The Canada Basin, : Upstream events and far-field effects of the Barents Sea

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. C7, 3082, /2001JC000904, 2002 The Canada Basin, : Upstream events and far-field effects of the Barents Sea Fiona McLaughlin, Eddy Carmack, and Robie Macdonald Institute of Ocean Sciences, Sidney, British Columbia, Canada Andrew J. Weaver School of Earth and Ocean Sciences, University of Victoria, Victoria, British Columbia, Canada John Smith Bedford Institute of Oceanography, Dartmouth, Nova Scotia, Canada Received 4 April 2001; revised 5 September 2001; accepted 26 October 2001; published 26 July [1] The effects of upstream events on southern Canada Basin waters were examined using physical and geochemical data collected at one location between 1989 and These events included Atlantic layer warming, relocation of the Atlantic/Pacific water mass boundary, and increased ventilation of boundary current waters. Early signals of change in the Canada Basin first appeared in 1993 along the continental margin and by 1995 were evident in the basin interior and farther downstream. Differences in physical and geochemical properties (nutrients, oxygen, 129 I, and CFCs) were observed between 150 and 1600 m of the water column. By 1995 the boundary between Pacific- and Atlantic-origin water was shallower and Atlantic-origin water occupied more of the Canada Basin water column. Atlantic-origin lower halocline water was colder and more ventilated, Fram Strait Branch water was colder, fresher, and more ventilated, and Barents Sea Branch water was warmer, fresher, more ventilated, and occupied a larger volume of the water column than in previous years. Water mass analysis showed that the composition of Atlantic-origin water in the Canada Basin in 1995 included 20% more water from the Barents Sea. Two related events upstream appear to be the source of this change: an atmospheric regime shift in 1989 that increased cyclonic circulation, a transition unprecedented within the latter half of the 20th century in magnitude, geographic reach, and apparent oceanographic impacts, and a large volume of dense water flowed from the Barents Sea to the Arctic Ocean between 1988 and These two phenomena illustrate the close relationship between atmospheric and oceanic events whose combined influence was evident 6 years later and 5100 km downstream in the Canada Basin at depths >1000 m. INDEX TERMS: 4207 Oceanography: General: Arctic and Antarctic oceanography; 4808 Oceanography: Biological and Chemical: Chemical tracers; 4215 Oceanography: General: Climate and interannual variability (3309); 4283 Oceanography: General: Water masses; KEYWORDS: Canada Basin, Pacific-origin water, Atlantic-origin water, Atlantic/Pacific water mass boundary, Barents Sea Copyright 2002 by the American Geophysical Union /02/2001JC Introduction [2] The historical view of the Arctic Ocean as an ocean in a steady state [cf. Treshnikov, 1959; Carmack, 1986] has given way over the past decade to a newer view of the Arctic as an ocean in transition. Beginning in 1990, temperatures higher than the climatological record were observed in the Atlantic layer in the Nansen Basin, geographically the first Arctic Ocean basin to receive Atlantic water inflow [Quadfasel et al., 1991, 1993] (Figure 1). By 1993, higher Atlantic layer temperatures were observed downstream in both Amundsen [Morison et al., 1998] and Makarov [Carmack et al., 1995] basins. Between 1991 and 1994, Atlantic layer temperatures near the Lomonosov Ridge in the Amundsen Basin increased 0.5 C [Swift et al., 1997], confirming that Arctic Ocean change was underway. [3] Recent research has also documented change in the water mass composition of the Makarov Basin, signaling variability in Arctic Ocean circulation. The historical view of the Arctic Ocean held that the Eurasian Basin halocline consisted of Atlantic-origin water characterized by cold temperatures and low nutrients [Coachman and Barnes, 1962; Moore et al., 1983]. Likewise, the Canadian Basin halocline was composed largely of Pacific-origin water, characterized by a local temperature maximum and minimum and a nutrient maximum, that overlay lower halocline water of Atlantic-origin [Coachman and Barnes, 1961; 19-1

2 19-2 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure 1. Map showing bathymetric features of the Arctic Ocean and general circulation of inflowing Atlantic-origin (red arrows) and Pacific-origin (blue arrows) waters. The location of station A and of upstream stations E (Larsen-93) and 20 (AOS-94) in the Makarov Basin, station 35 in the Amundsen Basin, and stations 21 and 70 in the Nansen Basin (ARKIX/4) are identified on the larger Arctic Ocean map. The location of Canada Basin stations occupied in 1993 (red circles) and 1995 (blue circles) are shown in the inset. The color associated with each station is used to identify data from these stations in the following figures. Data from stations 21 and 70 provide description of Atlantic-origin waters that enter via Fram Strait and the Barents Sea. Kinney et al., 1970]. Observations showed that these two haloclines were separated by a boundary located above the Lomonosov Ridge. This view was first challenged in 1991 when silicate concentrations markedly lower than previously reported were found in the upper layer of the Makarov Basin near the Lomonosov Ridge, suggesting variability of this halocline front [Anderson et al., 1994]. Observations in 1993 of low nutrients in the southern Makarov Basin halocline signaled that the Atlantic/Pacific water mass boundary had shifted eastward, from a location over the Lomonosov Ridge to one over the Mendeleyev-Alpha Ridge [McLaughlin et al., 1996]. Southern Makarov Basin waters were also well ventilated (i.e., elevated CFC-11 and oxygen concentrations) between 400 and 1600 m, suggesting rapid transport of upstream shelf water by topographically steered boundary currents. The basin-wide extent of the Atlantic/Pacific water mass transition was observed in 1993 [Morison et al., 1998] and 1994 [Carmack et al., 1997; Swift et al., 1997]. During this time, Eurasian Basin surface waters were found to be more saline than in the past, and this was attributed to an atmospherically driven change in Laptev Sea outflow [Steele and Boyd, 1998]. [4] Observations from the Nansen, Amundsen, and Makarov Basins (hereinafter referred to as upstream basins because of the order in which they receive Atlantic inflow) raised a question about whether similar changes had occurred downstream in the Canada Basin, the fourth and largest arctic basin and the one that most directly supplies low salinity outflow to the Canadian Archipelago and Labrador Sea. Time series study of the Canada Basin has been limited to two investigations. Melling [1998] reported that temperature and salinity changes in the southern Canada Basin were progressive between 1979 and He attributed upper halocline warming to a decrease in strength

3 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS 19-3 of the Siberian High and lower halocline cooling and increased thickness of the Atlantic layer s warm core to variability in Barents Sea outflow. Macdonald et al. [1999], using 18 O data collected between 1987 and 1997, reported that sea-ice melt in the Beaufort Sea increased 4 6 m after As this increase coincided with changes in atmospheric circulation and ice extent, they suggested it was a thermal and mechanical response to the 1989 atmospheric regime shift. [5] The following discussion compares physical and geochemical data collected in the southern Canada Basin between 1989 and 1995 to determine the extent of change and its timing. We also compare these data with comparable data from the three upstream basins to establish correspondence between upstream properties and downstream change. Our discussion then explores how upstream events reported in the literature provide a fuller understanding of the larger Arctic Ocean system when viewed through the perspective of Canadian Basin change. 2. Background [6] The Arctic Ocean s salt-stratified water column consists of three main layers: a cold low-salinity upper layer that includes the halocline, a warm Atlantic layer, and a cold saline deep layer. A strong inverse thermocline, where temperature increases rapidly with depth, separates upper and Atlantic layers, and a weak thermocline divides Atlantic and deep layers. Characteristics of the upper layer s halocline are not uniform across the Arctic Ocean and are established by the presence or absence of Pacific-origin water. [7] Pacific water enters the Arctic Ocean via the Bering Strait and, while crossing the wide Chukchi Shelf, is modified seasonally by productivity, heat exchange, and ice formation and melting. Inflowing waters include nutrient-rich water from the Gulf of Anadyr and fresher, lower nutrient Alaskan coastal water [Walsh et al., 1989]. Seasonal modification on the Chukchi Shelf produces two forms of Pacific-origin water that comprise the upper and middle halocline of the upper layer [Coachman and Barnes, 1961]. The upper halocline, due to summer modification, is characterized by a local temperature maximum between 31 < S < 32, low nutrient and high oxygen concentrations. The middle halocline, due to winter modification, is characterized by a temperature minimum near S = 33.1, high nutrient and lower oxygen concentrations. The temperature minimum of the middle halocline is reinforced by intermittent contributions of dense water from the Mackenzie Shelf [Melling and Moore, 1995]. [8] Within the Canada Basin water column the boundary between Pacific and Atlantic-origin water lies between the middle and lower halocline of the upper layer. Lower halocline water is formed on the Barents Sea Shelf [Jones and Anderson, 1986; Steele et al., 1995; Melling, 1998] and north of the Barents Sea [Rudels et al., 1996] and is modified by shelf input from Eurasian or Chukchi Sea Shelves [Salmon and McRoy, 1994; McLaughlin et al., 1996]. Low nutrient concentrations and an oxygen minimum characterize this water. [9] The Atlantic layer lies beneath the lower halocline and is traditionally identified by a temperature maximum at its core. However, temperature and salinity measurements in the eastern Nansen Basin [Schauer et al., 1997] identified two distinct branches of Atlantic water in the Arctic Ocean: one entering through Fram Strait (Fram Strait Branch (FSB)), and one entering through St. Anna Trough after seasonal modification on the Barents Sea shelf (Barents Sea Branch (BSB)). These observations confirmed the twobranch hypothesis proposed by Rudels et al. [1994]. Schauer et al. [1997] demonstrated that BSB water was colder and fresher than FSB water in the top 700 m and warmer and fresher below 800 m. They also showed that vigorous mixing occurred between these two branches downstream of the confluence zone. To reach the Canada Basin, Atlantic layer water passes through three upstream basins during which time it becomes colder and fresher. It also lies deeper in the Canada Basin than in upstream basins because Pacific-origin water thickens the Canada Basin halocline. 3. Methods [10] Oceanographic samples were collected at station A (73 N 144 W, 3300 m) in the southern Canada Basin in 1989, 1990, 1992, 1993, and Samples were also collected along two Beaufort Sea sections: in 1993 and 1995, northward from Tuktoyatuk Peninsula to station A and in 1995, westward from Banks Island toward the basin interior (Figure 1). Temperature, pressure, and conductivity measurements were collected during downcasts with either a Guildline ( ) or a Falmouth Scientific Instruments (1993, 1995) conductivity-temperature-depth (CTD) profiler. Instruments were calibrated before and after each cruise. Potential temperature (q) and potential density (s q ) were calculated using UNESCO [1983] algorithms. [11] Water samples were collected in Niskin bottles deployed from a wire or mounted on a General Oceanics rosette and tripped on the upcast. Salinity (S), oxygen, nutrient, and CFC-11 (1992, 1993, 1995) samples were analyzed onboard; 129 I samples were collected (1993, 1995) for laboratory analysis. Analytical methods are reported by McLaughlin [2000], and the pooled variance was as follows. S, Sp = , n = 33; O 2, S p = 1.57 mmol m 3, n = 85; NO 3, S p = 0.15 mmol m 3, n = 493; Si(OH) 4, S p = 0.16 mmol m 3, n = 496; PO 4, S p = 0.06 mmol m 3, n = 502; and CFC-11, S p = 0.07 nmol m 3, n = 26. The standard deviation for 129 I was 5 10% [Edmonds et al., 1998]. To remove interyear offsets, all deep casts below 1700 m were internally calibrated on the assumption that modification occurs extremely slowly in the 450- to 500-year-old deep layer [Macdonald et al., 1993; Schlosser et al., 1997] and is therefore undetectable over a 6-year period. Data were corrected by S < 0.001, q < C except in 1992, when q = 0.01 C, NO 3 < 0.4 mmol m 3,PO 4 < 0.04 mmol m 3, Si(OH) 4 < 0.3 mmol m 3, and O 2 < 1 mmol m Observations [12] Time series data from station A were examined together with comparable data from upstream source waters in the western and eastern Nansen Basin and from eastern Amundsen and southern Makarov basins (see Figure 1). Data were collected on the following expeditions: 1993, Polarstern ARK IVX/4, stations 21, 70, and 35 [Schauer

4 19-4 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure 2. Conductivity-temperature-depth (CTD) q/s plots from Canada Basin (CB) station A (solid lines), , and upstream stations (dashed lines). (a) S = (b) S = (c) S = FSB water is observed at s q = 27.9 at upstream stations and at s q > at station A; BSB waters are observed at s q = 28.0 at all stations. Abbreviations are NB, Nansen Basin; AB, Amundsen Basin; and MB, Makarov Basin. et al., 1997; Frank et al., 1998]; 1993, Canadian Coast Guard Ship (CCGS) Henry Larsen, station E [McLaughlin et al., 1996]; and 1994, CCGS Louis S. St. Laurent, station 20 [Swift et al., 1997]. Here station A measurements from the Pacific and Atlantic-origin components of the water column are reported separately Pacific-Origin Component [13] The mixed layer at the top of the water column exhibited significant variability between years but lacked a consistent temporal trend in q/s (Figure 2a). Surface salinity varied from S < 25 when temperatures were 0.5 to 1.3 C tos > 28 when temperatures were near 1.5 C. Measurements reflected annual variation from both local (Mackenzie River) and upstream (Alaska Coastal Current, Bering Sea, and Pacific Ocean) freshwater sources, as well as annual variations in atmospheric conditions affecting ice cover and surface drift. No temporal trend in geochemical properties at station A was observed. Nitrate concentrations were below detection in the mixed layer, silicate and phosphate were low, and oxygen and CFC-11 concentrations were high. [14] In the upper halocline, annual variability in temperature, salinity, and depth of the local q max was also observed (Figures 2b and 3). Variation was partly due to annual differences in the properties of inflowing Alaska Coastal Current and Bering Sea waters and partly due to variation in winter ice production and extent, which affect the depth and character of the polar mixed layer. Nutrient concentrations were higher, and oxygen and CFC-11 concentrations were lower than in the overlying mixed layer and exhibited no obvious temporal trend.

5 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS 19-5 Figure 2. (continued) [15] In the middle halocline the local temperature minimum was observed at S Temperatures were below 1.50 C in 1989 and 1990 and above 1.48 C in 1992, 1993, and Between 1989 and 1992 the depth of the local q min was m but shifted higher in the water column to 140 m in 1993 and 1995 (Figure 3a). Likewise, the nutrient maximum was located 50 m shallower in both 1993 and 1995, shown here in silicate (Figure 4), but was also evident in phosphate and nitrate profiles. Concentration of the nutrient maximum was also lower in 1995, decreasing from to 33.5 mmol m 3 in Shallowing of the middle halocline suggested that the vertical extent of Pacific-origin water in the Canada Basin water column had decreased. [16] One measure of this, the depth interval between q min (i.e., winter Pacific-origin water) and local q max (i.e., summer Pacific-origin water), showed that the vertical extent of Pacific-origin water decreased from m in to 65 m in Another measure is the location of the Atlantic/Pacific boundary, which is approximated by the intersection of silicate and 129 I profiles. Atlantic-origin water contains high 129 I concentrations from European nuclear reprocessing plant discharges [Smith et al., 1998], whereas Pacific-origin water contains practically none. In 1995 at station A the presence of Pacificorigin water in the middle halocline was identified by the silicate maximum near S = 33.1 at 140 m (Figure 5). Likewise, the 129 I maximum observed near S = 34.8 at 360 m identified the presence of Atlantic-origin water in the lower halocline. Accordingly, the Atlantic/Pacific boundary was located at 240 m. Shallowing of the silicate maximum in 1993 implies that the Atlantic/Pacific boundary depth also decreased, thereby decreasing the extent of Pacific and increasing the extent of Atlantic-origin water. [17] Although changes in nutrient concentrations at S = 33.1 could have derived from variability in Bering Sea winter water transport and properties, this could not be confirmed by available data. Comparison with silicate measurements from the Nansen, Amundsen, and Makarov basins at S = 33.1, however, showed that a decrease in

6 19-6 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure 2. (continued) silicate concentration could be explained by increased mixing with Atlantic-origin water (see Figure 4) Atlantic-Origin Water Observations [18] Changes in Atlantic-origin water were first encountered below S > 33.6 (180 m), where waters were colder in 1993 than previously. In the lower halocline at station A the temperature at S = 34.4 decreased slightly between 1989 and 1992 and then decreased markedly in 1993, where it remained in 1995 (Figure 2b). Despite its increased density, lower halocline water was found 40 m higher in the water column (Figure 3a). [19] Below the lower halocline, differences in the upper FSB component of the Atlantic layer at station A were also apparent (Figure 2c). In 1993 and 1995 the FSB was slightly colder (q max = 0.03 C), fresher (S = 0.03), and situated higher in the water column (z = 110 m) than previously. Temperatures were also lower below q max between < S < , and fine structure was apparent in the 1993 temperature-salinity plot. [20] The lower BSB component of the Atlantic layer, characterized by a local salinity minimum, was slightly fresher in 1993 at station A between 1200 and 2500 m and slightly warmer between 1400 and 2000 m than before. By 1995, however, the Atlantic layer s BSB component appeared markedly different. The q/s plot showed that freshening, evident only in the FSB component in 1993 between < S < , extended in 1995 from < S < and included both FSB and BSB components of the Atlantic layer. The water column was distinctly fresher between 500 and 2500 m in 1995 (Figure 3b) and warmer between 700 and 2200 m, indicating that warmer, fresher water had displaced colder, saltier water. Also, water at S = was found 150 m deeper in 1995, signifying that the BSB component occupied more of the water column than in previous years. [21] To investigate potential sources of this Atlanticorigin water change, measurements obtained upstream in the Nansen Basin in 1993 were used to identify Fram Strait (station 21) and Barents Sea (station 70) inflow character-

7 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS 19-7 Figure 3. CTD (a) potential temperature and (b) salinity profiles at Canada Basin station A, The shallow temperature minumim lay higher in the water column in 1993 and 1995, and waters below 500 m were much fresher in 1995 than in previous years. istics. Waters from both regions exhibited a range of temperature and salinity values, signaling the presence of water masses and not water types (see Figures 2b and 2c). Measurements from downstream stations in the Amundsen (station 35) and Makarov basins (stations E and 20) indicated that water from both branches mixed with ambient basin waters. The FSB component began warm and salty and then became progressively cooler and fresher along an isopycnal surface (s = 27.9) as it traveled from the Nansen to the Amundsen and Makarov subbasins. Conversely, the lower BSB component was initially colder and fresher and then became warmer and saltier en route along an isopycnal surface (s = 28.0). [22] In contrast, q/s curves within the Atlantic layer were rounded and smooth at station A, significantly different than the intrusion-filled curves observed upstream. The lack of fine structure suggested that Atlantic layer waters were mixed to a greater extent by the time they reached the Canada Basin and that warmer FSB water had not arrived. Also, the basin-to-basin progression of FSB water properties along s = 27.9 did not continue into the Canada Basin, where q max was found at s > Higher densities are likely derived from long-term temporal variability in inflow transport and properties of Fram Strait [Grotefendt et al., 1998] and Barents Sea [Loeng et al., 1997] waters. Smethie et al. [2000] observed that

8 19-8 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure 4. Silicate data at Canada Basin station A ( ) and upstream stations (a) in profile and (b) versus salinity. The silicate maximum at station A was shallower in 1993 and 1995, and concentrations at S > 34.4 were lower. both FSB and BSB waters in the Canadian Basin had been modified by mixing en route but found that FSB waters were more altered and must have mixed with ventilated waters. They suggested that East Siberian or Chukchi shelves were sources because waters of similar density were observed along the Chukchi shelf break at m. Nutrient and iodine measurements (station 7) [Swift et al., 1997; Carmack et al., 1997] identify that those Chukchi shelf break waters were of Atlantic origin. This suggests that lighter varieties of Barents Sea waters were transported eastward along the shelf break where they could be ventilated by direct shoaling or by mixing with briny shelf waters. [23] Comparison of Canada Basin and upstream q/s contours indicates that changes, observed simultaneously in all three components of Atlantic-origin water at station A, could be linked to Barents Sea outflow variability because it is the most saline of the shelf seas and has been shown to produce waters with this range of densities [Schauer et al., 1997; Steele and Boyd, 1998]. In 1993 and 1995, lower halocline water became colder and shallower; FSB water became fresher, colder, and shallower; and BSB water became fresher, warmer, and occupied more of the water column at station A. CFC-11 measurements showed that the water column was also much more ventilated in 1995 (Figure 6). The largest increase in CFC-11 occurred in the lower halocline and both components of the Atlantic layer, from 300 m (S = 34.4) to 1000 m (S = 34.9). Comparison of Canada Basin and upstream basin CFC-11 profiles clearly identified the Barents Sea as the primary source of this increased ventilation at depth. Downstream of Fram Strait inflow (station 21), waters were ventilated (4.8 nmol m 3 )

9 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS 19-9 Figure 5. Silicate (diamonds), 129 I (triangles) and salinity (solid line) profiles at Canada Basin station A, High silicate concentrations signify the presence of Pacific-origin water; conversely high 129 I concentrations identify the presence of Atlantic-origin waters. Their intersection approximates the location of the Atlantic/Pacific water mass boundary within the water column. to 400 m, but downstream of the Barents Sea outflow (station 70), waters were more ventilated (5.3 nmol m 3 ) and were evident much deeper in the water column to 1100 m. Although 1992 and 1993 station A CFC-11 profiles indicated that BSB waters were present in the Canada Basin, the increase in CFC-11 concentrations between 300 and 1000 m in 1995 provided compelling evidence that Atlantic-origin water composition had been changed by variability of Barents Sea outflow. [24] Oxygen concentrations at station A were also higher between 200 and 800 m (34.4 < S < 34.9) in 1993 and 1995 than in previous years. The oxygen minimum of the lower halocline was 50 m shallower and 20 mmol m 3 higher in concentration (Figure 7). Upstream, oxygen concentrations were high at both stations, and concentrations at S > 34.4 were 25 mmol m 3 higher at station 70 than station 21. Silicate (Figure 4b), nitrate, and phosphate concentrations were lower at 34.4 < S < 34.9 in 1993 and 1995, and over this salinity range, nutrient concentrations upstream were also lower at station 70 than at station I measurements made at two Canada Basin stations also showed that concentrations were much higher between 200 and 800 m of the water column at station A in 1995 than upstream at station B in 1993 (Figure 8). Although no 129 I data were available from stations 21 and 70, the appearance of elevated 129 I concentrations to 800 m in the Canada Basin

10 19-10 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure 6. CFC-11 profiles from Canada Basin station A ( ) and upstream stations. CFC-11 concentrations at station A were much higher in 1995 from m. Outflow of ventilated shelf water from the Barents Sea is evident at station 70 and at downstream stations in the Amundsen and Makarov basins. and 1400 m in the Makarov Basin indicated Barents Sea outflow. Together these data corroborated that Atlanticorigin water in the Canada Basin included a greater amount of Barents Sea water after [25] The difference in Canada Basin Atlantic-origin water composition between 1992 and 1995 was quantified using water mass analysis [cf. Macdonald et al., 1989]. Table 1 lists the values at s q = 28.0 from stations A (1992), 21, and 70 that were used in the inverse least squares calculation. Results showed that 1995 station A waters at this density contained 80% Canada Basin water, as defined in 1992, and 20% BSB water. This signaled that the composition of Atlantic water now included more water from the Barents Sea Canada Basin Sections [26] Observations made on two sections in 1993 and 1995 (see Figure 1) were used to study the transport of water masses within the Canada Basin. Comparison of q/s contours illustrates temporal and spatial change in Atlanticorigin water composition. At station FM, located over the continental margin (1700 m), halocline water between 33.1 < S < 34.4 was nearly identical in 1993 to water observed farther offshore in 1995 at station A (Figure 9a). Downstream at stations BB12 (700 m) and BB13 (1240 m), interleaving and mixing between ambient and colder water were evident over the same salinity range. [27] In 1993 the FSB q max component (Figure 9b) was colder and fresher at both inshore stations, AM10 (548 m, q = C, S = ) and FM (q = C, S = ), than offshore at station A (q = C, S = ). In 1995, q max occurred at a lower salinity. Downstream at station BB12, q max was C ats and the q/s contour was similar to station AM10 in 1993, located 410 km upstream. Farther offshore at station BB13, q max was higher and observed at a higher salinity (q = C, S = ). The 1995 salinity profile (not shown) identified that waters at station BB12 were fresher

11 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure 7. Oxygen versus salinity from Canada Basin station A ( ) and upstream stations. Concentrations at station A were higher at S > 33.5 in 1993 and than at station BB13 below 300 m (S > 34.7) and that the largest salinity difference (S = 0.015) was observed at 500 m. [28] Fresher BSB water was evident in 1993 at station FM at salinities S > but was not observed that year at station A. By 1995, salinities at station A were lower and almost identical to values observed at station FM. Downstream, BSB water was colder and fresher at station BB12, whereas offshore at station BB13, temperature and salinity values paralleled those observed upstream 2 years earlier at station A. [29] Barents Sea enriched Atlantic-origin water served, in effect, as a tracer of the progression of upstream waters through the Canada Basin, with the leading edge first appearing over the continental margin and then, 2 years later, both offshore and downstream. Transport offshore indicated lateral isopycnal spreading from the continental margin to the basin interior. Transport downstream signaled arrival of the leading edge at one station located over the continental margin. [30] Comparison of temperature-salinity plots along the 1995 section to the deep Canada Basin showed that the FSB component at S = was warmest at station AR4 (525 m) and that the BSB component at S = was freshest at stations AR6 (1233 m), AR7 (2030 m), and AR8 (2806 m). These observations identified the core of Atlantic-origin water in the boundary current; the lower halocline and FSB core were located over the continental margin, and the denser BSB core was located basinward and over the continental slope. Finding the Atlantic layer s FSB core above the continental margin was consistent with the location of a topographically teered boundary current [Holloway, 1987]. Location of the BSB core farther offshore, removed from the region of maximum topographical curvature, implied a weaker flow, but this was not so. [31] CFC-11 data were employed to estimate, in relative terms, the arrival of Barents Sea enriched Atlantic-origin water (Figure 10). In the lower halocline at S = 34.4, high subsurface CFC-11 concentrations, indicating fast delivery, were observed at the shallower stations AR4, AR5, and

12 19-12 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure I versus salinity from Canada Basin station B, 1993 (see also Figure 1); station A, 1995; and Makarov Basin station E, I concentrations were much higher between 200 and 800 m at station A in 1995 than upstream in the Canada Basin at station B in High surface water concentrations at station E identify that it was located on the Atlantic side of the Atlantic/Pacific front. AR6. Lower CFC-11 concentrations, indicating slow delivery, were observed at offshore stations AR7, AR8, and A. These observations suggested that flow was intensified over a region of sloping topography and weakened offshore. This was not the case for underlying FSB and BSB components. [32] At station AR7 in 1995, CFC-11 concentrations increased appreciably between 34.6 < S < 34.7 ( m) and then resembled concentrations at nearshore stations. CFC-11 concentrations likewise increased below S = 34.7 at station AR8, with a subsurface maximum observed at S = 34.8 near 400 m. In the FSB region of the water column, CFC-11 concentrations were similar between stations AR4 and AR8, suggesting a uniform rate of delivery. In the BSB region ( m) the highest Table 1. Source Water Mass Properties a Source Salinity q, C Si(OH) 4, mmol m 3 NO 3, mmol m 3 PO 4, mmol m 3 O 2, mmol m 3 CFC-11, mmol m 3 CFC-12, mmol m 3 CB FSB BSB a Abbreviations are CB, Canada Basin; FSB, Fram Strait Branch; and BSB, Barents Sea Branch.

13 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure 9. CTD q/s curves from 1993 and 1995 Canada Basin section stations. (a) S = (b) S = Freshening apparent at station FM in 1993 was evident offshore at station A in 1995 and downstream at station BB12. It was not apparent at station BB13 where properties were similar to those observed at station A in CFC-11 concentrations were found offshore at station AR8 and could not be attributed, as temperature and salinity measurements showed, to eddy transport. Thus the boundary current transporting denser BSB water was located farther offshore and its flow rate was not diminished. Lower CFC-11 concentrations at station A indicated that the boundary current did not extend this far offshore. [33] Downstream at stations BB12 and BB13 in 1995, CFC-11 measurements at S = 34.0 were comparable to 1993 station A measurements, indicating that increased ventilation observed upstream in the lower halocline was not apparent downstream. Similarity between CFC-11 concentrations in FSB and BSB components at stations BB12 and A in 1995, however, suggested that Barents Sea enriched Atlantic-origin water had arrived at station BB12 and that nutrient and oxygen concentrations were also similar at stations BB12, and A provided corroboration. Offshore at station BB13, CFC-11 concentrations in the FSB component were similar to those observed at station A in 1993 and ranged between 1993 and 1995 values in the BSB component below 600 m, suggesting that Barents Sea enriched waters had not fully arrived offshore. [34] Physical and geochemical data also provide a description of the size and structure of the boundary current in the Canada Basin. High CFC-11 concentrations, apparent in lower halocline water from stations AR4 to AR6, suggested that the boundary current over the continental margin was 120 km wide. High CFC-11 concentrations were similarly observed in both Atlantic layer components to 1500 m from stations AR5 AR8, indicating a width of 230 km over the continental slope. These findings differ from 1987 Canada Basin observations of a deep boundary current located over the slope, a few tens of kilometers wide at the core and negligible below 1000 m [Aagaard, 1989]

14 19-14 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure 10. CFC-11 versus salinity from 1993 and 1995 Canada Basin section stations. High CFC-11 concentrations from m at stations AR4 AR6 identify the location of the boundary current transporting lower halocline waters. Offshore, at stations AR7 AR8, lower halocline waters are less ventilated but underlying Atlantic layer waters are well ventilated, showing that the boundary current transporting FSB and BSB waters lies further offshore. and suggest that a change in Canada Basin circulation occurred between 1987 and Downstream along the Banks Island section, both components of the Atlantic layer were well ventilated, although the lower halocline was not. Thus the boundary current transporting both branches of the Atlantic layer traveled at a faster rate than the shallower current transporting lower halocline waters, a finding supported by the vertical velocity profile proposed by Aagaard [1989]. [35] Temperature and salinity profiles from the 1995 station A section revealed two other transport mechanisms. At stations AR5 and AR6, located above the continental margin, temperature decreased and salinity increased in a bottom boundary layer 40 m thick (Figure 11). Temperature and salinity values in this layer at station AR5 were similar to those observed farther offshore and 160 m deeper, suggesting that dense water had been lifted upslope, perhaps as a result of upwelling or tidal effects or as a manifestation of the Neptune effect modeled by Holloway [1987]. Also at station AR5, cold high-nutrient middle halocline water (S = 33.1) was observed 60 m deeper than at other stations, signaling the presence of a shallow (250 m) eddy. 5. Discussion [36] Data collected between 1989 and 1995 in the Canada Basin demonstrated that effects of upstream change were first observable downstream in 1993 in both Pacific- and Atlantic-origin components of the water column. They showed that the extent of Pacific-origin water decreased and was supplanted by Atlantic-origin water, which by 1995, contained a larger contribution from the Barents Sea than in the past. Downstream effects, first apparent in 1993 at station FM along the continental margin, were detectable 2 years later offshore at station A as well as 410 km downstream at station BB12. The following discussion examines what these

15 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Figure 11. CTD (a) temperature and (b) salinity profiles from 1995 Canada Basin section stations, illustrating the presence of a bottom boundary layer containing dense water at stations located over the continental slope. changes in the Canada Basin between 1989 and 1995 disclosed about changes in the larger Arctic Ocean system and investigates factors that prompted an increase in Barents Sea outflow Displacement of Pacific-Origin Water [37] Pacific-origin water occupied less of the upper layer in the water column in 1993 and 1995, and the Pacific/ Atlantic water mass boundary was shallower than in previous years. Shallowing may be explained by reduced Pacific water inflow through Bering Strait; however, this was not supported by direct measurements [Roach et al., 1995; Münchow and Carmack, 1997]. Freshening of Pacific-origin inflow during the 1990s (A. Aagaard, personal communication, 2000) and increase in Atlantic-origin inflow to the Canada Basin from 1993 offered the best explanation for shallowing of the Pacific/Atlantic water mass boundary. Within the larger context of the Arctic system an increase in Atlantic-origin inflow was also consistent with Pacific-origin water displacement from the Makarov Basin in the early 1990s. [38] The increase in Atlantic-origin inflow can be estimated from the reduction in Pacific-origin water. Disappearance of Pacific-origin water from the upper 200 m of the Makarov Basin required an Atlantic-origin inflow of 145,000 km 3. Midway around the Canada Basin at station A, a decrease of 40 m in the thickness of Pacific-origin water required an inflow of 42,000 km 3.

16 19-16 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS Assuming that these developments occurred between 1990 and 1993 (see section 5.3), the transport of Atlantic-origin water to the Makarov and Canada basins appears to have increased by 1.2 Sv (1 Sv = 10 6 m 3 s 1 ) and 0.3 Sv, respectively. [39] A decrease in the extent of Pacific-origin water also influenced the freshwater content of the Makarov and Canada basins. Freshwater content of the Makarov Basin, relative to a mean Arctic Ocean salinity S = 34.8 [Aagaard and Carmack, 1989], was estimated to be 12 m from salinity data collected near the Lomonosov Ridge in 1979 [Moore et al., 1983]. Fifteen years later at a nearby location [Swift et al., 1997] (station 29), it had decreased significantly to 7 m. In the Canada Basin the freshwater content decreased slightly from 19 to 17 m between 1992 and These changes coincide with observations of along-shelf transport of Eurasian River water into the Makarov Basin [Ekwurzel et al., 2001; Guay et al., 2001], whose effects on the freshwater content were overshadowed by the effects of the shift in the Atlantic/Pacific front. [40] Other studies provide clues about where this displaced Pacific-origin water went. From data collected in the Lincoln Sea, Newton and Sotirin [1997] observed that the temperature maximum/minimum of the upper layer was absent in , weak in 1991, and well defined in They attributed these changes to increased transport of waters from the Canadian Basin. Their observations coincided with reports of an absence of Pacific-origin water in the Makarov Basin, near the Lomonosov Ridge in 1991 [Anderson et al., 1994] and at the southern boundary near the Mendeleyev Ridge in 1993 [McLaughlin et al., 1996]. Pacific-origin outflow was also reported farther downstream in 1991 above the Morris Jesup Plateau at stations located along the continental slope [Anderson et al., 1994], but the silicate-salinity curve appeared incomplete, implying the presence of middle but not upper halocline waters. Unlike the classic Pacific-origin nutrient maximum curve, silicate concentrations at these locations were mmol m 3 between S = 32 and 32.8 and then increased sharply to 30 mmol m 3 at S = 33.2 (cf. Figure 4b). [41] Observations from the Makarov Basin, Lincoln Sea, and Morris Jesup Plateau connect upstream to downstream events. Disappearance of Pacific-origin water from the Makarov Basin by 1993 corresponded downstream with the arrival of progressively warmer Pacific-origin water in the Lincoln Sea that began in 1991 and peaked in The offset (nonvertical) structure of the Atlantic/Pacific front upstream near the Mendeleyev Ridge was mirrored downstream in the order that Pacific-origin water arrived: first, middle halocline water, identified in 1991 by increasing temperatures between 32.5 < S < 33 and Pacific-origin silicate concentrations at salinities S > 33.2 and then upper halocline, identified in 1993 by higher temperatures between 31.5 < S < Moreover, the time series temperature observations in the Lincoln Sea indicated that Pacific-origin water was displaced from the Makarov Basin over a period of 3 years. Where these waters went beyond the Lincoln Sea was unclear. Belkin et al. [1998] proposed that the great salinity anomalies (GSA) of the 1990s were enhanced by freshwater outflow from the Archipelago. Perhaps the large volume of freshwater that exited the western Arctic Ocean between 1991 and 1993 contributed Change in Atlantic-Origin Water [42] Atlantic-origin water in the Canada Basin was also marked by a change in composition and included more waters from the Barents Sea in 1995 than in the past. On the basis of the increased depth (150 m) of BSB water in the Canada Basin, an estimate of the increased outflow from the Barents Sea can be made. Given that the distance from the Mendeleyev Ridge to station A was 2300 km, assuming that the BSB boundary current s width was 230 km and that change occurred over 3 years, an outflow of 0.9 Sv from the Barents Sea was required. Including the volume needed to reduce the extent of Pacific-origin water by 40 m, an additional 0.9 Sv of Atlantic-origin water, primarily from the Barents Sea, entered the Canada Basin between 1992 and Upstream Events [43] Atmospheric and ocean studies offer evidence to explain an increase in Barents Sea outflow. Walsh et al. [1996] showed that a seasonally robust decrease in central Arctic sea level pressure occurred between 1987 and 1994, an event they associated with an increase in cyclonic wind forcing. This event also coincided with the highest values of the century-long North Atlantic oscillation record [Dickson et al., 2000]. Modeling studies have identified that Arctic atmospheric circulation is bimodal, that it influences upper ocean circulation, and that in 1989 a shift toward increased cyclonic circulation occurred [Proshutinsky and Johnson, 1997; Thompson and Wallace, 1998; Maslowski et al., 2000]. According to Johnson et al. s [1999] study of regime transitions during the period , the 1989 transition exhibited the largest decrease in sea level pressure and affected a region far greater than the three earlier transitions to cyclonic circulation. [44] Increased cyclonic circulation in the atmosphere produced certain attendant effects, including a warmer and stronger inflow of Atlantic water to the Arctic Ocean via both branches [Dickson et al., 2000]. Johnson et al. [1999] also associated increased cyclonicity with strengthening of the Icelandic Low in winter, extending it eastward over the Barents, Kara, and Laptev Seas. This atmospheric state favors Atlantic water transport into the Barents Sea over transport into Fram Strait [Zhang et al., 1998]. Using observational and model data, Polyakov et al. [1998] reported that the 1990s cyclonic regime was accompanied by an intense inflow of Atlantic water into the Arctic Ocean that entered primarily via the Barents Sea. [45] Although few direct measurements of Atlantic water inflow to the Barents Sea exist, Loeng et al. [1992] inferred that the transport rate was high in 1989 because they observed a rapid increase in temperature that year over the entire southwestern Barents Sea. Similarly, few direct measurements of Barents Sea outflow exist [Loeng et al., 1997], and variability can only be inferred from hydrographic data. Three quarters of a layer of cold dense bottom water 100 m thick, observed in the eastern Barents Sea in 1982, was replaced the following year by warmer water [Midttun, 1989]. This sequence of events was observed again between 1988 and 1989 and prompted Loeng et al.

17 MCLAUGHLIN ET AL.: CANADA BASIN AND FAR-FIELD EFFECTS [1997] to speculate that dense water outflow in 1982/1983 and 1988/1989 could have exceeded 4 Sv. [46] Variability in Barents Sea outflow may be linked to two sources. One source is dense water formed in polynyas near the St. Anna Trough [Martin and Cavalieri, 1989] and is likely characterized by annual outflow. The second source is dense water produced in the eastern Barents Sea, which has been shown to leave the shelf episodically. Eastern basin outflow may be related to atmospheric conditions or to production of a critical volume of dense bottom water. Although the 1982/1983 event is a possibility, both atmospheric and oceanographic literatures have identified the late 1980s through the early 1990s as a time of unprecedented change in the Arctic and that in 1989, Atlantic water inflow increased to the Barents Sea, and a significant volume of dense water exited the Barents Sea via the St. Anna Trough. With this date the rate at which upstream changes were delivered downstream to the Canada Basin can be estimated Boundary Current Spreading Rate [47] The 1995 observation of a colder, fresher, and more ventilated Atlantic layer signaled that waters incorporating the 1989 Barents Sea outflow were starting to appear in the Canada Basin, having traveled 5100 km in 6 years. Thus the boundary current speed in the Atlantic layer was 3 cms 1 during this time, similar to the 2 3 cm s 1 measured at 1400 m over the Amundsen Basin side of the Lomonosov Ridge in 1979 [Aagaard, 1989]. [48] Transient tracer ages can also be used to infer spreading rates. Assuming CFC-11 saturation to be 85% [Frank et al., 1998] and applying a boxcar mixing model [Wallace et al., 1992], the ages of FSB and BSB waters in the Canada Basin were estimated to be 22 and 26 years, respectively. These ages imply FSB and BSB mean spreading rates of 1.0 and 0.6 cm s 1, values similar to Smethie et al. s [2000] estimate of 0.5 cm s 1 for BSB waters in the Canadian Basin calculated from 1996 tritium/ 3 He measurements. Renewal estimates based on CFC concentrations thus differ from those derived by tracking changes in water mass structure. One plausible explanation for this difference is that water mass observations identify the arrival of the leading edge of Barents Sea enriched Atlantic-origin in the Canada Basin, whereas tracer estimates, due to mixing and dilution along the transport pathway, represent a mean spreading rate Arctic Atmosphere-Ocean System [49] Changes in the Canada Basin have been traced upstream to a 1989 coupling between an increase in Barents Sea outflow and an atmospheric transition of unprecedented magnitude. This finding prompts discussion of whether the 1989 transition to cyclonic circulation also produced broader Arctic Ocean changes. These changes include: warmer Atlantic layer temperatures observed first in the Nansen Basin in 1990 and then in the Amundsen and Makarov basins in 1993, a shift in the Atlantic/Pacific water mass boundary location by 1993, and increased ventilation at depth in Makarov Basin waters between 1993 and [50] One effect of cyclonic circulation is transport of warmer Atlantic water northward. Although entry via the Barents Sea was favored, observations in the Nansen Basin confirmed that warmer Atlantic water also entered the Arctic Ocean via Fram Strait. Other effects of cyclonicity are an increased volume of Atlantic water delivered to the Arctic Ocean and a redirection in the transport of Laptev Seawaters from across-shelf to along-shelf. Together these two effects explain the relocation of the Atlantic/Pacific front eastward from the Lomonosov Ridge to the Mendeleyev Ridge. Displacement of Pacific water resulting from relocation of the water mass front could also force a greater amount of freshwater from the Canadian Basin through the adjacent Archipelago to the Labrador Sea. The 1989 transition to cyclonic atmospheric circulation also coincided with an outflow of dense water from the Barents Sea into the Nansen Basin. This ventilated cold water was observed in 1993 downstream to depths >1500 m in the Makarov Basin and was observed to a similar depth in the Canada Basin in In addition, the thickness of the boundary current in the Canada Basin was extended by 500 m in The fact that CFC-11 concentrations and the depth of ventilation in the Makarov Basin increased between 1993 and 1994 (see Figure 6) foreshadows similar development in the Canada Basin in years to come. Thus cyclonicity serves to explain both the far-field effects of the Barents Sea on the Canada Basin and changes observed in the larger Arctic Ocean in the 1990s. [51] The interplay of atmosphere and ocean outlined above points to a hypothesis that portrays two modes of atmosphere-ocean interaction, extending Proshutinsky and Johnson s [1997] concept of two atmospheric regimes to encompass the ocean s response (Figure 12). One part of the hypothesis describes an anticyclonic regime characterized by reduced cyclonic circulation, which in turn engenders a specific ocean response, namely, the reduced transport of Atlantic water, which upon arrival in the Arctic Ocean, is colder. Weaker Atlantic inflow is likewise associated with location of the Atlantic/Pacific front over the Lomonosov Ridge and freshwater export primarily through Fram Strait. During the anticyclonic mode the Arctic Ocean boundary current appears thinner and may be slower. [52] Alternatively, the cyclonic regime is characterized by increased cyclonic circulation, which acts upon the ocean to increase delivery of Atlantic water that is warmer on arrival. Increased Atlantic inflow is likewise associated with eastward location of the Atlantic/Pacific front over the Mendeleyev Ridge and with increased freshwater export through the Archipelago. During the cyclonic mode the Arctic Ocean boundary current appears to be thicker and transport is possibly faster. Although the modeling community has chronicled atmospheric oscillations over the past half-century, the ocean s response has been recorded recently and only after an atmospheric transition of extreme magnitude. Whether the post-1989 shift from an anticyclonic to cyclonic regime will be followed by an equally dramatic shift to an anticyclonic regime remains to be seen, and the ocean s response has yet to be documented. [53] To date only one of the changes observed upstream is yet to be recorded in the Canada Basin. Warm Atlantic water observed in the Nansen, Amundsen, and Makarov basins was not evident by 1995 in the Canada Basin. However, given that BSB outflow effects took 6 years to reach the Canada Basin and that the FSB s speed was half that of the BSB [Frank et al., 1998], warmer FSB water