image © Enjoylife2/iStock/Thinkstock 

Dynamic Ocean Model - NEMO


The dynamic ocean model used for medium-range and seasonal forecasts of ocean structure is the Nucleus for European Modelling of the Ocean (NEMO).  It is coupled with all IFS forecast models (HRES, ENS, Extended-range and Seasonal forecast models).  NEMO is a three-dimensional general circulation ocean model and can reproduce the general features of the circulation and the thermal structure of the ocean and their seasonal and inter-annual variations.

The Numerical Structure

Whereas the atmospheric model covers the whole globe, the ocean model has the additional problem of lateral boundaries along coasts causing effects such as boundary currents (e.g. the gulf stream).  Also near the continents the sea depth becomes abruptly more shallow at the continental shelfs.

The ocean-atmosphere coupling is carried out every hour and is achieved by a two-way interaction:

  • the atmosphere affects the ocean through its wind, heat and net exchange of moisture by precipitation and/or evaporation
  • the ocean affects the atmosphere through its sea-surface temperature, ocean surface current and ice concentration.

The HRES and ENS use the atmosphere-wave-ocean coupling framework from the start of the forecast.  This is because it is important to capture two-way feedback between the atmosphere and the sea-surface temperatures, sea-ice extent and ocean waves (e.g. a slow-moving tropical cyclone can cool the sea surface).  

The ocean model has approximately 0.25 degree (~28km) horizontal resolution.  There are 75 ocean levels and the vertical resolution varies with depth.   Levels are closest together near the sea surface (18 levels in the first 50m, the top level is <1m depth) to capture the temperature and salinity structure of the uppermost layers and the thermocline.

For both atmospheric and ocean models it is important to assimilate not only the sea surface temperatures but also the salinity and temperature structure within the ocean itself. 

The sea surface temperature has a strong effect upon the development, evolution and persistence of tropical storms.  Surface wind stress associated with tropical storms, and even that associated with mid-latitude depressions, can induce upwelling that brings cooler water up to the sea surface.  This can affect the subsequent evolution of weather systems.

Therefore the dynamic ocean model requires detailed measurements of temperature and salinity at depth within the ocean.  Up to a decade ago such sub-surface information was infrequent at isolated locations.  However, the availability of these data has become extensive since the introduction of ARGO floats.  These floats take measurements as they sink to about 2000m before returning to the surface to download the information via satellite before repeating the cycle.  There are about 3800 ARGO floats in operation distributed throughout all the oceans. 

Additionally Airborne eXpendable BathyThermographs (AXBTs) are small floats or buoys that are dropped to the sea surface on parachutes. Once afloat, these instruments deploy long wires with temperature sensors, transmitting ocean temperature versus depth as the sensors sink through the water column to about 500m.  AXBT data is frequently used to increase knowledge of the ocean structure along the forecast track of a tropical storm and is an important factor in determining the intensity a hurricane may reach.

These data are assimilated by NEMOVAR.


Handling of Sea Ice

Throughout the forecast period the changing extent of sea-ice and the variation of the ice shelf with time have important effects upon the energy and moisture balance at the atmosphere/surface boundary. The Louvain-la-Neuve Sea Ice Model (LIM2) is a prognostic sea-ice model that deals with the dynamic and thermodynamic evolution of the sea surface so that sea-ice cover evolves dynamically.  It is incorporated into the dynamic ocean model.  The ice extent will change through the forecast period in response to sea temperatures and air temperatures, ocean currents and wind.

Three-Dimensional Data Assimilation  - NEMOVAR

NEMOVAR is a three-dimensional variational assimilation (3D-Var) system adapted to the NEMO model.  The observed characteristics (temperature, salinity) vary only slowly and a 3D-Var system is used since there is little need to fit observations to a precise time.  NEMOVAR assimilates:

  • temperature and salinity profiles,
  • sea-surface temperature (SST),
  • altimeter-derived sea level anomalies,
  • ice information.

The Operational Sea Surface Temperature and Sea-Ice Analysis (OSTIA), taken from an external source, provides sea-ice information.  This is combined with the NEMOVAR ocean assimilation to give initial conditions for coupled model and also to provide a first guess for the next NEMOVAR assimilation cycle.   The ocean analysis system consists of a reanalysis stream (ORAS5) and a real-time stream (ORTS5). 

Observed sea-surface temperatures are not assimilated directly but a strong relaxation towards the OSTIA sea-surface temperature data is applied during the outer loops of the data assimilation cycle.  

The model sea surface temperature is adjusted to be consistent with the sea-ice concentration.  If the sea-ice concentration is:

  • higher than a given threshold (55%) then the model sea surface temperature is set to freezing point.
  • less than the given threshold (including where there is no ice in the observations) but the model sea surface temperature is below freezing point then the strength of the relaxation term to observed sea surface temperature is increased.

Bathymetric observations are not used in regions where the total model depth is less than 500m in order to avoid assimilating data on the continental shelves where the model has poor representativeness.

The Ensemble of Data Assimilations for NEMO

It is important to represent uncertainty in the ocean initial conditions and in model structure.  An oceanic EDA system achieves this.  The perturbed analyses that result contribute through ocean-atmosphere coupling to the ensemble of forecasts used for probabilistic predictions at medium, monthly and seasonal ranges.

Additional Sources of Information

(Note: In older material there may be references to issues that have subsequently been addressed)