Amending North Atlantic Model Biases to Improve Arctic Predictions (NAtMAP)

NAtMAP is a European consortium project under the ERA-Net.RUS funding scheme, with partners from the Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research (AWI; BMBF grant number 01DJ16003), the Institute of Oceanology of the Polish Academy of Sciences (IOPAN; NCBR grant number 3/RUSPLUS-S&T/2016) and the Institute of Numerical Mathematics of the Russian Academy of Sciences (INMRAS; RFBR grant number 16-55-76004 ERA_a).

The project is funded from 2016 through 2018. The main goal of NAtMAP is to understand and reduce long-standing common biases in the simulated hydrography of the North Atlantic. To this end we analyse the impact of oceanic resolution in key regions, such as the Gulf Stream and North Atlantic Current regions, the Labrador Sea, and the Denmark Strait. The main tool for this research is the AWI unstructured-mesh model FESOM2. In addition, we study the influence of model parameterizations that are particularly relevant in this context, with FESOM2 as well as with the INM climate model. To assess the realism of the simulation, a particular focus is on oceanographic transects for which long-standing observational records exist, such as in the Fram Strait region.

Contacts:

• Agnieszka Beszczynska-Möller (IOPAN)
• Evgeny Volodin (INMRAS)
• Helge Goessling (AWI)

Some preliminary key findings of NAtMAP are summarised in the following.

Using long-term in situ and satellite observations and available climatologies we evaluate to what extent a higher resolution, allowing the explicit representation of eddies and narrow boundary currents in the North Atlantic and Nordic Seas, can alleviate the common model errors. The impact of increased resolution on the modeled characteristics of Atlantic water transport into the Arctic is examined with a special focus on separation of Atlantic inflow between Fram Strait and the Barents Sea, lateral exchanges in the Nordic Seas, and a role of eddies in modulating the poleward flow of Atlantic water. We also explore the effects of resolving boundary currents in the Arctic basin on the representation of the adjacent sea ice. Resolving the eddy field in the Greenland Sea is assessed in terms of reducing the deep thermocline bias. The effects of better resolving the Labrador Sea area on reducing the model bias in surface hydrography and improved representation of ocean currents is also addressed.

An example of our results is shown in the figure to the right. The top row shows temporally averaged observations of temperature (left) and salinity (right) as a function of longitude and depth along the Fram Strait. The remaining four rows show the same for model simulations with FESOM2 at 25km resolution (2nd row) and at 8km resolution (3rd), and with the INM model driven by atmospheric forcing (like FESOM2; 4th row) and coupled with an interactive atmosphere (5th row). The figure reveals that the strong observed vertical stratification, with maximum temperatures at the surface, is a challenge for the models. While the 25km-version of FESOM2 fails to represent this feature properly, the 8km-version produces a much more realistic pattern. The INM simulations perform comparably well in terms of the temperature profiles, but they exhibit much too saline waters in deep layers and fail to reproduce a thin fresh-water layer at the surface. Note that considerable differences exist also in the bathymetry (the shape of the white masked parts of the plots). This is also true for the two FESOM2 versions even though the vertical resolution is identical. Additional simulations with different bathymetries (not shown) that have been conducted as part of NAtMAP reveal that in fact the bathymetry plays a critical role that needs careful consideration in the construction of numerical ocean model meshes.

A more comprehensive comparison of these simulations with in-situ observations can be found on this poster, which was presented at the EGU General Assembly 2018.

Not only model resolution plays an important role when it comes to simulating the physics of the ocean, sea ice, and atmosphere: Many processes that take place on scales too small to be resolved explicitly need to be represented by model parameterisations. In NAtMAP we have studied the sensitivity of the simulations to some parameterisations that are particularly important for the climate of the North Atlantic and the Arctic. These include: The formation of salt plumes when new sea ice is formed; the occurrence of deep-ocean mixing, e.g. during cold-air outbreaks; and the diffusion of ocean properties along surfaces of constant density (isopycnals) or in the vertical (z-coordinate).

A result concerning the last of these processes is depicted in the figure to the right, which shows annual mean model errors for temperature (K) as simulated by two versions of the the INM climate model at the surface (top), at 1000m (middle) and averaged over 70W-10W in the Atlantic (bottom). In one version (left) there is additional diffusion only along isopycnal surfaces, whereas in another version (right) there is additional diffusion only in the vertical dimension. It becomes apparent that the additional diffusion in the vertical transports heat from the surface to the deeper ocean, resulting to enhanced cold biases close to the surface and warmer temperatures in the deep ocean. The surface cooling is particularly pronounced in the Northern North Atlantic - the NAtMAP study area - where the surface is too cold by more than 8K, and also the deep-ocean warming is particularly strong in the Atlantic. This error pattern is strikingly reminiscent of a very common bias seen in many climate models, suggesting that spurious vertical mixing might play an important role in this long-standing problem.

Some of our results on the influence of model parameterisations on the North Atlantic and Arctic hydrography are described in two papers that have been accepted for publication. Links to these papers will be provided here as soon as they become available.

A central idea of NAtMAP is to analyse the influence of local versus remote model resolution on the simulate ocean hydrography in the North Atlantic and Arctic. To this end we exploit the unique properties of FESOM2 which allows to use unstructured meshes with resolution changing locally in a highly flexible manner. We have designed a number of FESOM2 meshes specifically for NAtMAP to contrast the influence of key regions of the North Atlantic (bottom figure to the right). In a first step we ran simulations with homogeneous resolution in the whole North Atlantic and Arctic ranging from 25km to 4km. The simulation results indicate that the medium-resolution version with 8km grid-point spacing is already almost as realistic as the 4km version when it comes to representing fundamental aspects of the North Atlantic current systems (top figure to the right).

Based on this finding we designed specific meshes with local coarsening from 8km to 25km (the background resolution) to assess the importance of the respective key regions to shape the North Atlantic hydrography. Not surprisingly we find that the hydrography is deteriorated in the respective coarsened regions. More interestingly, however, we also find that the hydrography of adjacent regions is affected as well, and that this effect is more pronounced towards the respective region to the south-west, which is not obvious given that the main currents - the Gulf Stream and the North Atlantic Current - flow in the opposite direction, so that downstream effects might be expected rather towards the north-east. This suggests that the representation of the rather narrow coastal boundary currents along the eastern coast of Greenland and associated with the subpolar gyre might play a critical role when it comes to a realistic simulation of the North Atlantic Ocean.

A more comprehensive comparison of the uniform-resolution simulations can be found on this poster, which was presented at the EGU General Assembly 2018.The link to a paper describing these results in detail, including the influence of local coarsening, will be provided here as soon as it becomes available.