University of Alaska Fairbanks	School of Fisheries and Ocean Sciences	Institute of Marine Science

Mean Circulation

Figure 1 and Figure 2 show the mean circulation field based on a composite of the record length averages from a variety of current meter records over the Chukchi shelf. The mean flow offshore Cape Lisburne is northward at ~3 cm s^-1 and is representative of a broader northward flow that extends at least 150 km offshore [Coachman and Aagaard, 1981; Woodgate et al., in prep.]. Some of this flow rounds Cape Lisburne and continues along the Alaskan coast and some of it continues into the Central Channel. Within the Central Channel, the mean flow is northward at ~8 cm s^-1 (identical to that from 1991-92; Weingartner et al., 1998). The northward flow through the Central Channel appears to be a permanent feature of the shelf circulation field and it constitutes a third branch by which water from Bering Strait crosses the Chukchi shelf. The other two are Barrow Canyon where the mean flow is about 20 cm s^-1 and along the east wall of Herald Shoal where the mean flow is about 12 cm s^-1 (although based on only a single year of measurements [Woodgate et al., in prep.]). On the outer shelf (C3) the mean flow was ~5 cm s^-1 northeastward while along the south flank of Hanna Shoal (C2) the mean flow was ~4 cm s^-1 eastward. It appears that Hanna Shoal diverts central shelf waters northeastward toward the slope and eastward along the shoal's southern flank. The AC moorings were within the Alaska Coastal Current and these indicate a mean northeastward flow of ~5 cm s^-1. The data further suggest that the coastal current converges with the eastward flow south of Hanna Shoal before entering Barrow Canyon, where the mean flow was northeastward (downcanyon) at ~20 cm s^-1. Thus the canyon drains a relatively broad area of the shelf including the Alaska Coastal Current and some portion of the central shelf.

Cross-Chukchi Transects: Currents and Water Properties

Locations of selected hydrographic stations from 1992, 1993 and 1995 cruises on the R/V Alpha Helix are presented in Figure 3. Heavy ice and poor weather prevented similar sampling in 1994.
 

Figure 4a-c shows 15-minute averages of ADCP current vectors at 18-m depth in 1993, 1994, and 1995 (vectors at other depths are similar and not shown). Several features are consistent in each year. Along the eastern wall of Herald Valley the flow was always northward, with maximum speeds of ~50 cm s^-1, but along the western wall the flow was always weak or variable. Most of the northward flow within Herald Valley presumably continues down this channel. However, north of ~71^oN and east of ~175^oW the observations suggest that some of the flow veers northeastward over the central shelf and eastward along the north side of Herald Shoal where it merges with the northward flow emanating from the Central Channel. Flow along the south side of Hanna Shoal was weaker and more variable than elsewhere. Ice conditions in 1993 permitted sampling over Hanna Shoal where the flow was very weak and without a consistent direction (Figure 4a). In 1993 and 1994, the flow in Barrow Canyon was northeastward at 50 - 100 cm s^-1. The 1995 cruise captured a wind-forced reversal of the Alaska Coastal Current, very similar to that described by Johnson [1989], that enveloped the northwest coast of Alaska.

Figures 5a, b, and c show the autumn water mass properties from 1992, 1993, and 1995 along the stations spanning the Chukchi shelf (Figure 3). Temperature, salinity, and fluorescence along these sections are contoured as a function of depth and horizontal distance. The sections are viewed looking northward with Herald Valley on the left and Barrow Canyon on the right of each panel.

Three different water masses occupied Herald Valley in each year. On the east side of the canyon, the water column was well mixed and consisted of moderately salty (~32.5) and warm (>0.0^oC) Bering Sea Water. This water mass is the predominant water mass in Hope Sea Valley in fall [Coachman et al., 1975; Weingartner et al., 1999]. In 1992 fluorescence values were maximal in Herald Valley and associated with the Bering Sea Water mass. The water column along the west wall of the canyon was always stratified. Near-surface waters were cold (<0.0^oC) and fresh (28.0 - <32.0 psu) and products of ice melt. In 1992 and 1995 heavy concentrations of rotting ice covered the west side of the valley, but the east side was ice-free. Although Herald Valley was ice-free in 1993 cold, low-salinity water extended eastward from Wrangel Island to about the center of the channel. This water might have been the remains of local ice melt locally or it could have been advected southeastward from the northern Wrangel Island where ice was still present. Subsurface waters along the west wall of Herald Valley were salty (32.5-33.0 psu) and near freezing (<-1. 0^oC). Either they were shelf waters remnant from the preceding winter [Coachman et al., 1975] or they were advected southward from offshore. The juxtaposition of these three water masses (ice-melt, winter water, and Bering Sea Water) within Herald Valley formed a cross-canyon density gradient, which, if in thermal wind balance, implies a southward baroclinic flow along the west side of the canyon. In contrast, the northward flow of Bering Sea Water along the east wall is forced by the secular pressure gradient. These opposing flows created a strong (~10^-5 s^-1) cross-canyon cyclonic shear in the along-canyon velocities evident in the ADCP data (Figure 4).

In 1992 and 1995 surface waters over Herald Shoal were fresher than the waters along the east wall of Herald Valley and within the Central Channel. Heavy concentrations of rotting ice covered the shoal in 1992 and water temperatures were <2.0^oC along the ice edge. Although ice was absent in 1995, warm (>4 °C), low salinity water (<31 psu) capped the shoal. We suspect that this water is a remnant of ice-melt that warmed during the mild summer and fall of 1995 [Weingartner et al., 1999]. Our results suggest that waters over Herald Shoal are replaced only slowly, consistent with Martin and Drucker’s [1997] conclusions that a Taylor column forms over the shoal. (There was no ice melt signature over Herald Shoal in 1993 but the ice retreated very early that summer so that any meltwater could have been advected away by the winds prior to our arrival.)

Within the Central Channel the stratification was always weak and the fluorescence substantially higher (except 1995) in comparison to adjacent stations. Central Channel bottom waters were moderately salty (32.0 – 32.5 psu) and cool (0 – 2.0^oC), and their nutrient, d¹⁸O [Cooper et al., 1995] and barium signatures [Falkner et al., 1994] were different from adjacent stations and suggestive of Bering Sea Water.

Returning to Figures 5a-c note that between the Central Channel and Barrow Canyon the shelf was always stratified with meltwater occupying the surface layers and the bottom water containing mixtures of Bering Sea Water, Alaska Coastal Water and remnant winter water. The latter was also prevalent atop Hanna Shoal in 1993 where cold (<0.0^oC), salty (>32.2) water occurred.

Finally, the 1992 and 1993 sections show the thermohaline structure at the head of Barrow Canyon. The canyon was stratified in 1992 with meltwater capping warm (2 – 3^oC), and dilute (~31.0) Alaska Coastal Water, while in 1993 the water column within the canyon was unstratified and contained warm (~5.0^oC), low-salinity (~31.0) Alaska Coastal Water.

Central Channel Transport

The topography, property distributions, and ADCP data from the Central Channel suggest that the northward flow here is ~50 km wide (cf. Figures 4 and 5) and ~45 meters deep. Along with the mean flow of 8 cm s^-1, these geometric scales suggests a northward transport of ~0.2 Sv on annual average through the Central Channel, which is ~¹/₄ of the average Bering Strait transport. The table shows the mean (+ the 95% confidence level) monthly north component of velocity (first entry in each row) and transport (second entry) in the Central Channel for both years. The confidence limits were computed from the variance of the 6-hourly low-pass filtered data and an effective number of degrees of freedom/month of ~7 (based on an integral time scale of ~4 days). Red bold values are statistically significant at the P < 0.05 significance level.

Note that in both years the transport was a maximum in summer but not statistically different from zero from November through February.

Wind-forced Circulation Variability

Although the mean flows over the shelf are less than 10 cm s^-1 (outside of Barrow Canyon), current fluctuations can be substantial. For example, maximum current speeds are ~100 cm s^-1 in Barrow Canyon and 30 - 50 cm s^-1 elsewhere. At each mooring site, most of the current variability is aligned along the principal axes of variance, which are approximately parallel to the local isobaths and oriented in nearly the same direction as the mean flow. About 97% of the flow variance along the eastern wall of Barrow Canyon (EUBC) is aligned along the canyon axis and even over the weakly sloping central shelf most of the variance (67% at C2 and 82% at C3) is aligned along the principal axis.

Figure 7 shows time series of the 35-hour low-pass filtered current and wind components, each projected onto their principal axis of variance, at C1, C2, C3, and EUBC for 1994-95. Current reversals occurred most often between October and January and were more frequent in Barrow Canyon than at the central shelf sites. The current time series suggest a spatially coherent flow field varying with the winds. The first empirical orthogonal function (EOF) for the currents shown in Figure 7 accounts for ~76% of the variance indicating that current variations were coherent over a spatial scale of at least 300 km. (By comparison the spatial coherence scales are 25 - 60 km on the deep and narrow shelf along the U.S. west coast [Winant, 1983; Dever, 1997]).

The coherence squared (g²) and phase (f) spectra between the time amplitude function of the first EOF and the alongshore winds (Figure 8) shows that the currents are significantly coherent with the winds over most frequencies with the winds leading the currents by ~1 day.

Seasonal Evolution of Temperature and Salinity

The seasonal evolution of temperature and salinity in 1994-95 on the Chukchi shelf is discussed with reference to their time series (Figure 9). The annual temperature cycle includes: 1) cooling to near freezing temperatures in late fall – early winter, 2) nearly constant temperatures at the freezing point from winter through late spring, and 3) warming in spring and summer due to the arrival of warmer water from the Bering Sea. However, there are noteworthy differences among the various sites in both the shorter period variations and in the phasing and magnitude of the major seasonal thermal transitions.

For example, along the south flank of Hanna Shoal (C2) temperatures began decreasing in late December; more than a month later than at the moorings to the south or west. Moreover, temperatures here did not rise above 0^oC until late August whereas at the other locations (except for EUBC), temperatures rose to 0^oC 1 – 2 months earlier. The delay in winter cooling and in summer warming at C2 is consistent with the shipboard data that suggested that the area along the south flank of Hanna Shoal is relatively isolated from other portions of the shelf.

Note also that relatively warm water (~-1.0^oC) was observed during the latter half of December at AC2 and from mid-January to early February at C1, implying that portions of the shelf had not cooled to the freezing point until well into winter. This contrasts within 1991-92 [Weingartner et al., 1998] and 1993-94 (shown below) when the northeast Chukchi shelf cooled to the freezing point by early December and in a more spatially uniform manner. Reasons for these differences are presented in Section 5.

The Barrow Canyon (EUBC) record also differs from the others. In part some of these differences are because the Seacat was at a deeper depth than elsewhere. However, EUBC was also bathed on two occasions in November 1993 by upwelled Atlantic Water with temperatures of between –0.4 and –0.8^oC and salinities of ~34.5 psu. Upwelling of Atlantic Water into the canyon is not unusual [Bourke and Paquette, 1976; Johnson, 1989; Aagaard and Roach, 1990; Weingartner et al., 1998], especially in fall and winter. However, the absence of Atlantic Water at C2 and AC2 suggests that the upwellings do not extend far onto the shelf. Following the November upwelling events, temperatures at EUBC collapsed to the freezing point. Temperatures rose again in late December in conjunction with low salinity water (~32.2 psu) implying that the water in the canyon was derived from the shelf. Indeed, the temperature and salinity of this event suggest that it might have entered the canyon from along the south flank of Hanna Shoal (C2).

The seasonal changes in salinity are discussed in relation to the time series in Figure 9. At the beginning of each record, salinities ranged from 32.5 to 33.0 psu and decreased through fall, although the decrease was neither spatially uniform nor constant at a particular site. For example, at AC1 salinities decreased to 31 psu in late October as the temperature decreased to the freezing point. Salinity then began to increase, at first sporadically, and then more steadily from ~32 psu in mid-January to a maximum in late May of ~32.7 psu. Thereafter salinity decreased with the arrival of warmer water from the Bering Sea. In contrast, salinities at AC2 slowly decreased from 33.5 psu in October to 32.7 psu in December. From January through early March, salinities varied, sometimes rapidly (e.g., the latter half of January and in late February), between 31 and 33.5 psu. The rapid fluctuations could be intrusions of cold, saline water from coastal polynyas (which are discussed later). Following these rapid fluctuations, salinity steadily rose from ~32.0 to ~32.6 psu by May and thereafter decreased, with this decrease again coinciding with the appearance of warmer water. Salinities evolved more uniformly at C1, C2, and C3. They decreased more rapidly in the Central Channel (C1) and on the outer shelf (C3) than south of Hanna Shoal (C2), but were relatively low (31 – 31.5 psu) at all of these sites from December through February or March. Afterwards, salinities increased over a one month period, initially at C1 and later at C2 and C3, and then remained steady (~32.5 - ~32.8 psu) from May through June. Again the salinity declines that began in June (and August at C3) coincided with increasing water temperatures. The Barrow Canyon (EUBC) record also shows a fall decrease, albeit interrupted by the upwellings noted earlier. Throughout winter there were both low (~32 psu) and high salinity (~33.8 psu) outflows, reflecting additions of low salinity water interspersed with episodic releases of dense water from coastal polynyas. From late April onward, the salinity remained relatively constant varying from 32.5 to 33 psu.

In aggregate, the salinity records suggest a variety of dilute and moderately saline water masses over the Chukchi shelf from fall through winter. The circulation at this time was variable, but weak in comparison to other times of the year, so that the shelf was flushed slowly compared to other seasons (Figure 7 and Table 1). Consequently, shelf waters underwent considerable amount of local modification. This included horizontal mixing at the head of Barrow Canyon and vertical stirring by cooling and wind mixing in fall. The mixing was interspersed by occasional injections of more saline water from growing ice, and, within Barrow Canyon, by upwelled Atlantic Water. Beginning in late winter, when the shelf flow increased and was less variable, the shelf was replenished by a nearly uniform water mass with near-freezing temperatures and salinities between 32.5 and 33 psu. Beginning around June, this winter water began to be replaced by warmer, low-salinity summer water. In both cases, the replenishment proceeded systematically beginning first in the south and then spreading northward. The properties of both of these water masses were set on the northern Bering Sea shelf where ice formation occurs later and retreat earlier than on the Chukchi shelf. Data from Roach et al., [1995] show waters with these characteristics entering Bering Strait some 1 – 2 months prior to their arrival on the northern Chukchi shelf. Evidently the northern Bering winter and summer water masses crossed the Chukchi shelf without substantial modification in 1995. This is not surprising because the Chukchi shelf is effectively insulated from the atmosphere by ~2 m of ice from late winter through mid to late summer.

Interannual Variability

In many respects the characteristics of the shelf flow in 1994-95 were similar to those from 1991-92 discussed by Weingartner et al. [1998]. However, the fall-winter evolution of shelf thermohaline properties proceeded in a very different fashion in these years. In 1991-92, temperatures decreased to the freezing point by mid-December everywhere. That decrease was followed by a steady increase in salinity with maximum salinities of ~34 psu attained in February. In 1994-95, cooling was not complete until February and salinities remained relatively low throughout the winter. A further difference between these two years is shown in Figure 10, which compares the modal water mass properties from C2 and C2-93 along the south flank of Herald Shoal. In 1994-95, there were two modes, both at the freezing point. The dominant one had a salinity of ~32.7 psu and the secondary one was fresher (~31.5 psu). In 1993-94 five relatively distinct modes were observed at C2-93. All had temperatures close to or at the freezing point with salinities ranging from 31.3 - 34.5 psu. Each mode eventually enters the Arctic Ocean through either Barrow Canyon or across the shelf north of the Central Channel and ventilates slope waters at depths that depend upon the density of the shelf outflow and the slope stratification [Gawarkiewicz, 1998]. The depths denoted in Figure 10 are those that a particular mode would descend to along the Chukchi continental slope assuming that no mixing occurs as the parcel proceeds across the shelf.

The seasonal evolution in salinity also differs substantially between these years. This is illustrated for both years in Figure 11 which shows the salinity records from Central Channel (C1) and along the south flank of Hanna Shoal (C2). While fall and spring salinity values are similar in both years at both locations the differences in winter are striking. In 1993-94, minimum salinities occurred in December and maximum salinities occurred in January and February. The high salinity events occurred as distinct pulses of 2-weeks to a month duration. The salinities vary rapidly at the beginning and end of each pulse, changing by as much as 1 – 3 psu over a few days. In 1994-95, minimum salinities occurred from December through February and maximum salinities occurred in April and May with the salinity transitions being much more gradual.

The differences in water mass properties among these years adds to an expanding suite of observations that underscore the large interannual variability on arctic shelves particularly in regards to the formation of dense water [Aagaard and Roach, 1990; Roach et al., 1995, Melling, 1993; Weingartner et al., 1998]. We believe that the differences between the winters of 1993-94 and 1994-95 were largely related to the seasonal evolution of the shelf ice cover. Figure 12 shows changes in open water area as determined from SSM/I satellite imagery [Cavalieri, 1994] along with the mean monthly winds. In fall 1993, large expanses of the Chukchi shelf were ice-free through November. Open water area then diminished rapidly from late November through December as a result of both ice growth and the southward advection of ice over the shelf by the northeasterly winds. These were strong and persistent through January, creating polynyas along the Alaskan coast, wherein dense, hypersaline (>34 psu) formed [Schumacher et al., 1983; Martin and Cavalieri, 1989; Cavalieri and Martin, 1995; Weingartner et al., 1998]. In contrast, thick (~2 m), multi-year ice enveloped the Chukchi shelf north of 70^oN in October 1994 because persistently strong southward winds blew from September through October. The heavy ice conditions of fall 1994 effectively insulated the shelf and delayed cooling, particularly at the more northerly mooring sites, where ice concentrations were heaviest. This is one reason why temperatures at AC1 decreased to the freezing point in fall 1994 earlier than did the temperatures at the more northerly moorings (Figure 9).

Our data sets show that in some winters the inflow through Bering Strait undergoes very little additional salinization while crossing the Chukchi shelf, whereas in other years it is substantially modified on this shelf. Formation of hypersaline water on the northeast Chukchi shelf depends on three factors: fall ice extent over the shelf, the formation of coastal polynyas in winter, and the salinity of the Bering Strait inflow. The last of these sets the initial conditions of the shelf water and ultimately influences the salinity of the dense water produced during ice formation on the Chukchi shelf. The first two depend on the "local" winds, whereas the last is established by "remote" processes occurring over the Bering Sea (and possibly over the North Pacific).

Baroclinic Circulation Effects

The current meter data suggests a circulation field consistent with geostrophic, barotropic dynamics involving the (constant) secular pressure gradient and the (time-varying) shelf-wide response to regional wind forcing. A closer look at the salinity records suggest that baroclinic effects might be significant on occasion at least over certain portions of the shelf. Figure 13 shows salinity time series from 1994-95 pairing C1 with C3 and AC2 with C2. For each of these meridionally separated pairs we have indicated periods when the salinity gradients between them are relatively large. Two types of gradients can be distinguished. The first type, denoted by time periods A, B, and C in Figure 13, reflects the seasonal flushing of the shelf due to the Bering Strait inflow. The inflow salinity varies seasonally and similar to the advective time scale for water to cross the Chukchi shelf. It takes ~3 months for water from Bering Strait to reach the slope through Barrow Canyon [Weingartner et al., 1998] and probably twice as long by way of the Central Channel route. Consequently the meridional salinity gradients change in magnitude and sign throughout the year. For example, between C1 and C3, the salinity gradient is positive in December and January, reverses from February through March and is negligible thereafter. Although our moored array cannot quantify the spatial scale of these gradients we suggest that they are relatively broad and therefore drive a weak baroclinic circulation. For example, application of the vertically integrated thermal wind relation, (assuming no flow at the bottom) for the period of May through July between AC2 and C2 forces an eastward baroclinic flow of ~1 cm s^-1. Although small relative to the wind-forced flow variations, the baroclinic velocities are comparable in magnitude to the mean flow and when averaged seasonally might be therefore be significant.

The second type of gradient, noted as D in Figure 13, was probably formed locally during dense water production event associated with the relatively small polynyas that formed along the Alaskan coast in the winter of 1995. The rapidly varying nature of salinity at AC2 is in fact consistent with the meandering of a strong front or the passage of an eddy in a manner similar to that predicted by Gawarkiewicz and Chapman [1995]. Clearly the mooring array could not resolve the scale of this frontal feature and therefore application of the thermal wind relationship during this period is presumptive. However, the sign of the density gradient between AC2 and C2 implies a westward (and cross-isobath) baroclinic flow over this portion of the shelf in January and February. The pulses of high-salinity at C1-93 and C2-93 shown in Figure 11 are probably another example of this type of gradient and we believe that these are eddylike features probably spawned within the extensive polynyas that developed in winter 1993-94. Figure 14 shows time series of salinity, temperature, currents and winds between December 1993 and March 1994 at C2-93 when several high salinity events moved across the mooring. Brine rejection from ice is the probable source of the high salinity water given the massive polynya that formed and because temperatures tracked the freezing point throughout this time. Current fluctuations were very energetic O(10 -20 cm s^-1), similar in magnitude to the predictions of Gawarkiewicz and Chapman [1995] and show little relation to the winds. However, the data do suggest some correspondence between changes in salinity and currents especially in early and mid-January and during the first three weeks of March.

Barrow Canyon Transport

A 6-year transport time series for Barrow Canyon including daily (thin black line) and 33-day running mean (thick blue line) transports is presented in Figure 15. The mean transport and 95% confidence limits are ~0.3 + .07 Sv. The annual (Oct. - Sept.) mean transport for each year is shown along the top. The uncertainty in a daily estimate is ~30%. Note that while the mean annual transports are small the synoptic scale variability is high. Up and downcanyon transports can vary by as +/- 1 Sv. over a few days. The mean downcanyon flow is due to the sea level difference between the Pacific and Arctic oceans which forces water northward across the Chukchi shelf. However, the variations about the mean are associated with the regional wind-field.

The lower panel of Figure 16 shows the annual cycle of transport in Barrow Canyon (red) based on the 6-year time series shown in Figure 15 and the mean monthly transport through Bering Strait (blue) from Roach et al., [1995]. The annual transport through Barrow Canyon is in-phase with Bering Strait and is a maximum in summer and a minimum in winter. The upper panel of the figure shows the monthly standard deviations in Barrow Canyon transport (red) and the fraction of the Bering Strait transport accounted for by the Barrow Canyon transport. Variability is maximal in winter and minimal in summer. The blue curve implies that there is a seasonal partitioning of the Bering Strait transport across the Chukchi shelf, with very little of the strait transport exiting through the canyon in winter, but a substantial fraction exiting along this pathway in summer. Based on the transport results from the Central Channel (described above) we suggest that the bulk of the winter transport through Bering Strait is diverted through Herald Valley in the western Chukchi Sea. If true, then our results imply a seasonal partitioning in the fluxes of salt and nutrients across the Chukchi shelf. For example, most of the dense (and nutrient rich) winter water formed in the Bering Sea in winter is diverted through Herald Valley.