3. The Components of a Climate Model
A global climate model has to model each of the components of the climate system. Thus, the components of a global coupled climate model include an atmospheric model, an ocean model, a land model, and a sea ice model. In addition, some recent versions of models include a land ice model to simulate ice sheets and mountain glaciers. Since this is not in widespread use yet, these will not be described here.
Atmospheric models were the earliest models to be developed due to interest in numerical weather prediction. Both numerical weather prediction models and atmospheric general circulation models (AGCMs) used for climate modeling are designed similarly.
Atmospheric processes have to be divided into two classes—model dynamics and model physics —not to be confused with dynamics and physics. Model dynamics are processes that can be characterized by the grid-scale mean quantities, while model physics occur in spatial scales smaller than the grid box. Thus, dynamics can be resolved explicitly by the model, while physics cannot.
The part of the model code that describes the model dynamics is called the dynamical core. The dynamical core models the large-scale dynamics, or movement of air, and thermodynamics, or movement of heat, through the atmosphere. The dynamics are described by Sir Isaac Newton's second law which states that
the total acceleration of an object = the sum of forces acting on an object / the mass of the object
In the atmosphere, one can consider the object to be a parcel or small blob of air. The forces acting on a parcel include the force exerted by a horizontal pressure difference which would accelerate it towards lower pressure, gravity downwards towards the center of the Earth, and frictional drag. The rotation of the Earth also affects an air parcel. This can be explained by imagining a ball moving away from the center of a rotating plate. From above, the ball moves in a straight line, but an observer on the periphery of the plate would notice that it moved away from him/her to the right. This apparent misdirection of an object in a rotating environment is known as the Coriolis effect. A constraint on what the dynamical core does is the conservation of mass—no mass can be created or destroyed. For the atmosphere that means that the same amount of mass coming into a grid volume has to go out.
After the dynamics are completed, the model goes to the model physics. Even though these are the processes that occur at scales smaller than the model grid box, the grid mean quantities have to be used to find them. This is done through parameterizations, a set of equations that are usually empirical in nature to relate the sub-grid quantities to grid-scale quantities. The processes that are parameterized in this way include convection that generates cumulus clouds and thunderstorms, large-scale condensation and evaporation of water, conversion of cloud water droplets to precipitation, transmission and absorption of radiation through the atmosphere, the transfer of heat, water vapor, and momentum at the surface, and atmospheric turbulence.
Ocean and Sea Ice Models
Like the atmosphere, the oceans are a fluid, so dynamics plays a key role in the oceans as well. In fact, the same dynamical equations can be applied to the oceans (Washington and Parkinson 2005). Like in the atmosphere, these describe the large-scale motions. And there are processes that occur on scales smaller than the model grid box. These include eddies which mix temperature in the horizontal, vertical turbulence, and absorption of solar radiation.
However, there are some differences between ocean and atmospheric models. Besides upper and lower boundaries, there are also lateral boundaries. Plus at the upper boundary, the wind pushes the ocean currents. The ocean grid also poses a unique problem. The ocean dynamics would become unstable at the North Pole. So, the grid is adjusted so that the North Pole is located over a land mass like Greenland.
Wind also pushes sea ice around, but because sea ice floats on top of sea water, it also experiences a drag from ocean currents. The horizontal transport can cause sea ice to diverge or to come together. When sea ice collides together, it often compresses upward in ridges. Thus, there needs to be vertical transport of ice. Ice can also grow from the bottom due to freezing of sea water. The ice can also grow from above from snow falling on top. In the summer, the ice can melt from the top with the meltwater collecting in small pools called melt ponds. Thus, the thermodynamics within the ice are fundamentally important. This is complicated by the presence of brine pockets within the ice. The point of the sea ice model is to find two quantities: sea ice area or fraction and ice thickness.
The remaining model, the land model, has to model vegetation characteristics, the infiltration of soil water and the movement of groundwater horizontally, snow processes, and the diffusion of heat within the soil. The ultimate goal is to determine the exchange of heat, moisture, and momentum between the land surface and the air and the surface albedo, all of which are affected by vegetation as pointed out in the first chapter.
In order for the land surface to be modeled, it has to be characterized. How much of the surface within a grid cell is covered by vegetation? What type of vegetation is it? Such characterizations start with global land cover maps derived from satellite measurements like that of the European Space Agency (ESA). Such maps differentiate between needle-leaf and broadleaf forests, deciduous and evergreen forests, grassland, cropland, shrubland, deserts, tundra, and ice-covered areas. Each one of these is then assigned a characteristic roughness and albedo or the fraction of solar radiation incident on the surface that is reflected.
The ESA map is very high resolution, and model grid boxes are much larger. So, multiple land types have to be accounted for within a single grid box. Therefore, a land grid box has to be divided into subboxes for each land type that is contained within the whole grid box. Then, separate surface fluxes and surface temperatures are calculated for each subbox, and the grid box average is a weighted average of the subbox values.
There is added complexity in that the subboxes are inhomogeneous. Each vegetated subbox has a vegetated and a bare ground fraction. Again, separate fluxes and surface temperatures are calculate for the vegetated and bare ground fractions. The vegetation canopy is another complication. The air below the canopy is much different than the air above. Thus, the atmosphere really interacts with the canopy not the ground. Also, the seasonal cycle of the canopy has to be accounted for in deciduous forests.
While each of these components make up a coupled global climate model, each of these components except sea ice can be run offline or uncoupled. When they are run offline, the model is forced by data that represents some or all of the other components. For instance, the atmospheric model which is usually still coupled to a land model is usually forced by observed sea surface temperatures. The land and ocean models can be run completely uncoupled. They are forced from near-surface quantities that are taken from an atmospheric model run, from observations, or from reanalyses.
The purpose for running a model in offline mode is to ease a process study or sensitivity test. By taking the other components out of the picture, you standardize things, and any change in the results are due to changes made to the model component and not to the response from other components. Of course, the changes to the entire system have to be eventually simulated into the fully coupled model.
Washington, W. M., and C. L. Parkinson, 2005: An Introduction to Three-Dimensional Climate Modeling. University Science Books: Sausalito, Calif.