Identification of Thermohaline Sheet and Its Spatial Structure in the Canada Basin

Abstract

34 individual thermohaline sheets are identified at depths of 170–400 m in the Canada Basin of the Arctic Ocean, by using the hydrographical data measured with the Ice-Tethered Profilers (ITPs) between August 2005 and October 2009. Each sheet is well determined because the salinity within itself remains quite stable and the associated salinity anomaly is markedly smaller than the salinity jump between neighboring sheets. These thermohaline sheets are nested between the Lower Halocline Water (LHW) and Atlantic Water (AW) with lateral coherence of hundreds of kilometers and thickness varying from several to dozens of meters. While the physical properties, including temperature, heat flux, and vertical turbulent diffusivity in the sheet are found to be averagely associated with the AW propagation. Spatially, the thermohaline sheet is in a bowl-shaped distribution, which is deepest in the basin center and gradually becomes shallower towards the periphery. The interaction between the LHW and AW could be demonstrated through the properties variances in the sheets. The temperature variances in the upper and lower sheets are correlated with the LHW and AW, respectively, transited at the 15th sheet, while the depth variance in the sheet is strongly correlated with the LHW. It is proposed that the sheet spatial distribution is mainly dominated by the Ekman convergence with Beaufort Gyre, slightly modulated with the AW Intrusion.

Unsupervised clustering identifies thermohaline staircases in the Canada Basin of the Arctic Ocean

“…In Section 3, we introduce HDBSCAN and describe how we use it to identify staircases and how we choose the input parameters. Then, in Section 4, we apply this method to data from two different ITP experiments, focusing on results that reproduce those from the studies of Timmermans et al (2008) and Lu et al (2022), hereafter denoted as T08 and L22, respectively. Finally, in Section 5, we use the comparison between our results and those of T08 and L22 to evaluate the performance of the clustering algorithm and give recommendations on when and how it can be best used to identify staircases.…”

“…These structures have been observed throughout the world’s oceans (van der Boog et al, 2021a) as well as in saline lakes (Newman, 1976), while related double-diffusive staircases are thought to occur in gas giants such as Jupiter and Saturn (André et al, 2017; Pontin et al, 2021). In particular, they are well known to occur in the Arctic Ocean (Timmermans et al, 2008; Lu et al, 2022; Ménesguen et al, 2022). Around 250 to 800 meters below the surface of the Arctic Ocean, there is a layer of water originating from the Atlantic, which is warmer and saltier than the topmost layer that is in contact with the sea ice above (Timmermans et al, 2008).…”

“…Then, from the subset of data that meets that condition, only mixed layers with a maximum variation of density anomaly of or less, and whose neighboring interfaces had thicknesses of 30 dbar or less, were considered. Lu et al (2022) defined the intersections between layers and interfaces as locations where the difference in the potential temperature gradient between two neighboring points is greater than , then disregarded the points within the interfaces. After performing this detection, Lu et al (2022) then made cross-profile connections with the layer points that remained, following the work of Padman and Dillon (1988), and used valleys in histograms of salinity to guide their choice of boundaries between layers, some of which were manually adjusted.…”

“…Lu et al (2022) defined the intersections between layers and interfaces as locations where the difference in the potential temperature gradient between two neighboring points is greater than , then disregarded the points within the interfaces. After performing this detection, Lu et al (2022) then made cross-profile connections with the layer points that remained, following the work of Padman and Dillon (1988), and used valleys in histograms of salinity to guide their choice of boundaries between layers, some of which were manually adjusted. We find it useful to distinguish between these two steps: the detection of points within layers for each profile and the connection of these points to those in other profiles that are within the same layer.…”

“…This approach requires sufficient knowledge of the staircases properties to select appropriate thresholds, as well as data with sufficient vertical resolution to accurately estimate the gradients. The connection algorithm Lu et al (2022) describe, after its completion, still requires some amount of manual intervention to produce a final dataset of staircase layers. A commonality of all of these approaches is that they are purpose-built for the task of detecting staircases in specific contexts.…”