We examine the geostrophic pathways of Pacific and Atlantic Waters along isopycnal surfaces of the Arctic Ocean from Montgomery potentials using two observation-based climatologies: the World Ocean Atlas (WOA) and the Monthly Isopycnal & Mixed-layer Ocean Climatology (MIMOC). The decadal mean circulation (2005-2017) shows an anticyclonic circulation for both the summer Pacific Waters (sPW) and winter Pacific Waters (wPW, although somewhat weaker) in the Canada Basin with subduction along the Chukchi plateau for both water masses. On the wPW layer, the diagnostics also highlight a year-long persistent flow from Herald Canyon on Chukchi Shelf to the Canadian Arctic Archipelago around the Canada Basin. Outflows are found at Nares and Fram Sraits at the density of sPW and wPW. These outflowing waters, previously reported as Pacific Waters, are found here to originate from the Eurasian shelf. Deeper, the Atlantic Waters (AW) flow anticyclonically around Northwind Ridge and cyclonically along the Alaskan shelfbreak forming a boundary current. The subduction of warm water along the Alaskan and Chukchi shelves is found to occur in summer on the sPW isopycnal layer, thus contributing to transferring heat into the center of the gyre. Yet, an advection of virtual Lagrangian particles from Pt. Barrow suggests that processes operating on timescales shorter than a month or at interannual time scales are the main contributors to the heat build up observed over the past two decades. This analysis supports the hypothesis that high frequency processes are key in shaping the subsurface heat reservoir of the Canada Basin.

Modelled geospatial Lagrangian trajectories are widely used in Earth Science, including in oceanography, atmospheric science and marine biology. The typically large size of these dataset makes them arduous to analyze, and their underlying pathways challenging to identify. Here, we show that a Machine Learning unsupervised k-means++ clustering method can successfully identify the pathways of the Labrador Current from a large set of modelled Lagrangian trajectories. The presented method requires simple pre-processing of the data, including a Cartesian correction on longitudes and a PCA reduction. The clustering is performed in a kernalized space and uses a larger number of clusters than the number of expected pathways. During post-processing, similar clusters are grouped into pathway categories by experts in the circulation of the region of interest. We find that the Labrador Current mainly follows a westward-flowing and an eastward retroflecting pathway (20% and 50% of the flow, respectively) that compensate each other through time in a see-saw behaviour. These pathways experience a strong variability of up to 96\%. We find that two thirds of the retroflection occurs at the tip of the Grand Banks, and one quarter at Flemish Cap. The westward pathway is mostly fed by the on-shelf branch of the Labrador Current, and the eastward pathway by the shelf-break branch. Pathways of secondary importance feed the Labrador Sea, the Gulf of St. Lawrence through the Belle Isle Strait, and the subtropics across the Gulf Stream.