Modelling ocean surfaces

The majority of the surface of the earth is ocean and therefore the ocean/atmosphere interface is very important.  The wave model (ECWAM) provides information on sea-surface roughness and hence momentum loss in the boundary layer flow. The dynamic ocean model (NEMO) provides information on the sea-surface temperature.  Changes in these parameters as the forecast progresses impact strongly on monthly or seasonal forecasting.  This is particularly important with respect to El Niño/La Nina (ENSO) or other similar developments.

The state of the ocean surface can change on a daily time-scale.  Changes in the extent or movement of sea-ice, or modification of the structure of the upper ocean after the passage of hurricanes have an impact on the boundary energy and momentum interactions (e.g. a tropical cyclone can cool the sea surface through turbulent upwelling of colder water, particularly if the cyclone is slow-moving and/or the ocean's mixed layer is shallow).

Oceanic information is now derived in much the same way for all IFS model configurations.

  • Initial sea surface temperatures and sea ice concentration are given by NEMOVAR.  In practice the following procedures are adopted to deliver the T+0 fields used as the starting point for the coupled model forecasts:
    • For Sea-Ice, the ocean analysis assimilates OSTIA sea-ice fields and in effect is blended with the background (as happens in atmospheric assimilation).
    • For Sea Surface Temperature (SST), the latest OSTIA sea surface temperature analysis (re-gridded) is used but the approach depends on latitude.  Between 20S and 20N the NEMO ocean sea surface temperature analysis is relaxed towards the latest OSTIA sea surface temperature analysis.  North of 25N and South of 25S, the latest OSTIA sea surface temperature analysis is used unchanged.  Between 20 and 25 degrees a hybrid of these approaches is used.
  • Throughout the forecast period, NEMO provides the oceanic temperature structure near and at the surface.  ECWAM provides wave data, and therefore an indication of surface roughness.  From these, fluxes of heat, moisture and momentum are evaluated for passing to the lowest layers of the atmosphere by partial coupling (and later in the forecast by full coupling).  The formation, evolution and decay of ice over open waters is controlled by LIM2 (part of NEMO).  In effect, NEMO and LIM2 together move ice around (according to ocean drift etc.) and melt or form ice (according to sea-surface temperatures etc.).  The albedo over the sea-ice surface uses a climatology prescribed in the IFS rather than the model albedo of LIM2 (this is because LIM2 does not model some of the key processes important for albedo such as melt ponds).  Note: ECMWF uses LIM2 which is an earlier version of the Louvain-la-Neuve sea ice model currently available (Version 3.6)  


Partial coupling of sea surface temperature between IFS atmospheric and oceanic models.

Importance of atmospheric/ocean coupling

IFS needs to model the exchange of heat, moisture and momentum at the interface between the model atmosphere and the underlying model surface.  The ocean-atmosphere coupling is achieved by a two-way interaction:

  • the low-level atmosphere affects the ocean through its:
    • heat (causing changes in sea surface temperature and stability in the ocean).
    • moisture (by precipitation and/or evaporation).
    • low-level wind (affecting sea surface currents and causing wave development).
  • the ocean affects the atmosphere through its:
    • sea surface temperature (causing changes in stability in the lower atmosphere).
    • ocean surface current (causing changes in roughness and momentum exchanges). 
    • ice concentration (causing changes in roughness, momentum exchanges and boundary layer cooling).


Fig2.1.4.3-1: An example of the beneficial effect of using coupled atmospheric/ocean to realistically simulate the cool wake after a tropical cyclone (TC Neoguri in 2014).  The forecast changes in sea surface temperature agree closely with those measured by DRIBUs close to the wake of the cyclone.  Particularly well modelled is the sharp fall in sea surface temperature after passage of the tropical cyclone followed by successive pulses of warmer and colder water.

Importance of analysis of sea surface temperatures

IFS needs a representative and timely analysis of sea surface temperatures and sea-ice to derive radiances over the ocean surfaces.  These factors are important because:

  • they influence the representation of the lower atmospheric boundary condition for the model.
  • more up-to-date data improves the forecast.
  • the assimilation of satellite data depends upon the modelled sea surface temperature and distribution of sea-ice.

Difficulties in the assimilation of sea surface temperature:

  • OSTIA and EUMETSAT OSI-SAF data are sometimes quite delayed – possibly by some 60hr.
  • Uncertainty whether it’s better to adjust IFS model sea temps to sea ice extent or vice-versa.
  • Rapid changes in sea surface temperature (e.g. upwelling) take a long time to be absorbed into OSTIA.  Rapid changes are captured rather better with NEMOVAR and OCEAN5 than OSTIA.  Errors can lead to:
    • over prediction of intensity due to unrealistic heat availability from the ocean.
    • day-to-day consistency is not always good.
    • spurious or missing areas of ice.

Full atmospheric/ocean coupling

Full atmospheric/ocean coupling was used by IFS Cy44R and earlier.  

The ocean model (NEMO) takes sea surface temperature from the ocean data assimilation system (NEMOVAR and OCEAN5).  Sea surface temperature observations may be deficient both spatially and in timeliness.

Partial atmospheric/ocean coupling 

Partial atmospheric/ocean coupling is used by IFS Cy45R1 and later.

Sea surface temperature analysis is derived:

  • In the extra-tropics from:
    • sea surface temperatures of the initial conditions (e.g. OSTIA) plus sea surface temperature tendencies  (ΔSST in fig) from the ocean model (NEMO).
  • In the tropics from:
  • Sea-ice forecasts are derived from the ocean model (NEMOVAR and OCEAN5).

Beyond 5 days forecast the partial coupling is gradually switched off (if sea surface temperatures from atmospheric and ocean models are in line). See fig.


Fig2.1.4.3-2: Schematic diagram showing partial coupling is used for the first four days of the forecast then gradual reduction between day4 and day8 followed by full coupling after day8.

Advantages

  • There is assimilation rather than interpolation of OSI-SAF sea-ice.
  • Small-scale structures in the sea surface temperature field of OSTIA are preserved.
  • The coupled model is able to simulate realistically the cool wake after a TC (and possibly also after deep depressions).  Verification is often difficult as there are only few available relevant observations. See Fig.
  • It moves sea surface temperature and sea-ice analysis closer to the 4D-VAR analysis, and with better timeliness.
  • Tropical sea surface temperature analyses from dynamic ocean assimilation systems seem to be better than no-model assimilation systems.
  • Atmosphere/ocean interface are more dynamically consistent.
  • There is an increase in atmosphere/ocean coupling in the analysis.
  • Ensemble scores improve (e.g. root mean square errors in Geopotential heights) decrease.

Problems

  • The increased variability in sea surface temperatures can lead to difficulties in verification.


Modelling coastal waters

For a variety of reasons coastal regions are important for many customers.  Seas immediately adjacent to coastlines are difficult for the oceanic models (NEMO) to analyse or forecast, so coastal areas are dealt with by FLake as if they were salty water lakes.  Heat, moisture and momentum fluxes are evaluated according to the proportion of the area of the grid box that is covered by open water defined by the Land-Sea Mask.  Where there is:

  • more than 50% sea water cover then fluxes are dealt with by NEMO.  A mixture of land and ocean 'tiles' is not allowed, i.e. a grid box is considered 100% ocean.
  • more than 50% water cover, but the bodies of water are classed as lakes (e.g. large estuaries), then the HTESSEL 'tile' is "lakes and coastal waters" and fluxes are evaluated by FLake.
  • less than 50% water cover, then the HTESSEL 'tile' is "lakes and coastal waters" and fluxes are evaluated by FLake (as if it were a lake).

Tides and the covering or uncovering of coastal mudflats etc. are not considered.

See also Section on Lakes and Coastal Waters