The 2007 Bering Strait oceanic heat flux and anomalous Arctic sea-ice retreat

2.1. Data Sources

[3] The Bering Strait - ∼85 km wide, 50 m deep and divided into two channels by the Diomede Islands - is the only oceanic gateway between the Pacific and the Arctic Oceans. Year-round near-bottom oceanographic moorings (Figure 1a) have been deployed in the region almost continuously since 1990, generally at three locations – one at a mid-strait site (A3, ∼66.3°N 169°W) ∼60 km north of the Diomede Islands, and one at the centre of each channel (A2, eastern channel; A1, western channel).

[4] Although the annual mean transport through the strait is northward (∼0.8 Sv), there is substantial higher frequency variability: 0 to 1.5 Sv northward monthly; 2 Sv southward to 3 Sv northward daily (Figure 2a) [Woodgate et al., 2005b]. It appears that the velocity at site A3 correlates well with the total volume transport through the strait - mooring and ship-based data show high coherence in velocity at all sites in the strait region [Woodgate et al., 2005b], with velocities being higher in the Alaskan Coastal Current (ACC), a warm, fresh, ∼0.1 Sv current present along the Alaskan coast from midsummer until ∼ December (Figure 1a) [Woodgate and Aagaard, 2005]. This strait-scale coherence is expected since the suspected forcing for the flow – a pressure-head difference between the Pacific and the Arctic, mediated by local wind-driven effects – is on length-scales greater than the strait width (see Woodgate et al. [2005b] for discussion). Lacking year-round velocity cross-sections, we assume the flow field is homogeneous and barotropic and use a cross-section area of 4.25 × 10⁶ m² to convert A3 velocity to transport. This probably underestimates the transport (as it neglects the ACC), and may introduce systematic errors (likely < 20%).

[5] Temperatures in the strait vary substantially, especially seasonally – from freezing in winter to 4–10°C (near-bottom/surface data) in late summer (Figure 2c) [Woodgate et al., 2005a]. Other than in winter, there is a cross-strait temperature gradient with warmer temperatures in the east. Yet mooring data suggest that site A3, being central to the strait, yields a reasonable average of the mean near-bottom temperatures (good to ∼0.1°C/0.5°C in annual/monthly means [Woodgate et al., 2007]). Thus, we use site A3 to estimate mean near-bottom temperatures. Due to ice-keels, most in situ measurements in the strait are deep, 10 m above the bottom, and thus underestimate water column mean temperatures since summer/autumn CTD sections indicate a 10–20 m thick surface layer, 1–2°C warmer than below. To estimate temperatures in this layer, we use 7-day averages of AVHRR [Vazquez et al., 1998] Sea Surface Temperature (SST) from within ∼40 km of A3 (spatial and temporal averaging selected to reduce time-series gaps in this typically cloudy region). We estimate this assumption is good to ∼1°C, considering stated errors and a comparison with available coincident in situ CTD data. Since we seek the heat available to melt sea-ice in the Arctic, our temperature reference is −1.9°C, the freezing point of Bering Strait waters, reflecting that these waters lose most of their heat before leaving the Arctic [Steele et al., 2004].

[6] Using mooring data alone (and assuming a barotropic, homogeneous water column), we compute heat fluxes at the finest time resolution available (typically hourly). Then we estimate a simplistic stratification correction, using SST and two nominal upper layer thicknesses of 10 and 20 m (Figure 2e). Overall, uncertainties are of order 0.1 Sv, 0.8 × 10²⁰ J/yr. To include the ACC, we add ∼0.1 Sv, and ∼1 × 10²⁰ J/yr to the northward fluxes [Woodgate et al., 2006]. Since A3 data are incomplete before 1998, we also present results from eastern channel (A2) temperature and velocity data, to allow comparison back to 1991.

2.2. Results From 1991 to 2007

[7] Annual mean values (Figure 2) are calculated for calendar years. For volume transport, the interannual variability (0.6–1.0 Sv) masks any long term trend (Figure 2b). The record low (0.6 Sv, 2001) reflects weaker northward flow in many periods of the year, while the second lowest (0.7 Sv, 2005) is mostly due to anomalous southward flow in late 2005 (Figure 3). The highest transports (∼1Sv, 2004 and 2007) have strong northward flows in winter and for most of the summer.

[8] The most dramatic change in A3 (i.e., near-bottom) annual mean temperatures is a ∼1°C increase between 2001 and 2002 (Figure 2d). Although 1993 was possibly warm (A2 data), A3 means from 2002 –2007 are 0.1–0.4°C, compared to −0.2 to −0.5°C pre 2002. There is no obvious match between high transports and high temperatures, although it proves important that both are high in 2007. The SSTs, also with a record high in 2007 (Figure 2d), show slightly larger interannual variability than near-bottom temperatures.

[9] Annual mean heat fluxes (Figure 2e) increase almost monotonically from 2001 to 2007. Heat flux variations are driven almost equally by changes in volume transport and in temperature, and the highest heat flux years (2004 and 2007) are those where both transport and temperature are high. Year 2007 yields a clear record-length maximum, estimated at 3.5 × 10²⁰ J/yr from A3 data alone and at 4–4.7 × 10²⁰ J/yr including a 10–20 m surface layer. Adding ∼1 × 10²⁰ J/yr for the ACC yields a total heat flux of 5–5.7 × 10²⁰ J/yr. This is almost a doubling of the total 2001 heat flux (∼2.6–2.9 × 10²⁰ J/yr), and 1 × 10²⁰ J/yr greater than the previous high in 2004 (∼4.3–4.8 × 10²⁰ J/yr).

3.1. Bering Strait Throughflow as a Trigger to Start the Seasonal Melt Back of Ice

[16] The ice-albedo feedback – less ice allows more shortwave energy to be absorbed by the water, warming the water and thus melting more ice – is often cited as the main cause of sea-ice retreat. However, other than in leads and surface melt-ponds, the albedo feedback generally only takes over once a region of open water has been created.

[17] In 2007, a small open water area appeared south of the strait in early April, when the water flow was southward. When the flow turned north about 2 weeks later, open water appeared north of the strait, coincident with slight (0.1°C above freezing) near-bottom warming at A3. Through May the flow remained northwards and the sea-ice retreat occurred in patterns consistent with known Chukchi flow pathways via Herald Valley, the Central Channel and Barrow Canyon [Weingartner et al., 1998; Woodgate et al., 2005b] and, by the end of June, these pathways appear dramatically in the ice-melt pattern (Figure 1b).

[18] Even though initial opening may be due to ice-motion and/or melt, the ice retreat pattern strongly suggests that water pathways through the Chukchi are acting as a conduit of heat (gained either locally or further south) into the Arctic and, furthermore, are providing the initial heat (and/or motion) necessary to open up ice in the Arctic Basin. Conservation of volume dictates an increase of Bering Strait transport must be rapidly compensated with increased outflow to the Arctic Basin. (A 1 Sv imbalance would change the Chukchi sea-level by 1 m in about 4 days). Thus, strong northward transport in the Bering Strait (usual in summer, Figure 3) must correspond to strong northward flow of waters into the Arctic Basin.

3.2. Pacific Waters as a Time-Delayed Source of Subsurface Heat in the Arctic

[19] In terms of delivering heat to the north, we must also consider the time taken to transit the Chukchi, estimated at several months by Woodgate et al. [2005b]. Mooring data from 2002–2004 at ∼73.5°N 166°W (http://www.eol.ucar.edu/projects/sbi/mooring.shtml) indicate that near-bottom seasonally warm waters arrive/leave later in the northern Chukchi (December/February) than in the Bering Strait (May/December). Thus, subsurface heat – albeit modest (maximum 2°C, typically −0.5 to 0°C) – is present under the surface ice cover and is provided to the Arctic in the middle of winter.

[20] Within the western Arctic Ocean, the PW form a subsurface temperature maximum that weakens away from the Chukchi. The heat stored in PW is ∼40–140 MJ/m² depending on location [Steele et al., 2004]. Since the PW have cooled almost to freezing when they leave the Arctic, this heat must be lost somewhere, likely either to melt ice (100 MJ/m² could melt 0.3 m of ice) or to the atmosphere via leads. Thus again, the PW heat appears to be a modest, but significant quantity.

[21] Note this implies a memory to the system. The heat supplied to the Arctic Basin in January 2008 relates to the summer 2007 Bering Strait heat flux. Similarly, the subsurface PW temperature maximum in the western Arctic is almost ubiquitously the remnant of an earlier summer's heating in the shelf seas, and thus significant interannual variability in heat input may have a delayed effect on Arctic sea-ice.

3.3. Interannual Variability of the Pacific Influence

[22] One of our goals is to inform the discussion of interannual sea-ice variability in the Arctic. How large is the interannual variability in Bering Strait fluxes?

[23] The 2007 transport is more than 1.5 times the 2001 transport (1.0 Sv compared to 0.6 Sv). Since the Bering Strait inflow is the largest input to the Chukchi, this implies significant change in resident time of waters in the Chukchi (∼5.5 months, compared to 9 months, using Chukchi Sea volume, 14 × 10³ km³, divided by transport as a simplistic estimate). This has many possible implications for the physics and biogeochemistry of the Chukchi, especially since Bering Strait annual mean temperatures also rise from −0.4°C in 2001 to +0.4°C in 2007.

[24] The Bering Strait heat flux doubles between 2001 and 2007 (2–3 × 10²⁰ J/yr to 5–6 × 10²⁰ J/yr). Is this variability large compared with other sources? Using daily data (courtesy of B. Light) from the Chukchi south of 71°N as per Perovich et al. [2007], we estimate the annual shortwave input to the Chukchi between 1998 and 2007 varies between ∼3 and 4.5 × 10²⁰ J/yr. Thus, although both are relevant in terms of total heat input to the Arctic Basin, the Bering Strait heat flux variability appears to be slightly larger.

[25] The Bering Strait heat flux variability is likely so high as it reflects changes both in volume transport and in temperature. What sets the variability in these two terms? There is little understanding of Bering Strait temperatures, which are presumably a complex result of atmospheric and solar effects on the northward moving waters of the Bering Sea. For transport, however, it is generally thought that the flow has two main drivers – a Pacific-Arctic pressure-head (of debated origin, often assumed constant) and local wind forcing (see Woodgate et al. [2005b] for discussion).

[26] Assuming most simply “transport = PH + C × Wvel” where PH represents transport due to the pressure-head term, C is an unknown constant, and Wvel is the wind component with highest correlation to the flow (∼330°), we seek PH and the constant, C, by a linear best fit of 6-hourly NCEP (National Centers for Environmental Prediction) 10 m wind to estimated A3 transports in 1-year segments of data. Woodgate et al. [2006] tentatively related interannual transport variability to changes in the annual mean northward winds, admitting that the time-series was inconclusively short. The extended analysis (Figure 2f, wind velocity converted to equivalent transport assuming a 2001–2007 mean for “C’) suggests, overall, the wind acts to slow the flow by between 0.4 and 0.2 Sv, although exact numbers vary strongly with wind point used. However, interestingly, the interannual variability in the PH transport is found to be comparable in magnitude to the wind-driven variability, with the total PH transport greater in 2007 (Figure 2g). This suggests that recent changes in the Bering Strait transport relate not just to changes in the local wind, but also to significant variability (> 0.2 Sv equivalent) in the Pacific-Arctic pressure difference driving the flow (as suggested by Woodgate et al. [2005b]).