A Polar Surface Eddy Obscured by Thermal Stratification

Mesoscale and submesoscale eddies play an important role in the distribution of heat and biogeochemical properties throughout the global oceans. Such eddies are important in the Arctic Ocean, particularly in the frontal regions, but are difficult to detect using traditional satellite‐based methods. Here we use high‐resolution in situ data from an underwater glider to identify a surface eddy that was masked from remote‐sensing observations. We hypothesize that this masking was driven by thermal stratification driven by surface heat fluxes. The eddy was likely generated north of the Polar Front, before crossing the front and traveling south. We estimate that the observed eddy contained 4 × 1010 m3 of Arctic Water. The observation of this eddy, masked in satellite observations of sea surface temperature, suggests a historical underestimation of the prevalence and importance of eddies in this key mixing region. The water column of the Barents Sea, one of the circumpolar Arctic seas has a seemingly simple structure. In the south, warm Atlantic Water dominates; in the north, cold Arctic Water dominates; while at their boundary, the Arctic Water overlies the Atlantic Water. In the summer, the Arctic Water is largely devoid of the nutrients required to fuel the growth of phytoplankton, which is key to maintaining life in the ocean. In contrast, the Atlantic Water is one of the primary sources of nutrient‐rich water into the Arctic. In this study, we have used an underwater robotic instrument to identify a patch of Arctic Water which has been shed from the Arctic sector of the Barents Sea into the Atlantic sector. This patch of water is seen to have lower phytoplankton concentrations than the surrounding water. Due to atmospheric heating of the surface, this patch would be indistinguishable from the surrounding Atlantic Water and so would be absent for satellite observations of sea surface temperature. We suggest that this temperature masking has meant that we have previously underestimated how much water is moved within these patches in the Arctic seas.


Introduction
The Barents Sea is a key gateway for advection of warm Atlantic Water (AW) into the Arctic Ocean (Loeng, 1991) and the export of Arctic Water (ArW) and ice from the Arctic Ocean (Kwok, 2009). As the meeting point of these two oceanic pathways, the Barents Sea supports high biological productivity (Erga et al., 2014). This region has undergone substantial warming over recent years, with significant changes in its structure and heat content (Lind et al., 2018) as well as local ice dynamics (Comiso & Hall, 2014). In order to predict the impacts of these changes on the physical and biological processes in the region, we need to ensure that we fully understand the structure and spatial variability of the water column.
Heat, salt, and nutrients are advected into the Barents Sea, in AW ( Figure 1). On the southern shelf, the AW is present at all depths, remaining mixed throughout the winter before becoming thermally stratified during summer (Harris et al., 1998). Because the thermal stratification is weak, nutrient-rich deep waters are frequently mixed to the surface, replenishing surface nutrient inventories and sustaining primary production by phytoplankton (Reigstad et al., 2002). Further north, AW is confined to intermediate depths and is overlain by cold, fresh ArW (Harris et al., 1998). This northern water is mixed during the winter resulting in nutrient-replete surface waters (Wassmann & Reigstad, 2011). However, ice melt during the summer drives early, strong salinity stratification. The surface euphotic zone becomes depleted of nutrients as a result of the export of primary production out of the surface mixed layer, while strong stratification inhibits the vertical mixing required to resupply the surface with nutrients sourced from underlying waters (Wassmann & Reigstad, 2011).
The contrasting regions of the northern and southern Barents Sea are delineated by the Polar Front (PF) (Figure 1), which manifests as strong lateral gradients in temperature and salinity (Fer & Drinkwater, 2014;Harris et al., 1998). This boundary between water masses is subject to strong vertical mixing (Mahadevan, 2006), which enhances the supply of deep-sourced nutrients, making the PF a region of high primary production (Erga et al., 2014). Lateral transport can overcome the front, advecting nutrients, heat, salt/fresh water, and plankton between the water masses (Crews et al., 2018). Such transport relies on the breakdown of geostrophic control at the front (Huthnance, 1995) and the development of eddies and meanders that convert potential energy of the front into kinetic energy (Pingree, 1978).
Despite being localized features, eddies can have disproportionately wide impacts, influencing global circulation (e.g., Thompson et al., 2014) and biogeochemical cycles, driving enhanced carbon export (Omand et al., 2015) through their role in initiating spring blooms (Mahadevan et al., 2012). In the Arctic, horizontal advection by eddies affects nutrient transport and leads to localized phytoplankton patchiness, enhancing or limiting productivity depending on the impact on bloom conditions (Kushnir et al., 2010). Vertical transport within eddies further contributes to their significant role in Arctic biogeochemical cycles (Watanabe et al., 2014). Eddies are also important for redistributing heat across the Arctic (Hattermann et al., 2016) and ventilating subsurface waters (Crews et al., 2018).
The primary methods for quantifying the occurrence of eddies are models (Crews et al., 2018) and satellite observations of either sea surface temperature (SST) (Hausmann & Czaja, 2012), ocean color (Hausmann & Czaja, 2012) or sea surface height (SSH) (Itoh & Yasuda, 2010). However, the prevalence of cloud and rain in Arctic regions imposes limitations on using remote-sensing techniques to detect eddies from their characteristic SST and ocean color signatures. In the Barents Sea, the potentially small size of these eddies, due to the short Rossby radius of deformation (Crews et al., 2018), further exacerbates these observational difficulties, given the biased spatial resolution of the satellite data. Recently, synthetic aperture radar (SAR) has been used to address these resolution-based limitations (Atadzhanova et al., 2018;Kozlov et al., 2019). Beyond these satellite observations, in situ studies of eddies in the Barents Sea are limited, if not absent from the literature.
The use of satellite observations to estimate eddy abundance may also be limited by the presence of subsurface eddies (Kostianoy & Belkin, 1989;Manley & Hunkins, 1985) that are not detectable by remote sensing. We suggest that masking of eddies from SST and SAR imagery could be apparent in regions where there are strong atmosphere-ocean heat fluxes (Smedsrud et al., 2010), such as the Barents Sea. Thus, in regions where estimates of eddy activity are derived largely from remote sensing, the number of eddies may be underestimated substantially due to these masked or hidden eddies. Augmenting our observations with these hidden eddies, particularly their hydrographic and biogeochemical properties, will facilitate a more complete understanding of transport pathways and their impacts on phytoplankton dynamics and allow for improvements in our estimates of primary productivity in this key biological region.
In this study, we present a unique data set obtained using an ocean glider that captures an eddy traveling southward from the PF. We combine fields of temperature, salinity, and chlorophyll-a fluorescence (chl-a) measured by the glider operating south of the PF satellite observations to evaluate the properties of the eddy in a larger spatial and temporal context. The effects of this eddy on the local chlorophyll concentrations are discussed, as well as the implications for our current understanding of eddy prevalence in the Barents Sea PF region.

Data sets and Methods
To investigate submesoscale-mesoscale variability and instabilities in the PF region of the Barents Sea, we deployed a glider south of the PF to sample a south-north transect along 30°E. During the period of the glider deployment, the transect was occupied by the RRS James Clark Ross. Observations of hydrographic properties and chl-a were made in order to allow for discussion of observed hydrographic features and to provide calibration profiles for the glider.

Hydrographic Data
We deployed a Slocum Glider (Webb et al., 2001), Unit 306, on 17 July 2017 to complete a transect along 30°E between 74.5°N and 76.5°N and recovered it on 31 July 2017. The glider profiled to 200 m on each dive, collecting a total of 744 profiles in a zig-zag pattern with an average profile spacing of 1 km.
The conductivity, temperature, and depth (CTD) data from the glider were processed in accordance with Garau et al. (2011) before being corrected to the calibrated temperature and salinity profiles collected from the ship at either end of the transect, on deployment, and recovery. Profiles of buoyancy frequency (N 2 ) along the glider transect were calculated, with the depth of maximum N 2 giving a biologically relevant proxy for the lower limit of the mixed layer (Carvalho et al., 2017).
The glider-based chl-a data were derived from fluorescence at wavelengths of 470/695 nm, measured by a Wetlabs triplet ECO Puck. While these values do not quantitatively represent phytoplankton biomass, they allow for the identification of spatial and temporal patterns and gradients in the phytoplankton distribution. The chl-a data were checked for consistency using water samples collected during the calibration profiles. Triplicate samples for chl-a analysis were measured onboard using the fluorometric method (Holm-Hansen et al., 1965). Between 100 and 200 ml of seawater were filtered onto 0.7 μm Whatman GF/F filter papers then extracted overnight in the dark in 90% acetone in a −20°C freezer. Pigment extracts were then analyzed using a calibrated Trilogy fluorometer (TurnerDesigns, www.turnerdesigns.com, accessed: 2020).
The corrected, quality-controlled glider data set was gridded along the transect using a Barnes optimal analysis (Barnes, 1994) with a search radius of 4 km, representing the local Rossby radius. Due to surface and deep apogee maneuvers of the glider, we have only considered data deeper than 10 m and shallower than 190 m.

Satellite Data
Satellite-derived SST and SSH data were acquired for the duration of the glider deployment. The SST data are MURSST (Multi-scale Ultra-high Resolution Sea Surface Temperature) (Chin et al., 2017; NASA, 2019, https://podaac.jpl.nasa.gov/Multi-scale_Ultra-high_Resolution_MUR-SST, last accessed: 2020). The SSH data are delayed-mode absolute dynamic topography from Copernicus Marine Environment Monitoring Service (CMEMS, 2019, https://cds.climate.copernicus.eu/cdsapp#!/dataset/sea-level-daily-gridded-data-for-the-global-ocean-from-1993-to-present?tab=overview, last accessed: 2020). These data sets are merged products and as such are available regardless of the cloud cover. The merged SST data are a combination of low-resolution microwave data (during cloud cover or rain) and high-resolution radiometer data (during clear skies) with resolution dependant on the local weather conditions. SST and SSH data are used to investigate surface signals of mesoscale variability and to identify the location of the PF during the glider mission. As a proxy for its location, we calculated the magnitude of the SST gradient (Barton et al., 2018) following 3. Results

The Surface Location of the PF
Throughout this study, the latitude of the surface expression of the PF in the Hopen Trough at 30°E was between 77°N and 78°N ( Figure 2) although its strength and location were variable. Based on sea surface gradients, we calculate that the surface expression of the PF was located more than 55 km from the northern end of the glider transect.

Water Column Variability South of the PF
The glider observations showed AW (1°C < T < 8°C and 35.17 g/ kg < S < 35.37 g/kg (adapted from Harris et al. (1998)) dominating throughout the glider profiles (Figure 3a,b), with a shallow surface layer depth (defined as the depth of maximum N 2 ) of up to approximately 40 m, and strong surface stratification. Within the surface layer, measured chl-a concentrations were found to be higher than 2 μg/l reflecting significant phytoplankton biomass (Figure 3c). A deep incursion of colder and fresher water was found north of 75.4°N. This modified AW is evidence of mixing between ArW (T < 0°C and 34.46 g/kg < S < 35 g/kg (adapted from Harris et al. (1998))) and AW and is indicative of the glider entering the region influenced by the PF.
Along the transect, between approximately 75.3°N and 75.7°N, the glider encountered a notable anomaly 0.02 g/kg fresher than the adjacent surface water (Figure 3b). This surface salinity feature was also identified in the temperature field but with a weaker signal 0.4°C cooler than adjacent water (Figure 3a). This surface anomaly was tied to a colder fresh anomaly down to 100 m depth 0.7°C and 0.02 g/kg lower than adjacent water (Figures 3a and 3b). The water within the anomaly is fresher than elsewhere on the transect, suggesting it did not have a local source. Instead, this water (T = 2.4°C, S = 35.22 g/kg, at 100 m) lies on a mixing line between ArW and AW, indicating that it likely originated from north of the PF. The cold, fresh water covered an area with a diameter of~30 km and a depth of~100 m.
The anomaly shows increased N 2 in the upper water column, when compared to the rest of the transect (Figure 3d), reflecting the additional freshwater in the system. In the 0-20 m layer of the water column, the increased N 2 is driven by increased thermal stability as well as reduced salinity, while low salinity is the sole driver of increased N 2 between 20 and 60 m. The greater stability in the 0-20 m layer compared to the 20-60 m layer (Figure 3d) can be attributed to the additional thermal stratification at the surface which enhances water column stability.
Within the low-salinity, low-temperature anomaly, the surface chl-a concentration (0.8 μg/L) is lower than the ambient water (2.1 μg/L), and a subsurface chl-a maximum is observed, centered on approximately 50 m (Figure 3c). This contrasts with the rest of the glider transect, where chl-a concentrations were highest at the surface and decreased with depth.

Discussion
We have identified a cold, fresh anomaly south of the Barents Sea PF with an approximate radius of 15 km. This anomalous feature had low chl-a and was fresher in surface waters and colder and fresher at depth than the ambient water masses, indicating its source region as within or north of the PF. The location of this feature in otherwise warm and saline AW leads us to describe it as a cold-core eddy.
This eddy showed a surface temperature signal up to 1°C cooler than the surrounding waters but is not detectable within the satellite SST data (Figure 4a) which has a spatial resolution of 1-35 km and a thermal resolution of 0.3-1°C (NASA, 2019). We assume that given the structure visible in the available SST data that the time period of interest is toward the lower end of the spatial resolution of the MURSST data set. However, the eddy is observable as a slight depression in satellite SSH data (Figure 4b) which allows us to track its development/movement from the PF region toward the area where we measured it ( Figure 5). We show that this eddy is thermally capped such that thermal stratification decouples the thermal properties of the eddy from the surface signal (Figure 3), confounding its identification through SST.
The eddy we have described here has stronger thermally driven stability at its surface that at depth, however, we do not see a similar difference between the surface and deep haline stability (Figures 3d and 3e). If the local oceanographic conditions were driving the thermal stratification, we would expect to see a concurrent increase in salinity and a decrease in surface haline stability. We hypothesize that the change in surface thermal stability and the thermal capping of the eddy is due to local atmospheric heat fluxes.
Surface heat fluxes act over the entire region; hence, this heat input will have occurred over the entire observational area. However, in the region of the eddy, the surface fluxes act on a particularly stable patch of low salinity water, further stabilizing it. Here the energy is confined to the shallowest waters, increasing their stability. Using NCEP/NCAR reanalysis values (Kalnay et al., 1996) for latent heat (Q LH ), sensible heat (Q SH ), longwave radiation (Q LW ), and shortwave radiation (Q SW ), we can estimate the local net surface heat flux (Q T ) following We use this estimate to compare the thermal properties of the observed eddy with those of a model eddy whose temperature is initially isothermal (set equal to the subthermocline temperature of 2.75°C) and subject to the The similarity of the heat contents indicates that surface warming of the observed eddy through atmospheric heat fluxes after being shed at the PF is a plausible mechanism leading to the creation of a thermal cap which would obscure the eddy from observation by satellite-derived SST. It is clear that this process will occur in tandem with restratification processes such as discussed by Bosse et al. (2019), who describe lateral heat fluxes and exchanges.
Given that satellite SST observations are limited by both resolution and depth penetration (Chin et al., 2017), we suggest that the current estimates of surface eddies in this oceanographically-and biologically-important region are low, due to both inadequate spatial resolution to capture these features and the effects of thermal capping of cold-core eddies. The stability anomalies (Figures 3a and 3d) observed by the glider delineate the edges of the eddy, where we crossed it. If we assume that we crossed through the center of the eddy, this offers a conservative suggestion that it has a radius of approximately 15 km and a depth of 60 m, giving a total water volume of 4 × 10 10 m 3 . This is much larger than the local Rossby radius which is closer to 4-5 km (Nurser & Bacon, 2014), suggesting a mesoscale eddy.
Estimates from Atadzhanova et al. (2018) show that approximately 25% of the eddies in the Barents Sea occur in the region around the PF. Their estimates of the average number of cold-core eddies in 2007 and 2011, with an average diameter of 3.6 km, combined with our estimates of eddy depth (a conservative estimate of 60 m) suggest that up to 1.1 × 10 13 m 3 of low-chl-a water could be transported to the south from the PF region each year. Given that the eddy identified in this study is both masked from detection by thermal capping and larger than those suggested by Atadzhanova et al. (2018), this is likely to be an underestimation of the total volume of this cold, fresh, potentially low nutrient water transported in eddies south of the PF during summer. Such underestimation will impact our understanding of water mass transport across the PF and its biogeochemical implications with respect to larvae distribution (Schlüter & Rachor, 2001), nutrient concentrations, and primary productivity (Kushnir et al., 2010).
The northward progression of Atlantic conditions in the Barents Sea (Lind et al., 2018) may weaken or displace the PF (Oziel et al., 2016), influencing the location and frequency of eddy genesis. The results of the present study show that such changes in the PF region may not be observable using traditional SST based approaches due to capping of surface eddies. To understand the impacts of any change in eddy activity on the transport and redistribution of heat, freshwater, nutrients, and biomass, we need to quantify the effect of surface masking on our estimates of eddy activity. The phenomenon of thermally capped eddies observed here in the Barents Sea may be applicable to all regions subject to high surface heat fluxes.

Summary
We used a glider, satellites, and ship-based profiles to identify and describe a thermally capped PF eddy in the Barents Sea. The water within this eddy was characterized by low surface chl-a. Using glider-derived estimates of the extent of the measured eddy and previous estimates of eddy size and abundance in this region (Atadzhanova et al., 2018), we have calculated a minimum annual volume transport of 1.1 × 10 13 m 3 of cold, fresh low-chl-a water from the PF region into the Atlantic sector of the Barents Sea.
The thermal capping of this eddy and its size makes its detection by SST observations or traditional sampling methods difficult. This gives rise to the potential for historical underestimates of the occurrence of such features leading to a significant gap in our understanding of the physical biogeochemical and ecological importance of eddies within this PF region. This work has clear implications for our understanding of hydrographic variability around Polar Frontal regions and the transport of heat, salt, and nutrients across these hydrographic barriers.