Last updated: April 4,  2014

Pathways and transformation of the Denmark Strait Overflow Water in the Irminger Basin

Inga Koszalka


Thomas Haine


Earth & Planetary Sciences


Johns Hopkins University

Marcello Magaldi





The study was presented at the 19th Conference on Atmospheric and Oceanic Fluid Dynamics in Newport, Rhode Island, 20 June 2013, link, poster.

The manuscript on fates and travel times of the Denmark Strait Overflow is now published! I. Koszalka, T. W. N. Haine and M. G. Magaldi: Fates and travel times of Denmark Strait Overflow Water in the Irminger Basin, J. Phys. Oceanogr., 43, 2611--2628: doi, pdf

Watch Lagrangian particles released in the Denmark Strait Overflow: MOVIES.

Fig.1. The regional view of the study area with the location of the Denmark Strait Overflow marked with a star.


The Denmark Strait Overflow (DSO) is composed of dense water masses with various origin (Atlantic Water transformed in the Arctic and in the Nordic Seas, convection in Greenland and Iceland Seas). DSO leaves the Arctic through the Denmark Strait and subsequently transits southward along the continental shelf break in the Irminger Basin. DSO supplies one third of the North Atlantic Deep Water and is a key component of the global thermohaline circulation (Dickson and Brown, 1994). So far, the knowledge of the pathways of DSO through the Irminger Basin and its transformation there has been derived solely from sparse Eulerian (fixed-point) observations: repeated ship-borne hydrographic measurements and moored current meters at few locations at the Denmark Strait sill and along the route of DSO. The picture emerging from this data is shown below (e.g., Rudels et. al., 2002):

Fig.2. . Schematic pathways of the Denmark Strait Overflow Water (DSOW, plotted in magenta) from Rudels et. al. (2002). The volume transports of DSOW are given in Sverdrups (1 Sv = 10^6 m3/sek). The Irminger Current (IC) is marked with yellow.

There are many open questions considering the DSO however. For example, one would like to know what are the transit times for DSO to reach the North Atlantic and where along its route the largest transformation (change in temperature, salinity, density of the water mass) takes place. Moreover, the dense waters have also been repeatedly observed in various locations on the East Greenland Shelf, close to the coast. Where do these waters come from and what is their fate can only be speculated based on hydrographic data alone.

The Lagrangian (current-following) framework is ideal for addressing these questions. In the ocean, the Lagrangian observations of deep water flows are usually collected by acoustically-tracked subsurface floats (for example, RAFOS floats, see RAFOS Float Group). But no Lagrangian observations of DSO exist.

To fill this gap, we use a high resolution ocean model and deploy over 10,000 Lagrangian particles in DSO at the Denmark Strait to study its pathways and transformation in the Irminger Basin.

High resolution circulation model of the Irminger Basin

We use the MIT Global Circulation Model (MITgcm) configured for the Denmark Strait/Irminger Basin area. The configuration features:

-- Partial bottom cells and rescaled coordinate (for a better representation of flows in a complex bathymetry)

-- Highest spatial resolution to date of the Irminger Basin: dx~2km, dz=2-15m (210 layers), dt=30s

-- Simulation period: summer 2003 (7/1 - 9/1)

-- 3rd tracer advection with implicit diffusion; KPP for vertical diffusion, Leith horizontal viscosity

-- Forcing by NCEP fluxes & SeaWinds

-- We verified that the model transports and hydrography agree with the observations

For more details on the model configuration, see the published manuscript:doi.

Dense water distribution in the model

First, we looked into the distribution of dense waters in the model domain (Fig. 3). The "dense water" is water with a potential density sigma>=27.8, sigma=rho-1000 kg/m3 (Dickson and Brown, 1994).
Fig.3. Left: Occurrence frequency of model dense waters in the Denmark Strait/East Greenland Shelf (EGS)/Irminger Basin (IB). Superimposed are time- and depth-averaged dense water current vectors. The hydrographic sections are: Denmark Strait Extended (DSE), DS Sill, Spill Jet (SJ) and the Angmagssalik array (ANGM). The Dohrn Bank (DB) and the Kangerdlugssuaq Trough (KT) are marked. Right: A snapshot of the depth-averaged density in the dense water layer on 2 August 2003.

Note on fig.3. that the dense waters spread beyond the Denmark Strait sill. They reside on the adjacent shelf and in the Kangerdlugssuaq Trough. The time-averaged dense currents on the shelf are complex; the most prominent are the anticyclonic recirculation on the Dohrn Bank, and the cyclonic flow in the Kangerdlugssuaq Trough. The snapshot in the right panel of fig. 3 shows a "bolus" of dense water cascading off the Denmark Sill. The spilling of DSO off the sill is a hydraulically-controlled process, but with a pronounced variability. The dense water cascades the sill in form of boluses at 2-5 day intervals, seen in both, models and observations (Kase et. al., 2003, Magaldi et. al., 2011). Note also that the snapshot captured a spilling of dense waters from the shelf to the basin downstream off the sill, just ahead off the dense water bolus.

In order to study the pathways and transformation of the dense waters that reside on the shelf, we extend the standard hydrographic section of the DS Sill so that it crosses the entire Denmark Strait (DSE) and deploy numerical particles in the dense waters along this section (Fig. 4).

Fig.4. Left: Occurrence frequency of model dense waters, same as in Fig. 3. Right: A snapshot density section along DSE with marked particle deployments.

To capture the variability due to dense water boluses, the particles were deployed in waters of potential density sigma>=27.8 kg/m3 in 10 sets separated by 12h, starting 7/1. The particles were released on a regular grid (2km x 25m). They are classified in 3 groups: SILL (3301 particles released at the sill part of the section), SHELF (1843 particles released on the shelf adjacent to the sill) and KANGER (6669 particles released in the Kangerdlugssuaq Trough). These deployment groups are assigned color labels: dark blue, cyan blue and red, respectively.

Particle code

We integrate the Lagrangian particles off-line with a code written in MATLAB. At each time step, the simulated velocity fields (u,v,w components) are linearly interpolated on current particle locations. The particles are then advanced with MATLAB ode23t routine (2nd order predictor, 3rd order corrector). The time step is 15 minutes, chosen after a sensitivity study. For the boundary conditions, the code identifies events of particles hitting the boundary (inevitable since the time step for the Lagrangian code is larger than for the model simulation) and resumes the time stepping with a particle put on the boundary with the velocity component normal to the wall equal zero.

Lagrangian particles

The snapshots of particle positions are shown in Fig.5. The Animations of Lagrangian particles released in the Denmark Strait Overflow are available here: MOVIES.

Fig.5. Snapshots of particle distributions on 1, 8 and 33 day of the simulation. The particles are color-coded after the deployment group: SILL, SHELF and KANGER (dark blue, cyan blue and red, respectively).

The pathways of the dense water particles are shown in Figure 6.

Fig.6. The distributions of dense particle occurrences showing the pathways of the Denmark Strait Overflow originating at different locations along the DSE.

Note that the dense waters have indeed complex pathways on the shelf (compare with Fig. 3). There is a prominent influence of the anticyclonic recirculation on the Dohrn Bank governing the spreading of the SHELF particles, and the cyclonic flow in the Kangerdlugssuaq Trough redistributing particles widely over the East Greenland Shelf, and even carrying them close to the coast and into the fjords. Note that particles from SHELF and KANGER sets spill off the shelf break near the Dohrn Bank and Spill Jet sections. We find that these shelf particles contribute with 25% to the pool of dense particles recorded at the Angmagssalik section during 2 month-long simulation.

Transit times of the DSO through the Irminger Basin

The particle transit time distributions, constructued by counting the individual arrival times for particles from the deployment site at DSE to the Spill Jet and Angmagssalik sections, are shown in Figure 7.
Fig.7. Left: Particle transit times distributions from Denmark Strait to the Spill Jet section. Right: Particle transit times distributions from Denmark Strait to the Angmagssalik section.

The modal transit times for SILL particles to these sections are 5-6 days and 2-3 weeks, respectively. The SHELF particles recirculate on the Dohrn Bank before spilling over the shelf break and joining the sill overflow; their modal transit is longer by about a week. The KANGER particles recirculate in the Kangerdlugssuaq Trough for several weeks before spilling, their transit time distributions are broad. For a wider discussion of the transit time distributions, see the published manuscript: doi.

Ensemble-mean density evolution and comparison with the data

When the dense water particles travel through the Irminger Basin, they mix with the ambient waters so that their water mass properties (temperature, salinity and density), change. The evolution of the ensemble-mean particle density is shown in Figure 8.

Fig.8. Evolution of the mean particle potential density with distance from the DSE for the three deployments groups (SILL, SHELF and KANGER) as well as for dense particles recorded over a seabed deeper than 600m (SLOPE). Superimposed with circles are the observations from Girton & Sanford (2003) that should be compared to SLOPE particles.

In Figure 8 we also compare the particle observations with the hydrographic data. To this end, we average the densities on only those particle positions that are over the sea bed deeper than 600m (i.e., excluding the particle observations on the shelf) to match the locations of hydrograhic observations had been taken (Girton and Sanford, 2003). The mean density on the SLOPE particle shows good agreement with observations by Girton and Sanford, 2003. For more comparison with the hydrographic data, see the published manuscript: doi.

Water property transformation

Fig.9. Left: Particle positions corresponding to the density tendency (d(sigma)/dt) of less than -0.025, -0.05 and -0.1 kg/m3/day. Right: The histograms of the salinity tendencies (dS/dt) corresponding to the strong density transformation locations on the shelf and on the slope (shown in the left panel).

We map water mass transformation locations by plotting particle positions corresponding to high rates of the density drop on trajectories (Figure 9). This occurs where the dense waters cascade off the sill, and where dense waters from the shelf spill into the Irminger Basin between the Dohrn Bank and the Spill Jet section. From the distributions of the salinity tendency corresponding to these high transformation locations we infer that the transformation along the shelf break is due to mixing with warm and salty waters of Atlantic origin (the Irminger Current). The transformation locations on the shelf, on the other hand, correspond to freshening events suggests mixing with Polar Waters.

The transformation makes the densities of some of the particles lower than the threshold value of 1027.8 kg/m3, and hence they become classified as intermediate, rather than dense, waters. To further quantify this transformation, in Figure 10 we show the distributions of particle densities recorded at the Spill Jet section.

Fig.10. The average potential densities on particles at the Spill Jet section, calculated by binning all particle observations at the section during the simulation period. The inset shows the dominant contributions from the deployment sets to the bins with the potential densities sigma >=27.8 kg/m3.

Note a significant transformation to intermediate waters (gray bins in the inset). Note also that the SHELF particles, rather than those deployed at the Denmark Strait sill, dominate in bins with densest and deepest waters. The SILL particles are in close contact with warm and salty Atlantic Waters carried by Irminger Current and by this, are subject to strong mixing with the Atlantic Water at the sill and when they spill off it at Dohrn Bank toward the Irminger Basin. The SHELF particles avoid a prolonged contact with the Irminger Current and therefore have larger chance to keep high densities than the SILL particles. Finally, note the bins with dense KANGER particles are a signature of the repeated spilling process at the Spill Jet section.

Dense water pathways revealed by the Lagrangian study

A schematic summary of the results is shown in Figure 11.

Fig.11. The DSO pathways emerging from trajectories of the Lagrangian particles. The distribution of dense particles in the domain is shown in gray. At the Denmark Strait, dense water is found in the sill (blue), on the adjacent shelf (cyan), and in the Kangerdlugssuaq Trough (KT; red). Over 60 days these different water masses spread according to the arrows. There is cyclonic recirculation in the KT and anticyclonic recirculation on the Dohrn Bank (DB). Some of this recirculating water spills over the continental shelf break and join the overflow from the sill in an along-slope flow. The modal transit times to the Spill Jet and Angmagssalik sections are shown.


-- Pathways of dense waters on the shelf are mapped providing a context for hydrographic observations. They reveal importance of the anticyclonic recirculation on the Dohrn Bank (DB) and the cyclonic flow in the Kangerdlugssuaq Trough (KT).

-- The strongest transformation occurs: 1) upon the DSO cascading the sill and 2) at dense water spilling locations along the shelf break. The transformation in these locations is due to mixing with Atlantic-origin waters of the Irminger Current. Following the transformation, the DSO particles exhibit a wide spectrum of densities in the Irminger Basin.

-- Mixing with Polar Surface Waters causes the transformation on the shelf and may be a mechanism to transmit the fresh water variability to the Atlantic.

-- The modal travel times for the sill-DSO are 5-6 days to the Spill Jet section and 2-3 weeks to Angmagssalik. Particles seeded and recirculating on the shelf arrive one week later, and the dense waters from KT spill near Spill Jet section at approximately 25-day intervals.

-- The particle releases on the continental shelf (SHELF, KANGER) contribute to the pool of dense particles at Angmagssalik with 25%. The particles released on the shelf are separated from mixing with Irminger Current, thus preserving their densities and contributing to the densest water downstream at the Spill Jet section.

For a more exhaustive discussion, see the published manuscript: doi.

Ongoing work

-- Assessment of mixing rates in the Irminger Basin from the Lagrangian trajectories

-- Backward deployments at the Spill Jet section to study the origin of the Spill Jet and to estimate the entrainment rates

-- Observational study to corroborate the model results (pathways on the shelf)


R. R. Dickson and J. Brown (1994), J. Geophys. Res, 99 (C6), 12,319-12,341.

A. Falina et. al. (2012), J. Phys. Oceanography, 42, 2254-2267.

J. B. Girton and T. B. Sanford (2003), J. Phys. Oceanography, 33, 1351-1364.

R. H. Kase, J. B. Girton and T. B. Sanford (2003), J. Geophys. Res., 108, 3181.

M. G. Magaldi, T. W. H. Haine and R. S. Pickart (2011), J. Phys. Oceanography, 41, 2307-2327.

B. Rudels et. al. (2002), ICES J. Mar. Sci., 219, 319-325.

More information

The manuscript containing these and more results, published in J. Phys. Oceanography: I. Koszalka, T. W. N. Haine and M. G. Magaldi: Fates and travel times of Denmark Strait Overflow Water in the Irminger Basin, J. Phys. Oceanogr., 43, 2611–2628: doi, pdf


Inga Koszalka, Johns Hopkins University:,

Thomas Haine, Johns Hopkins University:,



This work was supported in part by NSF grants OCI-0904640 (Petascale Arctic Atlantic Antarctic Virtual Experiment), OCI-108849, OCE-0726640 and OCI-0904338. Data-intensive computations have been performed on the Johns Hopkins Data-Scope funded by OCI-1040114.