• Bakalian, F., S. Hameed, and R. Pickart, 2007: Influence of the Icelandic low latitude on the frequency of Greenland tip jet events: Implications for Irminger Sea convection. J. Geophys. Res., 112, C04020, doi:10.1029/2006JC003807.

    • Search Google Scholar
    • Export Citation
  • Barnier, B., and Coauthors, 2007: Eddy permitting ocean circulation hindcasts of past decades. CLIVAR Exchanges, No. 42, International CLIVAR Project Office, Southampton, United Kingdom, 8–10.

  • de Jong, M. F., H. M. van Aken, K. Våge, and R. S. Pickart, 2012: Convective mixing in the central Irminger Sea. Deep-Sea Res. I, 63, 3651, doi:10.1016/j.dsr.2012.01.003.

    • Search Google Scholar
    • Export Citation
  • Doyle, J. D., and M. A. Shapiro, 1999: Flow response to large-scale topography: The Greenland tip jet. Tellus, 51A, 728748, doi:10.1034/j.1600-0870.1996.00014.x.

    • Search Google Scholar
    • Export Citation
  • Dupont, F., and Coauthors, 2015: A high-resolution ocean and sea-ice modelling system for the Arctic and North Atlantic Oceans. Geosci. Model Dev., 8, 152, doi:10.5194/gmdd-8-1-2015.

    • Search Google Scholar
    • Export Citation
  • DuVivier, A. K., and J. J., Cassano, 2015: Evaluation of WRF Model resolution on simulated mesoscale winds and surface fluxes near Greenland. Mon. Wea. Rev., 141, 941963, doi:10.1175/MWR-D-12-00091.1.

    • Search Google Scholar
    • Export Citation
  • Falina, A., A. Sarafanov, and A. Sokov, 2007: Variability and renewal of Labrador Sea Water in the Irminger basin in 1991–2004. J. Geophys. Res., 112, C01006, doi:10.1029/2005JC003348.

    • Search Google Scholar
    • Export Citation
  • Fan, X., U. Send, P. Testor, J. Karstensen, and P. L’Herminier, 2013: Observations of Irminger Sea anticyclonic eddies. J. Phys. Oceanogr., 43, 805823, doi:10.1175/JPO-D-11-0155.1.

    • Search Google Scholar
    • Export Citation
  • Fichefet, T., and M. A. M., Maqueda, 1997: Sensitivity of a global sea ice model to the treatment of ice thermodynamics and dynamics. J. Geophys. Res., 102, 1260912646, doi:.10.1029/97JC00480.

    • Search Google Scholar
    • Export Citation
  • Harden, B. E., and I. A. Renfrew, 2012: On the spatial distribution of high winds off southeast Greenland. Geophys. Res. Lett., 39, L14806, doi:10.1029/2012GL052245.

    • Search Google Scholar
    • Export Citation
  • Harden, B. E., I. A. Renfrew, and G. N. Petersen, 2011: A climatology of wintertime barrier winds off southeast Greenland. J. Climate, 24, 47014717, doi:10.1175/2011JCLI4113.1.

    • Search Google Scholar
    • Export Citation
  • Hunke, E. C., 2001: Viscous–plastic sea ice dynamics with the EVP model: Linearization issues. J. Comput. Phys., 170, 1838, doi:10.1006/jcph.2001.6710.

    • Search Google Scholar
    • Export Citation
  • Hunke, E. C., and W. H. Lipscomb, 2010: CICE: The Los Alamos Sea Ice Model documentation and software user’s manual, version 4.1. Los Alamos National Laboratory Tech. Rep. LA-CC-06-012, 76 pp. [Available online at http://csdms.colorado.edu/w/images/CICE_documentation_and_software_user's_manual.pdf.]

  • Lavender, K. L., R. E. Davis, and W. B. Owens, 2000: Mid-depth recirculation observed in the interior Labrador and Irminger Seas by direct velocity measurements. Nature, 407, 6669, doi:10.1038/35024048.

    • Search Google Scholar
    • Export Citation
  • Lipscomb, W. H., E. C. Hunke, W. Maslowski, and J. Jakacki, 2007: Ridging, strength, and stability in high-resolution sea ice models. J. Geophys. Res., 112, C03S91, doi:10.1029/2005JC003355.

    • Search Google Scholar
    • Export Citation
  • Madec, G., P. Delecluse, M. Imbard, and C. Lévy, 1998: OPA 8.1 ocean general circulation model reference manual. Institut Pierre-Simon Laplace Notes du Pôle de Modélisation 11, 91 pp.

  • Marshall, J., and F. Schott, 1999: Open-ocean convection: Observations, theory and models. Rev. Geophys., 37, 164, doi:10.1029/98RG02739.

    • Search Google Scholar
    • Export Citation
  • Moore, G. W. K., and I. A. Renfrew, 2005: Tip jets and barrier winds: A QuikSCAT climatology of high wind speed events around Greenland. J. Climate, 18, 37133725, doi:10.1175/JCLI3455.1.

    • Search Google Scholar
    • Export Citation
  • Myers, P. G., and C. Donnelly, 2008: Water mass transformation and formation in the Labrador Sea. J. Climate, 21, 16221638, doi:10.1175/2007JCLI1722.1.

    • Search Google Scholar
    • Export Citation
  • Oltmanns, M., F. Straneo, G. W. K. Moore, and S. H. Mernild, 2014: Strong downslope wind events in Ammassalik, Southeast Greenland. J. Climate, 27, 977993, doi:10.1175/JCLI-D-13-00067.1.

    • Search Google Scholar
    • Export Citation
  • Outten, S. D., I. A. Renfrew, and G. N. Petersen, 2009: An easterly tip jet off Cape Farewell, Greenland. II: Simulations and dynamics. Quart. J. Roy. Meteor. Soc., 135, 19341949, doi:10.1002/qj.531.

    • Search Google Scholar
    • Export Citation
  • Pickart, R. S., M. A. Spall, M. H. Ribergaard, G. W. K. Moore, and R. F. Milliff, 2003a: Deep convection in the Irminger Sea forced by the Greenland tip jet. Nature, 424, 152156, doi:10.1038/nature01729.

    • Search Google Scholar
    • Export Citation
  • Pickart, R. S., F. Straneo, and G. W. K. Moore, 2003b: Is Labrador Sea Water formed in the Irminger basin? Deep-Sea Res. I, 50, 2352, doi:10.1016/S0967-0637(02)00134-6.

    • Search Google Scholar
    • Export Citation
  • Renfrew, I. A., S. D. Outten, and G. W. K. Moore, 2009: An easterly tip jet off Cape Farewell, Greenland. I: Aircraft observations. Quart. J. Roy. Meteor. Soc., 135, 19191933, doi:10.1002/qj.513.

    • Search Google Scholar
    • Export Citation
  • Sampe, T., and S. P. Xie, 2007: Mapping high sea winds from space: A global climatology. Bull. Amer. Meteor. Soc., 88, 19651978, doi:10.1175/BAMS-88-12-1965.

    • Search Google Scholar
    • Export Citation
  • Smith, G. C., F. Roy, P. Mann, F. Dupont, B. Brasnett, J.-F. Lemieux, S. Laroche, and S. Bélair, 2014: A new atmospheric dataset for forcing ice–ocean models: Evaluation of reforecasts using the Canadian global deterministic prediction system. Quart. J. Roy. Meteor. Soc., 140, 881894, doi:10.1002/qj.2194.

    • Search Google Scholar
    • Export Citation
  • Speer, K., and E. Tziperman, 1992: Rates of water mass formation in the North Atlantic Ocean. J. Phys. Oceanogr., 22, 93104, doi:10.1175/1520-0485(1992)022<0093:ROWMFI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Speer, K., H.-J. Isemer, and A. Biastoch, 1995: Water mass formation from revised COADS data. J. Phys. Oceanogr., 25, 24442457, doi:10.1175/1520-0485(1995)025<2444:WMFFRC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Spronson, D. A. J., I. A. Renfrew, and K. J. Heywood, 2008: Atmospheric conditions associated with oceanic convection in the south-east Labrador Sea. Geophys. Res. Lett., 35, L06601, doi:10.1029/2007GL032971.

    • Search Google Scholar
    • Export Citation
  • Stackhouse, P. W., Jr., S. J. Cox, S. K. Gupta, M. Chiacchio, and J. C. Mikovitz, 2001: The WCRP/GEWEX surface radiation budget project release 2: An assessment of surface fluxes at 1 degree resolution. IRS 2000: Current Problems in Atmospheric Radiation, W. L. Smith and Y. Timofeyev, Eds., A. Deepak, 485–488.

  • Treguier, A. M., S. Theetten, E. P. Chassignet, T. Penduff, R. Smith, L. Talley, J. O. Beismann, and C. Böning, 2005: The North Atlantic subpolar gyre in four high-resolution models. J. Phys. Oceanogr., 35, 757774, doi:10.1175/JPO2720.1.

    • Search Google Scholar
    • Export Citation
  • Våge, K., R. S. Pickart, G. W. K. Moore, and M. H. Ribergaard, 2008: Winter mixed layer development in the central Irminger Sea: The effect of strong, intermittent wind events. J. Phys. Oceanogr., 38, 541565, doi:10.1175/2007JPO3678.1.

    • Search Google Scholar
    • Export Citation
  • Våge, K., and Coauthors, 2009a: Surprising return of deep convection to the subpolar North Atlantic Ocean in winter 2007–2008. Nat. Geosci., 2, 6775, doi:10.1038/ngeo382.

    • Search Google Scholar
    • Export Citation
  • Våge, K., T. Spengler, H. C. Davies, and R. S. Pickart, 2009b: Multi-event analysis of the westerly Greenland tip jet based upon 45 winters in ERA-40. Quart. J. Roy. Meteor. Soc., 135, 19992011, doi:10.1002/qj.488.

    • Search Google Scholar
    • Export Citation
  • Våge, K., and Coauthors, 2011: The Irminger Gyre: Circulation, convection, and interannual variability. Deep-Sea Res. I, 58, 590614, doi:10.1016/j.dsr.2011.03.001.

    • Search Google Scholar
    • Export Citation
  • Walin, G., 1982: On the relation between sea-surface heat flow and thermal circulation in the ocean. Tellus, 34A, 187195, doi:10.1111/j.2153-3490.1982.tb01806.x.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Model bathymetry (m) over the Irminger Sea region. The black diamonds represent the location of the LOCO moorings. The LOCO2 mooring is located at 59°12′N, 39°30′W, and the LOCO3 mooring is at 59°12′N, 39°00′W. The three boxes represent areas used for averaging the total turbulent heat fluxes and winds. The dark gray line shows the section used for studying the modeled hydrography. The red (magenta) star represents the location of Cape Farewell (Tasiilaq).

  • View in gallery

    Averaged March mixed layer depth (m) from the CREG12 model: the top-left panel shows the multiyear average for 2004–10, while subsequent panels present individual years (with the associated wintertime period as subtitles). The magenta lines represent the average 50% sea ice cover contour.

  • View in gallery

    (a) Log10 of the 2-day averaged potential vorticity at the LOCO2 mooring site from the (top) model and (middle) McLane moored profilers. The observed data are binned to the model vertical levels for comparison. Black dots show the MLDs. The modeled MLD is defined as the depth where the density exceeds its surface values by 0.01 kg m−3. Model MLD is calculated at each time step and averaged over a 2-day period. The observed MLD is determined by visual inspection as described by de Jong et al. (2012). (bottom) The values of the MLD from the model (blue circles) and observations (red x symbols). (b) As in (a), but at the LOCO3 site.

  • View in gallery

    Time series of the mixed layer depth averaged over the area where March-averaged mixed layer depth deeper than 800 m has been simulated in any year (red contour on insert).

  • View in gallery

    Winter (December to March) total turbulent heat flux (latent + sensible): the top-left panel shows the multiyear average (W m−2) during 2004–10, while subsequent panels present yearly anomalies (W m−2) relative to the multiyear average. The color axis of the (left) multiyear averages and the (right) anomaly are shown in the bottom right. The magenta lines represent the average 50% sea ice cover contour.

  • View in gallery

    Winter (December to March) 2-m air temperature (colors, °C) and winds (vectors, m s−1): the top-left panel shows the multiyear average (°C) during 2004–10. Subsequent panels present winter averaged winds and the 2-m air temperature anomalies (°C) relative to the multiyear average. The color axis of the (left) multiyear average and the (right) anomaly are shown in the bottom right. The magenta lines represent the average 50% sea ice cover contour.

  • View in gallery

    (left) Scatterplots of 2-day averaged wind components and total turbulent heat fluxes (W m−2, blue shading). Composites of (center) surface wind speed (m s−1) and (right) total turbulent heat flux (W m−2). These composites are created by averaging winds and heat fluxes for events with total heat fluxes > 300 W m−2. The top, middle and bottom panels are for regions 1, 2, and 3 presented as black boxes in the center and right panels. The composites for the three regions use 29, 28, and 25 events, respectively.

  • View in gallery

    October model potential density section across the deep convection area from Greenland to the Reykjanes Ridge (approximately at 61°N; Fig. 1) for 2006, 2007, 2008, and 2009. The magenta triangles show the beginning and end of the maximum extent of the MLD deeper than 800 m in the winter of 2008–09. Density contours (fine black lines) are at every 0.02 kg m−3.

  • View in gallery

    Modeled winter (December to March) pressure anomaly at 643 m (hPa, colors) and March averaged mixed layer depth contours of 700 (black) and 1000 m (white).

  • View in gallery

    (left) Modeled current vectors (m s−1) and (right) mean EKE (m2 s) at 320-m depth. The area of MLD greater than 800 m is delimited by the red contour in both panels.

  • View in gallery

    Annual-mean surfaced-forced water mass transformation per density class (Δσ = 0.05 kg m−3) over the area of convection deeper than 800 m (see insert on Fig. 4). Positive (negative) values represent an input (removal) of buoyancy.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 96 53 4
PDF Downloads 62 39 1

Modelled Variations of Deep Convection in the Irminger Sea during 2003–10

View More View Less
  • 1 Department of Oceanography, Dalhousie University, Halifax, Nova Scotia, Canada
  • | 2 Bedford Institute of Oceanography, Dartmouth, Nova Scotia, Canada
  • | 3 Meteorological Service of Canada, Dorval, Québec, Canada
  • | 4 Mercator-Océan, Ramonville-Saint-Agne, France
© Get Permissions
Full access

Abstract

Results from a high-resolution ice–ocean model are analyzed to understand the physical processes responsible for the interannual variability of ocean convection over the Irminger Sea. The modeled convection in the open Irminger Sea for the winters of 2007/08 and 2008/09 is in good agreement with observations. Deep convection is caused by strong atmospheric forcing that increases the ocean heat loss through latent and sensible heat fluxes. Greenland tip jets are found to be the only strong wind events that directly affect the deep convection area and explain up to 53% of the total turbulent heat loss during active convection years. Deep convection is modeled where there is favorable preconditioning of the water column due to isopycnal doming inside the semienclosed Irminger Gyre. The region of deep convection is also characterized by weak eddy kinetic energy. Finally, an estimation of the surface-forced water mass transformation confirms the Irminger Sea as a region of intermittent production of Labrador Sea Water, with annual averages between 0.9 and 1.9 Sverdrups (Sv; 1 Sv ≡ 106 m3 s−1) of water denser than 27.7 kg m−3 for years of active convection.

Corresponding author address: Jean-Philippe Paquin, Department of Oceanography, Dalhousie University, 1355 Oxford Street, P.O. Box 15000, Halifax, NS B3H 4R2, Canada. E-mail: paquin.jeanphilippe@gmail.com

Abstract

Results from a high-resolution ice–ocean model are analyzed to understand the physical processes responsible for the interannual variability of ocean convection over the Irminger Sea. The modeled convection in the open Irminger Sea for the winters of 2007/08 and 2008/09 is in good agreement with observations. Deep convection is caused by strong atmospheric forcing that increases the ocean heat loss through latent and sensible heat fluxes. Greenland tip jets are found to be the only strong wind events that directly affect the deep convection area and explain up to 53% of the total turbulent heat loss during active convection years. Deep convection is modeled where there is favorable preconditioning of the water column due to isopycnal doming inside the semienclosed Irminger Gyre. The region of deep convection is also characterized by weak eddy kinetic energy. Finally, an estimation of the surface-forced water mass transformation confirms the Irminger Sea as a region of intermittent production of Labrador Sea Water, with annual averages between 0.9 and 1.9 Sverdrups (Sv; 1 Sv ≡ 106 m3 s−1) of water denser than 27.7 kg m−3 for years of active convection.

Corresponding author address: Jean-Philippe Paquin, Department of Oceanography, Dalhousie University, 1355 Oxford Street, P.O. Box 15000, Halifax, NS B3H 4R2, Canada. E-mail: paquin.jeanphilippe@gmail.com

1. Introduction

Deep convective mixing in the North Atlantic Ocean plays a key role in water mass exchange between the surface and the deep ocean and is thus a major contributor to the meridional overturning circulation (MOC). In the North Atlantic Ocean, the Greenland Sea, and the Labrador Sea are well known regions of deep convection. Both regions meet the following conditions that are favorable for deep convection to occur (Marshall and Schott 1999). First, the water column is weakly stratified. Second, a cyclonic circulation in the upper-layer domes the isopycnals toward the surface, bringing higher density waters upward in the water column. Third, strong atmospheric forcing causes extensive cooling of the ocean over the winter period to erode the surface stratification and the summertime thermocline.

Besides the Greenland Sea and the Labrador Sea, the southwestern Irminger Sea has been identified as an area with conditions favorable for deep convection (Pickart et al. 2003a,b). Recent observational evidence obtained from Argo floats (Våge et al. 2009a) and mooring data (de Jong et al. 2012) suggests that deep convection occurred intermittently within the western part of the Irminger Sea in the late 2000s. Våge et al. (2009a) reported convective mixing reaching depths below 1000 m in the Irminger Sea during the winter of 2007/08 and demonstrated the role of the strong atmospheric forcing occurring over the region for that particular winter. Additional evidence was also provided by de Jong et al. (2012) using records from moorings deployed by the Royal Netherlands Institute for Sea Research Long-Term Ocean Climate Observations program (NIOZ LOCO; www.nioz.nl/loco-en) at two different locations in the southwestern Irminger Sea from 2004 to 2010. Their records showed that convective mixing could reach depths below 800 m but with considerable interannual variability. Convective mixing was most active in the winter of 2007/08, in good agreement with Våge et al. (2009a), and also in the following winter of 2008/09 before becoming anomalously shallow in the winter of 2009/10.

In terms of contribution to the global MOC, Pickart et al. (2003a,b) suggested that deep convection in the Irminger Sea may be of secondary importance compared to that in the Labrador Sea and Nordic Seas. Based on previous estimates of the surface-forced water mass transformation, the Labrador Sea deep convection mainly contributes to the MOC at densities greater than 27.65 kg m−3 (Myers and Donnelly 2008). Over a limited area in the Labrador Sea, between 50° and 64°N and 45° and 60°W, Myers and Donnelly (2008) estimated the long-term average (1960–99) of the surface-forced water mass transformation (WMT) rate to be between 2.1 and 3.9 Sverdrups (Sv; 1 Sv ≡ 106 m3 s−1) for water with densities greater than 27.65 kg m−3, depending on the correction applied to the atmospheric turbulent heat flux data. They also showed a large interannual variability with a transformation rate up to 9.5–10.8 Sv some years and no production of dense waters in other years.

To understand the deep convection in the Irminger Sea and its interannual variability, the first factor to consider is the atmospheric forcing that can cause significant cooling at the ocean’s surface. The southwestern Irminger Sea is one of the windiest locations in the world’s ocean (Sampe and Xie 2007) mostly due to the interaction of low-level atmospheric synoptic systems with Greenland’s steep topography. The vast majority of the high wind speed events around Greenland are produced when a low pressure system transitions between Greenland and Iceland along the North Atlantic storm track (Moore and Renfrew 2005; Våge et al. 2008, 2009b; DuVivier and Cassano 2015). The high wind speed events in the vicinity of Greenland can be classified in four broad categories: tip jets, barrier winds, reverse tip jets, and localized downslope wind events.

Tip jets can be described as an eastward acceleration of the near-surface winds and are caused by a combination of conservation of the Bernoulli function during orographic descent and acceleration due to flow splitting as stable air passes around the southern tip of Greenland (Doyle and Shapiro 1999; Moore and Renfrew 2005). Wind speeds in excess of 25 m s−1 are present up to 15% of the time in winter (Moore and Renfrew 2005) with peak wind speeds over 30 m s−1 and turbulent heat fluxes over 600 W m−2 sustained for less than 1 day (Våge et al. 2008). Pickart et al. (2003a) showed that high wind speeds during tip jet events are associated with high turbulent heat fluxes and strong wind stress curl. A 2-day backward trajectory analysis showed the continental origin of the air parcels in the tip jets (see Fig. 7 in Våge et al. 2009b). Using an idealized regional ocean model forced by historical records of tip jet events, Pickart et al. (2003a) showed that these events can be sufficiently strong to trigger deep convection in the southwest Irminger Sea.

Barrier winds compose a high-speed flow along the southeast coast of Greenland. To first order, barrier winds are produced when the air is unable to ascend the topographic barrier, developing a pressure gradient perpendicular to the barrier and an associated geostrophic flow (Harden et al. 2011). Wind speeds greater than 20 m s−1 were shown to occur on a weekly frequency in winter (Moore and Renfrew 2005) and the spatial distribution of the maximum wind speed is primarily dictated by the position of promontories along the southeast coast of Greenland (Harden and Renfrew 2012). Despite the intensity and location of the barrier winds, their influence on deep convection in the Irminger Sea has not been investigated.

Reverse tip jets are characterized by strong flow from Cape Farewell (59°46'N, 43°55'W) to the Labrador Sea (Moore and Renfrew 2005; Renfrew et al. 2009; Outten et al. 2009). Reverse tip jets have been shown to play a secondary role in causing deep convection over the Labrador Sea where the main mechanism is the advection of cold air outbreaks from the coast of Labrador (Spronson et al. 2008). Therefore, reverse tip jets are not considered in this study.

Localized downslope wind events occur near Tasiilaq (65°36'N, 37°38'W). These short-lived events (approximately 1 day) cause strong winds oriented offshore that bring cold air down the ice sheet due to thermal and gravitational acceleration and flow convergence in the Ammassalik Valley (Oltmanns et al. 2014). One-fifth of the total wintertime ocean heat loss over the northern Irminger Sea can be explained by these rare events (approximately 1 event per month during winter) with peak values above 900 W m−2 for strong individual events (see Fig. 9 of Oltmanns et al. 2014). While these events all induce significant ocean heat loss, they may not all trigger deep convection in the Irminger Sea.

The second factor influencing the interannual variability of the deep convection is the preconditioning related to the ocean circulation. The cyclonic circulation of the Irminger Gyre causes doming of the isopycnals (Lavender et al. 2000; Pickart et al. 2003b; de Jong et al. 2012). The presence of the Irminger Gyre also tends to isolate the water column at its center, allowing the previous year’s convection to become an important factor in preconditioning the following winter. This circulation pattern determines the location of deep convection and which category of atmospheric forcing can trigger deep convection.

This study aims to identify the aspects of the atmospheric forcing and ocean circulation that may contribute to deep convection and its variability in the Irminger Sea. An estimate of the water mass transformation rate in the Irminger Sea, and hence its contribution to the MOC relative to that of the Labrador Sea, is also presented. The study is based on an analysis of a simulation using a high-resolution ocean model and atmospheric forcing data. After introducing the model and atmospheric data in section 2, the model-simulated interannual variability of deep convection is assessed in section 3. In section 4, the atmospheric data are analyzed to identify the impact of the different strong wintertime wind events on the turbulent heat fluxes. Section 5 assesses the hydrography and circulation in the Irminger Sea followed by an evaluation of the water mass transformation in section 6. Discussion and conclusions are presented in section 7.

2. High-resolution ocean model and atmospheric forcing

a. Model description

The ocean model is based on version 3.1 of the Nucleus for European Modeling of the Ocean (NEMO) with some additions from Mercator-Océan, the Met Office, and the DRAKKAR community (Dupont et al. 2015). The ocean engine of NEMO is the primitive equation model Océan Parallélisé (OPA; Madec et al. 1998) adapted to regional applications. OPA is coupled to the Los Alamos Sea Ice Model (CICE; Hunke 2001; Lipscomb et al. 2007; Hunke and Lipscomb 2010) version 4, which simulates the thermodynamic and dynamic processes of sea ice freeze/melt, advection, and deformation. Freshwater or brine exchanges between the ice and the ocean uses a virtual salt flux approach.

The ocean and sea ice domain use the ORCA tripolar grid (Barnier et al. 2007) at a nominal resolution of ⅝1/12° in latitude/longitude with the “northfold” discontinuity of the global grid being removed. The Canadian Operational Network of Coupled Environmental Prediction Systems (CONCEPTS) regional domain (CREG12) extends from 27°N to the Bering Strait, effectively covering the North Atlantic and Arctic Oceans. The horizontal grid size ranges from 8 km at the North Atlantic open boundary down to an average of 5 km over the Arctic Ocean and with high resolution around 2 km in the Canadian Arctic Archipelago. The model uses 50 vertical levels with spacing increasing from 1 m at the surface to 450 m at 5000 m. The high vertical resolution is mainly concentrated in the top layers, with 25 vertical levels located in the top 150 m. Partial steps of the z levels are used to increase the accuracy of the bathymetry.

The model is run from 1 January 2003 to 31 December 2010 without temperature or salinity restoring. The initial conditions and monthly lateral open boundary conditions for temperature, salinity, ocean currents, and sea surface height are taken from a global ORCA ⅝1/12° free run (T321) from Mercator-Océan. The ORCA ⅝1/12° free run is initialized on 1 January 1999 and also runs without restoring. The CREG12 ocean and sea ice initial conditions are extracted from the global ORCA ⅝1/12° run for 1 January 2003. Since the ORCA T321 run uses the Louvain-La-Neuve Sea Ice Model (LIM2; Fichefet and Maqueda 1997), the sea ice initial conditions are assigned to the corresponding CICE ice thickness category.

More detailed information about the model and evaluation of the simulation results with oceanic and sea ice observations are documented in Dupont et al. (2015).

b. Atmospheric forcing

The 3-hourly atmospheric forcing used for the simulation comes from the Canadian Meteorological Centre Global Deterministic Prediction System Reforecast (CGRF; Smith et al. 2014). The CGRF was produced through daily historical reforecasts with a high horizontal resolution of 0.45° longitude and 0.3° latitude (corresponding to approximately 33 km at 60°N). The CGRF surface temperature, humidity, and winds were shown to have equivalent biases to those found in the European reanalysis product ERA-Interim (Smith et al. 2014). The largest mean errors were found in the CGRF radiation fields. Significant errors in both shortwave and longwave radiation were found due to a combination of the initialization scheme used for clouds in the CGRF, insufficient cloud cover in some areas, and overly transparent clouds. To address that issue, a set of shortwave and longwave monthly bias correction fields have been produced based on the differences between CGRF and the Global Energy and Water Cycle Experiment Surface Radiation Budget (GEWEX SRB release 3.0; Stackhouse et al. 2001) over the period 2002–07 as a function of the hour of the day. Although the CGRF dataset is not a reanalysis and thus is expected to be less well constrained by available observations, its high spatial and temporal resolution permits a more detailed representation of atmospheric structures and topographic steering. This results in finer-scale coastal features and wind stress curl with improvement in the vicinity of Greenland (Smith et al. 2014). The maximum wind speed of individual tip jet events is still somewhat underestimated compared to QuikSCAT data, but they are better represented in CGRF compared to the ERA-Interim reanalysis.

3. Modeled variability of Irminger Sea deep convection

The bathymetry of the Irminger Sea region is shown in Fig. 1. Figure 2 shows the mixed layer depth (MLD) from the CREG12 model. The MLD is defined as the depth where the potential density exceeds its surface value by 0.01 kg m−3. MLD calculations are performed every time step (180 s) and averaged over the 2-day period of the model output. The multiyear average of the simulated March MLD shows a maximum over the Reykjanes Ridge, close to Iceland’s shelf break, and southwest of Cape Farewell (Fig. 2). In addition to these locations, a narrow region over the western Irminger Sea, oriented southwest to northeast just off the Greenland shelf break, shows multiyear average of MLD deeper than 500 m and large interannual variability. In March of 2008 and 2009, the MLD in this region is as deep as 800–1000 m. The modeled deep convection in these two winters is in good agreement with observations from Argo floats (Våge et al. 2009a) and the LOCO mooring sites (de Jong et al. 2012). In comparison, the period prior to 2008 shows relatively shallow MLDs, not deeper than 600 m. In March of 2010, the modeled MLD shows no deep convection over the western Irminger Sea and south of Cape Farewell.

Fig. 1.
Fig. 1.

Model bathymetry (m) over the Irminger Sea region. The black diamonds represent the location of the LOCO moorings. The LOCO2 mooring is located at 59°12′N, 39°30′W, and the LOCO3 mooring is at 59°12′N, 39°00′W. The three boxes represent areas used for averaging the total turbulent heat fluxes and winds. The dark gray line shows the section used for studying the modeled hydrography. The red (magenta) star represents the location of Cape Farewell (Tasiilaq).

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

Fig. 2.
Fig. 2.

Averaged March mixed layer depth (m) from the CREG12 model: the top-left panel shows the multiyear average for 2004–10, while subsequent panels present individual years (with the associated wintertime period as subtitles). The magenta lines represent the average 50% sea ice cover contour.

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

As part of the model evaluation, we compare the time–depth variations of hydrography from the model and observations at the NIOZ LOCO2 and LOCO3 mooring sites, described in de Jong et al. (2012).The LOCO2 and LOCO3 mooring sites are located close to the 3000-m isobath at 59°12′N, 39°30′W and at 59°15′N, 36°24′W in the Irminger Sea, respectively (Fig. 1). The LOCO2 data cover a 7-yr period between 2003 and 2010, while the LOCO3 data cover a 5-yr period from 2003 to 2007. Daily profiles of temperature and salinity are available at a vertical resolution of 1 db between ~150 and ~2400 db. Data recovery from the LOCO2 moored profiler achieves 80% of the expected number of values except during 2004/05 and 2009/10 when 53% and 62% of the expected values are obtained. The LOCO2 mooring dataset represents one of the best time series available for the region. The data recovery for the LOCO3 profiler achieves 71% of the expected values and does not cover the winter of 2006/07 due to an instrument failure. Nevertheless, the data cover the active winter of 2007/08 and are therefore useful for model evaluation. All data identified as “uncertain” during the quality control procedure have been removed. The LOCO data are averaged every 2 days to be consistent with the model output frequency.

Figure 3 compares the time evolution of planetary potential vorticity (PV) at the LOCO2 and LOCO3 mooring sites to an average of the nine closest model grid points. The planetary potential vorticity is calculated as
e1
where f is the Coriolis parameter, ρ is the density, and ∂ρ/∂z is the adiabatic vertical density gradient. The PV data are also binned vertically to the model levels for an easier visual comparison. The LOCO MLD data, provided by M. F. de Jong (2015, personal communication), are determined by visual inspection of all profiles using salinity, potential temperature, and potential density (de Jong et al. 2012). It is worth noting that the instruments do not cover the upper-ocean layer. The PV minimum propagates down through the water column as winter convection occurs and is still clearly visible even after restratification has taken place in the upper layers (de Jong et al. 2012).
Fig. 3.
Fig. 3.

(a) Log10 of the 2-day averaged potential vorticity at the LOCO2 mooring site from the (top) model and (middle) McLane moored profilers. The observed data are binned to the model vertical levels for comparison. Black dots show the MLDs. The modeled MLD is defined as the depth where the density exceeds its surface values by 0.01 kg m−3. Model MLD is calculated at each time step and averaged over a 2-day period. The observed MLD is determined by visual inspection as described by de Jong et al. (2012). (bottom) The values of the MLD from the model (blue circles) and observations (red x symbols). (b) As in (a), but at the LOCO3 site.

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

The LOCO2 observations show a distinct Labrador Sea Water (LSW)–type layer between 700 and 1300 m, indicated by a PV minimum, with stronger vertical stratification compared to the model. The underestimation of the vertical stratification of the LSW in the model may be caused by the coarser vertical resolution at these depths (model levels are spaced approximately 100 m apart between 600 and 1000 m). However, the model reproduces the seasonal cycle and the strong interannual variability in the upper 800 m of the water column. Both the observations and model show shallow winter mixed layers in the beginning of the period (2003 to 2007, except 2005) with the maximum MLD of about 400 m consistent with previous studies (de Jong et al. 2012; Våge et al. 2008). Winter 2008 shows deeper convection in both the model and the observations. The maximum MLD is about 800 m, occurring in late February and early March in the model and earlier (February) in the observations. Winter 2009 shows the deepest convection in the LOCO2 mooring data, while the model produces a maximum MLD similar to the previous winter. The signal of the minimum PV, at depths of 300–800 m, is also more persistent in the observations than in the model, corresponding to stronger restratification in the model. It should be noted that the LOCO2 mooring site is located south of the region of maximum MLD in the model, and therefore this comparison does not capture the maximum depth of the simulated convection. The model solution shows shallower MLD in early 2010, with weaker values of PV being present closer to the surface (400–600 m). Although the instrument failed during this period, de Jong et al. (2012) estimated the MLD to be around 400 m, even less than that obtained by the model. The shallow MLD for the winter of 2009/10 compared to the two previous winters is seen in both the model and observations.

The LOCO3 data show stronger stratification compared to the LOCO2 site for both the observations and model results (Fig. 3b). The model also reproduces the seasonal and interannual variability of the PV profiles observed in the top 700 m of the water column. The winter of 2007/08 shows deep MLDs (~1000 m) in the observations, while the modeled MLDs do not exceed 700 m. The observed MLDs also exhibit larger time variability with abrupt shoaling and deepening events, while the deepening in the modeled MLD over a winter season is more progressive.

Figure 4 shows a time series of the simulated MLD spatially averaged over the area in the Irminger Sea where convection deeper than 800 m has been modeled in any year (see insert on Fig. 4). The maximum MLD occurs mostly in February and March and is subject to significant interannual variability. The MLD signal shows significant variability of the 2-day averages. This might reflect the convection being a transient and intermittent phenomenon, as suggested by Marshall and Schott (1999), and highly sensitive to the atmospheric forcing. For example, in February of 2005, strong southerly winds due to an atmospheric anticyclonic circulation produced strong orographic precipitation (>30 mm day−1) around Cape Farewell. This event decreased the surface salinity over the Irminger Current but not over the ice-covered Greenland Shelf, and the increase in near-surface temperature caused by the northward advection of warm air resulted in a net gain at the ocean surface over the southwest Irminger Sea. These combined effects led to warming and freshening in the upper 100 m of the water column, disrupting the convection in the model solution.

Fig. 4.
Fig. 4.

Time series of the mixed layer depth averaged over the area where March-averaged mixed layer depth deeper than 800 m has been simulated in any year (red contour on insert).

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

The early deepening of the mixed layer during December and January for the winters of 2004/05, 2007/08, and 2008/09 is influenced directly by strong turbulent heat fluxes (see section 4). The winters of 2007/08 and 2008/09 show rapid deepening of the mixed layer and large variability in the MLD in February and March. Part of that variability at depth may be caused by the limited vertical model resolution. The modeled MLD in the winter of 2009/10 stays relatively shallow, less than 400 m, in good agreement with the analysis at the LOCO2 site. The shallow winter MLD in early 2010 suggests weak atmospheric forcing (see next section).

4. Variability of atmospheric forcing

This section aims to establish links between the modeled mixed layer and the atmospheric forcing provided by the CGRF. Interannual variability of the wintertime heat fluxes and atmospheric circulation is assessed first, and then the relative importance of the different high wind speed events is considered.

a. Interannual variability of surface turbulent heat fluxes, winds, and air temperature

Figure 5 shows the total turbulent heat flux (THF), defined as the sum of the latent and sensible heat fluxes and is positive for ocean heat loss. The wintertime (December to March), 2004–10, averaged THF reaches values above 300 W m−2 along the ice edge in the northern Irminger Sea and over the Irminger Current. A local minimum is present over the Irminger Gyre with THF below 100 W m−2. The pattern of average THF is consistent with colder air temperatures and winds generally blowing in a cyclonic circulation centered over the western Irminger Sea (Fig. 6), advecting cold air southward along the east coast of Greenland.

Fig. 5.
Fig. 5.

Winter (December to March) total turbulent heat flux (latent + sensible): the top-left panel shows the multiyear average (W m−2) during 2004–10, while subsequent panels present yearly anomalies (W m−2) relative to the multiyear average. The color axis of the (left) multiyear averages and the (right) anomaly are shown in the bottom right. The magenta lines represent the average 50% sea ice cover contour.

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

Fig. 6.
Fig. 6.

Winter (December to March) 2-m air temperature (colors, °C) and winds (vectors, m s−1): the top-left panel shows the multiyear average (°C) during 2004–10. Subsequent panels present winter averaged winds and the 2-m air temperature anomalies (°C) relative to the multiyear average. The color axis of the (left) multiyear average and the (right) anomaly are shown in the bottom right. The magenta lines represent the average 50% sea ice cover contour.

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

The THF shows a large interannual variability over the simulation period. The winter of 2007/08 shows a large positive anomaly in the THF, with the largest anomalies located east of Cape Farewell (>100 W m−2) and offshore. A northward shift of the Icelandic low produced a stronger pressure gradient extending over the Labrador Sea and Irminger Sea, with stronger westerly winds and increased cold air advection over the Irminger Sea (Fig. 6). This circulation anomaly produced a large heat loss at the surface, creating favorable conditions for the onset of deep convection. The 2007/08 anomaly in the CGRF is consistent with the North American Regional Reanalysis (NARR) anomalies presented by Våge et al. (2009a). The THF anomaly for the following winter of 2008/09 shows smaller anomalies, yet the model produces strong deep convection (Fig. 2). This is due to the preconditioning of the water column. Weak stratification from the previous year requires less heat loss at the surface to trigger the deep convection. This aspect is analyzed in detail in section 5. One might note that the pattern in the wind anomaly for the winters of 2007/08 and 2008/09 bears resemblance to the tip jet, as will be shown in section 4b.

The situation is radically different for the winter of 2009/10 when the THF anomalies are negative. During the 2009/10 winter, the Icelandic low was almost absent in the December–March average field associated with an extreme negative phase of the North Atlantic Oscillation1 (and Arctic Oscillation2). As a result, Fig. 6 shows westward winds associated with a positive temperature anomaly over the Irminger Sea. The advection of warmer air originating from the North Atlantic reduced the atmosphere–ocean temperature and humidity gradients, thus decreasing the THF. This suggests that the condition of strong atmospheric forcing for the onset of deep convection was not met for that particular winter.

b. Strong wind events, air temperature anomalies, and ocean heat loss

This section assesses the causes of high THF over the Irminger Sea and their links to the strong wind events described earlier. In the following analysis, 2-day averaged wind components and THF in winter (December–March) are spatially averaged over each of the three regions defined in Fig. 1. The resulting values are presented in the left panel of Fig. 7, showing the correspondence between the THF and wind components. Occurrences of spatially averaged THF greater than 300 W m−2 over each region are selected, and composites of wind vectors and THF (center and right panels) are constructed by averaging the fields for these selected events.

Fig. 7.
Fig. 7.

(left) Scatterplots of 2-day averaged wind components and total turbulent heat fluxes (W m−2, blue shading). Composites of (center) surface wind speed (m s−1) and (right) total turbulent heat flux (W m−2). These composites are created by averaging winds and heat fluxes for events with total heat fluxes > 300 W m−2. The top, middle and bottom panels are for regions 1, 2, and 3 presented as black boxes in the center and right panels. The composites for the three regions use 29, 28, and 25 events, respectively.

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

For region 1, high turbulent heat fluxes are mainly associated with westerly winds. The composite of the winds clearly shows the signature of the tip jets with an eastward acceleration of the winds close to Cape Farewell caused by low pressure systems traveling between Greenland and Iceland. The associated turbulent heat fluxes show maximum ocean heat loss offshore of the Greenland shelf break (located over the warm Irminger Current) with values in excess of 600 W m−2. Significant heat loss also occurs downwind from that location, over the southern limit of the deep convection area, with values above 400 W m−2. As reviewed in the introduction, tip jets are often associated with continental cold air outbreaks over the Labrador Sea. Cold temperature anomalies reach Cape Farewell and propagate over the Irminger Sea, where the air–ocean temperature gradient increases during the events.

Strong winds from other directions are not associated with high THF over region 1. Easterly and southerly winds advect relatively warm and moist air of Atlantic origin and hence are not associated with high THF events. Northerly winds close to Greenland (barrier winds) are associated with high THF events locally (see below for region 2), but they are not significant when averaged across region 1. Northerly winds over region 1 are generally associated with the large-scale circulation. The air mass travels over the relatively warm Irminger Sea, and hence these winds are not associated with high THF events.

The frequency of tip jets (detected using 2-day averaged wind speeds greater than 10 m s−1 with 30° departure from the east) shows large interannual variability (Table 1). High numbers of events, 20 and 12, respectively, occurred in the winters of 2007/08 and 2008/09, contributing to 53% and 40% of the total wintertime THF over region 1. In contrast, in the winter of 2009/10, when the model shows no deep convection, only two tip jets occurred and contributed to 6% of the winter THF in region 1. This supports the hypothesis that the lack of sufficient atmospheric forcing is directly linked to the shallow mixed layer for that winter. The winter of 2004/05 also shows a high number (17) of tip jets, explaining 52% of the winter total heat flux over region 1. No deep convection is simulated for this particular year (Fig. 2) despite the intense atmospheric forcing, likely due to the midwinter restratification of the ocean surface layer in early February (Fig. 4). This restratification is caused by an episode of intense orographic precipitation and net ocean heat gain associated with a large-scale atmospheric anticyclonic system causing advection of warm air from the North Atlantic northward toward the southern tip of Greenland. These combined effects led to a warmer and fresher near-surface layer in the model, effectively increasing the stratification and disrupting the convection. The 2004/05 time series of THF, precipitation, and near-surface air temperature (figures not shown) shows that the detected tip jets mainly occurred in the early winter (December and January) associated with the rapid deepening of the MLD (Fig. 4). In February and March, the restratification remains intact, as the number of strong wind events is low.

Table 1.

Number of strong wind events and contribution of wintertime (December–March) total turbulent heat flux (THF; %) over each region of the Irminger Sea (see Figs. 1 and 7). The strong wind events’ detection uses a threshold of 10 m s−1 for the 2-day averaged winds. Tip jets are detected over region1 where averaged winds show a 30° departure from the east. Barrier winds (downslope wind) are detected where winds show a 30° departure from −45 (−135) over region 2. Northeasterly winds are detected where winds show a 30° departure from −135 over region 3.

Table 1.

Over region 2, high THF are related to two different wind regimes: (i) northeasterly barrier winds along the coast and (ii) occasional downslope winds events from Greenland blowing offshore near Tasiilaq (Oltmanns et al. 2014). The composite (Fig. 7) shows that the signal is dominated by the barrier winds, with an average of 15 events per year. The downslope wind events are less frequent with only four events per winter on average. Although the THF composite reaches values above 500 W m−2, the region affected by these wind events is limited to the continental shelf. Therefore, these strong wind events do not contribute directly to the deep convection over the southwestern Irminger Sea.

Over region 3, high values of THF are related to strong northeasterly winds advecting cold air from the ice-covered to the ice-free ocean. The strongest winds in the composite are located offshore and closer to Iceland. Stronger winds occur closer to Greenland, but because they are blowing over an ice-covered area, they have a very limited effect on the THF. As in region 2, the signal of high total turbulent heat fluxes does not directly influence the region of deep convection in the southwestern Irminger Sea.

Although high turbulent heat fluxes can be found along the sea ice margin over the western part of the Irminger Sea, only the tip jets directly influence the area where deep convection is simulated. However, deep convection does not always occur simultaneously with tip jets. This suggests that preconditioning, associated with the ocean circulation, plays an important role.

5. Irminger Sea hydrography and circulation conditions

This section investigates the interannual variability of the water column preconditioning by assessing the modeled variations of the hydrographic conditions and the circulation of the Irminger Sea.

a. Variations of the hydrographic conditions of the Irminger Sea

This section examines the hydrographic conditions in the fall to assess the preconditioning of the water column. Figure 8 shows the model potential density in October from 2006 to 2009 along a section across the deepest convection area from Greenland to the Reykjanes Ridge (Fig. 1).

Fig. 8.
Fig. 8.

October model potential density section across the deep convection area from Greenland to the Reykjanes Ridge (approximately at 61°N; Fig. 1) for 2006, 2007, 2008, and 2009. The magenta triangles show the beginning and end of the maximum extent of the MLD deeper than 800 m in the winter of 2008–09. Density contours (fine black lines) are at every 0.02 kg m−3.

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

A strong horizontal gradient is present at the Greenland shelf break between the light (cold and fresh) waters of the East Greenland Current and the denser (warmer and saltier) waters of the Irminger Current. Over the Reykjanes Ridge, the relatively light Atlantic water is clearly visible, extending from the surface to approximately 800 m. This is in agreement with observations (Pickart et al. 2003b; de Jong et al. 2012), although the model overestimates the depth of the 27.8 kg m−3 isopycnal. This model bias is partly inherited from the model’s initial conditions taken from the ORCA ⅝1/12° run and remains fairly constant over the simulation period (figure not shown). Comparison with the AR7E section (figures not shown) also suggests that the model tends to underestimate the depth of the 27.7 kg m−3 isopycnal, and this is related to a salty bias at middepth. Despite this bias, the interannual variability of the hydrography across the selected section (Fig. 8) is analyzed to understand the physical mechanisms leading to the occurrence of the deep convection in the model.

The doming of the isopycnals is visible in the western Irminger Sea, a favorable condition for the onset of deep convection (Marshall and Schott 1999). However, the stratification and doming is less pronounced in October of 2007 than October 2006. This indicates that preconditioning of the water column at the end of the summer of 2007 was less favorable to the onset of deep convection than the previous year. The deep convection for the winter of 2007/08 is the result of the strong THF anomalies that efficiently eroded the surface stratified layer, despite the weaker preconditioning.

For October 2008, the doming of the isopycnals is more pronounced than 2007, with the 27.7 kg m−3 isopycnal reaching the bottom of the summer stratified surface layer. This weaker surface stratification helps explain the presence of deep mixing over the winter of 2008/09, even though the wintertime heat loss is smaller compared to the winter of 2007/08 (Fig. 4).

The preconditioning of the water column in October 2009 is very similar to the previous year and still favorable for the onset of deep mixing. Again the 27.7 kg m−3 contour is present at the bottom of the surface summer stratified layer. The favorable preconditioning but the absence of deep mixing for the winter of 2009/10 suggests that the very weak atmospheric forcing was not sufficient to erode the stratified surface layer.

b. Irminger Sea circulation and mesoscale eddy variability

Figure 9 shows the winter-averaged pressure anomaly at a depth of 643 m and the associated March-averaged MLD for the years from 2006 to 2010. The pressure anomaly shows a minimum extending from the Labrador Sea south of Cape Farewell into the western Irminger Sea. A region of positive pressure anomaly is found over the Reykjanes Ridge. The model pressure anomaly has stronger gradients along the shelf break compared to data from Lavender et al. (2000, see their Fig. 2b), indicating either a stronger Irminger Current in the model or the limited spatial coverage of the observations. The simulated Irminger Gyre structure is consistent with observations from the early 2000s, showing a narrower but stronger Irminger Gyre (see Fig. 12 of Våge et al. 2011). Våge et al. (2011) showed that although the strength of the atmospheric forcing generally decreased in the early 2000s, increases in sea surface height (SSH) over the subpolar gyre were not spatially uniform. The increase in steric height in the eastern half of the basin created steeper gradients, resulting in a stronger Irminger Gyre. The model simulation is carried out for the period of 2003–10 and reproduces this narrower and more intense Irminger Gyre.

Fig. 9.
Fig. 9.

Modeled winter (December to March) pressure anomaly at 643 m (hPa, colors) and March averaged mixed layer depth contours of 700 (black) and 1000 m (white).

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

The location and strength of the simulated Irminger Gyre show little interannual variability between 2006 and 2010 (Fig. 9). As expected, the deep MLD coincides with the area of weak horizontal flow at the center of the Irminger Gyre (Fig. 10, left panel). This is also the region where the model shows weak eddy activity (Fig. 10, right panel). The weak circulation and eddy mixing help to isolate the water at the center of the gyre. As a result, the water mass properties from one winter may influence the strength of the convection in the following winter. This may explain the occurrence of deep convection in the winter of 2008/09 despite the moderate THF anomaly. On the other hand, in the winter of 2009/10, when the THF anomaly is strongly negative, deep convection did not occur although favorable preconditioning was present (Fig. 8).

Fig. 10.
Fig. 10.

(left) Modeled current vectors (m s−1) and (right) mean EKE (m2 s) at 320-m depth. The area of MLD greater than 800 m is delimited by the red contour in both panels.

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

The simulated eddy activity over the eastern Irminger Sea (Fig. 10, right panel) is in good agreement with observations (Fan et al. 2013). These eddies are formed over the Reykjanes Ridge and propagate toward the northeast along the eastern part of the Irminger Gyre. Fan et al. (2013) described a second eddy formation region over the east Greenland Irminger Current associated with baroclinic instabilities produced from steep topography. These smaller eddies, with radii less than 30 km, are not resolved by CREG12 due to the insufficient model resolution (5–6 km; see also Fig. 2 in Dupont et al. 2015). The possible influence of these unresolved eddies on the deep convection in the central Irminger Sea needs to be investigated in the future.

6. Water mass transformation over the Irminger Gyre

The previous sections identified the southwestern Irminger Sea as an area where intermittent deep convection occurs under the right atmospheric forcing and water column preconditioning. Observations show that the characteristics of the water in the center of the Irminger Gyre are similar to LSW (Pickart et al. 2003b). In response to unrealistic transit times for the LSW waters from the Labrador Sea to reach the Irminger Sea, Pickart et al. (2003b) suggested that a fraction of the LSW water needs to be produced locally.

To assess the local LSW production in the Irminger Sea, we consider in this section the variability of deep convection in terms of the surface-forced WMT. This method evaluates the formation rate of a water mass at the surface in specific density ranges based on the modification of surface temperature and salinity by the atmosphere–ocean fluxes of buoyancy (Speer and Tziperman 1992; Speer et al. 1995). Walin (1982) and Speer and Tziperman (1992) showed that the WMT of surface waters in an outcrop region is equivalent to the volume transport across the outcropping isopycnal in a steady state. Contributions from surface radiative and turbulent fluxes, precipitation, and evaporation are taken into account. Because of the limited area covered by sea ice in the deep convection region of the southwestern Irminger Sea, contributions from sea ice freeze and thaw are neglected here. The cross-isopycnal volume flux due to surface buoyancy forcing is given by integrating the buoyancy flux through the outcrop area between isopycnals σ and σ + Δσ for some small Δσ:
eq1
where QH is the heat flux (W m−2) and QF is the freshwater flux (kg m−2 s−1); α and β are the thermal and haline expansion/contraction coefficients; S is the sea surface salinity; cw is the specific heat capacity of water (J kg−1 K−1); and T is the time over which the average is performed. The outcrop region A is defined as the area where convection deeper than 800 m is simulated for any year (see insert on Fig. 4). Monthly averages of the modeled sea surface salinity, heat, and freshwater fluxes are used in this analysis, and density bins are defined for Δσ = 0.05 kg m−3.

Figure 11 presents the annual-mean WMT over the deep convection area. The WMT spreads between densities of 26.1 and 27.80 kg m−3, but the maximum WMT is confined to a narrower range between 27.45 and 27.75 kg m−3. WMT for the 27.6 and 27.75 kg m−3 density classes confirms the formation of water with similar density to the LSW. The positive WMT for density classes lower than 27.3 kg m−3 is mostly driven by the summertime absorption of shortwave radiation and a positive freshwater flux (figure not shown). The WMT for densities higher than 27.3 kg m−3 mainly occurs in the winter period and is caused, in decreasing level of contribution, by latent heat flux, longwave radiation, and sensible heat flux. The multiyear winter average of the WMT shows a predisposition for waters with a density of 27.6–27.75 kg m−3 to be produced by the model. These model results support observational evidence that the Irminger Sea is a secondary location for the production of LSW (Pickart et al. 2003a,b; Falina et al. 2007; Våge et al. 2009a).

Fig. 11.
Fig. 11.

Annual-mean surfaced-forced water mass transformation per density class (Δσ = 0.05 kg m−3) over the area of convection deeper than 800 m (see insert on Fig. 4). Positive (negative) values represent an input (removal) of buoyancy.

Citation: Journal of Physical Oceanography 46, 1; 10.1175/JPO-D-15-0078.1

The WMT shows significant interannual variability. The winters in the early period, between 2003 and 2006, and the winter of 2009/10 mostly produce lighter waters in the 27.6 to 27.7 kg m−3 density class. For the deep convection winter of 2007/08, a higher volume of denser water, around 1.9 Sv with densities above 27.70 kg m−3, is produced. The winter of 2008/09 also shows production of denser waters with density above 27.7 kg m−3, although the total formation rate is reduced to approximately 0.9 Sv. This decrease in the volume formation is likely due to the weaker positive anomaly in the surface forcing compared to the previous winter (Fig. 5). The higher WMT for density above 27.75 kg m−3 is related to the preconditioning of the previous winter (Fig. 8). The model overestimates the density of the MLD water compared to the observed density of 27.721 ±0.005 kg m−3 in the mixed layer at the LOCO2 mooring site (de Jong et al. 2012). In 2010, the model simulates a lower WMT (~0.37 Sv) for densities less than 27.7 kg m−3 because of the negative anomaly in the atmospheric forcing (Figs. 5, 6) despite the favorable preconditioning of the water column.

7. Discussion and conclusions

In this study, we investigate the interannual variability of deep convective mixing in the Irminger Sea for the period between 2003 and 2010 using a high-resolution ocean–sea ice model. Simulated deep convection shows large interannual variability with active convective winters for 2007/08 and 2008/09, in good agreement with observations from mooring data and Argo floats. The model reproduces to a large extent the observed interannual variability of the water column at the LOCO mooring sites, showing similar potential vorticity structure and MLDs. The model also reproduces the high temporal variability of the MLD suggested by observations (Marshall and Schott 1999; de Jong et al. 2012). This variability cannot be captured in simple one-dimensional mixed layer models as they do not consider lateral advection of different water properties in the region of deep convection (Pickart et al. 2003b; Våge et al. 2008).

The interannual variability of the deep mixing is related to the variability of the wintertime atmospheric forcing. To assess the relative contribution of high wind speed events on the deep convection region, we analyzed the impact of four different strong wind events: the tip jets, barrier winds, downslope winds, and the strong northeasterly winds blowing from the Denmark Strait toward the northern Irminger Sea. Our analysis shows that only the tip jets directly influence the deep convection area and can be responsible for a large fraction (up to 53%) of the total wintertime ocean heat loss during active convective years. The number of tip jets is positively correlated with the wintertime North Atlantic Oscillation (NAO) index (Pickart et al. 2003a; Våge et al. 2009b), and increased frequency of the tip jets is associated with a northward shift of the position of the Icelandic low (Bakalian et al. 2007). Despite similar positive NAO indices, the Icelandic low was located farther north during the winter of 2007/08 compared to the previous winter and contributed to the increased frequency of the tip jets. Wind speed anomalies averaged for the winter of 2007/08 resemble the signature of the tip jets, increasing ocean heat loss through turbulent heat fluxes in the model. This enhanced heat loss was sufficient to erode the water column stratification and trigger deep convection.

The representation of the Irminger Gyre in the model is similar to observations in the early 2000s (Våge et al. 2011). The increased SSH over the eastern part of the basin led to a strong and narrow cyclonic gyre that favors the presence of deep mixing by lifting the isopycnals within the gyre. The weak circulation and eddy activity in the center of the gyre isolates the water from exterior influence. Combined with strong ocean heat loss to the atmosphere, all conditions were favorable for deep mixing to occur for the winters of 2007/08 and 2008/09. Since little change occurred within the Irminger Gyre structure and circulation, the shallow MLDs for the winter of 2009/10 are a direct consequence of the anomalously weak atmospheric forcing associated with an extreme low phase of the NAO.

The surface-forced water mass transformation diagnosed using the model results shows the creation of Labrador Sea Water type in the center of the Irminger Sea with densities between 27.6–27.8 kg m−3. The formation rate coherently shows large interannual variability with the active convection years of 2007/08 and 2008/09 generating, respectively, 1.9 and 0.9 Sv of water with density above 27.7 kg m−3. In other years, the WMT rate is lower with the maximum transformation occurring at lighter densities, below 27.7 kg m−3.

The rate of WMT in the Irminger Sea, according to our estimate, is significantly lower than that of the Labrador Sea estimated in previous studies and obtained in our model solution. Over 50°–64°N and 45°–60°W in the Labrador Sea, Myers and Donnelly (2008) estimated the annual-mean WMT rate for density greater than 27.65 kg m−3 to be 2.1–3.9 Sv, for the long-term average, and reaching 9.5–10.8 Sv during years of very active convection. For the CREG12 solution, we evaluate the WMT over the central Labrador Sea, defined as MLD greater than 800 m and bathymetry greater than 3000 m. The model’s annual-mean WMT rate is 6.9 Sv averaged over 2003–10 and 14.3 Sv and 7.5 Sv, respectively, for 2008 and 2009 for waters with densities of 27.75–27.85 kg m−3. According to the CREG12 model solution, the interannual variation of the WMT rates in the Labrador and Irminger Seas are correlated. Furthermore, if normalized by the areas of deep convection for each region, the WMT rates per unit area for the two regions are comparable (figure not shown). In other words, the smaller area of deep convection determines the smaller contribution to the MOC by the Irminger Sea compared with the Labrador Sea.

While the CREG12 model well reproduces the interannual variations of the deep convection in the Irminger Sea as observed by the LOCO moorings, the model’s bias, in particular in salinity, is a potential concern and needs to be minimized in future work. The salinity bias is more significant in the Labrador Sea, increases with time, but is also present in the initial conditions from the ORCA ⅝1/12° free run. This model bias is common to multiple ocean models and is likely caused by strong transport of warm and salty water over the Reykjanes Ridge in the subpolar gyre, combined with an underestimation of the freshwater in the East Greenland Current and its transport in the central Labrador Sea by eddies (Treguier et al. 2005). In agreement with this hypothesis, comparison with Argo floats shows that CREG12 underestimates the seasonal cycle of freshwater in the surface layer of the central Labrador Sea (figure not shown). Reducing the model drift in the Labrador Sea is important but remains to be addressed in future work.

Finally, the present analysis focuses on the period 2003–10 covered by the CREG12 hindcast simulation. It is of interest to study longer time scale variability of the deep convection by extending the duration of the model simulation. Increasing the model’s vertical resolution at the depth of the Labrador Sea Water will help to assess the transport of the newly formed Irminger Sea dense waters and their interaction with the dense waters formed in the Labrador Sea.

Acknowledgments

This work is funded by the Marine Environmental Observation Prediction and Response Network (MEOPAR). We acknowledge the Canadian government CONCEPTS program in facilitating the collaborations among government departments, MEOPAR and with Mercator-Océan of France. We acknowledge the Long-Term Ocean Climate Observations (LOCO) program (https://www.nioz.nl/loco-en) for providing data at the two mooring sites in the Irminger Sea and Dr. M. F. de Jong for generously providing the mixed layer depth data at the LOCO sites. Finally, we are grateful to the two anonymous reviewers who provided constructive suggestions that helped improve the manuscript.

REFERENCES

  • Bakalian, F., S. Hameed, and R. Pickart, 2007: Influence of the Icelandic low latitude on the frequency of Greenland tip jet events: Implications for Irminger Sea convection. J. Geophys. Res., 112, C04020, doi:10.1029/2006JC003807.

    • Search Google Scholar
    • Export Citation
  • Barnier, B., and Coauthors, 2007: Eddy permitting ocean circulation hindcasts of past decades. CLIVAR Exchanges, No. 42, International CLIVAR Project Office, Southampton, United Kingdom, 8–10.

  • de Jong, M. F., H. M. van Aken, K. Våge, and R. S. Pickart, 2012: Convective mixing in the central Irminger Sea. Deep-Sea Res. I, 63, 3651, doi:10.1016/j.dsr.2012.01.003.

    • Search Google Scholar
    • Export Citation
  • Doyle, J. D., and M. A. Shapiro, 1999: Flow response to large-scale topography: The Greenland tip jet. Tellus, 51A, 728748, doi:10.1034/j.1600-0870.1996.00014.x.

    • Search Google Scholar
    • Export Citation
  • Dupont, F., and Coauthors, 2015: A high-resolution ocean and sea-ice modelling system for the Arctic and North Atlantic Oceans. Geosci. Model Dev., 8, 152, doi:10.5194/gmdd-8-1-2015.

    • Search Google Scholar
    • Export Citation
  • DuVivier, A. K., and J. J., Cassano, 2015: Evaluation of WRF Model resolution on simulated mesoscale winds and surface fluxes near Greenland. Mon. Wea. Rev., 141, 941963, doi:10.1175/MWR-D-12-00091.1.

    • Search Google Scholar
    • Export Citation
  • Falina, A., A. Sarafanov, and A. Sokov, 2007: Variability and renewal of Labrador Sea Water in the Irminger basin in 1991–2004. J. Geophys. Res., 112, C01006, doi:10.1029/2005JC003348.

    • Search Google Scholar
    • Export Citation
  • Fan, X., U. Send, P. Testor, J. Karstensen, and P. L’Herminier, 2013: Observations of Irminger Sea anticyclonic eddies. J. Phys. Oceanogr., 43, 805823, doi:10.1175/JPO-D-11-0155.1.

    • Search Google Scholar
    • Export Citation
  • Fichefet, T., and M. A. M., Maqueda, 1997: Sensitivity of a global sea ice model to the treatment of ice thermodynamics and dynamics. J. Geophys. Res., 102, 1260912646, doi:.10.1029/97JC00480.

    • Search Google Scholar
    • Export Citation
  • Harden, B. E., and I. A. Renfrew, 2012: On the spatial distribution of high winds off southeast Greenland. Geophys. Res. Lett., 39, L14806, doi:10.1029/2012GL052245.

    • Search Google Scholar
    • Export Citation
  • Harden, B. E., I. A. Renfrew, and G. N. Petersen, 2011: A climatology of wintertime barrier winds off southeast Greenland. J. Climate, 24, 47014717, doi:10.1175/2011JCLI4113.1.

    • Search Google Scholar
    • Export Citation
  • Hunke, E. C., 2001: Viscous–plastic sea ice dynamics with the EVP model: Linearization issues. J. Comput. Phys., 170, 1838, doi:10.1006/jcph.2001.6710.

    • Search Google Scholar
    • Export Citation
  • Hunke, E. C., and W. H. Lipscomb, 2010: CICE: The Los Alamos Sea Ice Model documentation and software user’s manual, version 4.1. Los Alamos National Laboratory Tech. Rep. LA-CC-06-012, 76 pp. [Available online at http://csdms.colorado.edu/w/images/CICE_documentation_and_software_user's_manual.pdf.]

  • Lavender, K. L., R. E. Davis, and W. B. Owens, 2000: Mid-depth recirculation observed in the interior Labrador and Irminger Seas by direct velocity measurements. Nature, 407, 6669, doi:10.1038/35024048.

    • Search Google Scholar
    • Export Citation
  • Lipscomb, W. H., E. C. Hunke, W. Maslowski, and J. Jakacki, 2007: Ridging, strength, and stability in high-resolution sea ice models. J. Geophys. Res., 112, C03S91, doi:10.1029/2005JC003355.

    • Search Google Scholar
    • Export Citation
  • Madec, G., P. Delecluse, M. Imbard, and C. Lévy, 1998: OPA 8.1 ocean general circulation model reference manual. Institut Pierre-Simon Laplace Notes du Pôle de Modélisation 11, 91 pp.

  • Marshall, J., and F. Schott, 1999: Open-ocean convection: Observations, theory and models. Rev. Geophys., 37, 164, doi:10.1029/98RG02739.

    • Search Google Scholar
    • Export Citation
  • Moore, G. W. K., and I. A. Renfrew, 2005: Tip jets and barrier winds: A QuikSCAT climatology of high wind speed events around Greenland. J. Climate, 18, 37133725, doi:10.1175/JCLI3455.1.

    • Search Google Scholar
    • Export Citation
  • Myers, P. G., and C. Donnelly, 2008: Water mass transformation and formation in the Labrador Sea. J. Climate, 21, 16221638, doi:10.1175/2007JCLI1722.1.

    • Search Google Scholar
    • Export Citation
  • Oltmanns, M., F. Straneo, G. W. K. Moore, and S. H. Mernild, 2014: Strong downslope wind events in Ammassalik, Southeast Greenland. J. Climate, 27, 977993, doi:10.1175/JCLI-D-13-00067.1.

    • Search Google Scholar
    • Export Citation
  • Outten, S. D., I. A. Renfrew, and G. N. Petersen, 2009: An easterly tip jet off Cape Farewell, Greenland. II: Simulations and dynamics. Quart. J. Roy. Meteor. Soc., 135, 19341949, doi:10.1002/qj.531.

    • Search Google Scholar
    • Export Citation
  • Pickart, R. S., M. A. Spall, M. H. Ribergaard, G. W. K. Moore, and R. F. Milliff, 2003a: Deep convection in the Irminger Sea forced by the Greenland tip jet. Nature, 424, 152156, doi:10.1038/nature01729.

    • Search Google Scholar
    • Export Citation
  • Pickart, R. S., F. Straneo, and G. W. K. Moore, 2003b: Is Labrador Sea Water formed in the Irminger basin? Deep-Sea Res. I, 50, 2352, doi:10.1016/S0967-0637(02)00134-6.

    • Search Google Scholar
    • Export Citation
  • Renfrew, I. A., S. D. Outten, and G. W. K. Moore, 2009: An easterly tip jet off Cape Farewell, Greenland. I: Aircraft observations. Quart. J. Roy. Meteor. Soc., 135, 19191933, doi:10.1002/qj.513.

    • Search Google Scholar
    • Export Citation
  • Sampe, T., and S. P. Xie, 2007: Mapping high sea winds from space: A global climatology. Bull. Amer. Meteor. Soc., 88, 19651978, doi:10.1175/BAMS-88-12-1965.

    • Search Google Scholar
    • Export Citation
  • Smith, G. C., F. Roy, P. Mann, F. Dupont, B. Brasnett, J.-F. Lemieux, S. Laroche, and S. Bélair, 2014: A new atmospheric dataset for forcing ice–ocean models: Evaluation of reforecasts using the Canadian global deterministic prediction system. Quart. J. Roy. Meteor. Soc., 140, 881894, doi:10.1002/qj.2194.

    • Search Google Scholar
    • Export Citation
  • Speer, K., and E. Tziperman, 1992: Rates of water mass formation in the North Atlantic Ocean. J. Phys. Oceanogr., 22, 93104, doi:10.1175/1520-0485(1992)022<0093:ROWMFI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Speer, K., H.-J. Isemer, and A. Biastoch, 1995: Water mass formation from revised COADS data. J. Phys. Oceanogr., 25, 24442457, doi:10.1175/1520-0485(1995)025<2444:WMFFRC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Spronson, D. A. J., I. A. Renfrew, and K. J. Heywood, 2008: Atmospheric conditions associated with oceanic convection in the south-east Labrador Sea. Geophys. Res. Lett., 35, L06601, doi:10.1029/2007GL032971.

    • Search Google Scholar
    • Export Citation
  • Stackhouse, P. W., Jr., S. J. Cox, S. K. Gupta, M. Chiacchio, and J. C. Mikovitz, 2001: The WCRP/GEWEX surface radiation budget project release 2: An assessment of surface fluxes at 1 degree resolution. IRS 2000: Current Problems in Atmospheric Radiation, W. L. Smith and Y. Timofeyev, Eds., A. Deepak, 485–488.

  • Treguier, A. M., S. Theetten, E. P. Chassignet, T. Penduff, R. Smith, L. Talley, J. O. Beismann, and C. Böning, 2005: The North Atlantic subpolar gyre in four high-resolution models. J. Phys. Oceanogr., 35, 757774, doi:10.1175/JPO2720.1.

    • Search Google Scholar
    • Export Citation
  • Våge, K., R. S. Pickart, G. W. K. Moore, and M. H. Ribergaard, 2008: Winter mixed layer development in the central Irminger Sea: The effect of strong, intermittent wind events. J. Phys. Oceanogr., 38, 541565, doi:10.1175/2007JPO3678.1.

    • Search Google Scholar
    • Export Citation
  • Våge, K., and Coauthors, 2009a: Surprising return of deep convection to the subpolar North Atlantic Ocean in winter 2007–2008. Nat. Geosci., 2, 6775, doi:10.1038/ngeo382.

    • Search Google Scholar
    • Export Citation
  • Våge, K., T. Spengler, H. C. Davies, and R. S. Pickart, 2009b: Multi-event analysis of the westerly Greenland tip jet based upon 45 winters in ERA-40. Quart. J. Roy. Meteor. Soc., 135, 19992011, doi:10.1002/qj.488.

    • Search Google Scholar
    • Export Citation
  • Våge, K., and Coauthors, 2011: The Irminger Gyre: Circulation, convection, and interannual variability. Deep-Sea Res. I, 58, 590614, doi:10.1016/j.dsr.2011.03.001.

    • Search Google Scholar
    • Export Citation
  • Walin, G., 1982: On the relation between sea-surface heat flow and thermal circulation in the ocean. Tellus, 34A, 187195, doi:10.1111/j.2153-3490.1982.tb01806.x.

    • Search Google Scholar
    • Export Citation
Save