DRAKKAR Workshop 2018 Submitted Abstracts Updated: 18 January 2018

1. Robustness of low latitude western boundary current dynamics in the Indian Ocean in ORCA12.

Bernard Barnier, Jean Marc Molines, Quam Akuetevi, Jacques Verron. IGE - Grenoble ORCA12 hindcast simulations differing by numerical settings (free slip, partial slip, momentum advection), atmospheric forcing (DFS or ERAi), or data assimilation (analyses by Mercator Ocean) are used to describe the interactions between the large anticyclonic eddies generated by the Somali Current system during the Southwest Monsoon. All simulations show that during the period when the Southwest Monsoon is well established, the Southern Gyre (SG) may moves northward along the Somali coast and encounters the Great Whirl (GW). When this occurs, the interaction between the SG and the GW is most frequently a collision without merging, in a way that has not been described in observations up to now. This process is found in all the simulations, but its occurence varies significantly depending on the numerical parameters chosen for the simulation. The presentation will show that this process of interaction is not inconsistent with the satellite altimetry observations and will question the impact of model numerics on the representation of this dynamical process.

2. Near-Surface Horizontal Wavenumber Spectra of the Greater Agulhas Region from a 1/60° Model and from Satellite Observations

René Schubert, Arne Biastoch and Franziska U. Schwarzkopf GEOMAR Helmholtz-Centre for Ocean Research, Kiel. Contact: rschubert@geomar.de The Greater Agulhas Region is one of the regions with highest mesoscale activity in the world ocean. Among other drivers, the interactions of the mesoscale eddies drive energetic submesoscale dynamics in the surface boundary layer. Here we derive near-surface horizontal wavenumber spectra from high-resolution satellite products and from the output of a double nested (1/20°, 1/60°) NEMO configuration with 120 vertical levels. The nests are embedded via AGRIF in a ORCA025 base model. The regional distribution of inertial range and mesoscale spectral slopes can be attributed to the different dynamical regimes prevalent in the Greater Agulhas Region. Further, we emphasize the use of horizontal wavenumber spectra to identify an appropriate choice of the subgrid closure for the model.

3. Simulating the Atlantic Ocean at 1/20° resolution. Klaus Getzlaff, Franziska Schwarzkopf, Arne Biastoch and Claus Böning GEOMAR, Kiel – Session 1 Two AGRIF configurations based on NEMO3.6 using LIM2 with high-resolution focus onto the Atlantic Ocean at 1/20°, embedded in ORCA025, have been established. VIKING20X is the successor of VIKING20 with extended high-resolution area over the whole Atlantic from about 68°N to the southern tip of Africa at 34°S. The nest region of INALT20 extends from 10°N, reaching into the southern Ocean at 63°N, also covering the western Indian Ocean. The two configurations are set up as complementary pair to help to disentangle the influences resulting from an improved representation either of the North Atlantic with its deep water formation contributing to the lower limb of the AMOC or the South Atlantic with its impact from the Agulhas system to the upper limb. Several sensitivity experiments have been conducted focusing on the influence of lateral boundary conditions (free/no-slip) as well as the applied wind stress formulation (relative/absolute winds).

4. Exploring viscosity space in a 1/4° global NEMO Alex Megann Marine Systems Modelling, National Oceanography Centre, Empress Dock, Southampton SO14 3ZH UK Contact: apm@noc.ac.uk The standard ORCA025 NEMO configuration uses a biharmonic viscosity scheme for lateral momentum diffusion, with a default value for rn_ahm_0_blp of -1.5 x 10-11 m4s-1. This is widely suspected to be too

small to provide an ideal closure for the momentum equation discretisation, and in practice leads to grid-scale noise in regions where mesoscale features are poorly resolved. A suite of simulations is described, all using an identical NEMO configuration, but with different viscosity parametrisations. Two have values of the bilaplacian viscosity increased relative to that in the control experiment, and a further pair use instead the biharmonic Smagorinsky deformation-dependent viscosity, which enhances the momentum diffusion in eddy-rich regimes, while decreasing it for more laminar flows. We use a method based on watermass transformation to analyse the numerical mixing in these experiments, and show that increasing viscosity tends to reduce spurious mixing by suppressing noise vertical velocity at the grid-scale, and we discuss the relative advantages of the various viscosity schemes.

5. Schwarz-Christoffel Conformal Mapping based Grids for High-Resolution Ocean Models Shiming Xu Tsinghua University, China We propose a grid generation method for high-resolution ocean general circulation models. Contrary to conventional dipolar or tripolar grids, the proposed methods are based on Schwarz-Christoffel (SC) conformal mappings. By the SC mappings, the continental areas of irregular boundaries can be effectively removed, and the grid lines are aligned to the large-scale coastlines. This improves the simulation of key processes such as coastal Kelvin waves, results in enhanced spatial resolution in coastal regions, and enables easier spatial refinement in these areas. Furthermore, easier computational load balancing is attained. Since the grids are orthogonal curvilinear, they can be easily utilized by the majority of OGCMs such as NEMO. The proposed methods can also be applied to the grid generation task for regional ocean modeling where complex land-sea distribution is present.

6. Insights from energetics on how ocean mixing parameterisations should evolve as model resolution increases

Remi Tailleux University of Reading As model resolution increases, parameterised physical processes such as mesoscale eddies start to become resolved or partially resolved, raising the issue of how to make mixing parameterisations scale aware. In this talk, I’ll provide a simple energetics framework to shed light on the issue, which also clarifies the links between model dissipation due to physical parameterisations and actual dissipation due to irreversible molecular viscous and diffusive processes, while also naturally explaining why backscattering energy is a good idea.

7. Semi-Lagrangian Advection in NEMO François Roy, Christopher Subich, Pierre Pellerin, Greg Smith, Frédéric Dupont Environment Canada Global eddy permitting simulations are presented focusing on the impact of using a semi-Lagrangian (SL) advection scheme in NEMO. The SL method permits the use of higher-order, less-diffusive interpolation for the advection of both tracer and momentum quantities, improving the model's ability to capture rapidly-varying features. The SL scheme is relatively more stable than other schemes commonly used in ocean modelling, allowing a potential increase of the model time step under certain considerations. The implementation of the SL scheme in NEMO 3.6 is summarized. Results from multi-year simulations are presented focusing on eddy kinetic energy. We discuss the possible application to an environmental prediction context (e.g. ice-ocean and coupled weather forecast), considering issues like computational cost, model time-step and dealing with ice-ocean coupling.

8. Lagrangian Ocean Analysis: answers to the ‘why’, and questions on the 'how' Erik van Sebille Utrecht University, Netherlands Lagrangian Ocean analysis, where virtual particle trajectories are computed from OGCM velocity fields, has proven to provide a valuable way to analyse the output of models such as NEMO. The Lagrangian analysis of model velocity fields has helped to answer questions ranging from the transport of heat, nutrients, and plastic litter, to dynamical analysis of the flow itself.

The question is how we can make optimal use of these Lagrangian tools, and what is then required from the OGCMS? For example, which fields need to be saved, at what temporal resolution? How do we treat parameterised physics such as convection? How do we implement advection and diffusion of offline virtual particles in a similar way as within the OGCM schemes?

9. The challenges of tracking Lagrangian particles within high-res OGCMs: the case of Parcels Philippe Delandmeter Utrecht University, Netherlands Lagrangian particle tracking is an increasingly popular tool to analyse ocean general circulation model (OGCM) outputs. Just as OGCMs deal with up to hundreds of terabytes data, so the new analysis tools need to process such enormous amount of data. Parcels is a new Lagrangian Ocean Analysis framework, designed to process the output of state-of-the-art models such as from NEMO. To combine modularity, user-friendliness and efficiency, the model API is implemented in Python before being compiled in C and executed as a library. During this presentation, the main challenges of this model design will be addressed: Which architecture should be used to keep high modularity for the user? How to deal with different grid types? How to scale to massively parallel particle simulations on HPC clusters? Should Parcels be coupled with OGCMs such as NEMO for online particle simulation?

10. Ensemble ocean modelling: new results from the OCCIPUT and PIRATE projects Thierry Penduff IGE Grenoble The OCCIPUT ANR/PRACE project led to the production of 2 main NEMO-based, 1/4°, 50-member ensemble simulations: a 1960-2015 global ORCA025 hindcast, and a 1993-2012 regional NATL025 hindcast. These ensembles are being analyzed and valorized in the framework of the OST-ST PIRATE project, the CMEMS GLO-HR project, and the AtlantOS H2020 project for applications in observational and operational oceanography. This talk will present new results concerning the structure and dynamics of the forced (ensemble-mean) and chaotic (ensemble spread) ocean variabilities simulated by these ensemble hindcasts.

11. Probabilistic interpretation of global SSH variability Sally Close (1), Thierry Penduff (2) and Sabrina Speich (2) (1) IGE Grenoble, (2) ENS Paris Outputs from the OCCIPUT experiment, a 50-member, 0.25° ocean-sea ice model ensemble, are analysed with the aim of providing probabilistic context relevant to the interpretation of available observations. In a first step, the form of the probability density functions of the simulated SSH and their temporal stability are described in various frequency ranges, yielding an estimate of the likelihood of obtaining a measurement approaching the climatological state for any given time. Further decomposition of the SSH signal into forced/intrinsic components permits investigation of the temporal variability of the intrinsic part, elucidating the extent of atmospheric influence on the non-deterministic part of the variability. Ongoing work using the same experiment is focused on the meridional heat transport at 34S, the location of the SAMBA array, and again aims to evaluate the relative role of intrinsic variability and its likely implications for the observation-based record.

12. Simulating the Agulhas system in global ocean models – nesting vs. multi-resolution unstructured meshes

Arne Biastoch (1), Dmitry Sein (2), Jonathan V. Durgadoo(1), Qiang Wang (2), Sergey Danilov (2) (1) Helmholtz-Zentrum für Ozeanforschung Kiel, Kiel, Germany, (2) Alfred-Wegener Institut, Helmholtz-Zentrum für Polar- und Meeresforschung (AWI), Bremerhaven, Germany Many questions in ocean and climate modelling require the combined use of high resolution, global coverage and multi-decadal integration length. For this combination, even modern resources limit the use of traditional structured-mesh grids. Here we compare two approaches: A high-resolution grid nested into a global model at coarser resolution (NEMO with AGRIF) and an unstructured-mesh grid (FESOM) which allows to variably enhance resolution where desired. The Agulhas system around South Africa is used as

a testcase, providing an energetic interplay of a strong western boundary current and mesoscale dynamics. Its open setting into the horizontal and global overturning circulations also requires global coverage. Both model configurations simulate a reasonable large-scale circulation. Distribution and temporal variability of the wind-driven circulation are quite comparable due to the same atmospheric forcing. However, the overturning circulation differs, owing each model’s ability to represent formation and spreading of deep water masses. In terms of regional, high-resolution dynamics, all elements of the Agulhas system are well represented. Owing to the strong nonlinearity in the system, Agulhas Current transports of both configurations and in comparison with observations differ in strength and temporal variability. Similar decadal trends in Agulhas Current transport and Agulhas leakage are linked to the trends in wind forcing. Although the number of 3D wet grid points used in FESOM is similar to that in the nested NEMO, FESOM uses about two times the number of CPUs to obtain the same model throughput (in terms of simulated model years per day). This is feasible due to the high scalability of the FESOM code.

13. The Copernicus Marine Environment Monitoring Service global ocean eddy-resolving physical analysis, forecasting and reanalysis: description and quality evaluation

Yann Drillet, Jean-Michel Lellouche, Olivier Le Galloudec, Florent Gasparin, Eric Greiner*, Gilles Garric, C. Regnier, Marie Drévillon, Romain Bourdallé-Badie, Clément Bricaud Mercator Océan, Ramonville Saint Agne, France ; *CLS, Ramonville Saint Agne, France Over the past years, Mercator Ocean has been regularly upgrading its global ocean physical reanalysis through improvements in the ocean model, assimilation scheme and assimilated data sets. The last upgrade concerned the eddy-permitting reanalysis GLORYS2V4 (¼° horizontal resolution and 75 vertical levels) covering the altimetry era (1993-2016). R&D activities have been conducted at Mercator Ocean in 2016/2017 in order to propose, in the framework of Copernicus Marine Environment Monitoring Service (CMEMS), an eddy-resolving physical reanalysis called GLORYS12V1, covering the same time period and based on the current real-time global forecasting CMEMS system (1/12° horizontal resolution and 50 vertical levels). In parallel to the operational system, a twin free-numerical simulation (without any assimilation) has been performed on the period 2007-2015. The model component is the NEMO platform driven at the surface by ECMWF ERA-Interim for the reanalysis and operational IFS analysis for the real time. Observations are assimilated by means of a reduced-order Kalman filter. Along track altimeter data (Sea Level Anomaly – SLA), satellite Sea Surface Temperature (SST), Sea Ice Concentration and in situ temperature and salinity (T/S) vertical profiles are jointly assimilated. Moreover, a 3D-VAR scheme provides a correction for the slowly-evolving large-scale biases in temperature and salinity. This presentation will provide an overall assessment of this first global 1/12° ocean reanalysis highlighting the level of performance and the reliability of this new eddy-resolving physical reanalysis. In addition, monthly-averaged fields of ocean state estimates based on free simulation and operational system are compared with observation products on the period 2007-2015, to examine the consistency of model fields with related observations for large-scale variability and to provide a baseline mainly focused on in situ comparisons for validation/qualification of on-going system developments.

14. Update on ECMWF latest reanalysis ERA5 Jean-Raymond Bidlot, Kristian Mogensen, Hans Hersbach, European Centre for Medium range Weather Forecasts, UK ERA-5 is a climate reanalysis dataset, covering the period 1950 to present. ERA-5 is being developed through the Copernicus Climate Change Service (C3S). ERA5 data are open access and free to download for all uses, including commercial use. ERA5 is still under development. The definitive release will be done in batches: data covering the period 2010 to 2016 were released in July 2017. Data covering prior periods back to 1979 will be released in early 2018 with the rest (1950-1978) in 2019. The data processing for ERA5 is carried out by ECMWF. All parameters available in ERA-Interim plus some additional ones are also available in ERA5. ERA-5 constitutes a substantial improvement with respect the ERA-interim, with increase resolution and hourly output. Ten ensemble members, and also ensemble mean and spread for all parameters are also being

produced. The temporal resolution is 3-hourly, rather than hourly for the deterministic ERA-5 product, though. Mean and spread should be made available from the outset. A brief description of the data will be made with emphasis on parameters with relevance to ocean modelling. Differences with respect to ERA-Interim, the previous reanalysis from ECMWF will be highlighted.

15. Sea surface height variability in ORCA025 forced with COREv2 and JRA-55-do Patrick Wagner, Claus Böning, Markus Scheinert GEOMAR Helmholtz Centre for Ocean Research, Kiel. Contact: pwagner@geomar.de Sea surface height (SSH) variability in the tropical Pacific is determined by surface fluxes of which momentum flux is the most important one. The new atmospheric forcing data set JRA-55-do (Tsujino, H. et al. (submitted)) offers the chance of an improved representation of tropical SSH variability. We present a comparison of SSH variability of the tropical Pacific in two global ORCA025 (Nemo3.6) configurations, forced with COREv2 and JRA-55-do atmospheric datasets. While the amplitude of SSH variability in COREv2-runs is in close agreement with altimeter observations, a reduced interannual variability of wind stress in JRA-55-do leads to weaker SSH-variability in the tropical Pacific. A lagged correlation analyses of SSH with climate indices such as ENSO or Southern Oscillation shows a higher agreement between JRA-55-do and altimeter observations than between COREv2 and observations, suggesting an improved representation of the processes that determine SSH variability.

16. FOCI - A Flexible Ocean and Climate Infrastructure with Nesting Capability Jan Harlaß, Torge Martin, Arne Biastoch GEOMAR, Helmholtz Centre for Ocean Research Kiel, Kiel, Germany (jharlass@geomar.de) – Session 1 Mesoscale ocean dynamics are well known elements for a variety of processes on both regional and global scale. Global high-resolution ocean models still are computationally expensive for multi-decadal simulations or ensemble integrations. Here, a regional nesting approach proves as an alternative. Furthermore, fixed atmospheric forcing in ocean-only configurations limits the capability of such models to represent feedbacks associated with atmosphere-ocean fluxes. To include the full feedback, we introduce a coupling strategy for a global ocean/sea-ice model, with a high-resolution nesting capability, coupled to a global atmosphere model. This enables a refinement to 1/10˚ in a distinct region within the global 1/2˚ ocean model with 46 levels (ORCA05, NEMO3.6), while the global atmosphere model runs at 1.8˚ resolution and 95 levels (T63, ECHAM6), including the stratosphere. With this setup, we are able to analyze the impact of resolving the oceanic mesoscale in a specific region and atmosphere-ocean-feedbacks as well as to conduct long-term sensitivity experiments with and without a high-resolution nest. Accompanying ocean-only integrations further allow us to isolate the mechanisms of atmosphere-ocean feedbacks. We present our coupling procedure and discuss its implications, such as flux calculations in the framework of different grid resolutions, effective surface boundary forcing and model performance. Initial results from two independent nest configurations covering the North Atlantic Ocean (VIKING10) and the South Atlantic and Indian oceans (INALT10), respectively, show a reduction of prominent surface biases in coupled mode.

17. High-resolution ocean-atmosphere modelling using OpenIFS and NEMO Joakim Kjellsson, Wonsun Park, Mojib Latif GEOMAR, Kiel - Session 1 We discuss the development of a new coupled climate model at GEOMAR, capable of resolving the mesoscale. The ocean component is a ORCA05 NEMO configuration with regional refinements in e.g. the tropical Atlantic (INALT), Southern Ocean (ORION), or North Atlantic (VIKING) where the horizontal resolution is 10 km or finer. The atmosphere component is OpenIFS which can be run at T511 (39 km), T1023 (20 km) or T2047 (10 km) resolution. A climate model that resolves the mesoscale can capture oceanic eddies and tropical storms, and thus be used to understand how these phenomena respond to global warming.

Latest results include experiments using NEMO ocean and ECHAM atmosphere, as well as atmosphere-only experiments with OpenIFS to understand the impact of horizontal resolution in the atmosphere. We report on the latest results and discuss the challenges and goals of future model development.

18. Frequency-domain analysis of energy transfer in an idealized, eddy-resolving ocean-atmosphere model

Paige Martin (1), Brian Arbic (1), Andy Hogg (2), Bill Dewar (3), Jeff Blundell (4), Amanda O’Rourke (1) (1) University of Michigan, (2) Australian National University, (3) Florida State University, (4) National Oceanography Centre Southampton There has been recent interest in deciphering whether ocean and atmosphere variability is primarily due to forced or intrinsic motions in each fluid. To answer this question, we use an idealized, regional, eddy-resolving, coupled ocean-atmosphere model - the Quasi-Geostrophic Coupled Model, or QGCM. The simplified nature of the model lends itself well to the study of how ocean eddies impact atmospheric dynamics, and how the feedback from the atmosphere can affect ocean dynamics. Furthermore, the model allows for partial coupling of the ocean and atmosphere; that is, I can “turn off” certain processes in the model in order to isolate the effects of other processes, or I can run it in purely ocean- or atmosphere-only regimes. This flexibility in the model will prove very useful when trying to decipher exactly how eddies affect and are affected by ocean-atmosphere interaction. Model data is then analyzed via a frequency-domain energy transfer diagnostic, which reveals the relative magnitudes of energy sources and sinks, in both the ocean and the atmosphere. Although QGCM is a relatively basic model, its behavior can improve our understanding of ocean-atmosphere interaction and thus help guide our decisions in how to best represent these processes in larger, more realistic models.

19. Impact of ocean resolution on the response of the AMOC and ocean heat uptake to anthropogenic forcing

Helene Hewitt, Malcolm Roberts, Daley Calvert, Pierre Mathiot, Pat Hyder Met Office Hadley Centre, Exeter Are future projections of the AMOC and ocean heat uptake likely to be sensitive to the resolution of the ocean component? To address this question, experiments with instantaneous quadrupling of CO2 in the GC3 coupled model hierarchy (with ocean resolutions of ORCA1, ORCA025 and ORCA12) have been analysed. These experiments follow the HighResMIP protocol as part of the EU PRIMAVERA project. The timescale for reduction in the AMOC strength is shorter in the higher resolutions which is related to a rapid shutdown of Labrador Sea convection at the higher resolutions. Since Labrador Sea convection is too strong in both coupled control experiments and forced ocean experiments at higher resolutions, reducing this systematic bias may increase confidence in projections. We also discuss the pattern and magnitude of ocean heat uptake. This work will help to inform the choice of ocean resolution for future Hadley Centre climate models.

20. Evaluation of the Irminger basin properties in the GO6 configurations Pierre Mathiot (1), Helene Hewitt (1), Dave Storkey (1), Adam Blaker (2), Tim Graham (1), Leon Hermanson (1), Pedro Colombo (3) and Bernard Barnier (3) (1) Met Office Hadley Center Exeter (2) NOC Southampton, (3) IGE Grenoble The representation of the Nordic overflows is poor in z-coordinate configurations. Recent studies suggest that poor representation of overflows could be the driver of excessive Labrador Sea convection. One of the main objectives of the next Met Office global configuration (GO8) is to improve the overflow representation. As a reference, we evaluated the circulation and water mass properties into the Irminger basin in the eORCA1/025/12 GO6 configurations. First order analysis showed that, in addition to poor representation of the overflow, the mean water masses in the basin as well as coming into the basin are not well represented and thus could be part of the story as this water is entrained into the overflow plume.

21. Overflows in 1/12° in the NEMO context Pedro Colombo, Bernard Barnier, Thierry Penduff IGE Grenoble The representation of overflows at 1/12° resolution is known to be poor in z-coordinate models. A modelling challenge would be to have a vertical coordinate system that is locally adapted to the most critical ocean process. In this work we approach this problem with the study of the Denmark Strait Overflow (DSO) in a regional z-coordinate configuration of NEMO. First, numerical simulations are carried out that scans through different horizontal and vertical resolutions. It is shown that improvements are obtained only at very high horizontal and vertical resolution (1/60° 300L) for a high computational cost. A different approach is then implemented by the use of enveloping s-coordinates (terrain following), in which significant improvements are found in the properties of the DSO even in 1/12° (good comparison with hydrographic sections and observation based transport estimates). Finally, a hybrid z-s vertical coordinate is proposed that relies on a local implementation of s-coordinates within the z-coordinate grid, limited to the area where overflow waters are produced in a North Atlantic configuration. This local implementation is such that it minimizes the effects of pressure gradient errors linked to this type of coordinate, that is smoothly connects to the global z-coordinate and that it does not add any significant computational cost. The improvement of the DSO is found to be drastic.

22. Evaluation of cascading water formation and pathways from NEMO-shelf Arctic Ocean model Maria Luneva, James Harle and Jason Holt National Oceanography Centre, Liverpool Cascading (or shelf convection) is a type of density-driven current in which dense water is formed over continental shelves due to cooling/freezing events and descends down the slope to a greater depth. In the Arctic Ocean cascading contributes to the transformation of warm and saline intermediate Atlantic Waters (AW), which underlies a fresh and cold layer, and the renewal of deep, salty and cold waters below AW (1000-4000m). These dense waters propagate as overflows from the Arctic to Atlantic Ocean, with open ocean deep convection contributing to the Meridional Overturning Circulation. Cascading is thought to be one of the major players in shelf-ocean interaction in the Arctic Ocean and to carbon export from continental shelves to the open ocean. However, as it is an episodic process, happening in wintertime, it is difficult to observe. To our knowledge, there are no observational based estimates of deep water mass formation fluxes at this time. We use outputs from pan-Arctic NEMO-shelf model to examine the effects of cascading on water mass transformations on multi-decadal timescale (1980-2010). We identify key locations on the Arctic shelf where cascading is most probable; evaluate vertical and cross –shelf dense water fluxes and pathways; examine spatial and temporal variability and their relation with summer ice decline and atmospheric forcing (e.g. Arctic Ocean Oscillation). The model has a moderate horizontal resolution corresponding to ¼° ORCA grid and high vertical resolution with terrain following s levels on the shelf, resolving the benthic layer, and z-partial steps in the deep ocean. The model explicitly resolves tides. The GLS vertical mixing scheme is used to reproduce the structure of benthic and ice-ocean boundary layers. We discuss further improvements of vertical coordinates,

23. Evaluation of North Atlantic Subpolar Gyre simulations in coordinates Mathieu Le Corre, Jonathan Gula, Anne-Marie Tréguier LOPS Brest. The aim of the project is to study the interaction between the flow and the topography along the Reykjanes Ridge. To do so we use the Regional Oceanic Modelling System (ROMS) at different resolution (6-2 km) with an increasing number of -level (50-80) to better represent bottom processes. Here we present a first comparison between our simulations and observations in the subpolar gyre. We focus on the water masses properties, the fluxes across the Greenland-Scotland Ridge, and the structure and variability of currents in the vicinity of the Ridge. We use in particular observations along the OVIDE section and results from the RREX2015 cruise. We would like to share about the main differences (and commonalities) between simulations using NEMO and ROMS in the area.

24. Where does the sinking of the Mediterranean thermohaline circulation take place?

Robin Waldman (1), Nils Brüggemann (2), Samuel Somot (1), and Florence Sevault (1) (1) CNRM / Meteo France, Toulouse, France, (2) TU Delft, Delft, The Netherlands Ocean deep convection is a major process of interaction between surface and deep ocean. It feeds the ocean overturning circulation through its downwelling branch. In conceptual schemes of the thermohaline circulation, both deep convection and downwelling are assumed to take place at the same location. However, idealized modelling suggests there is no dynamical source for a net sinking in deep convection areas. Furthermore, oceanic GCMs reveal that the deep convection and overturning rates do not generally match. In this study, we diagnose the sources of downwelling of the Mediterranean thermohaline circulation using a hindcast simulation (1979-2013) of the eddy-permitting Mediterranean model NEMOMED12 (7km resolution) forced by a regional downscaling (12km resolution) of ERAINTERIM reanalysis. We focus on the two main Mediterranean overturning cells, the deep western and the superficial eastern cell. We diagnose the physical sources of downwelling by determining the sources of vortex stretching in the vorticity equation, which is deduced from the online momentum trend diagnostic. Our results confirm expectations from idealized modelling, with no net sinking at intermediate and deep convection sites, whereas boundary currents dominate the downwelling. The dominant physical source of downwelling is lateral dissipation near boundaries, whereas lateral advection ultimately determines the sinking location.

25. High Resolution Modelling in the Labrador Sea and the eastern Canadian Arctic Paul G. Myers, Xianmin Hu, Laura Castro de la Guardia, Clark Pennelly, Laura C. Gillard, Juliana Marson, Charlene Feucher, Nathan Grivault, Yarisbel, Garcia Quintana, Amanda Kahn, Natasha Ridenour University of Alberta, Edmonton, Canada We focus on the evolution of freshwater exchange and water masses process in the Labrador Sea and environs. Using a 1/12 degree Arctic and North Atlantic configurations as well as AGRIF nests at 1/12 degree around the Canadian Arctic, and 1/60 degree in the Labrador Sea, we look at freshwater exchange into the interior of the basin, the impact on mixed layer depth and water formation, and links to the AMOC. The role of melt from the Greenland Ice Sheet is considered, both liquid and solid form. We also examine the role resolution plays in permitting warm waters to reach Greenland and the Canadian Arctic. Experiments using the biogeochemical model BLING examine the importance of freshwater exchange on oxygen and carbon evolution in this region.

26. Evaluation of the OSMOSIS ocean boundary layer model in ORCA1_LIM3 George Nurser (1), Alan Grant (2) and Stephen Belcher (3). (1) NOC Southampton, (2) UKMO/U. of Reading, (3) UKMO. A set of 20-yr runs of ORCA1_LIM3 have been performed driven by the inter-annually varying CORE-2 atmospheric fields. Mixed-layer depths (MLDs) and SSTs in runs using the new OSMOSIS ocean boundary layer (OBL) model model are compared with those using the standard TKE model and against the ARGO climatologies. The effects of switching on and off the diurnal cycle of solar buoyancy forcing and the Fox-Kemper sub-mesoscale mixing parameterization are also considered—these are substantial, especially for the OSMOSIS-OBL model. The OSMOSIS-OBL gives generally deeper austral summer MLDs and cooler SSTs over the ACC region than does the standard TKE model. The different options available (either (0) constant Langmuir number, (1) Pierson-Moskowitz wind-sea or (2) Stokes drift and penetration depth taken from output from the ECMWF wave model) give slightly different results; these results are compared in terms of the impact of the various Stokes drift and Stokes penetration depth fields.

27. An improved parameterization of tidal mixing Casimir De Lavergne University of New South Wales (UNSW), School of Mathematics and Statistics Generated by tidal currents over sloping topography, internal tides are a prominent type of internal waves whose breaking feeds much of the observed small-scale turbulence in the ocean interior. In this talk, I will present a mixing scheme which accounts for the local and remote breaking of internal tides and which obviates the need for a fixed background diffusivity. The scheme uses four static 2D maps of internal tide

dissipation, each associated with a breaking process and a corresponding vertical structure. In spite of the substantial limitations attached to the estimated dissipation maps, favourable comparison with some fine- and micro-structure observations and successful implementation in NEMO suggest that the scheme has potential for improving understanding and modelling of ocean mixing.

28. High Resolution Biogeochemical Modelling in Canadian Coastal Waters Amber M. Holdsworth, Jim Christian, Youyu Lu, Nadja Steiner Institute of Ocean Sciences, Dept. Fisheries and Oceans Canada, Sidney, B.C. High-resolution regional configurations of NEMO 3:6 are being developed for the North-east Pacific and the Canadian Polar Shelf incorporating the Canadian Ocean Ecosystem Model (CanOE). The Northeast Pacific (NEP36) model has a resolution of 1/36° (2-3) km. For the Canadian Polar Shelf, regions of high resolution (Kitikmeot and Inuvialuit Settlement Region) are being nested within the Northern Atlantic and Arctic (NAA) 1/4 (9 - 15)km grid using adaptive grid refinement in Fortran (AGRIF). Both models include the effects of tidal mixing and the latest version of the Louvain-la-Neuve sea-ice model (LIM3) will be included in our high resolution Arctic regions. For both regions we will show that resolution of small scale processes including upwelling and tidal mixing has a large effect on nutrient distributions and biological productivity.

29. High-resolution modelling of a coastal harbor in the presence of strong tides and significant river runoff

Youyu Lu (1), Jean-Philippe Paquin (1), Stephanne Taylor (1), Xiamin Hu (1), Li Zhai (1), Simon Higginson (1), Jerome Chanut (2) (1) Bedford Institute of Oceanography, Fisheries and Oceans Canada, (2) Mercator-Ocean, France Version 3.6 of NEMO is adopted to model the Saint John harbor, Bay of Fundy in eastern Canada, characterized by strong tides and significant river runoff. Multi-level, one-way nesting is applied to link a series of model configurations including basin-scale, shelf, coastal and nearshore regions, with increasing horizontal resolutions of approximately 7.5 km, 2.5 km, 500 m and 100 m. At the mouth of the river, the open boundary routine of NEMO is modified to use observed time series of water level. Evaluation with observational data shows the model’s satisfactory performance in simulating tidal elevation and currents, non-tidal water and currents, temperature, salinity, and surface drifter trajectories. The model solution also includes oscillation tidal fronts, non-stationary tidal eddies, and movement and mixing of “double salt wedges” in the estuary. While the above results are obtained with the “z-level” and one-way nesting, the preliminary results of using the “S-coordinate” and two-way nesting (AGRIF) are also presented.

30. Wetting & drying schemes for shelf seas and partial cells with sloping bottoms for global models Mike Bell (1), Enda O’Dea (1), Dave Storkey (1), Andrew Coward (2), Jason Holt (2), Hedong Liu (2) (1) Met Office UK (2) National Oceanography Centre (NOC), UK. Contact: mike.bell@metoffice.gov.uk Two wetting and drying schemes are being introduced into NEMO version 4.0: an iterative limiter, conceived at NOC; and a directional limiter based on ideas implemented in POM and ROMS. Both schemes conserve volume and do not change tracer fields that are initially uniform. Their performance is illustrated in simple test case and real-world configurations. A simple implementation of a finite volume discretization of pressure forces for partial cells with sloping bottoms is being explored. The bottom slopes are currently restricted by the cell geometry so the ocean bottom is still vertical where it is steep. The scheme runs stably in ORCA025 but its impact appears to be limited. Its impact in idealised “shelf and bowl” configurations is briefly described.

31. Global modeling of internal tides and the internal gravity wave continuum spectrum Brian Arbic University of Michigan, Ann Arbor In this talk I will discuss the challenges and opportunities of global modeling of internal tides and the internal gravity wave continuum spectrum. I will discuss the components that are needed in such models--i.e. simultaneous tidal and atmospheric forcing--and will describe the fundamental technical issues involved in including tides within an ocean general circulation model. I will discuss recent research results on global modeling of internal tides and internal gravity waves, drawing from our ~20 papers on the

subject. I will emphasize comparison to observations, applications to satellite altimetry and operational oceanography, and remaining challenges for this new field of ocean modeling.

32. Simulating and parameterizing the influence of tides on the Amundsen Sea ice shelves Nicolas C. Jourdain (1), Jean-Marc Molines (1), Julien Le Sommer (1), Pierre Mathiot (2), Jérôme Chanut (3), Casimir de Lavergne (4), Gurvan Madec (4) (1) IGE Grenoble, (2) UKMO Exeter (3) Mercator Ocean International, Toulouse, (4) LOCEAN, Paris We undertake harmonic analyses of pseudo-barotropic simulations to show that our regional model configuration of the Amundsen Sea (AMU12.L75) is able to produce tides with similar characteristics as in the barotropic tide simulation imposed at the domain lateral boundaries. Diurnal tides induce topographically trapped vorticity waves along the continental shelf break in the vicinity of the Dotson-Getz Trough. This slightly strengthens vertical mixing in that area, and is likely responsible for the residual circulation of 0.2Sv that flows southward on the eastern flank of the Dotson-Getz Trough. While diurnal tides have a larger amplitude than semi-diurnal tides over the Amundsen Sea continental shelf, semi-diurnal tides are found to resonate in the shallowest ice-shelf cavities, so that both diurnal and semi-diurnal tides produce strong velocities along the ice drafts. The dominant effect of tides on ice-shelf melt in our simulations is by far increased velocities along the ice-shelf drafts rather than vertical mixing or residual currents over the continental shelf. Tides increase melt rates in all the simulated cavities of the Amundsen Sea, with weakest effects for Pine Island and Thwaites ice shelves (<+10%) and strongest effects for Dotson, Cosgrove and Abbot ice shelves (>+30%). A dynamical/thermodynamical decomposition indicates that such simulated increase is mostly due to tide-induced turbulence that enhances heat fluxes across the top ocean boundary layer of ice-shelf cavities. Approximately a third of this effect is counterbalanced by the resulting release of cold melt water. We also bring evidence for a positive feedback whereby tide-induced circulation produces more melt, which strengthens the buoyancy-driven ocean circulation underneath ice-shelves, which in turns produces stronger melting. As tides moslty affect melt rates through increased velocities in the top boundary layer (TBL) of ice-shelf cavities, it is possible to parameterize their effect by including a "tidal TBL velocity" in the calculation of the friction/exchange velocity along the ice draft. In the Amundsen Sea, prescribing a uniform tidal TBL velocity leads to overestimated melt rates near deep grounding lines where the thermal forcing is high due to the presence of modified CDW, but where actual tidal currents are weak. In the absence of spatially-distributed observations of tidal currents along the ice draft, we therefore recommend to derive spatially-varying tidal velocities from the outputs of tide models and to inject them in the three equations of the ice-shelf melt module. In our simulations, prescribed TBL velocities are 20% weaker than barotropic velocities from a tide model because of the vertical profile due to the interaction with the ice draft. Furthermore, if it was possible to observe spatially distributed currents underneath an ice shelf and to perform a harmonic analysis, we would still have to apply a correction factor before prescribing the resulting mean-square TBL velocity. Indeed, the associated tide-induced melting enhances observed TBL velocities, so that directly applying observed tidal velocities would be equivalent to represent two times the feedback related to the buoyancy-driven ocean circulation.

33. Canadian Arctic Archipelago sea-ice motion and consequences on ocean transport: numerical errors and potential improvements from the addition of tides

Nathan Grivault and Paul G. Myers

University of Alberta, Edmonton, Canada

An accurate representation of sea-ice, with realistic concentration and thickness, is one of the most

challenging aspects of numerical simulations of the Arctic. Particularly in the Canadian Arctic Archipelago

(CAA), where narrow straits and shallow basins limit ice motion.

We use NEMO version 3.6 coupled to LIM2 with the ANHA4 regional configuration. We set up two

experiments, one with tides, and one without. Both experiments are forced with high spatial and temporal

resolution atmospheric forcing, CGRF, and run from 2002-2016.

We examine sea-ice motion without tides in the CAA from 2004-2016 and evaluate the impact on volume

and freshwater fluxes through and out of the CAA. We then analyze how the inclusion of tides might

improve the sea-ice representation and transports in this region.

34. The marginal sea ice zone in the subarctic region in ORCA12, forced and coupled to the

atmosphere Anne-Marie Tréguier (1), Jean Sterlin (2), Tim Graham (3), Pierre Mathiot (3), Helene Hewitt (3), Camille Lique (4), Claude Talandier (1)

(1) CNRS, LOPS, Plouzané, (2) ELIC, U.C. Louvain la Neuve, (3) UK Met Office, Exeter, (4) Ifremer, LOPS, Plouzané.

The marginal sea ice zone (MIZ) is a complex interface between the open ocean and the pack ice, where ocean-ice-atmosphere interactions are extremely complex. The global 1/12° ORCA12 resolves to some extent the spatial structure of the MIZ, whose width is about 100 km. When performing ocean-ice simulations forced by a prescribed atmospheric state, there is a possible mismatch between atmosphere and ice/ocean due to the absence of feedback between sea ice distribution and atmospheric temperature. This could lead to spurious air-sea fluxes and water mass formation in the MIZ in forced DRAKKAR simulations. This issue is examined by comparing two ORCA12 simulations performed by the U.K. Met Office, using the G06 configuration (ORCA12+CICE ice model). One simulation is forced by CORE forcing, and the other is coupled to a high resolution atmosphere. Our analysis is focused on the subarctic region (Nordic seas and the entrance of the Arctic Ocean). We find that the biases in the MIZ position and width are of the same order of magnitude in the coupled and forced model. Surface heat fluxes in the MIZ are similar in amplitude in both cases. The comparison of the forced and coupled simulations demonstrates the key role of the ocean circulation in setting the location of the sea ice edge.

35. Challenges of modelling a realistic Arctic Ocean stratification Verena Haid, Claude Talandier, Camille Lique Laboratoire d'Océanographie Physique et Spatiale, Brest, France – Session 1 or 4. In the Arctic Ocean, as in any cold environment, the density and thus the stratification is primarily determined by the salinity. Schematically, the vertical structure of the water column is as follows: a thin fresh and cold surface layer, the Polar Mixed Layer, is separated from the underlying warm Atlantic Water by a strong cold halocline. Thus, the heat from the Atlantic inflow that is not lost to the atmosphere upon entrance into the Arctic Basin is prevented from interacting with the sea ice cover. This stratification is stable, since vertical mixing is much weaker in the Arctic Ocean than in the other parts of the world ocean, due the presence of sea ice hampering the momentum input from the atmosphere and effectively attenuating surface waves, and the lack of strong tidal mixing. When it comes to models, representing a realistic Arctic stratification remains a major challenge. Here, we present results from a series of simulations performed with a regional Arctic-North-Atlantic configuration at ¼° resolution, based on NEMO 3.6_LIM3, in which we explore the sensitivity of the stratification to the choice of ocean and sea ice parameters, as well as parameterizations of the vertical mixing. On short timescales (up to a few years), the stratification is mostly controlled by the 1D vertical balance between freshwater input and vertical mixing. In this context, a proper vertical distribution of the level of energy as well as a correct river runoff distribution and sea ice melt and formation processes are crucial. On longer timescales, horizontal advection plays a bigger role for the stratification. In particular, we show the importance of how we represent the subduction of the warm and salty Atlantic-originated water mass at the entrance of the Arctic basin, which is mostly controlled by sea ice parameters, such as the sea ice strength and ice-ocean drag coefficient.

36. Southern Ocean and AMOC sensitivities in a traceable hierarchy of global configurations. Dave Storkey (1), Adam Blaker (2), Pierre Mathiot (1), Alex Megann (2), Yevgeny Aksenov (2), Ed Blockley (1), Daley Calvert (1), Tim Graham (1), Helene Hewitt (1), Pat Hyder (1), Till Kuhlbrodt(3), Jamie Rae (1) and Bablu Sinha (2) (1) Met Office Exeter, (2) NOC Southampton, (3) University of Reading The latest UK versions of the NEMO ORCA configurations form a largely traceable hierarchy of models at 1 degree, 1/4 degree and 1/12 degree resolution. Recent developments have focussed on the Southern Ocean. Sensitivity experiments with the 1/4 degree configuration show that the tuning of vertical and isopycnal mixing coefficients has a greater impact on Southern Ocean hydrography and sea ice than

changes to the sea ice parameters and the representation of icebergs. This is consistent with other recent papers in the literature. Opening the main ice shelf cavities mostly reduces temperature and salinity biases on the shelf. The evolution of the Atlantic overturning circulation in the first 30 years of simulation is very different for the different resolutions.

37. High resolution modelling of the ocean and sea ice in Adelie Land: Pierre-Vincent Huot ELI - TECLIM UCL, Louvain-la-Neuve, Belgium – Session 4 Polar climate results from complex interactions between the ocean, the cryosphere and atmosphere. It differs from other regions in several ways : small oceanic eddies due to small Rossby number, strong seasonality of ocean stratification and deep water formation, sea ice coverage modulating air-sea fluxes, katabatic winds and polar lows. Current global climate models do not suitably represent associated processes because of the fine resolution needed to resolve them. Some processes are indeed parametrized or even neglected. Yet, they might play a substantial role in defining the way ocean, ice and atmosphere interacts. Our goal here is to evaluate the importance of small scale processes on the state of the Southern Ocean and its sea ice. For this purpose, a set of regional simulations will be performed off Adelie Land, East Antarctica, where polar processes are particularly strong and observational data relatively abundant. The simulations are run NEMO3.6-LIM3 at two horizontal resolutions, namely 2 km and 10 km. We particular focus on the sensitivity of the modelled sea ice state and upper ocean stratification to: ocean mesoscales eddies; modifications of the air-ice drag coefficient and inclusion of tides.

38. Modelling dynamics of sea ice and ocean in the marginal ice zone: effects of rheology,waves and mixing

Yevgeny Aksenov (1), Stefanie Rynders (1), Lucia Hosekova (2), Danny Feltham (2), A.J. George Nurser (1), Gurvan Madec (3), Andrew Coward (1) (1) National Oceanography Centre, Southampton, (2) University of Reading, (3) Institut Pierre Simon Laplace (IPSL) Paris. Contact: yka@noc.ac.uk. Exposure of large, previously ice-covered areas of the Arctic Ocean to the wind and surface ocean waves results in the Arctic pack ice cover becoming more fragmented and mobile, with large regions of ice cover evolving into the Marginal Ice Zone (MIZ). The need for better climate predictions, along with growing economic activity in the Polar Oceans, necessitates climate and forecasting models that can simulate fragmented sea ice with greater fidelity. We use sea ice-ocean general circulation model NEMO (stands for Nucleus for European Modelling of the Ocean) coupled with the ocean wave model WAM output from model of the European Centre for Medium-Range Weather Forecasts (ECMWF) and examine several key mechanisms through which the waves effect ocean and sea ice. The wave-ice interactions include ice fragmentation due to break–up by waves in the vicinity of the ice edge; wave attenuation due to multiple scattering and non-elastic losses in the ice, wave advection and evolution of ice fragmentation (floe sizes). In the model combined collisional and Elastic-Viscous-Plastic (EVP) rheology is implemented, reflecting granular behaviour of sea ice in MIZ and continuum pack ice dynamics in the areas of the compact ice. The effect of surface waves on ice motion is included in the turbulent kinetic energy or ‘granular temperature’ of ice floes. Breaking waves also cause mixing of the upper water column and present mixing schemes in ocean models take this into account through surface roughness, which is commonly parameterised from wind speed. We present results from simulations using modelled significant wave height instead, which accounts for the presence of sea ice and the effect of swell. We analyse the impact of the waves on sea ice and the upper ocean, focusing on the marginal ice zone (MIZ) where the wave impacts are the most. However, it is found that the combined rheology has impact beyond the marginal ice zone, influencing ice motion and sea ice thickness. In the Arctic, ice thicknesses increase overall, with higher increases in the Western Arctic and decreases along the Siberian coast. In the Southern Ocean the new mixing parameterisation improves sea ice volumes in the simulations. The study is a part of the EU FP7 Project ‘Ships and waves reaching Polar Regions (SWARP)’ under the European Union's Seventh Framework Programme (FP7/2007-2013), grant agreement N°607476 and also contributes to the Copernicus system and is linked to several ongoing UK national research initiatives.