For the past 15 years, the development of Ocean Forecasting Systems (OFS) have been focusing in providing a continuous and routinely updated description of the ocean physical parameters for the past (hindcast1 and nowcast2 products), as well as in prediction mode (forecast products). Principal physical parameters of interest

1 Hindcast refers in the assimilation oceanographic community to ocean estimates obtained with an assimilated run where all observations are available, usually in delayed mode and numerical simulations performed over a past period.

2 Nowcast refers in the assimilation oceanographic community to ocean estimates obtained with an assimilated run in real time or near-real time where all possible observations are not yet available. This is the nominal "past estimates" that are provided by operational system over the previous days before the forecast.

F. Hernandez (El)

Mercator Océan/IRD, Parc Technologique du Canal, 8-10 rue Hermes,

31520 Ramonville St. Agne, France e-mail: [email protected]

A. Schiller, G. B. Brassington (eds.), Operational Oceanography in the 21st Century, 633

DOI 10.1007/978-94-007-0332-2_23, © Springer Science+Business Media B.V. 2011

being description of the water masses (temperature and salinity), of the currents (in the three dimensions), of sea level, of sea-state, of near-surface properties (like mixed layer depth, fronts) and of sea-ice. Heat and momentum exchanges with the atmosphere are also interesting meteorologists. More recently, by using coupled biogeochemical models, the ocean description is extended to ecosystem parameters from low to high trophic levels.

Due to the sparseness of available ocean observations and due to errors attached to numerical models, the OFS development have tried to integrate observation description together with modelling approaches using assimilation methods. OFS are thus composed of numerical models of ocean dynamics, possibly coupled with sea-ice dynamics models and biogeochemical models, including forcing fields, together with ocean observations collecting systems, and assimilation procedures. The performance3 of such system depends on the robustness4, accuracy5 and reliability6 of these different components. This performance is thus appreciated from a user point of view by the accuracy and usefulness of ocean products delivered routinely by OFS (hindcasts, nowcasts, forecasts) for their respective applications.

The OFS developed during the past years have first considered the ocean physical description. In many countries, local initiatives started to develop regional or coastal forecasting systems. In parallel, in the framework of GODAE (Global Ocean Data Assimilation Experiment, see, some groups and countries worked to propose basin scale, or global description of the ocean dynamics. This second kind of forecasting systems is discussed here. More specifically, are discussed here the methodology proposed to evaluate the performance of eddy-permitting to eddy-resolving systems, where diurnal cycle and ocean high frequencies are not considered. Most of these systems rely on primitive equation ocean models, where tides dynamics are usually neglected (Dombrowsky et al. 2009). During the recent years, these systems benefitted from an ocean observability never reached before: the satellite altimetry together with ARGO, buoys and drifter programs strongly enhanced the mesoscale description since 2002 (Clark et al. 2009). This observability promoted the development of state-of-the-art assimilation tools, and the implementation of mature multivariate methods (Cummings et al. 2009).

The GODAE system performance can be degraded by several causes, listed for their different components in Table 23.1. The four component listed here are different fields of ocean studies that have been usually studied separately. Thus, ocean modelling, as well as assimilation developments are usually associated with vali-

3 Performance has the same meaning than the title of this chapter, and is considered here in terms of usefulness and efficiency for users of ocean products provided by the OFS. In the framework of operational oceanography validation, a more specific definition is given later in this lecture notes.

4 Robustness (the quality of being able to withstand stresses, pressures, or changes in procedure or circumstance) is considered here in terms of OFS capacity to provide a consistent behavior and results under similar circumstances.

5 Accuracy is considered here as the degree of closeness of ocean estimates provided by the OFS to its actual true value. In the framework of operational oceanography validation, a more specific definition is given later in this lecture notes.

6 Reliability is considered here as the ability of the OFS to perform its required functions and provide ocean estimates under stated conditions while it is routinely operated.

Table 23.1 List OFS components errors, that reduce the performance and increase ocean products

Ocean model

External inputs


Assimilation method

Numerical errors

Physical parametrization and approximations (e.g., sub-grid parametrization) Explicitly not represented ocean processes (e.g., tides, diurnal cycle, surface gravity waves...) Errors on initial conditions

Errors in forcing fields (atmospheric fluxes, river run-off errors)

Bathymetry errors

Climatology errors

Boundary condition errors

Data accuracy level

Data sparseness, aliasing effects

Level of robustness of the multivariate estimation/correction Mismatch with the data representativeness Analysis shock

Level of consistency for variational techniques (linear tangent model) in highly non-linear flows errors dation studies to evaluate strength/drawbacks of new improvements. Most of the performance assessment methodologies applied in operational mode to OFS are derived from validation/evaluation techniques used separately by the research community on these components.

For years, ocean modellers were solely evaluating their numerical results by (1) internal check, looking at consistency of ocean dynamics, or sensitivity studies to some parameters; and (2) external check, through the comparison of model results to reference studies or existing observations. Then intercomparison studies were scheduled, following example from the atmospheric modelling community.

0 0

Post a comment