The Beaufort Gyre’s Diffusive Staircase: Finescale Signatures of Gyre-Scale Transport

Abstract

Arctic Ocean waters sourced from the Atlantic contain a vast amount of heat. In the Arctic’s Beaufort Gyre, diffusive convection is the primary mechanism by which this heat is transported vertically. This mixing process is characterized by a “staircase” where convective layers are separated by interfaces in temperature and salinity. It is not well-understood what governs layer thickness, which is an important parameter in heat transport. Here we relate staircase properties to the background water-mass structure of the Beaufort Gyre via analysis of Ice-Tethered Profiler observations. We find that staircase layer thicknesses vary with intrusive features below the staircase and the stratification overlying the staircase. We relate these features to the pathway of anomalously warm Atlantic Water in the Beaufort Gyre. Results suggest that intrusive features in context with the Gyre’s large-scale geostrophic flow may be key to understanding layer thicknesses and the propagation of warm waters into the Gyre.

Key Points

• Atlantic Water layer diffusive-convective staircases in the Beaufort Gyre exhibit a west-east gradient in layer thickness

• Layer thicknesses are correlated with the stratification overlying, and intrusive features surrounding, the Atlantic Water core

• This layer thickness gradient may be due to lateral propagation of a warm Atlantic water pulse that entered the central Arctic in the 2000s

Plain Language Summary

Heat from a warm Arctic Ocean water layer, known as the Atlantic Water Layer, is vertically transported by double-diffusive convection. This process frequently is evidenced by a staircase structure. We relate changes in staircase properties to both the background oceanographic setting as well as to an anomalously warm water pulse which entered the Arctic Ocean in the 1990s–2000s. We show that staircase properties are likely related to the propagation of warm water intrusions and subsequent dissipation of their property gradients. This work sheds light on what the western Arctic Ocean may look like under continued climate change.

1 Introduction

Heat contained in the Arctic Ocean has long been recognized for its major influence on Arctic sea ice and climate (Carmack et al., 2015; Maykut & Untersteiner, 1971). Of particular interest is a warm water layer, which originates in the Atlantic Ocean, and contains enough heat at depth to entirely melt the sea ice if all of this heat were somehow transported to the surface (Maykut & Untersteiner, 1971). Here, we analyze observations from the Canada Basin to understand how large scale circulation features there influence heat transport from this Atlantic Water Layer.

The relatively-warm and salty Atlantic Water enters the Arctic Ocean through Fram Strait and the Barents Sea (Figure 1a). In the Canada Basin, the Atlantic Water Layer core sits around 400 m depth, and is overlain by Pacific Summer and Winter water, sourced from the Pacific Ocean (Coachman & Barnes, 1961). Pacific Summer Water ventilates the Canada Basin in the summer and is warmer and fresher than Pacific Winter Water, which ventilates the basin during winter (Timmermans et al., 2014). Pacific Winter Water is cooler and fresher than the Atlantic Water, and sits at a depth of around 150 m (Figures 1c and 1d). Between Pacific and Atlantic waters in the halocline, there is a local stratification maximum, with the potential to insulate the deeper waters from energetic conditions above (Figure 1e).

Figure 1

Arctic observations of the diffusive staircase. (a) Map of the Arctic Ocean indicating the study region (Beaufort Gyre Region, BGR), the Northwind Ridge (NWR), as well as the Atlantic Water (AW, red lines) and Pacific Water (PW, blue lines) inflows. (b) BGR showing locations (dots with color indicating year) of all ITP profiles analyzed between 2004 and June 2020; gray and black squares mark the locations of the representative profiles (color coded accordingly) shown in (c–e) from the central and eastern BGR, respectively. (c) Potential temperature (°C) versus depth, with the diffusive-convective staircase shown in the inset (indicated by gray boxes). Prominent thermohaline intrusions around the core of the Atlantic Water Layer are indicated. (d) Salinity versus depth, with the staircase in the inset. (e) Smoothed buoyancy frequency N² (s⁻²) versus depth (smoothing over 40 data points, approximately 10 m). Black dots in panels (c–e) indicate the depth of .

The Arctic Ocean’s central basins are energetically quiescent (Dosser et al., 2021). In these regions, ocean heat transport from the warm Atlantic Water is dominated by a small-scale mixing process known as diffusive convection (e.g., Padman & Dillon, 1987). This convective-diffusive heat transport process can occur when cooler, fresher waters sit above warmer, saltier waters, allowing for positive temperature and salinity gradients with depth (Radko, 2013; Turner, 1965). Diffusive convection is marked by its distinct “staircase” structure, where convective cells of O(1) m thick are separated by sharp, diffusive interfaces of O(1) cm thick (e.g., Guthrie et al., 2015; Timmermans et al., 2008). These staircases are ubiquitous in the central Arctic (Guthrie et al., 2015; Padman & Dillon, 1988; Shibley et al., 2017).

Vertical heat fluxes from the Atlantic Water Layer are often inferred via 4/3-flux parametrizations (Kelley, 1990; Kelley et al., 2003) which depend upon the temperature jumps across staircase interfaces. These diffusive-convective heat fluxes are generally weak, of O(0.01–0.1) W m⁻² in the central Canada Basin (Padman & Dillon, 1987; Shibley et al., 2017; Timmermans et al., 2008). Staircases may only exist in relatively-quiescent regions (Shibley & Timmermans, 2019), and their presence and properties are a fingerprint of how energy is distributed at intermediate depths in the Arctic Ocean. For example, in Arctic boundary regions, which may experience more energetic flows (Polyakov et al., 2020) and interactions with rough topography (Fer et al., 2010; Rippeth et al., 2015), staircases are frequently absent (Shibley et al., 2017).

Below the staircase, prominent thermohaline intrusions (Figure 1c), are frequently found (Carmack et al., 1998; May & Kelley, 2001; Woodgate et al., 2007). These intrusions originate near the Northwind Ridge as relatively warm and salty waters propagating laterally into relatively cool and fresh water (the ambient water in the Canada Basin) (e.g., McLaughlin et al., 2009). This generates sharp vertical temperature and salinity gradients at the upper and lower boundaries of distinct intrusive features (e.g., Carmack et al., 1998; May & Kelley, 2001; Woodgate et al., 2007), which further appear to be related to the development of the staircase (Bebieva & Timmermans, 2017, 2019). Intrusions propagate horizontally, driven by divergences in vertical heat and salt fluxes across these gradients. The fluxes ultimately dissipate the gradients to render them vertically uniform (Bebieva & Timmermans, 2019). We refer to this state as that of “run-down” intrusions.

The Beaufort Gyre, an anticyclonic wind-driven ocean circulation feature in the Canada Basin (Proshutinsky et al., 2009), provides an opportune environment to explore how the temperature-salinity structure and associated heat fluxes of a diffusive-convective staircase may be influenced by features of the background setting. Atlantic Water transits to the Beaufort Gyre as part of the Atlantic boundary current or after propagating eastward from the Northwind Ridge (McLaughlin et al., 2009), while Pacific Water propagates into the central Gyre via the Chukchi Sea (Timmermans et al., 2014). A diffusive-convective staircase is prominent in the Beaufort Gyre; fluxes through the staircase are a heat source from the Atlantic Water to the overlying Pacific Water.

This paper investigates variability in the staircase across the Beaufort Gyre Region and relates it to the circulation features of the Gyre. In the next section, we describe the ITP data set and the methodology used to identify staircase properties. Then, we provide a general overview of the dominant features of the Beaufort Gyre, including the broad distribution of Atlantic Water properties in the Canada Basin. Next, we describe the properties of the staircase in the same region, and how these properties vary and relate to the large-scale background features of the water column. Finally, we speculate on how these might be related to a changing Atlantic Water inflow.

2 Data and Methods

The data analyzed here are from Ice-Tethered Profilers (ITPs; Krishfield et al., 2008; Toole et al., 2011). An ITP consists of a buoy that sits atop the surface sea ice, with a wire which extends into the ocean to about 750-m depth (Krishfield et al., 2008). A CTD profiler traverses the wire at a rate of ∼25 cm s⁻¹, acquiring water-column profiles of conductivity, temperature, and pressure approximately twice per day (Krishfield et al., 2008). A total of 3,286 profiles (only upcasts) with a vertical resolution of about 25 cm, spanning the region from approximately 160 to 130°W and 73 to 81.5°N are analyzed in this study (Figure 1b). This is a profile density of about 47 observations per 100 km².

We locate the halocline stratification peak (denoted ) by finding the maximum buoyancy frequency deeper than 103 m (to avoid the surface-ocean mixed layer base stratification maximum) in smoothed profiles (over every 40 samples, about 10 m). In the next section we describe how staircase layers and interfaces are identified; the procedure is similar to a combination of the methods outlined by Shibley et al. (2020) and Shibley et al. (2017).

2.1 Identifying Layers and Interfaces

We first examine the vertical profile of vertically-differenced potential temperature in the depth segment between the Atlantic Water potential temperature maximum and the depth at which salinity is 34 (provided that the pressure is ≥175 dbar and that the profiles span a pressure range of at least 350 dbar). Local maxima in the depth segment are taken to indicate the presence of interfaces if the following criteria are met: any local maximum must be larger than one standard deviation from either the largest adjacent local minimum or the endpoints of the segment; any local maximum must be at least 0.5 standard deviations above the mean of the potential temperature difference segment; if two maxima are found within 10 cm of each other, only the larger is considered to indicate an interface. The start and endpoints of each interface are taken to be the point at which the vertical potential temperature differences become 10% or less of the peak magnitude.

To identify mixed layers, we require at least three consecutive points. The potential temperature gradient across a mixed layer must be ≤0.009°C m⁻¹ and at least 5 times smaller than that across the overlying interface, and the potential temperature jump across mixed layers must be ≥−0.0025°C. In each profile, the first and last interface in the depth segment is removed, as are interfaces (and the mixed layers overlying them) with negative jumps in potential temperature or jumps greater than 0.11°C. Then, a profile is considered to have a staircase if at least 15 mixed layers are found, and the sum of mixed layer thicknesses in the profile is ≥25% of the total thickness of the depth segment.

In each profile with a staircase, we find the mixed layers closest to depths 12.5-m above and below the 1,027.7 kg m⁻³ isopycnal (this isopycnal generally corresponds to an intermediate depth of the staircase, based on visual inspection of the data), and consider mixed layers/interfaces only within this depth segment (which generally spans about 25 m). Three criteria are applied to the segment to restrict the analysis to well-defined staircases: (a) The standard deviation of the mixed layer thicknesses must be <40% of the mean mixed layer thicknesses in that segment; (b) The potential temperature differences between the middle and upper depths of the segment, and its middle and lower depths must differ by ≤0.05°C; (c) At least 5 mixed layers in this ∼25 m region must be present. Finally, profiles where the salinity at the depth of the is less than 32.5 are excluded. We quantify layer thicknesses, interface potential temperature jumps, and bulk density ratios across this depth segment.

3 Results

3.1 Properties of the Beaufort Gyre

We begin by outlining the hydrographic properties of the Beaufort Gyre based on ITP profiles (Figure 1b). The imprint of the anticyclonic Gyre circulation is observed as a deepening of isopycnal surfaces toward its center (Proshutinsky et al., 2002) (Figure 2a overlay). The potential temperature maximum of the Atlantic Water Layer is warmest in the central Beaufort Gyre closest to where it first enters, east of the Northwind Ridge, and is coldest near the eastern boundary (Figure 2a); potential temperature values range from 0.5°C to 1.2°C. Overlying the diffusive-convective staircase in the range around 150–250 m, there is a peak in stratification (i.e., buoyancy frequency, denoted by ), a local maximum in the Beaufort Gyre halocline stratification (Figure 1e). This peak is associated with the interface between the Pacific Winter Water and the Atlantic Water, and is largest near the Northwind Ridge and the eastern boundary of the Gyre, and smallest in the central Gyre (Figure 2b); values range from about 1.5 × 10⁻⁴  s⁻² to 3.7 × 10⁻⁴ s⁻².

Figure 2

Staircase variability in the Beaufort Gyre Region. Map of (a) the Atlantic Water potential temperature maximum (°C) with the depth at which salinity = 34 overlain (the Northwind Ridge, NWR, is labeled), (b) the peak buoyancy frequency (s⁻²), (c) mean layer thickness (m), (d) the potential temperature variance (°C²) in a 100-m depth region centered at the Atlantic Water potential temperature maximum, and (e) bulk density ratio R_ρ. (f) Mean layer thickness (m) against (s⁻²) (gray dots) and binned in seven 0.25 × 10⁻⁴ s⁻² increments, with error bars indicating ±1 standard deviation in binned mixed layer thickness. Only bins with at least 10 data points are included.

3.2 Staircase Properties in Context With the Beaufort Gyre

Similar to the large-scale Gyre properties, the structure of the Atlantic Water staircase varies across the Beaufort Gyre. An understanding of how staircase layer properties change across the Gyre may give insight into the processes regulating vertical heat transport.

In the central Gyre staircase, layer thicknesses are generally thinnest, and increase toward the eastern Gyre (Figure 2c) from an average of 1.5 ± 0.5 m west of 145°W to an average of 2.2 ± 0.6 m east of 145°W. There is also a west-east gradient in mean potential temperature jump across interfaces (increasing from an average of 0.037 ± 0.008°C west of 145°W to an average of 0.042 ± 0.009 m east of 145°W, not shown); an increase in temperature jump toward the east coincides with thicker layers there. Moreover, there is a section of the northwestern Beaufort Gyre, close to where water enters the Canada Basin via the Northwind Ridge, which exhibits both larger mean layer thickness and larger interface potential temperature jumps than the general west-east pattern (i.e., 1.8 ± 0.5 m and 0.042 ± 0.008°C in the region west of 145°W and north of 80°N).

To consider how intrusions vary across the Gyre, we calculate the potential temperature variance of a profile over a 100-m segment centered at the Atlantic Water potential temperature maximum; the mean considered here is the mean of the potential temperature values in this subset of the data. The temperature variance gives an approximate indication of the strength of thermohaline intrusions; high variance generally indicates prominent intrusions (black profile, Figure 1c), whereas lower variance generally indicates run-down intrusions (gray profile, Figure 1c). Run-down intrusions have already fluxed their heat and salt vertically and are characterized by homogenized potential temperature gradients. Similar to the spatial pattern in staircase layer thicknesses, temperature variance displays lower values (approximately <0.001°C²) in the central Gyre, and generally higher values (∼0.001–0.004°C²) on the eastern side and near the Northwind Ridge (Figure 2d), although low values may be found throughout. Several lower values (generally <0.0003°C²) appear to be due to nearly homogeneous temperature gradients in the 100-m segment surrounding the Atlantic Water potential temperature maximum, largely from the latter half of the temporal record. We also note that the large value found near 76°N, 160°W is due to a single point of anomalously high variance surrounding the Atlantic Water core, likely the signature of a warm-core eddy.

Finally, we calculate an approximately 25-m bulk density ratio, R_ρ = (βΔS)/(αΔT), across the Gyre. Here β is the coefficient of haline contraction, and α is the coefficient of thermal expansion. The density ratio gives an indication of the stability of the water column; higher values of R_ρ signify a more stabilizing influence of salinity in the water column than in regions with lower values of R_ρ. We find that R_ρ values are higher on the eastern Gyre boundary (around 6–7) and lower in the central Gyre (around 4–5), although there are isolated regions of higher density ratio interspersed (Figure 2e).

3.3 Interpretation of Relationships

An investigation of bulk properties, such as density ratio, Atlantic Water potential temperature maximum, and stratification, along with potential temperature variance around the Atlantic Water core, suggests that diffusive-convective staircase properties are related to property gradients and the large-scale geostrophic flow of the Beaufort Gyre. In particular, thick layers appear to arise in regions with both prominent intrusions and strong overlying stratification, such as the eastern Gyre. Thin staircase layers are found in the central Gyre, where intrusive features are less prominent (i.e., smaller temperature variance around the Atlantic Water Layer core) and the overlying stratification is generally smaller. A robust linear relationship (with an r² value of 0.99) is apparent between binned staircase layer thickness and overlying stratification (Figure 2f). We note that there is not a clear relationship between layer thickness and bulk R_ρ, a quantity which has at times been invoked as a key parameter in setting layer thickness (Kelley, 1984). We hypothesize that the lateral change from thick to thin staircase layers (prominent intrusions/large to run-down intrusions/small ) is generally related to Atlantic Water circulation in the Beaufort Gyre. We next outline this possible relationship, which is schematized in Figure 3.

Figure 3

Schematic indicating the west to east gradient in staircase layer thickness in context with Atlantic Water propagation (including via thermohaline intrusions) and Beaufort Gyre circulation. Thin staircase layers, run-down intrusions, and low values of overlying stratification generally occur together, while thick layers, prominent intrusions, and high are generally observed together.

In the late 1990s, a warm Atlantic Water pulse entered the Arctic Ocean, which by the early 2000s had entered the Canada Basin (McLaughlin et al., 2009; Polyakov et al., 2011). There are two pathways for Atlantic Water: one is via the cyclonic boundary current at the geographic margins of the basin. The other is via thermohaline intrusions that propagate from west to east into the Gyre, from the AW core where the cyclonic boundary current encounters the Northwind Ridge. These intrusions are further advected anticyclonically following the large-scale geostrophic flow of the Beaufort Gyre (see e.g., McLaughlin et al., 2009), although the precise advection pathways are unclear. Intrusions run down as they generally propagate from west to east via buoyancy fluxes. The strong signature of intrusions on the eastern side of the Beaufort Gyre (east of 140°W), manifest by relatively large vertical temperature (and salinity) variance through the Atlantic Water core, is likely related to the propagation of this warm pulse across the Beaufort Gyre (McLaughlin et al., 2009).

In the central Gyre (west of 140°W) where intrusions are less prominent, we expect that intrusions have already propagated through the central basin, and the intrusive features there are now in a run-down state (Figure 3). This may be reasoned as follows. A characteristic intrusion propagation speed is ∼1 mm s⁻¹ (Bebieva & Timmermans, 2019). Thus, it would take around a decade for intrusions to propagate the few hundred kilometers from the vicinity of the Northwind Ridge to the central Beaufort Gyre region. Therefore in the central Gyre, for most of the time over which we have samples (2004–2020), the intrusions have likely already run down, consistent with our inferences from the ocean finestructure. The presence of active (i.e., not in a run-down state) intrusive features on the eastern side of the (approximately 800 km wide) basin is also consistent with intrusion propagation timescales. There is uncertainty in these estimates, as it is not well known how the propagation and pathways of the intrusions vary as they are advected by the Gyre geostrophic circulation.

Regions of prominent intrusions generally appear to correspond with regions of strong overlying stratification, and vice-versa, although this relationship is not necessarily causal. The fact that weaker intrusive features generally correspond to regions of lower stratification (characterized by ) may be due to either a freshening of the inflowing Atlantic Water (Glessmer et al., 2014) and/or a freshening of the Atlantic Water in the Canada Basin associated with a downwards salt flux as intrusions run down (Bebieva & Timmermans, 2019), and/or mixing between Atlantic and Pacific Water layers. Any of these processes would influence only the central Gyre (where run-down intrusions are present), but not yet the eastern side (where prominent intrusions are present), and would reduce the stratification overlying the staircase. Further, the strong stratification and thick staircase mixed layers evident in the northwestern part of the domain is observed in the vicinity of Atlantic Water first entering the Gyre under the influence of the northern extent of its anticyclonic circulation (McLaughlin et al., 2009). We note an exception in the southeastern part of the domain where thick staircase mixed layers exist in the presence of smaller (Figure 2b). These measurements are predominantly from the years 2012–2014, and it is possible that uneven temporal sampling by ITPs over the 17-year duration of the data set may be interrupting mean spatial patterns.

This then raises the question of why prominent intrusions and strong overlying stratification appear to be linked to thicker layers, and run-down features appear to be linked to thinner layers. A possible explanation is that after thermohaline intrusions transition to homogenized layers (Bebieva & Timmermans, 2019), processes such as interface splitting further evolve the staircase leading to thinner layers overall (Kelley, 1988). In addition, it may be possible that stronger overlying stratification helps to insulate the staircase from more energetic motions in the upper-ocean, allowing layers to grow thicker.

The relationship between staircase layer thicknesses, intrusions, and overlying stratification suggests that the background Gyre structure and dynamics are a key factor in setting the properties of the diffusive staircase, and thereby a control on the vertical heat transport. However, vertical diffusive-convective fluxes in the Beaufort Gyre are small (O(0.01–0.1) W m⁻², Shibley et al., 2017), and it is unlikely that changes in these small fluxes will result in appreciable differences in the heat budget. On the other hand, it is useful to understand staircase layer thicknesses in that they provide a marker of the changing water mass structure of the Gyre, as described here. Moreover, staircase properties, or the presence or absence of a staircase, give insight into local levels of turbulence in a region (Guthrie et al., 2017; Shibley & Timmermans, 2019); such turbulence levels in turn affect heat transport.

Past studies relating staircase layer and interface properties to bulk scale parameters have largely focused on the background stratification and temperature gradient (Fedorov, 1988; Kelley, 1984, 1988) outside the context of the ocean circulation. In the context of the Arctic Ocean, these comparisons have largely involved the Atlantic Water potential temperature maximum and the bulk background density ratio (Shibley et al., 2017), but have lacked a sufficient spatiotemporal record to fully investigate circulation features as they pertain to the staircase. Here, we suggest that background ocean structure and dynamics are influential in setting and understanding the Atlantic Water staircase properties.

4 Discussion

Properties of the Atlantic Water Layer’s diffusive-convective staircase appear to be tied to both the intrinsic propagation of intrusive features across the Gyre as well as their advection by the large-scale geostrophic flow of the Gyre. Future analyses are required to understand the precise pathways of thermohaline intrusions in the basin and how shifting Gyre dynamics influence the spreading pathways. We have demonstrated the broad result that staircase layers appear to be thinner where the Atlantic Water intrusions are less prominent, or in a run-down state (Figures 2c, 2d and 3). This result may seem intuitive because run-down intrusions have already given up a significant portion of their anomalous heat and salt to the overlying and underlying water, in contrast to the larger-amplitude active intrusions (e.g., Bebieva & Timmermans, 2017). Under continued warming of the Atlantic Water in the Beaufort Gyre, lateral temperature gradients between incoming and Gyre waters are likely to be smaller due to the warming of background waters, and thermohaline intrusions are likely to propagate shorter distances before running-down (see Figure 6 of Bebieva & Timmermans, 2019). Since we expect that run-down intrusions correspond with thinner staircase layers, we anticipate thinner layers in a warming Beaufort Gyre. This raises the interesting question of how and when the staircase might disappear altogether, making way for a turbulent transfer of heat. Such a change is not decoupled from the stratification overlying the staircase, as we have seen here in the generally positive relationship between staircase mixed layer thicknesses and the overlying stratification.

If the Atlantic Water inflow freshens in the future (see, e.g., Glessmer et al., 2014), the stratification overlying the staircase may decrease. At the same time, overlying Pacific Winter Water may also freshen (see Woodgate & Peralta-Ferriz, 2021), which could increase the overlying stratification. In a turbulent setting, if the stratification becomes weaker overall, we expect this to be associated with enhanced surface-ocean energy propagating to the depths of the staircase, and therefore a likely increased transfer of ocean heat. This discord highlights the need for analyses (such as direct numerical simulations) to probe the relationships between turbulence and staircase layer thickness for the range of energy levels commensurate with the observations. If the Arctic Ocean becomes more energetic under continued sea-ice losses, as has been hypothesized at least in the Eurasian Basin (Dosser et al., 2021; Polyakov et al., 2020), the effect on diffusive-convective heat fluxes from the Atlantic Water Layer remains an open question. Evolving staircase properties may offer an opportunity to further understand the complex pathways and modes of heat transport in the Arctic Ocean.

Acknowledgments

Support was provided by the National Defense Science and Engineering Graduate Fellowship and the Princeton Center for Theoretical Science (NCS), as well as the National Science Foundation Division of Polar Programs (M-LT).

Open Research

Data Availability Statement

The Ice-Tethered Profiler data were collected and made available by the Ice-Tethered Profiler Program (Krishfield et al., 2008; Toole et al., 2011) based at the Woods Hole Oceanographic Institution (https://www2.whoi.edu/site/itp/). Ocean Data View (Schlitzer, 2015) was used to produce the maps.