Demystifying Climate Models

Andrew Gettelman, Ricky Rood


What is a climate model?

A climate model is a numerical representation of the earth, focusing on the parts that effect climate (the distribution of weather). The atmosphere, land, ocean and sea ice are all simulated and work together.

What does coupled mean?


A coupled climate model usually has different components of the climate system working together. The atmosphere, land, ocean and sea ice are all simulated with computer code for each ‘component’ and then they talk to each other (coupling).

How certain are climate model projections?

Climate model projections are conditional on uncertainty in estimates (or scenarios) how much total greenhouse gases humans will emit. This is called scenario uncertainty. They are also uncertain due to the current state of the climate system (initial condition uncertainty) and because of errors in models (model uncertainty). The uncertainty changes with scale and time. Next week (or even next year) depends on the initial conditions, the next few decades on model uncertainty, and the end of the century on scenario uncertainty.

Where are climate models projections good?

For large scale properties, like the global average temperature, there are good constraints of energy and mass that help us narrow the possible range of solution. Models are pretty good for large scale temperature (global and continential scale). We have some idea of how rainfall may change as well.

Where are climate models likely to be wrong (large uncertainty)?

Small scale effects that are infrequent, and that are not global are hard to constrain. So the frequency of extreme events (floods or droughts) or severe storms (tornadoes or hurricanes) and changes to that frequency are difficult. At large scales, we do not understand how ice sheets (Greenland and Antarctica) work very well, and this may contribute to global sea level rise.

What are the secrets to simulating the earth system?

We rely on basic laws of physics and chemistry, especially that mass and energy are conserved. This really limits the range of possible answers in many cases, and insures that each process in each part of a climate model is ‘physical’ (e.g. the mass of water is a positive number). Each part of the system has its own secrets. In the atmosphere, water moves heat around when it evaporates (takes up heat) and condenses (releasing that heat) somewhere else. In the ocean, the currents (motion of water) are driven by wind at the surface, and by the density of water. Density changes with temperature and salt content. On the land, plants are responsible for moving water from deep in the soil to their leaves and stems where some is lost (transpired) as a byproduct of photosynthesis. This is a big part of evaporation.

What is the difference between weather models and climate models?

A weather model and a climate model use similar methods. Weather models are run for short periods of time (days) and the most important part is getting the initial state right to start the model. Weather models are highly dependent on initial state. Climate models are more concerned with the distribution of weather states, so they are run for a long time to generate statistics. These statistics or distributions are not dependent on the initial state. Climate models must conserve energy and mass to a high degree.

What are some of the different applications for using climate models?

Climate models are used in ‘broad’ assessments of how regional climates might change (warmer, drier). Climate model output can also be used specifically: what is the projected future precipitation over a watershed? Using models also requires evaluation of the simulation of the current climate. The danger is that the coarse horizontal spacing in a model means one value represents a large area, and models usually do not capture local variations in the effects of mountains, or variations in precipitation on small scales (less than 20-100 miles).

How do we evaluate climate models?

Climate models are usually compared to observations. Either at a single location (a station with weather records or balloon profiles over many years) or compared to satellite data. Usually many aspects of a model are evaluated: temperature and precipitation patterns. Some evaluations are on the results of models (temperature) and some evaluations use processes in the model (how clouds form, the size of cloud drops) to compare to observations.

What are the biggest uncertainties in climate models?

Clouds and how they will react to climate change are the biggest uncertainty: clouds cool a lot, and small changes to clouds may have big impacts on climate. Cloud changes are not a direct response to climate change: cloud react as their environment changes.