SYMPHONIE-NH is the non-hydrostatic ocean model following the Boussinesq hydrostatic SYMPHONIE-2010 model developed by the Sirocco system team (CNRS & Toulouse University). Both are using an Arakawa type finite difference method for the C grid. The R&D team generally gives priority to a physically based approach of modelling (global conservation of the mechanical energy, consistency of pressure and density, accuracy of the bottom pressure torque,...) that tends to favour low order and robust numerical schemes. The principal equations of the physical engine are detailed in Auclair et al (2011), Marsaleix et al ( 2006, 2008, 2009). Most of the physical and numerical options (Non-Hydrostatic, free surface, generalised coordinates combined to an ALE method,...) are particularly suitable for the coastal area.
Webcontributors: get the background page
Contact us

SIROCCO events

November 2010 :



Notebooks       Forecast

Sources      Guide

Publications  Team

Training      PHD  

Fukushima Forecast Description
Webcontributors: get the htm file
Last update April 10

Who are we?

     We are scientists from two laboratories of Observatoire Midi-Pyrénées (Laboratoire d'Aérologie and Legos from CNRS and University ofToulouse) who develop regional oceanic models. INSU (an institute from CNRS for Universe Sciences) gave us the mission to distribute our models and to form scientists and engineers to their use. This service is named SIROCCO. SIROCCO also develops forecasting systems mainly as a decision-making support during scientific cruises. Meanwhile, we develop research activities in the frame of physical oceanography, sediment transport and marine biogeochemistry. Our research group is named POC (Pôle d'Océanographie Côtière).

      At the request of the International Atomic Energy Agency (IAEA, March 14, 2011), SIROCCO is delivering every day a real time 6-day forecast bulletin of the dispersion in seawater of radionuclides emitted by the Fukushima nuclear plant. The simulations are based on
the S2010.18 release of the 3D SIROCCO ocean circulation model. The system is operational since March 24 and the bulletin is available on an "open-access" basis since March 28.

      The model uses a stretched horizontal grid with a variable horizontal resolution: from 600m x 600m at the nearest grid point from Fukushima, to 5km x 5km offshore. The initial fields (T,S,U,V,SSH) and the lateral open boundary conditions are provided by the Mercator PSY4V1R3 system (one field per day, horizontal resolution 1/12 x 1/12 ). At the sea surface, the ocean model is forced by the meteorological fluxes delivered every 3hours by ECMWF.i The tidal forcing at the lateral open boundaries is provided by the T-UGO model, implemented for this purpose by the SIROCCO team on the Japonese Pacific coast.

Some details are given below on our methodology.


        The first and often critical step in building-up a new model implementation is to extract, rebuild, check an accurate bathymetry from which model bathymetry will be derived. In the case of the Japanese waters, the availability of high resolution bathymetric data released by the Japan Oceanographic Data Center as more or less regularly distributed point-wise values, packed in 2x2 boxes, has been a great help to reconstruct a quality ocean terrain model (see Figure 1a). As these data do not cover the full modeling area, the JODC-derived database has been merged with the GEBCO-08 30" database (30 arc-seconds resolution).

       This bathymetry has been projected on the curvilinear grid of the Sirocco model (Figure 1b)

3D view of the continental shelf bathymetry
- click to enlarge the picture -

3D view of the continental shelf bathymetry
- click to enlarge the picture -

Initialization and large scale forcing

           Our limited-area model needs initial fields and boundary conditions.We use daily outputs of sea surface height, temperature, salinity and currents given by Mercator which is a system providing analyses and forecasts in real time for the global ocean. For this application, we use fields at the resolution of 1/12° from the PSY4V1R3 prototype providing a two-week analysis and seven days of forecast.


            Global, high accuracy tidal atlases are available for the global ocean and could have provided directly the inputs needed by the 3-D circulation model. However, they usually offer tidal elevation only, i.e. tidal currents, needed for an efficient treatment of the open boundary conditions, are missing. In addition, the low spatial resolution of the global atlases usually triggers consistency issues if used at the open boundaries of an high resolution model, and thus is damaging to the model simulations. Therefore a specific, regional tidal model has been developed, which include the 3-D circulation model implementation (see Figure 2). It is based on the finite element model T-UGOm, routinely used in tidal applications. Thanks to an embedded quasi-linear spectral solver, T-UGOm can rapidly reconstruct adequate open boundary conditions (.i.e. consistent with the local bathymetry and mesh resolution of the regional model) to feed its sequential-in-time solver. As the region of study do not include shelf seas, a limited spectrum of about 20 tidal constituents has been modeled, from which the tidal forcing has been extracted for the 3-D circulation model ( Symphonie).

Regional tidal atlas validation
The T-UGOm tidal atlas, created to feed the SYMPHONIE coastal ocean circulation model with accurate tidal boundary conditions, has been validated against tide gauge data. In addition to this routine validation, a special effort has been made by the CTOH/INSU service ( to process satellite altimeter data from the joint CNES/NASA missions Topex/Poseidon and Jason-1/2 to build sea level time series with a 6km along-track sampling. These time series have been harmonically analyzed and corrected from solid and loading tides to extract the ocean tides constants. On Figure 6, a zoom of the M2 amplitude has been displayed with the misfits computed between the altimeter-derived data and the M2 regional solution. Misfits are less than one centimeters, except in the Kuroshio path where its meso-scale dynamics interfers with the tidal sea level signal. This is due to the fact that true tidal frequencies are aliased in the lower part of the spectrum because of the temporal under-sampling by the satellite instruments. Apparent period of M2 tide is about 62 days in the T/P and Jason (~10 days repetitivity). The methodologies and softwares deployed for tidal simulation and analysis have been mostly supported by the CNES through the OSTST science working group and COMAPI project funding.


Atmospheric forcing

      The 3D model is forced at the surface by air/sea fluxes calculated from air parameters as well as radiative fluxes and precipitation given by the ECMWF (European Centre for Medium-Range Weather Forecasts). Every day, a six-day forecast is downloaded. The spatial resolution is about 25km and the time resolution is 3 hours. As the contaminated water is emitted from the shoreline at Fukushima nuclear plant, its advection by the ocean dynamic is controlled by the coastal circulation, and thus might be sensitive to the short scales in the surface wind forcing. The Sirocco group has decided to examine the feasability of using a regional downscaling. A forecasting system has been rapidly implemented, based on the WRF model at a 10km resolution, with NCEP initialization and lateral boundary conditions. A forecasting exercise has been performed for the March 17th to March 24th period. As it can be seen in Figure 7, large scale patterns in ECMWF and in the regional WRF are very similar, but interestingly the regional atmospheric model displays significant short scales structures. Because further validations of the regional forecasts are still needed, it has been decided to maintain the ECMWF products for the atmospheric forcing in our operational system. However, in the perspective of contaminated water dispersion re-analysis, the question of atmospheric downscaling could be examined in collaboration with the atmospheric modeling expert groups.

Forecast protocol

The 3D model SYMPHONIE is at the heart of our modelling system. It receives the different forcing listed before. Every wednesday, a hindcast run of the previous seven days (red segments on Figure 8) is carried out followed by a six-days forecast. Every other days, a six-day forecast (blue segment) is run begining from the forecast of the day before. This protocol allows us using always the more accurate information available. Dynamic fields (temperature, salinity, currents) are archived every 6 hours. Afterwards, these three-dimensional currents are used to disperse radionuclides following different

Scenarios for radioactive tracers

The Fukushima nuclear plant has injected in the atmosphere and in the seawater radionuclides at different times and in variable quantities. A model of dispersion needs a good knowledge of the source terms to be able to correctly calculate the dispersion. Of course, very few information is available to build elaborate scenarios of radionuclides emssion. Besides, the evolution of dispersion in the sea also requires to know very well the behaviour of the radionuclides, for example the fraction which is dissolved in the seawater, the particulate fraction and its associated sedimentation velocity. Even if we knew all this information, radionuclides can aggregate with marine particles and then their sedimentation velocity can evolve. Finally, the oceanic currents computed by our model are not the reality: they are the result of mathematical equations too simple to fully represent the complexity of nature. The wind which strongly drives the oceanic currents is also a forecast whose accuracy is not known.

        We do not know how much radionuclides have been injected, when they have been injected and how they behave once they reach the sea. That is why we do not claim that our simulations are able to provide an accurate quantification of radioactivity in the sea. However, in order to build our scenarios, special attention has been paid to the measurements of Cesium 137 concentration taken several times every day by TEPCO at 30 and 300m in front of the nuclear plant.

Two sources of radionuclides are considered. One corresponds to a direct emission in the sea in front of the nuclear plant (migration of water contaminated by the reactors), the other one corresponds to fallout of atmospheric particles. In the first case, we introduce a flux at the grid point corresponding to the nuclear plant. This flux is adjusted to produce a concentration close to the values of Cesium 137 measured (see our page of validation). In the last case, we used 1-h outputs of the atmospheric transport model Polyphemus/Polair3D (0.25 horizontal resolution) (pers. comm. Marc Bocquet & Victor Winiarek, Ecole des Ponts ParisTech/CEREA).
For each source (direct release and atmospheric), we consider two cases: one corresponds to dissolved elements, the other one to particles that fall into the sea with a velocity of 10meters per day. Obviously dispersion in the first case will happen at larger scale than for the second case for which deposition of particles on the sea floor reduces the dispersion. Deposition of particles is cumulated over time and will be mapped later on.

Click to enlarge the picture

Figure 1a: Regional bathymetry. 500m resolution data have been collected on 2x2 boxes from the Japan Oceanographic Data Center.

Click to enlarge the picture

Figure 1b: bathymetry projected on the curvilinear grid of the SIROCCO model

Click to enlarge the picture

Figure 2: T-UGOm finite element mesh. Coastal resolution is a about 500 meters in Fukushima neighborhood and along small islands, 2 kilometers eslsewhere. Open ocean resolution is about 15 km, except on steep bathymetric slopes where it increases up to a few kilometers.

Click to enlarge the picture

Figure 3: M2 tide amplitude (cm) frome T-UGOm simulation. Amplitude tends to amplify from North-east to South-west. The 140 longitude seamounts barrier effects are clearly visible in the tidal amplitude.

Click to enlarge the picture

Figure 4: Mean tidal transport after 1 year of T-UGOm integration. The Northern mean tidal transport follows the coast toward the South of the island and disaggregate in the vicinity of the Tokyo Bay, where it merges with off shore transport cells due to the presence of the 140 longitude seamount chain.

Click to enlarge the picture

Figure 5: Mean tidal current after 1 year of T-UGOm integration (m/s). The Northern mean tidal current follows the coast toward the South of the island and disaggregates in the vicinity of the Tokyo Bay. Off shore cells are due to the presence of the seamount chain.

Click to enlarge the picture

Figure 6: M2 regional tide validation: map of the amplitude (in centimeters) and misfits with altimetry-derived tidal constants (proportional to red dots size). Data have been obtained from the harmonic analysis of the Topex/Poseidon and Jason-1 time series (~17 years). The altimetry data used in this work were developed, validated, and distributed by the CTOH/LEGOS, France ( The mean misfit level is about one centimeter, i.e. close to the data accuracy itself. It is less than a few millimeters in the Fukushima region. The larger misfits levels correspond to regions where the presence of strong meso-scale circulation tends to degrade the tidal analysis accuracy.

Figure 7: ECMWF surface wind forecast (left) and regional WRF forecast (right, courtesy of P. Marchesiello) March 18th, 00:00:00. ECMWF and regional WRF fields have similar large scales pattern, WRF downscaling shows small scales structures not present in ECMWF fields.

Click to enlarge the picture

Figure 8: Protocol of simulation

  DOWNLOAD the last SYMPHONIE release

REAL TIME FORECAST in North-Western Mediterranean (CASCADE projet)

REAL TIME FORECAST of the Pacific coast of Japan

PhD proposition: Budget of biogenic elements on the Gulf of Lion shelf and offshore transport. More details

Energy transfers in internal tide generation, propagation and dissipation in the deep ocean
Floor J.W., Auclair F., Marsaleix P., 2011 Ocean Modelling

The energy transfers associated with internal tide (IT) generation by a semi-diurnal surface tidal wave impinging on a supercritical meridionally uniform deep ocean ridge on the f-plane, and subsequent IT-propagation are analysed using the Boussinesq, free-surface, terrain-following ocean model Symphonie. The energy diagnostics are explicitly based on the numerical formulation of the governing equations, permitting a globally conservative, high-precision analysis of all physical and numerical/artificial energy transfers in a sub-domain with open lateral boundaries. The net primary energy balances are quantified using a moving average of length two tidal periods in a simplified control simulation using a single time-step, minimal diffusion, and a no-slip sea floor. This provides the basis for analysis of enhanced vertical and horizontal diffusion and a free-slip bottom boundary condition. After a four tidal period spin-up, the tidally averaged (net) primary energy balance in the generation region, extending ±20 km from the ridge crest, shows that the surface tidal wave loses approximately C = 720 W/m or 0.3% of the mean surface tidal energy flux (2.506 × 105 W/m) in traversing the ridge. This corresponds mainly to the barotropic-to-baroclinic energy conversion due to stratified flow interaction with sloping topography. Combined with a normalised net advective flux of baroclinic potential energy of 0.9 × C this causes a net local baroclinic potential energy gain of 0.72 × C and a conversion into baroclinic kinetic energy through the baroclinic buoyancy term of 1.18 × C. Tidally averaged, about 1.14 × C is radiated into the abyssal ocean through the total baroclinic flux of internal pressure associated with the IT- and background density field. This total baroclinic pressure flux is therefore not only determined by the classic linear surface-to-internal tide conversion, but also by the net advection of baroclinic (background) potential energy, indicating the importance of local processes other than linear IT-motion. In the propagation region (PR), integrated over the areas between 20 and 40 km from the ridge crest, the barotropic and baroclinic tide are decoupled. The net incoming total baroclinic pressure flux is balanced by local potential energy gain and outward baroclinic flux of potential energy associated with the total baroclinic density. The primary net energy balances are robust to changes in the vertical diffusion coefficient, whereas relatively weak horizontal diffusion significantly reduces the outward IT energy flux. Diapycnal mixing due to vertical diffusion causes an available potential energy loss of about 1% of the total domain-averaged potential energy gain, which matches to within 0.5%, for km linearly distributed grid-levels and constant background density ?0, vertical diffusivity (KV) and buoyancy frequency (N).

Approche communautaire:








Pôle Océan & Couplages- Observatoire Midi-Pyrénées - 14 Avenue Edouard Belin - 31400 Toulouse - France