Follow this link to skip to the main content NASA Jet Propulsion Laboratory California Institute of Technology JPL HOME EARTH SOLAR SYSTEM STARS & GALAXIES SCIENCE & TECHNOLOGY BRING THE UNIVERSE TO YOU JPL Email News RSS Podcast Video
JPL Banner
Ocean Surface Topography from Space
Multivariate data assimilation in support of forecasting of the North Atlantic and Mediterranean

figure 1


P. De Mey
LEGOS, France
N. Pinardi
INGV, Italy
M. Benkiran Coframi, France
P. Faucher
LEGOS, France
E. Demirov
INGV, Italy
S. Sparnocchia
ISTT, Italy
F. Raicich
ISTT, Italy
C. Maes
LEGOS, France
E. Dombrowsky
CLS, France
E. Greiner
CLS, France
F. Auclair
LEGOS, France
M. Gavart
SHOM/CMO, France
P. Bahurel
SHOM/CMO, France
J. Verron
LEGI, France

Pierre De Mey
14, av. Edouard Belin
31400 Toulouse - France


This Jason-1 proposal deals with short-range (days) and medium-range (weeks to months) prediction
of ocean dynamical variables (typically, currents and tracers). This is done by assimilation of
altimetry and in situ data and the definition of ad hoc diagnostics. Two major applications to
existing forecasting systems are included in the proposal: the Mediterranean Forecasting System
Pilot Project (an EU project), and the French MERCATOR project.


This Jason-1 proposal deals with short-range (days) and medium-range (weeks to months) prediction
of ocean dynamical variables (typically, currents and tracers).

  1. A multivariate data assimilation code has been developed to control the trajectory of

    ocean models with altimetry as well as with other data types in near-real-time conditions.

  2. Applications to forecasting the Atlantic and Mediterranean at time scales of days to

    weeks are underway.

  3. Studies aimed at assessing the impact of various modus operandi (such as near-real-time

    operation) on prediction quality are also underway.

Europe-wide and national projects such as the Mediterranean Forecasting System Pilot
Project (MFSPP) and the MERCATOR North Atlantic prototype (MNATL) have brought together
ocean physicists, engineers, environment specialists and potential customers of ocean
products in a common, integrated effort with a view to supporting programs such as GODAE
and CLIVAR. Both projects involve a full range of data assimilation tools in R&D and
production modes, some of them developed specially for the projects.

A multivariate data assimilation method for real-time applications

The SOFA 3.0 optimal interpolation scheme, a robust, flexible tool permitting large-scale
state estimation on the basis of vertical EOFs has been devised as part of this Jason proposal
[De Mey and Benkiran, 2001]. This tool is the assimilation scheme used by MFSPP and in MERCATOR.
Several types of EOFs are being considered for SOFA, both in the Mediterranean [data-versus
model-based: Sparnocchia et al., personal communication, 2001] and in the Atlantic [isopycnal
EOFs: Faucher et al., 2001].

The Mediterranean Forecasting System 2000 pilot forecasting experiment

The Mediterranean Forecasting System (MFS) is a European project aimed at forecasting
Mediterranean circulation and ecosystem parameters at ranges from days to weeks. As part
of its first phase, a pilot forecasting experiment TOP (Target Operational Period) was
conducted in the first half of 2000. A Voluntary Observing Ship (VOS) program was conducted
in real time, and SST as well as ERS and TOPEX/POSEIDON data were transmitted to the forecasting
center. There, a 1/8°-resolution configuration of the MOM model forced by ECMWF analyzed and
forecast atmospheric fields was used to simulate the Mediterranean basin-scale circulation.

For these purposes, two implementations of the SOFA assimilation scheme are being used and
compared in the MFSPP project:

  1. The Mark-I Data Assimilation System is only univariate and was used to assimilate

    temperature profiles during TOP.

  2. The Mark-II DAS is multivariate and permits assimilation of SSH, SST and temperature

    profiles. Order reduction on the vertical is implemented as a projection onto a reduced
    state of vertical EOFs.

The real-time system runs every week, producing ocean analyses and short-term forecasts
of velocities, temperature and salinity. As of February 2001, one to two VOS tracks are
still operational about once a month and the altimetry data keep the model on track even
in this degraded mode. The system has been proven to produce useful, physically-sound
analyses and forecasts (figure 1) for potential customers, which include the Mediterranean
ocean science community, as well as an initial list of interested commercial users. The MFS
is the current leading project of the EuroGOOS Mediterranean Task Team. More information
and the latest analyses and forecasts can be found on the MFSPP website at

Figure 2 The real-time forecasting was coordinated in Bologna, while the data assimilation R&D, in particular for altimetry, the calculation of EOFs and predictability studies was
conducted in Toulouse (De Mey, Benkiran) and Trieste (Sparnocchia, Raicich). Figure 2
shows the influence of assimilation on the improvement of temperature at depth as well
as the value of using regionally-defined EOFs.

Data assimilation R&D for MERCATOR forecasting in the North Atlantic

The MERCATOR project aims to forecast the ocean circulation at ranges from days to weeks
on the global scale. It started in January 2001 with real-time analyses and forecasts of
the North Atlantic circulation in a 1/3°-resolution model and will be extended to the
Mediterranean and to the global scale in the next two years with increased spatial
resolution and sophisticated data assimilation methods.

More information on the MERCATOR project can be found in [Bahurel et al, this issue].
Analyses and forecasts appear every week on the MERCATOR website at

The 2001 system assimilates satellite altimeter data with a univariate optimal
interpolation configuration of the SOFA assimilation scheme. It runs every week,
producing ocean analyses and short-term forecasts of velocities, temperature and
salinity. Verification of the products involve comparisons with in situ and satellite
observations not used in the system as well as dynamical and statistical diagnostics.
Investigators working on this Jason proposal are currently helping to define the
multivariate system that will be put on line later this year and will assimilate
temperature profiles and altimetry in a consistent manner. This is the data assimilation
system that will assimilate the first Jason data in MERCATOR. Major upgrades are planned:
in 2002, the data assimilation configuration will include the Mediterranean and will
increase in resolution; and in 2003, the North Atlantic/Mediterranean system will be
coupled with a new global ocean prediction system. Development of the assimilation
system for GODAE is planned and should include a major upgrade to a SEEK-based algorithm.

Predictability and consistency diagnostics

This Jason proposal is centered around ocean prediction and the assimilation of altimetry
data for that purpose. To this end, specific forecast analysis diagnostics are being added
to the online MERCATOR data assimilation system. These include: ImA (information minus
analysis) diagnostics, assimilation impact diagnostics (forecast vs. hindcast), statistics
on biases, as well as internal consistency diagnostics (such as the "p/2" method, e.g.
[Talagrand, 1997]). The investigators are also currently working on the ensemble methods,
which are powerful tools for studying predictability (Ayoub PEA project, Auclair's results).


De Mey P., M. Benkiran, 2001: A multivariate reduced-order optimal interpolation method and its application to the Mediterranean basin-scale circulation. In: Ocean Forecasting, Conceptual basis and applications, N. Pinardi and J.D. Woods, Eds., Springer-Verlag, (in press).

Maes C., M. Benkiran, P. De Mey, 1999: Sea-level comparison between TOPEX/POSEIDON altimetric data and a global ocean general circulation model from an assimilation perspective. J. Geophys. Res., 104(C7), 15575-15585.

Faucher P., M. Gavart, P. De Mey, 2001: Isopycnal EOFs in the North and Tropical Atlantic and their use in estimation problems. J. Geophys. Res., (submitted).

Link to
Site Manager: Margaret Srinivasan
Webmaster: Kristy Kawasaki
JPL Clearance: CL01-1707