1 Introduction

In the western Arctic, Pacific Waters (PW) enter the Chukchi Sea via the Bering Strait at an annual mean rate of 1 Sverdrup (Sv; 1 Sv = 10⁶ m³ s⁻¹; Woodgate, 2018). The inflow of PW was enhanced largely due to the increase in the far-field Pacific-Arctic 'pressure-head' force from 1990 onward (Woodgate and Peralta-Ferriz, 2021). Seasonally, the surface current dominates the summer Bering Strait throughflow, which includes relatively warm and fresh Alaska Coastal Water (ACW) and the cooler and saltier summer Bering Sea Water (sBSW, von Appen and Pickart, 2012). The ACW mainly flows via the Alaska Coastal Current. In Barrow Canyon, a portion of the ACW outflow feeds the eastward-flowing current to the Beaufort Sea, known as the Beaufort Shelfbreak Jet (Weingartner et al., 1998, 2005; Pickart et al., 2009), and the sBSW is advected poleward through the central pathway between Herald and Hanna Shoals (Woodgate et al., 2005). These two types of Pacific Summer Waters can significantly impact the marine environment of the Arctic region via substantial heat and freshwater input (Yang, 2006; Pickart et al., 2013). The ACW may experience more rapid heat loss than sBSW because it overlies sBSW and contains more heat (Steele et al., 2004). The spatial correlations between newly open water areas and regions of temperature increase of warm halocline indicate that ACW may play an important role in the loss of Arctic sea ice (Stroeve and Notz, 2018). From previous studies, the anomalously warm ACW intrusion into the Canada Basin via anticyclonic eddies can cause upward heat flux of O(1 – 10) W m⁻², which melts the bottom of sea ice at a rate of 4 mm d⁻¹ (Kawaguchi et al., 2012; Fine et al., 2018, 2022; Lin et al., 2021). Under various conditions of North Pacific wind fields, the temperature and salinity properties of ACW inflow also change, demonstrating the annual difference of the ACW heat and freshwater contents in the Arctic Ocean in general (Zhang et al., 2020).

The transport through Barrow Canyon accounts for roughly 94% of the ACW during July – September (Itoh et al., 2013). However, the export of ACW in this region is liable to flow directly northward to the Canada Basin under the circumstances of increased and strengthened easterly winds (Okkonen et al., 2009; Brugler et al., 2014). Synchronously, insolation warms the surface water through newly ice-free areas in the Chukchi Sea during the melting season (Timmermans et al., 2018). Subsequently, these warmed waters are mostly subducted in the northern Chukchi Sea to the Central Canada Basin, along the isopycnals that outcrop on the surface to feed the ACW (Timmermans et al., 2014). Steel et al. (2004) argued that the ACW and sBSW in the Canadian Basin may be geographically separated under the regime of the positive Arctic Oscillation (AO) index. Other studies suggested that the transport of the ACW and sBSW has no significant correlation with the AO index (Wang et al., 2009). In recent years, with reinforced easterly winds along the coast of Alaska, more ACW flow has progressed to a northerly path via northward Ekman transport and has been injected into the Arctic Ocean with sBSW (Watanabe, 2011; Brugler et al., 2014). Less coverage of sea ice, in combination with anomalous anticyclonic wind fields on the basin scale, provides an opportunity to alter the Beaufort Gyre intensively and extensively (Rampal et al., 2009; Yang, 2009). Thus, ACW and sBSW may not be geographically separated but can coexist widely in the entire basin.

The dramatic sea ice decline in the Pacific sector of the Arctic over the last few decades may be the consequence of the increased inflow and heat flux of the ACW (Itoh et al., 2015). Nevertheless, freshwater is accumulating at an average rate of (397 ± 116) km³ yr⁻¹ in the Beaufort Gyre region (Proshutinsky et al., 2019), which indicates the enhanced stratification in the western Arctic surface layers and damped disturbances in external forcing (Bertosio et al., 2022). Without forceful eddies or mixing, ACW experiences difficulty in producing sufficient upward vertical heat flux to impact sea ice due to the restriction of haloclines in addition to over-thin mixed layers (< 12 m) despite the considerable heat content carried by the ACW (Fer, 2009; Jackson et al., 2010; Toole et al., 2010; Lincoln et al., 2016). Downstream at the continental slope north of Alert, Ellesmere Island, ACW still appears despite the substantial loss of its heat content by that point (Newton and Sotirin, 1997); the ACW overcomes these barriers to release heat in certain conditions during flow. Since 2012, the freshwater content in the Makarov Basin has increased as a result of enormous freshwater northward intrusion out of the Beaufort Gyre, which thickened the halocline layer and strengthened stratification in the upper water column within this region (Bertosio et al., 2022). This hydrographic condition may create an opportunity for ACW to proceed farther downstream areas as the ACW heat loss responds to the strength of upper ocean stratification. Therefore, comprehending the mechanism and reasons for the changes in the ACW is the key to revealing the relationship between warming ACW and sea ice reduction in the western Arctic.

This study used the in-water temperature and salinity data sampled by ice-borne Drift-Towing Ocean Profilers (DTOP) over the Alpha Ridge during the 9th and 11th Chinese National Arctic Research Expedition to investigate the revolution of ACW north of Canada Basin and its connection to oceanic heat flux. In Section 2, we describe the overall data set applied in this study and the calculation of relevant physical quantities. Section 3 presents the hydrography results from the DTOP records, estimates of heat flux, and analysis of the Ekman upwelling related to the ACW heat release. Section 4 summarizes our results and discussion.

2 Data and Methods 2.1 DTOP Buoy Data

The temperature and salinity profiles from the DTOP buoys, which were designed by the Ocean University of China and manufactured by Tianjin Yuanxun Technology Co., Ltd., were analyzed here. Meteorologic sensors on the DTOP measured hourly the near-surface air temperature, humidity, and pressure using a module of the Global Positioning System recording the real-time coordinates of buoy sites. The underwater CTD profiling instrument sampled the temperature and salinity of the underlying ice-covered ocean on a 12 h basis with a weightless cable of 120 m in length (Fig. 1a). The long-term hydrographic properties beneath the ice bottom to the depth of nearly 120 m captured by DTOP can be analyzed to study the interactive processes between sea ice and ocean. The sampled data were transmitted to the domestic surficial receiving station through the iridium system in real time. Buoy DTOP18 collected profiles from August 22, 2018, to December 12, 2018, before it stopped transmitting data, and the data from DTOP21 over the period of June 1, 2021 – December 24, 2021, were also used. Both buoys were deployed over the Alpha Ridge north of the Canada Basin, with DTOP21 located farther east, closer to Ellesmere Island (Figs. 1b and 1c). In this study, the traditional potential temperature and practical salinity were utilized as the basis for the discrimination of water masses and the calculation of relevant physical quantities.

2.2 Other Data

Data inverted by microwave brightness temperature from the successor of Advanced Microwave Scanning Radiometer for Earth Observing System on the satellite 'Shizuku' (GCOM-W1), which was developed by Japan Aerospace Exploration Agency, were used to estimate daily-average sea ice concentration C_si on a 6.25 km grid (Melsheimer and Spreen, 2019). Monthly mean sea ice drift data calculated based on combinations of multiple satellite sensors and gridded onto a Lambert azimuthal equal area grid with 25-km grid resolution were obtained from National Snow and Ice Data Center (NSIDC, Tschudi et al., 2019). The objectively analyzed data set of temperature and salinity with a spatial resolution of 0.25˚ × 0.25˚ and 5 m in lateral and vertical, respectively, from World Ocean Atlas 2018's (WOA18) 13-year monthly mean (2005 – 2017) climatology was adopted here to compare with our DTOP results (Locarnini et al., 2018). The hourly wind speed at 10 m, downward shortwave radiation under clear skies, and cloud friction obtained from the 5th generation European Centre for Medium-Range Weather Forecasts reanalysis (ERA5) were analyzed to study the relations between atmospheric conditions and oceanic variation (Hersbach et al., 2018).

2.3 Methods of Calculation

From observations of heat flux and Reynolds stress during the Marginal Ice Zone Experiment, based on the turbulence theory, McPhee (1992) suggested a simple bulk parameterization of oceanic heat flux:

$ F_w^{{\text{OHF}}} = {\rho _{sw}}{c_p}{C_H}{u_*}\delta T, $

(1)

where ρ_sw is the sea water density, and c_p is the specific heat of sea water (3.98 × 10³ J kg⁻¹), C_H is a bulk heat transfer coefficient (5.7 × 10⁻⁴ ± 4 × 10⁻⁴; McPhee, 2002); u_* is the interface friction velocity as described in Eq. (2); δT = T_ml − T_f (S_ml) represents the surface mixed layer (SML) temperature elevation above freezing T_f, where T_ml and S_ml are the temperature and salinity of SML, respectively. u_* is described using the Rossby similarity theory (McPhee et al., 1999):

$ \frac{{\kappa V}}{{{u_*}}} = \log \frac{{{u_*}}}{{f{z_0}}} - A - iB, $

(2)

where κ is von Karman's constant (0.4); V is the ice velocity relative to the surface geostrophic flow. Here, V can approximate the sea ice drift velocity calculated by the coordinate data derived from DTOP, given that it considerably outweighs the surface geostrophic flow velocity. f is the Coriolis parameter, z₀ is the hydraulic roughness length of the underside of the ice (0.1 m); A and B are similarity constants with values of 2.21 and 1.91, respectively.

The net incident solar radiation SW_n that enters the upper ocean through leads or open water is calculated using the following equation (Stanton et al., 2012):

$ S{W_n} = S{W_d}(1 - 0.25{c^3})(1 - {\alpha _w})(1 - {C_{si}}), $

(3)

where SW_d is the downward shortwave radiation top of the atmosphere under clear skies; c is the cloud fraction, α_w is the albedo of open water and leads (0.07); C_si is the sea ice concentration. Here, the local C_si (0 ≤ C_si ≤ 1) at the buoy sites was estimated by averaging the concentration value from grid points found within circles with 50 km radii.

The Ekman pumping vertical velocity w_e relevant to the curl of wind stress is defined as follows (Polyakov et al., 2013):

$ {w_e} = \frac{1}{{{\rho _{sw}}}}(\nabla \times \frac{{\vec \tau }}{f}), $

(4)

where the wind stress $\vec \tau $ is calculated from the following equation:

$ \vec \tau = {\rho _\alpha }{C_d}|{\vec u_s}|{\vec u_s}, $

(5)

$ {C_d} = (1.5 + 2.233{C_{si}} - 2.333C_{si}^2){10^{ - 3}}, $

(6)

where ρ_α is the air density (1.25 kg m⁻³); ${\vec u_s}$ is the 10 m-high wind speed applied from ERA5; the drag coefficient C_d, which is considered for $\vec \tau $, is derived from a second-order polynomial function of sea ice concentration C_si as described in Eq. (6) (Andreas et al., 2010). The positive (negative) value of w_e manifests the divergence upwelling (convergence downwelling) of sea water in response to the cyclonic (anticyclone) wind field acting on the sea surface, which can represent the strength of the Beaufort Gyre.

3 Results and Discussion 3.1 Variability of Upper Ocean Hydrography Derived from DTOP Records

Normally, the freezing season of sea ice in the central Arctic ends during June and early July (Lei et al., 2018). The SML depth (MLD) exceeded 50 m due to the brine-driven convection mixing at this time with a salinity greater than 30.0 (Fig. 2). SML was generally heated to above −1.5℃ in July and August, affected by solar radiation and freshened by the meltwater out of sea ice base. The shoaling of MLD occurred significantly during mid-summer, with a minimum of approximately 20 m in August (Fig. 2b). SML became cooler and denser in September. After October, the SML temperature dropped to between −1.72℃ and −1.62℃, and MLD was deeper than 40 m. In September 2018, a subtle near-surface temperature maximum (NSTM) ranged from −1.58℃ to −1.55℃ at the MLD of approximately 30 – 40 m (Figs. 2a and 3a). Its formation was associated with the summer solar radiation that entered the upper ocean, where the water column was cooled down and freshened by the melting of sea ice, which resulted in a temperature maximum at an intermediate depth (Jackson et al., 2010; Steele et al., 2011). With the dramatic decline of sea ice extent in recent years, NSTM in the Canada Basin has appeared farther north and survived winters (Zhao and Cao, 2011). Beneath NSTM, remnant winter water (RWW) with a temperature below −1.6℃ and salinity ranging from 31.0 to 31.5 can be found; this water was the residual of the previous winter's mixed layer (Jackson et al., 2011; Steele et al., 2011; Figs. 2a, 2c, 2g, and 3a). The layer of RWW became thinner after October 2018 but remained below the SML (Fig. 3b). However, the RWW was barely visible over the same season in 2021. Therefore, the water with a temperature of nearly −1.5℃, which was warmed by summer solar radiation, was directly maintained at the upper part of ACW instead of in the shape of temperature maximum during October – December 2021 (Figs. 2b and 3d).

During summertime (June – August) in 2021, within the region of around 83.5˚ – 84.5˚N, 110˚ – 120˚W, the thick layer of ACW was significantly present between 50 and 90 m, with temperature and salinity generally ranging from −1.35℃ to −0.9℃ and from 30.2 to 32.5, respectively, and the temperature maximum $T_{\max }^{{\text{ACW}}}$ approximating −0.8℃ (Figs. 1b, 2b, 2d, 2g, and 3c). The ACW experienced slight thinning in mid-July (about 35 m) and immediately thickened to more than 40 m again. Starting from late August, when buoy DTOP21 drifted northward, the layer of the ACW dramatically thinned to approximately 20 m. $T_{\max }^{{\text{ACW}}}$ also decreased to approximately −1.25℃, and the salinity remained unchanged (Figs. 1b, 2b, and 2d). By contrast, during late summer and early autumn in 2018, the ACW with a $T_{\max }^{{\text{ACW}}}$ of above −1.3℃ was captured appreciably near 80 m only on the first 6 days in September. The temperature maximum of the ACW was lower than −1.5℃ for the rest of this period, and this finding was possibly correlated with the buoy being positioned farther northern and western (84.5˚ – 84.8˚N, 145˚ – 168˚W; Figs. 1b, 2a, 2c, and 3a). From October 2018 and 2021, $T_{\max }^{{\text{ACW}}}$ was mostly between −1.45℃ and −1.4℃, and its depth ranged from 60 m to 70 m. In addition, the buoyancy frequency N² in the ACW layer was the highest for the whole water column, and the value of N² was larger during the autumn of 2021, which was related to the overall freshening of the water column (Figs. 2d and 2f), in view of the reduction of salinity that can result in the shoaling of the ACW and strengthened stratification (Timmermans et al., 2014). Another saltier warm water called sBSW (S > 33.0) can be found underneath ACW at a depth greater than 100 m, following the work of Brugler et al. (2014). The temperature of the upper part of sBSW was exceeded by −1.3℃ in September 2018 (Figs. 2a, 2c, 2g, and 3a); however, the sBSWs have not been observed until October 2021, when their temperature maximum fluctuated between −1.5℃ and −1.3℃, and the salinity ranged from 32.5 to 33.0 (Figs. 2b, 2d, 3c, and 3d). The temperature-minimum water that separated ACW and sBSW was not defined as a distinct type of water mass (Figs. 2 and 3); the origin and mechanisms of the formation of ACW have not been confirmed, although some scholars argued that it likely originated from the winter ACW (Steele et al., 2004). Given that after October, the layer of this temperature-minimum water thickened to about 35 m, and the 33 isohaline sank down to about 120 m, sBSW may be at a greater depth, although it was almost invisible from our data (Figs. 2a and 3b).

3.2 Comparison with WOA18 Climatology

Figs. 4 and 5 show the lateral extent of ACW within the WOA18's 13-year monthly mean (2005 – 2017) climatology. The ACW with the highest temperature maximum was mainly distributed in the Chukchi Plateau and Beaufort Sea. North of Barrow Canyon, the ACW, with a temperature maximum exceeding 1.0℃, started to intrude into the Central Canada Basin from shelfbreak in September. The origin of this ACW intrusion shifted westward slightly to the Chukchi Plateau in October, and the extent of ACW with a temperature maximum exceeding −1.0℃ in the basin continued to expand. The climatological average $T_{\max }^{{\text{ACW}}}$ around 75˚ – 78˚N in the Chukchi Plateau was elevated to more than 2.0℃ in December (Fig. 4). The ACW in the Canadian Basin cooled gradually since January, mainly reflected in the shrinkage of isotherms and decreased $T_{\max }^{{\text{ACW}}}$. A large spatial difference was observed in the depth of the ACW temperature maximum. In the southern Beaufort Sea, increased Ekman pumping due to stronger Beaufort Gyre during autumn and winter deepened the ACW temperature maximum to around 70 – 85 m. By contrast, the depth of the ACW core upturned to approximately 50 – 65 m in the Beaufort Gyre region, where Ekman pumping was normally the weakest from June to August (Fig. 5; Proshutinsky et al., 2009).

The ACW temperature maximum outside the main region of Beaufort Gyre fluctuated between −1.5℃ and −1.2℃ throughout the year. $T_{\max }^{{\text{ACW}}}$ invariably decreased northward from the Northern Canada Basin to the Alpha Ridge (Fig. 4). Zonally, the ACW temperature maximum greater than −1.3℃ was commonly observed east of 160˚W over the Alpha Ridge. Our hydrographic data from DTOP18 coincided well with the climatologic results. However, the ACW temperature maximum anomaly recorded in the summer of 2021 in the region of 83.5˚ – 84.5˚N, 110˚ – 120˚W exceeded by nearly 0.4℃ – 0.5℃ those of WOA18 (Fig. 2). Year-round Bering Strait mooring data presented a long-term increase (about 0.01 Sv yr⁻¹) in the annual mean transport of PW into the Arctic due to the far-field (pressure-head) forcing of the flow from 1990 to 2019 (Woodgate, 2018; Bertosio et al., 2022). Clear warming was observed, especially in recent years, which yielded a net heat inflow of approximately 6 × 10²⁰ J yr⁻¹ in 2017 and 2018 (Woodgate and Peralta-Ferriz, 2021). These trends contributed greatly to the warming of the ACW in the entire basin. Steele et al. (2004) discovered that the expansion of Beaufort Gyre in the negative Arctic Oscillation (AO) phase can sweep ACW eddies into larger anticyclonic circulation and provide ACW signatures farther downstream. The AO index for winter 2020/2021 was −1.80, which demonstrates that more ACW with a temperature maximum exceeding −1.0℃ can be transported farther with the expansion of Beaufort Gyre. In addition, a northward intrusive warm tongue occurred north of 80˚N during June and July (Fig. 4), with the temperature maximum similar to the anomalous $T_{\max }^{{\text{ACW}}}$ recorded in summer 2021 but relatively farther west in position (about 145˚W). We must consider the possibility that the eastward shift of this warm tongue caused the anomalously warm ACW mentioned above. However, the physical mechanism that influenced the location of this warm tongue remains unidentified thus far. In general, the anomalously warm ACW showed great relations with the increased transport and heat of Bering Strait inflow in recent years and the expansion of Beaufort Gyre in 2021 (Woodgate and Peralta-Ferriz, 2021). As the DTOP21 drifted northward, the ACW lost most of its heat content from late August 2021. Meanwhile, the DTOP18 did not capture evident ACW signals persistently, with the temperature maximum exceeding −1.4℃ until south of 84˚N. This finding agreed with the pattern of ACW cooling north of Ellesmere Island proposed by Steele et al. (2004). Thus, a certain mechanism could have prevented the northward extension of the warm ACW. In other words, the sign that the warm tongue retreated to the interior of the Canada Basin in August (Fig. 4) also indicated that ACW was apt to lose its heat north of the Canada Basin at that time, which led to a significant reduction in the temperature maximum downstream.

3.3 Under-Ice Oceanic Heat Flux

The $F_w^{{\text{OHF}}}$ is related to the mixed layer temperature elevation above its freezing temperature ∆T and friction velocity u_* and is normally accelerated by storms (Eq. (1)). In the following discussion, the major storms are as defined by Cohen et al. (2017) and indicated by horizontal orange bars in Figs. 6a and 6b. The start and end of major storms correspond to the periods when the wind speed at 10 m height U_10m was greater than 8 m s⁻¹ for more than 3 h, with no interruptions exceeding 1 h and with the rate of pressure decrease exceeding 5 hPa in 6 h. From June to early July 2021, major storms occurred every few days, and each lasted for about 2 – 5 days. Therewith, $F_w^{{\text{OHF}}}$ peaks were also observed (Figs. 6b and 6d). In the last two cases of major storms in July, $F_w^{{\text{OHF}}}$ exceeded 25 W m⁻² due to the constant increase in ∆T. Solar radiation heated the upper ocean since June, and the net shortwave radiation flux was elevated conspicuously during the period of storm events. The wind speed descended between July and August, and $F_w^{{\text{OHF}}}$ was generally lower than 15 W m^{−2 in spite of} ∆T reaching above 0.1℃. $F_w^{{\text{OHF}}}$ approximated 25 W m⁻² again due to the highest ∆T (> 0.12℃) observed on the first 10 days in August. Owing to the solar radiation entering the upper ocean via leads (C_si ≈ 95%), the $F_w^{{\text{OHF}}}$ remained above 10 W m⁻² during mid-to-late August, with a maximum of about 20 W m⁻² (Figs. 6b and 6d).

The net shortwave radiation flux decreased to near zero, with ∆T lower than 0.02℃ since September. A major storm with the highest wind speed over 14 m s⁻¹ was recorded in late September 2018, and the NSTM was observed at the bottom of SML (Fig. 6a). The NSTM is normally trapped by a strong seasonal halocline. The heat stored in the NSTM was hardly released upward to the ice-ocean interface until the halocline structure was eroded by convection mixing through brine rejection during the freezing season or by turbulent mixing through abrupt storm events (Jackson et al., 2012). In late September 2018, the NSTM slightly descended, and the mixed layer temperature increased within 1 d. $F_w^{{\text{OHF}}}$ approximated 5 W m⁻² over this storm period (Figs. 6c and 6e). Starting in October, the heat flux was typically low (〈$F_w^{{\text{OHF}}}$〉= 1.2 W m⁻²), with the mixed layer approximating the freezing state (〈∆T 〉 = 9 × 10⁻³ ℃). Although two more strong atmospheric activities were recorded in October and November 2018 and 2021, respectively, the friction velocity was generally less than 10 mm s⁻¹, which corresponded to the poor mobility of the ice pack caused by the sea ice concentration approaching 100%. Thus, $F_w^{{\text{OHF}}}$ in these two major storm cases was 2.5 W m⁻² on average (Figs. 6c and 6d).

Given the above results, the $F_w^{{\text{OHF}}}$ was large when the ice floe where the DTOPs sat traveled with a fast velocity. The velocity was mainly affected by the surface wind field and sea ice concentration. The low sea ice concentration enhanced the mobility of ice packs by reducing the ice stress between them, thereby accelerating the ice drift. Meanwhile, the upper ocean must absorb more solar radiation to elevate the $F_w^{{\text{OHF}}}$ during conditions of low sea ice concentration. Furthermore, the heat stored in warm waters of the ocean interior was released efficiently because the wind-driven turbulent mixing became aggressive under the circumstances of less sea ice cover.

3.4 Estimates of the Impact on Oceanic Heat Flux by ACW

With the high sea ice concentration in summer (> 90%), the $F_w^{{\text{OHF}}}$ is not only related to the net shortwave radiation absorbed by the upper ocean but also originates from the internal ocean heat storage brought about by the vertical advection process (McPhee et al., 2003). The ACW contains more heat than sBSW and resides under the SML most of the time (Fig. 2b); therefore, it may experience rapid heat loss due to external forces, such as atmospheric disturbance (Steele et al., 2004). Especially in light of storm events, diapycnal heat fluxes will be transferred from the ACW in the form of upwelling or entrainment with enhanced turbulent mixing (Yang et al., 2001). Following the work of Shimada et al. (2001), the loss of heat out of the ACW layer $F_w^{{\text{ACW}}}$ is given by the following:

$ F_w^{{\text{ACW}}} = \frac{\partial }{{\partial t}}({\rho _{sw}}{c_p}T{\text{d}}z) = {\rho _{sw}}{c_p}k\frac{{\partial T}}{{\partial z}}, $

(7)

where k and ∂T/∂z are the vertical diffusivity and ACW temperature gradient, respectively. Macdonald et al. (1993) estimated the bulk diffusivity in the Canada Basin halocline (4 × 10^{−5 m2} s⁻¹). Referring to the research of Steele et al. (2004), ∂T/∂z is calculated as the average temperature gradient between the depths of $T_{\max }^{{\text{ACW}}}$ and 15 m shallower and between the depths of $T_{\max }^{{\text{ACW}}}$ and 15 m deeper.

The ACW regularly released diapycnal heat flux of approximately 3 – 4 W m⁻² during summer, and $F_w^{{\text{ACW}}}$ reduced to below 2 W m⁻² from September (Figs. 7a and 7b). Correspondingly, the content of ACW reached 80 – 90 MJ m⁻² and was reduced to roughly 40 MJ m⁻² in the months of June – August and September – December (Figs. 7c and 7d, respectively). The Canada Basin used to exhibit a low salinity (about 28 on average; Maykut and McPhee, 1995). The heat release out of the ACW upward to the ice-ocean interface must be activated by buoyancy sink (convection), whereas it is largely restrained by a great vertical density gradient most of the time (Shaw et al., 2009). In consequence, the mighty heat storage in the ACW exerts negligible influence on the seasonal evolution of sea ice in this region. We estimated the stratification index (SI) using the following formula (Ladd and Stabeno, 2012):

$ SI = - \int_{ - h}^0 {\left\{ {\rho (z) - \left\langle \rho \right\rangle } \right\}} gz{\text{d}}z, $

(8)

$ \left\langle \rho \right\rangle = \frac{1}{h}\int_{ - h}^{\text{0}} {\rho (z){\text{d}}z}, $

(9)

where 〈ρ〉 is the average density of vertical integration, and h is the depth, which was set to the depth of $T_{\max }^{{\text{ACW}}}$ in this study. In June, when $F_w^{{\text{ACW}}}$ was the highest and the salinity of the SML exceeded 30.0, the SI was generally under 0.5 × 10³ J m⁻², with a very low buoyancy frequency. The freezing season in the central Arctic Ocean ended in June, and thus, the low SI and N² over this period were considered to be caused by brine-driven mixing. From July, when freshening of the SML started, the SI increased considerably to approximately 1 × 10³ J m⁻², and N² also exhibited an increase (Figs. 7e and 7f). The mixing process remained very active between the SML and ACW layer during August (Fig. 3c), which resulted in a relatively low SI in this month. The SI and N² reached the maximum values of 2 × 10³ J m⁻² and 2 × 10⁻⁴ s⁻², respectively, in October of these two years. Starting in December 2021, the SI and N² declined to near zero again during the start of the freezing season. However, over the same period in 2018, the SI and N² remained high due to the halocline with a great salinity gradient (Fig. 3b).

The distance between the mixed layer base and the depth of $T_{\max }^{{\text{ACW}}}$ was mostly lower than 20 m by mid-August. Hence, the absolute value of the temperature gradient |δT_z| between MLD and the depth at 5 m deeper was greater than 0.05℃ m⁻¹, with a maximum of nearly 0.1℃ m⁻¹ (Figs. 7h and 7j, respectively). This result was at variance with the values of the central Canada Basin calculated by Toole et al. (2010), who observed that the |δT_z| in this region approximated zeros in summer because of the very thin SMLs (< 12 m) based on about 5800 ITP profilers. In winter, the SMLs in the Canada Basin were also limited to less than 40 m on average. Given the high sea ice concentration, only a small amount of ageostrophic energy (for instance, wind stress) was inputted into the interior ocean, which accounted for the overly thin SMLs. From mid-August 2021, the mixed layer base and ACW were slightly farther apart, and the |δT_z| approached zero during the shoaling of the SML. However, as mentioned above, the mixed layer overlying the ACW layer can be the bridge to transfer the upward heat release out of the ACW, which resulted in a $F_w^{{\text{ACW}}}$ of roughly 3 W m⁻² despite the dramatic reduction in the heat content of ACW (highlighted by gray shade in Fig. 7d). After September, with the low temperature maximum and heat content of ACW, the |δT_z| stabilized at near zero, and the mixed layer base can be found at a distance of 30 – 40 m from the ACW core, which gave rise to a lower $F_w^{{\text{ACW}}}$ during that period (Figs. 7g, 7h, 7i, and 7j). In addition, the subtle elevation of $F_w^{{\text{OHF}}}$ in October 2018 was possibly influenced by the increased |δT_z| with the deepening of SML during the storm event (Figs. 6a, 6c, and 7i). $F_w^{{\text{OHF}}}$ and $F_w^{{\text{ACW}}}$ also increased at higher ACW maximum temperature after buoy DTOP21 drifted southward to approximately 85˚N in late December (Figs. 7b and 7l). Meanwhile, the heat content and temperature maximum of the ACW decreased rapidly from mid-August to mid-September 2021 (gray shades in Figs. 7d and 7i, respectively), which will be analyzed in detail in the following section.

3.5 Potential Mechanism that Caused the Cooling of ACW in Summer 2021

Upwelling and entrainment during storm events are considered to be vital physical mechanisms for the heat loss out of the ACW (Yang et al., 2001). In our study, powerful Ekman suction spread widely all over the Arctic Ocean except for the Beaufort Gyre region in June 2021 (Fig. 8a). More ACW can be swept into the Transpolar Drift with the expansion of Beaufort Gyre in years of the negative AO index (−1.80 in winter 2020/2021). In the flowing process with the Transpolar Drift, the heat stored in the ACW was consumed continuously by vertical heat flux during the upwelling. In July 2021, the vertical Ekman velocity w_e decreased in the Arctic Ocean, apart from the region north of the Canada Basin (80˚ – 85˚N, 100˚ – 160˚W, Fig. 8b). The daily time series showed that in early July, w_e in this area peaked at over 40 cm d⁻¹ (Fig. 8d). Polyakov et al. (2013) noted that episodic atmospheric events, such as cyclones, can distort isopycnal surfaces substantially. This finding explains the abruptly thinned SML from 50 m to 30 m in mid-July (Figs. 2 and 7). The slight weakening of Ekman suction between July and August alleviated the upward heat loss out of the ACW, with the renewal of heat content of more than 80 MJ m⁻² (Fig. 7d). The intensified Ekman suction in August was widespread and covered almost the entire Canadian Basin (Fig. 8c). Since mid-August, w_e in the calculated area escalated again and stabilized at (9.4 ± 8.0) cm d⁻¹ in general (Fig. 8d), which caused a remarkable decline in the ACW temperature maximum and heat content at a mean rate of 1.4 MJ m⁻² d⁻¹ (Figs. 7d and 7i, respectively). Furthermore, the mixed layer correspondingly thinned again from about 40 m to 20 m under the action of cyclonic Beaufort Gyre, which can reshape isopycnal surfaces into an inverted bowl-shaped convex structure, as mentioned above.

The Ekman suction (or cyclonic Beaufort Gyre) can also produce northward Ekman transport in the Northern Canada Basin, which facilitates the release of freshwater that was previously converged in the basin (Giles et al., 2012). For instance, the reduced salinity observed in the upper ocean of the Eurasian Basin in 2010 was believed to be caused by the injection of freshwater as a result of a weakened Beaufort Gyre (Timmermans et al., 2011). As shown in Fig. 7n, with the emergence of the substantially strong Ekman suction, the SML salinity decreased by 0.5 twice in July and August, and it stabilized at approximately 29.0 by October. Notably, the freshening of SML could have partly resulted from the freshwater out of the melting sea ice, given the signs of SML temperature decrease were also observed from August.

4 Conclusions

This study analyzed the temperature and salinity data from the DTOPs deployed over the Alpha Ridge north of the Canada Basin during 2018 and 2021. The data were used to investigate the variabilities of the ACW and the under-ice $F_w^{{\text{OHF}}}$ in this area to estimate the impact on oceanic heat flux of the heat release process out of the ACW and discuss Ekman upwelling as a potential mechanism that caused the cooling of the ACW during the summer of 2021.

The ACW was observed substantially during the period when DTOP18 and DTOP21 drifted in the Arctic Ocean. The temperature maximum of the ACW was generally lower than −1.5℃ in August – September 2018, which was consistent with the climatologic feature of the ACW north of 80˚N. During the summer (June – August) of 2021, the temperature of the 40 m-thick ACW layer generally ranged from −1.35℃ to −0.9℃, with $T_{\max }^{{\text{ACW}}}$ approaching −0.8℃ and exceeding the climatologically average of the same region by approximately 0.4˚C − 0.5˚C. The AO index was −1.80 in winter 2020/2021. The anomalously warm $T_{\max }^{{\text{ACW}}}$ was observed farther northerly over the Alpha Ridge due to the expansion of the Beaufort Gyre in the negative AO phase (Steele et al., 2004). After October, $T_{\max }^{{\text{ACW}}}$ ranged from −1.45℃ to −1.4℃, and the salinity values still ranged between 30.2 and 32.5 in these two years.

The under-ice $F_w^{{\text{OHF}}}$, with the peaks exceeding 25 W m⁻² in the cases of major storms, was intimately correlated with the net shortwave radiation flux entering the upper ocean in summer. The elevation of $F_w^{{\text{OHF}}}$ and the warming of SML both resulted from increased absorption of solar radiation by the upper ocean under circumstances of relatively low sea ice concentration. Furthermore, when the sea ice concentration was reduced, the ice floes were liable to drift with a high velocity during storms due to the great mobility of ice packs. In June – August, the ACW contained more heat (> 80 MJ m⁻²), the stratification upon the ACW layer was weak, and the mixed layer base approached the ACW. All these factors caused an upward vertical heat flux $F_w^{{\text{ACW}}}$ of 3 – 4 W m⁻² out of the ACW. $F_w^{{\text{ACW}}}$ decreased to lower than 2 W m⁻² since autumn, and $F_w^{{\text{OHF}}}$ also showed a considerable reduction to 1.2 W m⁻² on average. During the freezing season, the ACW was apt to release heat upward due to storm disturbance or brine-driven mixing, which resulted in subtle rises in the $F_w^{{\text{OHF}}}$.

North of the Beaufort Gyre region in the Arctic Ocean, Ekman suction (upwelling) became pervasive under the influence of wide cyclonic wind fields. Intensive upwelling dominated north of the Canada Basin during summer, with the maximum Ekman vertical velocity exceeding 40 cm d⁻¹, which caused the thinning of the ACW layer and the decline in the ACW temperature due to the upward heat release process on its pathway of transport. This finding is consistent with the climatologic pattern of the cooling of the ACW in this region. North of the Canada Basin, the enhanced Ekman suction distorted the isopycnal surfaces and released the freshwater content that was previously converged in the basin with the northerly Ekman transport. This condition caused the thinning and freshening of SML, as observed in the summer of 2021. Similar to the studies on double-diffusive heat fluxes out of the ACW with calculated turbulent dissipation rates south of the Beaufort Gyre (Kawaguchi et al., 2012; Fine et al., 2018, 2022), further research is required to quantify more precisely $F_w^{{\text{ACW}}}$ caused by Ekman upwelling or other factors, with vertical diffusivity k determined from fine-scale observations in the central Arctic Ocean.

Acknowledgements

We thank Drs. Xiaoyu Wang and Yilin Liu for their contribution to the DTOP platform preparation and deployment, as well as the crew of the R/V Xuelong for their help. We also thank Linxu Zhai and Xinyun Liu for their valuable comments on the manuscript. This work was supported by the National Natural Science Foundation of China (Nos. 42276239 and 41941012), the National Key R & D Program of China (No. 2019YFC1509101), and the Fundamental Research Funds for the Central Universities (No. 202165005).