On this day in history...
...in 1996, a paper by Gurvan Madec and Maurice Imbard of the Laboratoire d'Océanographie Dynamique et de Climatologie at the Université Paris VI was published on May 1 in Climate Dynamics. Its title was modest: "A global ocean mesh to overcome the North Pole singularity." Its subject appeared narrow: a mathematical method for constructing a global ocean grid that avoids the coordinate singularity that arises when a regular latitude-longitude mesh converges at the North Pole. But the solution it described, known ever since as the ORCA grid, became the geometric foundation on which NEMO was built, and NEMO became the ocean model that today underpins the climate projections of the IPCC, the operational forecasts of the Copernicus Marine Service, and the ocean component of the Earth system models used by the Met Office, ECMWF, CMCC, CNRS, and institutions across Europe and beyond.
To understand what the paper solved, it helps to understand the problem. Any numerical ocean model requires a computational grid that divides the ocean into discrete cells. The most natural choice is a regular latitude-longitude grid, where lines of constant latitude and longitude form the cell boundaries. But such a grid has a fundamental geometric defect: as lines of longitude converge toward the poles, the grid cells become vanishingly small near the North Pole, and the time step required for numerical stability shrinks proportionally. For a model of the Southern Hemisphere ocean, this is not a serious problem: the South Pole lies on the Antarctic continent and is simply not part of the ocean domain. But the North Pole lies in the middle of the Arctic Ocean, which is dynamically connected to the North Atlantic and the global thermohaline circulation. Any global ocean model that takes the Arctic seriously cannot simply exclude the North Pole, and any model that includes it faces a severe computational penalty. Madec and Imbard's solution was to deform the grid: instead of placing the poles of the coordinate system at the geographic poles, they moved them onto land, one onto the North American continent and one onto Siberia, using an analytical construction in the stereographic polar plane that produced a smooth, continuous, orthogonal curvilinear mesh covering the entire global ocean without singularities. The equator remained a grid line, preserving the numerical accuracy of equatorial dynamics. The Bering Strait could be opened without special treatment. And the computational cells remained reasonably uniform in size across the entire domain, enabling a single consistent time step for global integrations. The result was what they called the ORCA grid.
The ORCA grid was not itself an ocean model: it was a geometric infrastructure on which a model could be built. The model was OPA (Océan PArallélisé), a primitive equation ocean general circulation model that had been under development at the Laboratoire d'Océanographie Dynamique et de Climatologie since the late 1980s. OPA had been used in a range of regional and global configurations, but it lacked the ability to simulate the full global ocean, including the Arctic, in a single consistent framework. The ORCA grid provided that ability. Over the years that followed, OPA evolved through successive releases, incorporating a sea ice component, a biogeochemical module, and interfaces for coupling with atmospheric models. In the early 2000s, a consortium of European institutions formalised the development around a common platform and gave it a new name: NEMO, for Nucleus for European Modelling of the Ocean. The consortium brought together CNRS and the Institut Pierre-Simon Laplace in France, the National Oceanography Centre and the Met Office in the United Kingdom, CMCC in Italy, and other European partners, creating the governance structure that has managed the model's development ever since. Gurvan Madec, the first author of the 1996 paper, has remained the central figure in NEMO's development throughout its history.
The scientific reach of NEMO is difficult to overstate. It is the ocean component of more than a dozen Earth system models that contributed simulations to the sixth phase of the Coupled Model Intercomparison Project (CMIP6), the model ensemble that underpinned the IPCC Sixth Assessment Report. It is the engine of the global ocean analysis and forecast system of the Copernicus Marine Service, which provides daily ocean state estimates and ten-day forecasts to users across Europe and worldwide. It drives the ocean forecasting systems of ECMWF's seasonal prediction suite. Its ORCA grid, at resolutions ranging from two degrees to one-twelfth of a degree, is the standard global ocean grid for a generation of European climate and operational ocean modelling. And it is the model whose output feeds OceanLive, which uses its global ocean fields to generate the daily satellite-derived maps of sea surface temperature, salinity, sea level anomaly, chlorophyll, and current speed that are the centrepiece of this ephemeris site.
NEMO's architecture has grown considerably from its OPA origins. The current model integrates the OPA ocean physics engine with the LIM and SI3 sea ice models, the PISCES and TOP biogeochemical models, and a flexible coupling interface for interaction with atmospheric, land surface, and wave models. It supports both z-coordinate and partial step vertical discretisations, a range of subgrid-scale mixing parameterisations, and configurations from idealised process studies to eddy-resolving global simulations. The development consortium releases new versions on a regular cycle, with each release incorporating advances in numerical methods, physical parameterisations, and high-performance computing performance.
The contributions of Madec and Imbard (1996) and of the NEMO framework that grew from their work span several interconnected dimensions:
- Solution of the North Pole singularity in global ocean models: The ORCA tripolar grid eliminated the numerical singularity that had constrained global ocean models to either exclude the Arctic or treat it with costly special handling, enabling the first smooth, computationally efficient global ocean simulations that include the full Arctic Ocean domain within a single consistent framework.
- Foundation of NEMO and the European ocean modelling consortium: The ORCA grid became the geometric core of OPA and subsequently of NEMO, the platform that has unified European ocean and climate modelling around a common technical infrastructure, enabling collaborative development at a scale impossible for any single institution.
- Ocean component of the CMIP6 model ensemble: More than a dozen Earth system models contributing to CMIP6, and through it to the IPCC Sixth Assessment Report, used NEMO as their ocean component, making the framework a direct contributor to the international scientific basis for climate policy.
- Infrastructure of operational ocean forecasting in Europe: NEMO underpins the Copernicus Marine Service global ocean analysis and forecast system, the ECMWF seasonal prediction suite, and the operational systems of national meteorological and oceanographic services across Europe, making it the primary tool through which European ocean science informs marine services, disaster preparedness, and resource management.
- Platform for coupled physical-biogeochemical Earth system modelling: The integration of PISCES and other biogeochemical models within the NEMO framework has made it one of the primary platforms for studying ocean carbon uptake, deoxygenation, acidification, and marine ecosystem dynamics under past and future climate scenarios.
The paper published on May 1, 1996 was eight pages long and introduced one mathematical idea: move the poles of the grid to land. It is not the kind of paper that wins prizes or generates public attention. But it is the kind of paper that makes things possible that were not possible before, and then becomes invisible precisely because what it enabled has become infrastructure. Every NEMO-based climate projection, every Copernicus Marine Service forecast, and every OceanLive map generated from NEMO output traces its global domain back to the curvilinear grid that Madec and Imbard described in those eight pages.
Sources