12 sites. Satellite maps of SSH reveal that a large meander trough of the PF protruded northward into the eastern LDA (Fig. 7g). When the PF retreated after shedding a cold-core ring, the bottom temperatures rapidly warmed by over 0.2°C to about 0.6°C (Fig. 7b). Concurrent measurements made by the CPIES (and spatially averaged over the eastern LDA) also exhibit prominent drops in sea surface elevation and bottom pressure, and increased eddy kinetic energy (EKE) followed by recoveries within 15 days to normal values (Figs. 7c–e). Bottom temperatures and SSH decreased again by about 0.1°C and 0.15 m, respectively, in early May 2008 when the PF protruded into the eastern LDA again. The 20-yr time series of satellite-derived SSH (Fig. 7f) shows that the meandering event in February 2008 was unusually strong, while the May 2008 event was more typical. Maps of bottom temperature and bottom pressure anomaly (Fig. 7h) show cold waters were carried northward by the near-bottom currents. During the subsequent 8-month period, bottom temperatures steadily warmed, reaching over 0.9°C by late January 2009, and the maps reveal that they were spatially uniform throughout the LDA. Satellite SSH maps (July 2008–January 2009) show that, during this prolonged warming period, the PF (Fig. 7g, blue shades) skirted the southern edge of the LDA. The PF did not protrude into the eastern region again until late March 2009 when the bottom temperatures cooled rapidly once more. In the western LDA, the cooling events during February 2008 and May 2008 were less dramatic, and the warming period was less steady (Fig. 3). Bottom temperatures cooled when the SAF (Fig. 7g, orange shades) shifted north of the western LDA, hugging the topography near Cape Horn. Gradual warming occurred as SAF shifted southward and meanders penetrated into the western LDA.
d. Variability at the base of the SFZ
The five temperature records in the H array at the base of the SFZ (mean bottom pressures were 3800–4400 dbar) are highlighted in Fig. 8. These sites were very closely spaced: ~10 km separated the sites straddling the SFZ and ~18 km separated those along it in the northwest–southeast direction. The records are grouped by location, either upstream (H01 and H05) or downstream (H02, H03, and H04) of the topographic ridge, which emphasizes the differences in their character. The 13-month-long means of 0.84°–0.90°C at the five sites are similar, and the ranges vary from 0.65°C on the upstream side to 0.55°C on the downstream side. The coldest temperatures observed were 0.55°C, but they occurred at different times on either side of the ridge. A narrow, deeper (~3200 m) gap in the SFZ (inset in Fig. 8c) separated sites H04 and H05 from the other three sites. Firing et al. (2016) suggest that in the mean, deep flow with a small eastward component passes through this gap. We demonstrate later in section 3f that, intermittently, deep flow through the gap can be westward.
The upstream temperatures exhibit large fluctuations with periods of about 50 days (Fig. 8c), whereas the dominant fluctuations at the downstream sites have either longer or shorter periods (mainly 60–100 and 25–33 days, respectively). The spectral peaks of H01 and H05 are larger than those obtained for LDA sites (Fig. 6), indicating that the variability is enhanced along the SFZ. As was shown in Fig. 4b, the H01 and H05 temperatures are highly correlated with the τ-derived temperatures (r > 0.6), and their 50-day oscillations are associated with large-amplitude meanders of the SAF bringing warmer bottom waters southward. The 50-day fluctuations are largely absent from the records at downstream sites H02 and H03, where the variability is only weakly correlated with the τ-derived temperatures (r < 0.4). The character at H04 is mixed, exhibiting features common with both the upstream and downstream sites. Differences in the observed variability across the ridge may be related to differences in the regional temperature gradients, which are stronger on the upstream side of the SFZ compared to the downstream side (Fig. 1b).
The longest-period signals are partially related to north–south shifts in position of the ACC. Temperatures at 3500 dbar estimated from the CPIES-measured τ (Fig. 8d) were warm between December 2010 and February 2011, cold during April–July, and warm again during August–September. The SSH maps in Fig. 8e provide regional views of the ACC fronts. Warmer bottom temperatures occurred during February when the SAF was farther south. Colder bottom temperatures occurred in May and July when the PF penetrated deep into the LDA near the longitude of the H array. Rapid warming occurred at sites H01 and H05 during August when a SAF meander crest protruded southward. Gradual warming occurred at the downstream sites between August and October when the ACC shifted to a more southerly position.
e. Variability at the southern continental margin
C16 and C17 were moored at shallower depths (2550 and 1280 dbar, respectively) on the Antarctic continental slope seaward of Livingston Island. The record-long mean temperature of 0.82°C at shallower C17 is 0.38°C warmer than that at C16 (Fig. 1). Their 4-yr time series are plotted in Fig. 9. The temperature variance (σ2 = 0.0065°C2) at the shallower site C17 is 2–3 times greater than at C16.
The mean bottom flow is mainly to the southwest at both locations and aligned with the local bathymetry (Fig. 1). Yet, there is a considerable difference between their mean speeds. Progressive vector diagrams (Figs. 9a,e) illustrate that very different flow regimes exist near the bottom at these two locations. The trajectories are constructed by vector addition of consecutive velocities and are color-coded by bottom temperature. Each trajectory begins at the plus mark and successive displacements are plotted at 2-day intervals. The net movement at both sites is to the southwest. However, whereas the C16 progressive vector exhibits little variation in direction over time, flow reversals lasting weeks to months occurred at C17. Plotting the eastward distance as a function of time (Fig. 9d) reveals that the longer-lasting reversals happened seasonally with northeastward flow typically observed during the austral winter. Colder bottom temperatures were measured at C17 in October of each year when the currents reversed again to flow to the southwest. Vertical profiles of temperature at C17 estimated from τ (not shown) indicate that colder waters occurred throughout the water column when the bottom flow was to the southwest. At C16, temperature fluctuations increased during the final 21 months. At the same time, the southwestward bottom currents strengthened and became more variable. Together these indicate a southward shift in the path of the ACC, in agreement with Donohue et al. (2016) who reported a marked change in the water properties at C17 between 2009 and 2010.
Temperatures and currents were also measured at mooring M03 (Fig. 9), 8 km inshore of C16, during the first 2 yr of cDrake at 100 (D-100) and 300 m (D-300) above the bottom at 1900 m. At D-300, the flow was primarily to the northeast, consistent with the northeastward flow of the ACC; the currents reversed direction for only brief intervals. Like C17, D-100 exhibited longer-lasting flow reversals during the first 10 months. For the remaining 14 months, the flow was mainly westward and occasionally northward. The D-100 and D-300 temperature records are highly correlated (r = 0.93), which is consistent with lateral shifting of a baroclinic temperature structure.
The temperature records from the three sites on the continental slope are uncorrelated with one another and exhibit signals at different periods. A clear annual signal is present at C17 with minima occurring between September and December each year, which is remarkable considering the 1280-m depth of the instrument. Meredith et al. (2003) observed a similar annual signal at depths below 1000 m off Elephant Island, about 300 km northeast of these cDrake sites. The annual cycle was not evident at either C16 or M03. Cold temperatures also occurred at M03 in October–November, but they were also found during March–May, while at C16 the coldest temperatures were observed in June 2011. The temperature spectra (Fig. 6k) for C16 and M03 (not shown) have peaks at periods of 90–100 days. A prominent spectral peak occurs at the fortnightly tidal period at C17 and to a lesser extent at M03, yet fluctuations with 14-day periods were absent at C16.
f. Vertical temperature gradients
Until now, we have been examining, for each site, a single bottom temperature record at daily intervals. Next, we will discuss differences that were observed between the D-1 and D-50 temperature records. In particular, the deeper measurements were occasionally a couple of tenths of a degree colder. These deep vertical temperature gradients dT/dz were produced by different mechanisms than those observed by Meredith et al. (2013) in the eastern portion of the Scotia Sea. In their case, the vertical gradients resulted from episodic dense overflow waters that became trapped in deep trenches for several years. As described below, during cDrake the vertical gradients occurred intermittently on time scales from hours to weeks and were associated with deep pressure anomalies and interactions with topography.
1) Cold intrusions
During the calibration process (section 2) occasional differences between the daily D-1 and D-50 temperature records were observed at most CPIES sites. On these occasions, the D-1 temperatures were several hundredths of a degree colder than the D-50 temperatures measured 50 m above. Generally, these large dT/dz events were observed 5%–10% of the time at sites in the LDA and less often elsewhere. Figure 10 illustrates three typical LDA sites with the daily D-50 temperature records shown by the gray curves and the corresponding D-1 temperatures by black dots whenever their differences exceeded 0.04°C (a factor of 2 greater than our calibration accuracy). Large dT/dz values are mainly associated with cold events, indicating that they are stably stratified. Additionally, these large gradients often occurred at more than one site at a time. This is more clearly illustrated by the case studies shown in the bottom three rows of Fig. 10. Sites with dT > 0.04°C (solid circles) occur around the periphery of areas with the coldest temperatures (darker blues). Weaker vertical gradients (dT > 0.02°C) are also observed at neighboring sites (open circles). These near-bottom vertical gradients happened throughout the LDA in association with cold pools initially advected into the region by upper jet meanders. Thus, the intrusions occurred more frequently in the eastern LDA, where the colder PF more commonly protruded northward, rather than in the western LDA, where the warmer SAF more commonly protruded southward. During the three case studies, cold pools that entered the LDA near the southeastern corner were subsequently advected to the north and west by deep eddies (near-bottom pressure anomalies). As the cold pools shifted locations, dT/dz gradients at the original sites vanished and new gradients formed at neighboring sites. The distribution of the dT/dz gradients suggests a finescale spatial structure, indicative of filaments or streamers created by interactions with local topography.
2) Internal tides
The D-1 and D-50 measurements exhibited interesting fluctuations at periods of 1 day and shorter. Of particular interest are the hourly records at sites H01 (Fig. 11a) and H05 (not shown) at the base of the SFZ, which exhibited two time intervals with very large (~0.3°C), high-frequency (periods ≤ 1 day) temperature fluctuations. The strong fluctuations were observed during mid-December 2010 and late July–early August 2011, when the SAF meandered southward over the sites at the SFZ. During the December event at H01, rapid cooling by 0.3°C and rewarming occurred in brief 5-h intervals that recurred every 20–25 h for roughly 9 days during 11–20 December. Simultaneously, large differences were observed between the D-1 and D-50 temperatures (Fig. 11i). At H05 (not shown), the cold spikes were of shorter duration (~3 h) and recurred more rapidly (10–13 h). The temperature spectra at both sites exhibited peaks at both the diurnal and semidiurnal tidal periods and increased energy in the internal wave band (which can be seen by comparing the D-1 spectrum at H01 to the line showing a slope of −3.3 in Fig. 11m). As described next, these vertical temperature gradients and cold spikes were formed by bottom waters spilling over the SFZ from the downstream side when the currents were weak and their directions variable.
A detailed description of the events at H01 is as follows. In mid-November, the D-1 and D-50 temperatures were equal valued (dT/dz ~ 0; Fig. 11c) and exhibited little high-frequency variability (Fig. 11b). At the same time, hourly near-bottom currents with speeds of 0.1–0.2 m s−1 flowed consistently to the southeast along the base of the SFZ (gray curves in Figs. 11d,e). By 1 December, as the hourly current speeds reduced to less than 0.05 m s−1 and the directions became more variable (Figs. 11g,h), the D-50 and D-1 temperatures decreased by nearly 0.2°C, and, in addition, vertical temperature gradients and high-frequency fluctuations began to develop (red and blue curves in Fig. 11f). The near-bottom stratification continued to intensify until 11 December when dT > 0.3°C. Meanwhile, both D-1 and D-50 gradually warmed as the ACC meandered south. The hourly τ-derived temperatures at 3500 dbar (green curve) also warmed during early December, but lacked the high-frequency fluctuations observed in the near-bottom records. Between 11 and 20 December, large cold temperature spikes (Fig. 11i) occurred in the D-50 and D-1 temperatures, coincident with the tidal currents rotating counterclockwise from north to southwest. After 20 December, the high-frequency fluctuations tapered off, and the vertical gradients vanished as steadier southeast flow reestablished.
To explain the events at H01, we needed to examine the temperature and current records from site H02 on the downstream side of the SFZ. The H02 records are shown by orange curves in Fig. 11. In late November, the bottom temperatures at H01 and H02 were approximately equal. The H01 τ-derived temperatures gradually warm prior to 10 December as ACC meanders steepened in the vicinity of the H array (Fig. 11k). The lower layer was squashed and stretched by the meandering ACC, with a deep anticyclone and cyclone developing on the upstream and downstream sides of the SFZ, respectively. Northwestward flow associated with these deep features opposed the persistent southeastward flow at H01 and enhanced the northwestward flow at H02. Without strong flows advecting warm waters from farther north along the ridge, mixing with adjacent colder bottom waters occurred at H01, and the D-50 and D-1 temperatures cooled to 0.65°C by the end of November. (Note the colder waters south of the H array in Fig. 1b.) With tidal currents accounting for a larger fraction of the near-bottom currents at H01, the current direction became more variable. When the currents at H01 and H02 had a westward component, warmer water at H02 spilled over a deeper sill in the SFZ (Fig. 11l), initiating the vertical temperature gradient formation at H01 (1–10 December; Fig. 11f); warm (upward) spikes were observed in the D-50 record, and stratification initially prevented the warmer water from reaching as deep as D-1. More consistent southeast flow of intermediate speed returned to H01 after 11 December, and the D-50 and D-1 temperatures warmed to 1.1°C, in agreement with the τ-derived temperatures and the southward meander of the ACC. At H02, however, the near-bottom flow (Fig. 11j) was still influenced by the counterclockwise-rotating tidal currents. As the H02 currents rotated between northward and southwestward, its bottom waters again spilled back over the sill. The now relatively colder and denser H02 waters descended on the upstream side of the SFZ, and cold (downward) temperature spikes were observed at both D-50 and D-1 at H01.
There are numerous intervals throughout the year-long record at H01 when vertical temperature gradients and high-frequency fluctuations coincide. Time intervals with high-frequency signals are clearly evident in the 1-day, high-pass filtered D-1 record (e.g., December 2010 and May, August, September, and December 2011 in Fig. 11b). These events all share the following characteristics: When short-period temperature fluctuations have larger amplitudes (|T| ≥ 0.05°C), near-bottom stratification develops (dT/dz ≥ 0.002°C m−1), hourly near-bottom currents are slower (<0.05 m s−1; Fig. 11d) with speeds similar to those of the tidal currents, and the flow rotates through angles (Fig. 11e) with a westward component. High-frequency fluctuations and vertical gradients were observed in the temperature records at H02 when the near-bottom flow had an eastward component, but those events were less energetic (Fig. 11m) compared to H01. In addition, the records at H05 and H04 exhibited similar characteristics as H01 and H02, respectively.
The 4-yr measurements obtained during cDrake and spanning the full passage provide new insights into the spatial and temporal variability of temperature at the seafloor. Large temperature changes on the order of 0.3°–0.4°C were observed on a wide range of time scales from as short as 3 h to as long as 8 months. Minimum-to-maximum ranges of 0.5°–0.9°C and 0.2°–0.3°C were observed at sites north and south of the SFZ, respectively. Temperature variance decreased across the passage from north to south by a factor of 4. While the variability south of the SFZ seems small when compared to northern sites, it is substantial when compared to the temperature variations beneath two western boundary currents in the Northern Hemisphere. Bottom temperatures measured by PIES under the Gulf Stream near 68°W in water depths of 3000–4000 m during SYNOP had ranges of 0.1°–0.2°C and σ = 0.01°–0.02°C, comparable to the southern cDrake sites. Only two PIES located along the continental margin near Cape Hatteras near 74°W exhibited ranges and variances comparable to the northern cDrake sites. CPIES temperature measurements under the Kuroshio Extension at depths of 5300–6300 m during KESS had ranges less than 0.04°C and σ ~ 0.005°C; thus, the temperature variance in southern Drake Passage is two orders of magnitude greater than that observed in the deep North Pacific.
Because of the high variability in Drake Passage, it is difficult to determine the presence or absence of long-term warming or cooling trends with these 4-yr-long records. Simulations were conducted to assess how well a specified warming trend could be resolved given the observed variability. A 50-yr-long record was simulated by replicating temperature time series at F02 and adding a trend of 0.002°C yr−1. Trends of that magnitude were determined over a 10-yr period for the deep Amundsen–Bellingshausen Basin by Purkey and Johnson (2010) and over a multidecadal period for the open-ocean regions around Antarctica by Azaneu et al. (2013). A suite of regressions were performed during which the record length and sampling interval were varied. The estimated trend for the 4-yr record with daily sampling was 20 times larger than the specified trend. A record length of 13 yr was required to estimate the trend to within a factor of 2. Changing the sampling interval increased the number of years of data required to determine the trend within a factor of 2 to 17 yr for 90- and 180-day sampling and to more than 30 yr for yearly sampling. Since the southern sites exhibited weaker temperature variance, the simulations were repeated using the record from C14. Record lengths of 9–12 yr were required when the sampling was 180 days or more frequent, and about 20 yr of data were required for yearly sampling.
In Drake Passage, waters with neutral density γn > 28.26 kg m−3 are classified as WSDW (Naveira Garabato et al. 2002; Provost et al. 2011). As noted in section 3a (Fig. 2), the in situ temperatures measured by CPIES were used as a proxy to distinguish the deep-water masses, and WSDW is identified by bottom temperatures below 0.38°C. During cDrake, WSDW was primarily confined to the southern end of the passage. It was the dominant water mass at C15 (61.1°S), occurring there nearly 75% of the time, and was observed 30% of the time at C14 (60.6°S). C14 and C15 were typically located beneath the SACCF, which shifts away from the southern margin just west of the cDrake line. The near-bottom currents there persistently flowed slightly east of due north (Fig. 1e). During 1979–1980, northward flow was also directly measured at a mooring (Fig. 1a) whose position was only 40 km west of C15 (Nowlin and Zenk 1988; Pillsbury et al. 1981). However, because their deepest current meter was located about 1000 m above seafloor, the recorded temperatures were considerably warmer (ranging 0.45°–0.8°C) and would be more indicative of the upper ACC variability.
Remarkably, on four occasions (February 2008, May 2008, March 2009, and March 2011) WSDW was observed as far north as 56.5°–58°S within the LDA. In particular, during the first two events, bottom temperatures below 0.38°C were observed at several sites in the eastern portion for 5–10 days when PF meander troughs protruded northward. During the latter two events, WSDW was observed only at E03, but the neighboring sites had temperatures near 0.4°–0.45°C. The bottom topography contoured in Fig. 7a reveals that a deep gap occurs between segments of the West Scotia Ridge just south of E03. This transform fault, oriented northwest–southeast, provides a relatively deep passageway for water to flow from the Ona Basin in southern Drake Passage, where WSDW is commonly observed (Sudre et al. 2011), into the Yaghan Basin, where the LDA was located. Based on a Russian hydrographic survey with deep current measurements, conducted in October–November 2008 in the region where the West Scotia Ridge and SFZ meet, Koshlyakov et al. (2010) identified this same gap as a pathway for Antarctic Bottom Water. Their measurements indicated, however, that after the cold bottom water flowed through the southernmost portion of the gap, it turned to flow westward toward the SFZ within the deep rift valley of the West Scotia Ridge. Bottom temperatures at D03 (Fig. 3) dropped by only 0.1°–0.15° to 0.75°C, confirming that cold WSDW did not enter the Yaghan Basin through the northern portion of the transform fault at the time of the Russian field program in late 2008. Additional evidence that this gap provides a pathway for the denser water masses to exchange between the basins comes from the hydrographic surveys conducted by Sudre et al. (2011) in 2006; they found that WSDW contributed a sizable fraction to the water masses found near the bottom in the gap.
The relationship between variations in bottom temperature and PF position were examined by Rubython et al. (2001) and Cunningham et al. (2003) to investigate the suggestion of Deacon (1937) that the location of the upward-sloping thermocline should be determined by the circulation of bottom water. Those two studies drew different conclusions: Rubython et al. (2001) found that the variability of the 4-yr bottom temperature record at MYRTLE was uncorrelated with the variability of the PF surface position as determined by satellite. On the other hand, Cunningham et al. (2003) found good agreement between the bottom temperature anomalies at six CTD stations near MYRTLE taken between 1993 and 2000 and the PF location determined from hydrography. The cDrake data can be used to investigate this relationship at a different location in the Drake Passage (nearly 400 km upstream of MYRTLE) and using temperature measurements from multiple sites. It should be noted, however, that WSDW patches found upstream of the SFZ are farther from the formation region and have undergone more mixing than those near MYRTLE. Foppert et al. (2016) used two methods to determine the position of the PF during cDrake: weekly surface positions from satellite altimetry and the maximum geopotential anomaly gradient determined from CPIES τ measurements. Because their study was restricted to times when the path of the PF was nearly perpendicular to the C line, the latter positions were irregularly distributed with time. No significant correlations were found between these PF positions and the temperature fluctuations from two different sites (C14 and C15) located south of the meandering PF where WSDW was observed at the bottom. The standard deviation σ of the temperatures at those sites was determined on a daily basis as a way to estimate changes in the volume of bottom water occupying the southern passage; smaller σ would indicate more uniform temperatures throughout the region and, thus, a larger volume of bottom water. Again, no significant correlations were found between σ and either estimate of the PF location. Thus, while locally variations in the position of the ACC and bottom temperature may be correlated (Fig. 4), the relationship does not apply to the region as a whole.
The cDrake instruments at the southern margin corroborate the findings of previous field programs. C16 and C17 were situated at similar locations and depths on the continental slope as two moorings (white circles in Fig. 1a) of Nowlin and Zenk (1988). Our new measurements also find westward flow near the bottom and confirm that different current regimes exist in the upper and deeper portions of the continental slope. The M03 currents also confirm that the westward flow along the deeper slope is confined to within a few hundred meters of the bottom. Meredith et al. (2003) observed a pronounced annual cycle at 1040-m depth off Elephant Island, more than 300 km northeast of C17. Their temperature fluctuations were about a factor of 2 larger and much colder than at C17, but the phasing was similar. Absent from the C17 record, however, are the large very cold spikes of water (about −0.9°C), which Meredith et al. (2003) observed and associated with downslope convection. The lack of very cold temperatures and poor correlations (regardless of time lag) between neighboring sites suggests this phenomenon did not occur north of Livingston Island during cDrake. The displacement of the progressive vector at C17 slightly to the northwest (Fig. 9a), however, indicates that slantwise cross-slope transport near the bottom did occur.
We examined the high-frequency signals in the hourly temperature records at all sites, not just the H array. Spectral peaks at the semidiurnal tidal period were observed at other sites near steep topographic features and absent from those in flatter regions. The 10 sites with semidiurnal spectral peaks included northern sites B02, C03, C05, and C19; southern continental slope sites C16 and C17; and sites H01, H02, H04, and H05 at the base of the SFZ. The semidiurnal spectral peak was absent for H03, which was in a slightly flatter region downstream of the ridge. At H01 and H05, when the along-slope flow subsided, the near-bottom current direction became more variable. At those times, bottom waters advected over the ridge created vertical temperature gradients in, at least, the bottom 50 m, permitting the generation of internal waves. That the τ-derived temperatures lacked high-frequency fluctuations indicate that the internal wave amplitudes were small and that the vertical temperature gradients did not extend far above the bottom.
The array of CPIES deployed during the cDrake experiment provides an opportunity to examine bottom temperature variability within Drake Passage over a 4-yr period. Two different temperature sensors separated vertically by 50 m yielded similar records; thus, they could be used interchangeably to provide continuous coverage at nearly all the instrumented sites. Deep CTDs taken on five cruises were used to calibrate the measurements, which enabled records from different instruments to be patched together to create a single time series at each location. The 16 sites that were continuously occupied by single instruments for 4 yr did not require patching. This paper is focused on the variability of the data collected in the deep passage (water depth > 3500 m) and at the southern continental margin. In all, we report on 4-yr-long bottom temperature records from 30 sites, and shorter-duration records from an additional 11 sites.
The temperature fluctuations differed across the passage. In the northern portion, the temperatures were highly variable with ranges of up to 0.9°C and standard deviations over 0.1°C. The variability decreased in the southern portion of the deep passage with the ranges and standard deviations of 0.2°–0.3°C and 0.05°C, respectively. Nearly half of the bottom temperature variance could be attributed to meandering of the deep-reaching thermocline. Near-bottom circulation around propagating pressure anomalies also advected the near-bottom temperature field and contributed to the observed variability. The high degree of variability on interannual time scales would require 13 to 17 yr of hourly records to estimate long-term trends reliably. During the first year, exceptionally cold water (in situ temperature 0.4°C) was observed in the northern portion of the passage. The data suggest that the cold water was advected northward into the Yaghan Basin through the West Scotia Ridge via a deep transform fault during a strong meandering event.
Occasionally large differences (0.1°–0.4°C) were observed between the temperatures separated 50 m vertically. These vertical temperature gradients were associated with different phenomena at different time scales depending on location. In the LDA, large dT/dz occurred at multiple sites for several days to weeks when cold waters were advected into the region by deep pressure anomalies under the meandering ACC. At the base of the SFZ, however, large dT/dz accompanied high-frequency (periods ≤ 1 day) temperature fluctuations when the near-bottom currents advected colder waters over the ridge from the downstream side.
Temperatures and near-bottom currents were measured at shallower depths on the continental slope along the western Antarctic Peninsula. Temperature records from closely spaced locations were uncorrelated, indicating short horizontal scales. The currents were generally to the southwest (but were more variable on the upper slope compared to the deeper slope) and were confined to the bottom 200–300 m.
We are grateful to Erran Sousa, Gerard Chaplin, and Dan Holloway at URI who carefully developed and prepared the CPIES. They were aided by C. Cipolla and G. Savoy at the URI Equipment Development Laboratory. We thank the captains and crew of the RV/IB Nathaniel B. Palmer and the staff of Raytheon Polar Services for their support during the cruises. Yvonne Firing wrote the objective mapping codes used for cDrake. The National Science Foundation Office of Polar Programs supported this work under Grants ANT-0635437 and ANT-1141802.