Forecasting models

The POSEIDON weather forecasting system

Development and implemenation: HCMR, Atmospheric Modeling and Weather Forecasting Group of the University of Athens
Chief Scientist: Dr. Anastasios Papadopoulos

The POSEIDON weather forecasting system has been developed in 1997-2000 in the framework of the POSEIDON-I project. The basic concept was to design a reliable and computationally efficient system that produces high accuracy weather forecasts, particularly useful for predicting local atmospheric conditions and forcing the wave, the ocean hydrodynamic and the ecosystem models of the POSEIDON system with surface fluxes of momentum, moisture, heat, radiation (short wave and long wave) and precipitation rates. The system with its nesting capability (Papadopoulos et al., 2002) is operational since October 1999 and provides 72-hour forecasts in two different model domains and resolutions (1/10°x1/10° - (10 km) and 1/4°x1/4° - (25 km) grid increment). The coarser model domain covers an extended area including most of Europe, Mediterranean Sea and N. Africa, while with the finer grid increment the model is integrated over the Eastern Mediterranean. Its central component is the SKIRON/Eta model (Kallos et al., 1997), which is a modified version of the Eta/NCEP model and is coupled with a dust cycle prediction model (Nickovic at al., 2001).

A major upgrade of the system has been carried out through the POSEIDON-II project (2005-2008) in collaboration with the Atmospheric Modeling and Weather Forecasting Group of the University of Athens and includes: (a) the implementation of the latest non-hydrostatic version of the SKIRON/Eta model (Janjic et al., 2001), (b) the introduction of state-of-the-art parameterisations of all the major phases of the atmospheric dust life such as production, diffusion, advection and removal related to particle size distribution (Kallos et al., 2006), (c) a 3D data assimilation package, the Local Analysis Prediction System (LAPS) to produce high resolution analysis fields. LAPS uses the 1/2°x1/2° GFS/NCEP global analysis to generate 3D first guess fields, then integrates all available real-time surface and upper air observations and finally produces a high resolution analysis for the definition of the initial conditions of the atmospheric model. For the boundary conditions the 1/2°x1/2° GFS/NCEP global forecasts are used.

The POSEIDON-II weather forecasting system (Papadopoulos et al., 2008) is operational since December 2007 and is applied on a horizontal resolution of 1/20°x1/20° (~5 km) over the domain covering the whole Mediterranean and Black Sea regions and the surrounding countries. In the vertical, 50 levels are available up to 25 mb (~25 km). It uses high resolution NCEP SST data (1/2°x1/2°) and high resolution snow depth and ice cover analysis data. The simulation period has been enlarged to 120 hours (5 days).


Janjic, Z.I., Gerrity J.P.Jr., & Nickovic, S., 2001. An Alternative Approach to Nonhydrostatic Modeling. Monthly Weather Review, 129: 1164-1178.

Kallos, G., S. Nickovic, A. Papadopoulos, D. Jovic, O. Kakaliagou, N. Misirlis, L. Boukas, N. Mimikou, G. Sakellaridis, J. Papageorgiou, E. Anadranistakis, and M. Manousakis, 1997, The regional weather forecasting system SKIRON: An overview, Proceedings of the International Symposium on Regional Weather Prediction on Parallel Computer Environments, 15-17 October 1997, Athens, Greece, 109-122.

Kallos, G., A. Papadopoulos, S. Nickovic, and P. Katsafados, 2006: “Trans-Atlantic North African dust transport: Model simulation”. Journal of Geophysical Research, 111, D09204, doi:10.1029/2005JD006207.

Nickovic, S., Kallos, G., Papadopoulos A. & Kakaliagou, O., 2001: A model for prediction of desert dust cycle in the atmosphere. Journal of Geophysical Research, 106: 18113-18129.

Papadopoulos, A., P. Katsafados, G. Kallos, and S. Nickovic, 2002. The weather forecasting system for POSEIDON-An overview. Global Atmosphere and Ocean System, 8 (2-3): 219-237.

Papadopoulos, A., P. Katsafados, E. Mavromatidis, and G. Kallos, 2008. Assessing the skill of the POSEIDON-II weather forecasting system. Abstracts books of the EuroGOOS 2008 conference.

The Mediterranean ocean circulation forecasting system

Development and implemenation: HCMR
Chief Scientist: Dr. Gerasimos Korres

The Mediterranean Sea ocean forecasting system is composed of a 1/10° resolution – 24 sigma layers Mediterranean implementation of POM model (Korres et al., 2008) and a data assimilation scheme based on the Singular Evolutive Extended Kalman (SEEK) filter (Pham et al., 1998; Hoteit et al., 2005). SEEK is an error subspace extended Kalman filter that operates with low-rank error covariance matrices as a way to reduce the computational burden. The filter uses covariance localization and partial evolution of the correction directions (Korres et al., 2008). The assimilation scheme corrects the forecast state of the model on a weekly basis. The assimilated observational data set is multivariate including AVISO sea level height, AVHRR sea surface temperature, MEDARGO floats T and S profiles and XBT data.

The Mediterranean hydrodynamical model is forced with hourly momentum, heat and freshwater fluxes derived from the POSEIDON weather prediction system based on a 1/20° ETA regional non-hydrostatic atmospheric model.

The Mediterranean model provides 5-days forecasts and additionally initial and open boundary conditions to the 1/30° POM model of the Aegean Sea.


Hoteit, I., G. Korres and G.Triantafyllou, 2005. Comparison of Extended and Ensemble based Kalman filters with low and high resolution primitive equation ocean models. Nonlinear Processes in Geophysics, 12, 755-765.

Korres, G., K. Nittis, I. Hoteit and G. Triantafyllou, 2008. A high resolution data assimilation system for the Aegean Sea hydrodynamics. J. Mar. Syst. (in press).

Korres, G., K. Tsiaras, K. Nittis, G. Triantafyllou and I. Hoteit. The POSEIDON-II system: Forecasting at the Mediterranean scale. 5th EuroGoos Conference. Exeter, UK, May 20-22, 2008.

Pham, D., Verron, J., Roubaud, M.C., 1998. A singular evolutive extended Kalman filter for data assimilation in oceanography. J. Mar. Syst. 16 (3-4), 323-340.

The Aegean Sea hydrodynamic model

Development and implemenation: HCMR
Chief Scientist: Dr. Gerasimos Korres

The Aegean Sea hydrodynamic model is based on the Princeton Ocean model (POM) and was initially developed as part of the Poseidon-I system (Nittis et al., 2006 ; Korres et al., 2002). POM is a primitive equations free surface ocean model which operates under the hydrostatic and Boussinesq approximations. The model equations are written in sigma-coordinates and discretized using the centered second-order finite differences approximation in a staggered “Arakawa C-grid” with a numerical scheme that conserves mass and energy.

The model domain covers the geographical area 19.5°E – 30°E and 30.4°N – 41°N with a horizontal resolution of 1/30° and 24 sigma layers along the vertical with a logarithmic distribution near the surface and the bottom. The model includes parameterization of the main Greek rivers (Axios, Aliakmonas, Nestos, Evros) while the inflow/outflow at the Dardanelles is treated with open boundary techniques. The Aegean Sea model is forced with hourly surface fluxes of momentum, heat and water provided by the Poseidon - ETA high resolution (1/20°) regional atmospheric model issuing forecasts for 5 days ahead.

Nesting procedures

Boundary conditions at the western and eastern open boundaries of the Aegean Sea hydrodynamic model are provided by

1) The HCMR Mediterranean model with a resolution of 1/10° and 24 sigma layers in the vertical on an hourly basis for VERSION-I of the forecasting system.

2) the Mediterranean Ocean Forecasting System model (MFS, GNOO-INGV) covering the whole Mediterranean Sea with a resolution of 1/16° and 72 levels in the vertical on a daily basis (daily averaged fields) for VERSION-II of the forecasting system.

The nesting between the two models involves the zonal/meridional external (barotropic) and internal velocity components, the temperature/salinity profiles and the free surface elevation following the nesting procedures described in Korres and Lascaratos (2003). Additionally, volume conservation constraints between the two models are applied at both open boundaries of the Aegean Sea model.

Model initialization

The Aegean Sea model is re-initialized from the HCMR Mediterranean model (Version-I) or MFS OGCM (Version-II) analysis once every week. In order to filter out spurious oscillations that may occur during the re-initialization procedure, the VIFOP optimization tool has been implemented (Auclair et al, 2000) in the forecasting system. VIFOP is a variational initialization technique based on the minimization of a cost function involving data constraints as well as a dynamical penalty involving the tangent linear model.

Basic system attributes are summarized in the following table:

Version name VI/VII
Type 3D primitive equation, finite difference, free surface (based on POM code)
Model region Aegean Sea(19.5°E -30°E; 30.4°N-41°N)
Horizontal coordinate system Type Orthogonal curvilinear – WGS84 geographical projection
Grid spacing 1/30°×1/30° - One way nested with the Mediterranean SYS2b model
Vertical coordinate system Sigma coordinates
Vertical Grid spacing 25 levels with logarithmic distribution near the surface and the bottom
Open boundary conditions In cases of outflow T,S are advected upstream while in cases of inflow T,S are directly prescribed from the coarse resolution  model. Internal velocities are directly prescribed from the parent model. Modified Flather condition for the external velocities. Zero gradient condition for sea level.

 Initialization procedure of new forecast cycle



Initialization method

Variational initialization method (VIFOP)

Surface forcings


POSEIDON non-hydrostatic atmospheric model

Grid spacing

1/20° × 1/20°


Wind speed at 10 m, air temperature at 2 m, relative humidity, net shortwave radiation, downward longwave radiation, precipitation


Every 1 hour

River runoff


Climatological runoff data for the major Greek rivers.




Auclair F., Casitas, S., Marsaleix, P., 2000. Application of an inverse method to coastal modeling. J. Atmos. Oceanic Technol., 17, 1368-1391.

Korres, G., A.Lascaratos, E. Hatziapostolou and P.Katsafados, 2002. Towards an Ocean Forecasting System for the Aegean Sea. The Global Atmosphere and Ocean System, Vol. 8, No. 2-3, 191-218.

Korres, G., and A. Lascaratos, 2003. A ïne-way nested eddy resolving model of the Aegean and Levantine basins: Implementation and climatological runs. Analles Geophysicae, MFSPP – Part I Special Issue, 21, 205-220.

Nittis, K., L.Perivoliotis, G.Korres, C.Tziavos and I.Thanos, 2006. Operational monitoring and forecasting for marine environmental applications in the Aegean Sea.

WAM based wave forecasting system

Development and Implementation: HCMR
Chief Scientist: Dr. Gerasimos Korres

The wave forecasting system was set-up as a nested configuration with a coarse grid covering the entire Mediterranean Sea at a spatial resolution of 0.1o×0.1o and a fine grid nested within the coarse grid. The domain of the fine grid covers the Aegean Sea between 30.4oN and 41oN, and between 19.5oE and 30oE at a spatial resolution of 1/30o ×1/30o  resolving the wave spectrum at each grid point in 24 directional and 30 frequency (0.05Hz->0.79316Hz) bins. The wave models are based on the WAM Cycle-4 code. The WAM model was the first model (followed by SWAN and WAVEWATCH-III models) to solve the complete action density equation, including non-linear wave-wave interactions. WAM (Cycle 4) is a third generation wave model, which computes spectra of random short-crested wind-generated waves. The WAM code can be used for shallow and deep-water calculations and can account for depth and current refraction. The following basic wave physics are accounted for in the WAM code:

  • Wave propagation in time and space

  • Wave generation by the wind

  • Shoaling and refraction due to depth

  • Shoaling and refraction due to current

  • White-capping and bottom friction

  • Quadruplet wave-wave interactions

The wave forecasting system issues wave forecasts for the next 5 days forced with hourly analysis and forecast winds produced by the POSEIDON weather prediction system.

WAVEWATCH based wave forecasting system

Development and Implementation: HCMR
Chief Scientist: Dr. Gerasimos Korres

The WAVEWATCH based Mediterranean wave forecasting system is a stand-alone model implementation covering the entire Mediterranean and the Black Sea at a spatial resolution of 1/20o×1/20o. The wave spectrum is resolved at each grid point in 24 directional and 30 frequency (0.05Hz->0.79316Hz) bins. The wave model is based on the WAVEWATCH-III code – Version V2.22 (Tolman 2002). WAVEWATCH III includes refraction and straining of the wave field due to temporal and spatial variations of the mean water depth and optionally of the mean current. The source terms include wave growth and decay due to the actions of wind, nonlinear resonant interactions, dissipation and bottom friction.

The wave forecasting system issues wave forecasts for the next 5 days forced with hourly analysis and forecast winds produced by the POSEIDON weather prediction system.



Tolman, 2002: User manual and system documentation of WAVEWATCH-III version 2.22. NOAA / NWS / NCEP / MMAB Technical Note 222, 133 pp.

Ecosystem model

Development and Implementation: HCMR
Chief Scientist: Dr. George Triantafyllou

The POSEIDON ecosystem simulation tool comprises two on-line coupled models: A hydrodynamic, based on the Princeton Ocean Model (POM) (Blumberg and Mellor, 1978) with a 1/10° resolution and 24 sigma layers, and a low-trophic biogeochemical model based on the European Regional Seas Ecosystem Model (ERSEM, Baretta et al., 1995; Blackford et al., 2004; Petihakis et al. 2002). A weekly data assimilation scheme is employed on physical parameters (Korres et al., 2008) while a second order conservative upstream advection scheme (Lin et al., 1994) is used for the biological tracers.

Following the prominent characteristics of the Mediterranean Sea ecosystem, three major functional groups are included (primary producers, heterotrophs and decomposers), providing all the necessary information for the description and analysis of carbon cycling processes, with each group further subdivided into a number of subgroups following size and/or feeding method differentiations creating a web of ten state variables. Four of those belong to phytoplankton (diatoms, nanoplankton, picoplankton, dinoflagellates), three to heterotrophs (heterotrophic flagellates, microzooplankton, mesozooplankton), one to decomposers (bacteria) and two represent non-living organic matter (particulate and dissolved). Carbon flow in the web is governed by processes operating both at physiological and community level such as growth, respiration, lysis, excretion, mortality and grazing, while nutrients are loosely coupled to carbon through a variable C:N:P:Si ratio scheme.

The simulation tool is setup with initial nutrient fields extracted from the Medatlas 2002 climatology data base ( and bilinearly interpolated into the model grid. To account for the effect of influx waters, all major riverine systems in the Mediterranean (Po, Rhone, Ebro, Nile) and N. Aegean (Axios, Strymon, Nestos, Evros, Pinios) are incorporated. Perpetual year inputs based on data provided by EU-SESAME project (dissolved inorganic nutrients) along with available data from the literature (Moutin et al., 1998 for Rhone; Degobbis et al., 1990 for Po; Skoulikidis et al., 1993 for N. Aegean rivers) are used. Additionally, Black sea water inputs at the Dardanelles straits are assigned through seasonally mean concentrations of dissolved inorganic nutrients (Tugrul el a., 2002) and annual mean values for dissolved organic matter, and ammonium (Polat et al. 1996).



Baretta, J.W., W. Ebenhoh and P. Ruardij. The European regional seas ecosystem model, a complex marine ecosystem model, Netherlands Journal of Sea Research, 33, (1995), 233-246.

Blackford, J.C., J,I,Allen, F.J. Gilbert, (2004). Ecosystem dynamics at six contrasting sites: a generic modelling study. Journal of Marine Systems, 52, 191-215.

Blumberg, A.F. and Mellor, G.L., 1978. A Coastal Ocean Numerical Model. In: J. Sunderman and K. Holtz (Editor), Mathematical Modelling of Estuarine Physics. Proceedings of the International Symposium. Springer-Verlag Berlin, Hamburg, pp. 203-214.

Degobbis, D., and M. Gilmartin. 1990. Nitrogen, phosphorus, and biogenic silicon budgets for the northern Adriatic Sea. Oceanologica Acta 13: 31-45.

Korres, G., K. Nittis, I. Hoteit and G. Triantafyllou, 2008. A high resolution data assimilation system for the Aegean Sea hydrodynamics. J. Mar. Syst. (in press).

Lin, S.J., W.C. Chao, Y.C. Sud, and G.K. Walker, 1994. A Class of the van Leer type transport schemes and its application to the moisture transport in a general circulation model, Mon. Wea. Rev., 122, 1575-1593.

Moutin T., P. Raimbault, H. L. Golterman and G. Coste, 1998. The input of nutrients by the Rhone into the Mediterranean Sea: recent observations and comparison with earlier data, Hydrobiologia 373/374, 237-246.

Petihakis, G., G. Triantafyllou, I.J. Allen, I. Hoteit, C. Dounas, Modelling the spatial and temporal variability of the Cretan Sea ecosystem, Journal of Marine Systems 36 (3-4) (2002) pp. 173-196.

Polat C. and S. Tugrul, (1996). Chemical exchange between the Mediterranean and Black Sea via the Turkish straits. CIESM Science Series No.2, Bull. De l'Institut Oceanog., 17, 167-186.

Skoulikidis, N.T., (1993). Significance evaluation of factors controlling river water composition, Env. Geo., 22, 178-185.

Tugrul, S., S.T. Besiktepe and I. Salihoglou, (2002). Nutrient exchange fluxes between the Aegean and Black Seas through the Marmara Sea, Med. Mar. Sci., 3/1, 33-42.

Oil spill fate and trajectory model

Development and Implementation: HCMR
Chief Scientist: Leonidas Perivoliotis

The POSEIDON Oil Spill fate and trajectory model is based on PARCEL model (Pollani et al. 2001) which is able to simulate not only the drift of the oil but also the chemical transformations under the specific environmental conditions. The oil slick is represented as “parcels” that have time dependent chemical and physical characteristics. The 3-D position of each parcel is calculated using advection and diffusion estimates while additional information such as wind and wave conditions and hydrological characteristics are requested for the simulation of the varying physicochemical properties.

The basic oil spill fating processes simulated by the oil-spill model are the evaporation, emulsification, beaching and sedimentation.

  • The evaporation process (transfer of oil from the sea surface to the atmosphere) influences mainly the lighter fractions of the hydrocarbons and can result to 20-40% loss of oil in the first few hours. It is affected by the surface area and thickness of the spill, the vapour pressure, the wind speed and the temperature. The method used to characterize the evaporation of the oil has been suggested by Stiver and MacKay (1984) and Stiver et al. (1989).
  • The emulsification process describes the mixing of water in the heavier fractions of the hydrocarbons and is affected by the wind and wave conditions as well as by temperature and spill characteristics (local thickness, degree of weathering etc). The numerical approach of the emulsification process that is used in the oil spill model has been proposed by Riemsdijk van Eldik et al. (1986).
  • The sedimentation process describes the trapping of oil particles that reach the sea bottom while the beaching process addresses the trapping of oil along the coast depending on the type of coastline (rocky cliff, tidal flats etc.). For beaching and sedimentation processes the model uses the Gundlach approach (1987).

The POSEIDON OSM was initially developed for the needs of the POSEIDON operational oceanography system and it was further developed and upgraded during the ROSES (2003-2004) and MarCoast (2005-2008) ESA funded projects (Perivoliotis et. al., 2006). The POSEIDON oil spill modelling system was the forecasting component of the MarCoast integrated oil spill service which has provided an operational service in the Aegean Sea during 2006 and 2007.



Gundlach E.R. (1987): Oil holding capacities and removal coefficients for different shoreline types to compute simulate spills in coastal waters, Proc. Oil Spill Conf., 1987, pp. 451-457.

Janjic, Z.I. (1994) The Step-mountain Eta Coordinate Model: Further Developments of the Convection, Viscous Sublayer and Turbulence Closure Schemes. Monthly Weather Review, 122, 927-945.

Pollani A., G.Triantafyllou, G.Petihakis , K.Nittis, K.Dounas and C.Koutitas, 2001. The POSEIDON Operational Tool for the Prediction of Floating Pollutant Transport, Marine Pollution Bulletin, Vol. 43/7-12, pp 270-278.

Perivoliotis L., K.Nittis, A.Charissi, “An integrated service for oil spill detection and forecasting in the marine environment”, European Operational Oceanography: Present and Future, Publication of the European Communities, ISBN- 92-894-9788-2, pp 381-387, 2006.

Riemsdijk van Eldik J., R.J. Ogilvie, W.W.Massie (1986): MS4: Marine spill simulation software set. Process descriptions. Dept. Civil Engineering, Delft Univ. of Technology, Delft, The Netherlands, 74p.

Stiver W. and D. Mackay (1984): Evaporation rate of spills of hydrocarbons and petroleum mixtures, Envir. Sci. Technology, 18, No 11.

Latest Data
Weather Forecast
Click for the latest data transmission
forecast over Greece


POSEIDON system funded by:

Financial Mechanism of the European Economic Area - EFTA &
Greek Ministry of National Economy


Poseidon Weather now on Android Market


Visit the Mediterranean marine world