15.6: Coupled Ocean and Atmosphere Models

Coupled numerical models of the atmosphere and ocean are used to study the climate, its variability, and its response to external forcing. The most important use of the models has been to study how Earth’s climate might respond to a doubling of CO₂ in the atmosphere. Much of the literature on climate change is based on studies using such models. Other important uses of coupled models include studies of El Niño and the meridional overturning circulation. The former varies over a few years, the latter varies over a few centuries.

Development of the coupled models tends to be coordinated through the World Climate Research Program of the World Meteorological Organization (WCRP/WMO), and recent progress is summarized in Chapter 8 of the Climate Change 2001: The Scientific Basis report by the Intergovernmental Panel on Climate Change (IPCC, 2007).

Many coupled ocean and atmosphere models have been developed. Some include only physical processes in the ocean, atmosphere, and the ice-covered polar seas. Others add the influence of land and biological activity in the ocean. Let’s look at the oceanic components of a few models.

Climate System Model

The Climate System Model, developed by the National Center for Atmospheric Research (NCAR), includes physical and biogeochemical influence on the climate system (Boville and Gent, 1998). It has atmosphere, ocean, land-surface, and sea-ice components coupled by fluxes between components. The atmospheric component is the NCAR Community Climate Model, and the oceanic component is a modified version of the Princeton Modular Ocean Model, using the Gent and McWilliams (1990) scheme for parameterizing mesoscale eddies. Resolution is approximately \(2^{\circ} \times 2^{\circ}\) with 45 vertical levels in the ocean.

The model has been spun up and integrated for 300 years, the results are realistic, and there is no need for a flux adjustment. (See the special issue of Journal of Climate, June 1998).

Princeton Coupled Model

This model consists of an atmospheric model with a horizontal resolution of \(7.5^{\circ}\) longitude by \(4.5^{\circ}\) latitude and 9 levels in the vertical, an ocean model with a horizontal resolution of \(4^{\circ}\) and 12 levels in the vertical, and a land-surface model. The ocean and atmosphere are coupled through heat, water, and momentum fluxes. Land and ocean are coupled through river runoff. And land and atmosphere are coupled through water and heat fluxes.

Hadley Center Model

This is an atmosphere-ocean-ice model that minimizes the need for flux adjustments (Johns et al, 1997). The ocean component is based on the Bryan-Cox primitive equation model, with realistic bottom features, vertical mixing coefficients from Pacanowski and Philander (1981). Both the ocean and the atmospheric component have a horizontal resolution of \(96 \times 73\) grid points, the ocean has 20 levels in the vertical.

In contrast to most coupled models, this one is spun up as a coupled system with flux adjustments during spin up to keep sea surface temperature and salinity close to observed mean values. The coupled model was integrated from rest using Levitus values for temperature and salinity for September. The initial integration was from 1850 to 1940. The model was then integrated for another 1000 years. No flux adjustment was necessary after the initial 140-year integration because drift of global-averaged air temperature was \(\leq 0.016 \ \text{K/century}\).

Comments on Accuracy of Coupled Models

Models of the coupled, land-air-ice-ocean climate system must simulate hundreds to thousands of years. Yet,

It will be very hard to establish an integration framework, particularly on a global scale, as present capabilities for modelling the earth system are rather limited. A dual approach is planned. On the one hand, the relatively conventional approach of improving coupled atmosphere-ocean-land-ice models will be pursued. Ingenuity aside, the computational demands are extreme, as is borne out by the Earth System Simulator — 640 linked supercomputers providing 40 teraflops [10¹² floating-point operations per second] and a cooling system from hell under one roof — to be built in Japan by 2003.— Newton, 1999.

Because models must be simplified to run on existing computers, the models must be simpler than models that simulate flow for a few years (WCRP, 1995).

In addition, the coupled model must be integrated for many years for the ocean and atmosphere to approach equilibrium. As the integration proceeds, the coupled system tends to drift away from reality due to errors in calculating fluxes of heat and momentum between the ocean and atmosphere. For example, very small errors in precipitation over the Antarctic Circumpolar Current leads to small changes the salinity of the current, which leads to large changes in deep convection in the Weddell Sea, which greatly influences the volume of deep water masses.

Some modelers allow the system to drift, others adjust sea-surface temperature and the calculated fluxes between the ocean and atmosphere. Returning to the example, the flux of fresh water in the circumpolar current could be adjusted to keep salinity close to the observed value in the current. There is no good scientific basis for the adjustments except the desire to produce a “good” coupled model. Hence, the adjustments are ad hoc and controversial. Such adjustments are called flux adjustments or flux corrections.

Fortunately, as models have improved, the need for adjustment or the magnitude of the adjustment has been reduced. For example, using the Gent-McWilliams scheme for mixing along constant-density surfaces in a coupled ocean-atmosphere model greatly reduced climate drift in a coupled ocean-atmosphere model because the mixing scheme reduced deep convection in the Antarctic Circumpolar Current and elsewhere (Hirst, O’Farrell, and Gordon, 2000).

Grassl (2000) lists four capabilities of a credible coupled general circulation model:

1. “Adequate representation of the present climate;
2. “Reproduction (within typical interannual and decades time-scale climate variability) of the changes since the start of the instrumental record for a given history of external forcing;
3. “Reproduction of a different climate episode in the past as derived from paleoclimate records for given estimates of the history of external forcing; and
4. “Successful simulation of the gross features of an abrupt climate change event from the past.”

McAvaney et al. (2001) compared the oceanic component of twenty-four coupled models, including models with and without flux adjustments. They found substantial differences among the models. For example, only five models calculated a meridional overturning circulation within 10% of the observed value of \(20 \ \text{Sv}\). Some had values as low as \(3 \ \text{Sv}\), others had values as large as \(36 \ \text{Sv}\). Most models could not calculate a realistic transport for the Antarctic Circumpolar Current.

Grassl (2000) found that many coupled climate models, including models with and without flux adjustment, meet the first criterion. Some models meet the second criterion (Smith et al. 2002, Stott et al. 2000), but external solar forcing is still not well known and more work is needed. And a few models are starting to reproduce some aspects of the warm event of 6,000 years ago.

But how useful are these models in making projections of future climate? Opinion is polarized. At one extreme are those who take the model results as gospel. At the other are those who denigrate results simply because they distrust models, or on the grounds that the model performance is obviously wrong in some respects or that a process is not adequately included. The truth lies in between. All models are of course wrong because, by design, they give a simplified view of the system being modelled. Nevertheless, many—but not all—models are very useful. —Trenberth, 1997.