Limits of Climate Modelling

Running climate models and interpreting their output has to be handled carefully and is beset with many challenges, such as:

Sensitivity to initial conditions: Weather prediction models are highly sensitive to initial conditions because they deal with short-term predictions. Climate models, however, deal with long-term average trends and are therefore less sensitive to initial conditions, something referred to as the butterfly effect or chaos theory. To get around this issue, climate models are run many times, or in an ensemble.

Non-linear nature of climate change: Climate change and climate models are non-linear in nature (which means that the relation between cause and effect is not proportional: a small change can lead to an abrupt and massive consequence). For example, the ice-albedo feedback, where a small rise in average global temperature leads to the melting of ice on land and in the seas, means that a darker Earth and ocean absorb more solar radiation, leading to further warming.

Parameterisations: As noted above, climate models deal with long-term prediction of averages. In a typical model, a grid is 100 km by 100 km, and a time step is 10–30 minutes. These models cover a large geographical area and therefore cannot include localised phenomena such as rainfall, ocean currents, and cloud formation. To include them in models, their aggregate impact on the grid is approximated through mathematical equations. In other words, the effect of sub-grid phenomena such as clouds and rainfall is approximated for the whole grid.

Micro-initialisation issues: These, as distinct from specifying initial conditions, require that a variable (e.g., temperature) be given a slightly different value in one grid, so that different pathways result as an outcome of ensemble climate modelling.

Initial macro-level conditions: These are basically the global properties of the Earth, ocean, and atmosphere (temperature, pressure, humidity, wind speed, surface temperature of oceans, etc.) and boundary conditions, or external ‘forcings’, such as a rise in greenhouse gases and volcanic activity, which change continuously and affect the climate system from the outside.

We have mostly discussed aspects of the atmosphere in climate models. However, a comprehensive climate model, such as Earth System Models, also has to include oceans, land, ice sheets, and glaciers. Again, the ocean and land are also divided into 3D grids, and a set of equations is solved for each grid. Here again, there are many aspects of climate that are dealt with through parameterisation. Incorporating the carbon cycle, ice sheets, and glaciers into GCMs gives us more comprehensive ESMs, as discussed above.

Even after we take into account the above issues, there is uncertainty in climate predictions from these models. Many climate scientists run hundreds of simulations with a view to reducing uncertainty. One of the largest simulation efforts was conducted by David Stainforth and his team in the Climateprediction.net project, which carried out a hundred thousand simulations. Stainforth, however, says that probability distributions suggested by such “model versions” or simulations have limited value in predicting long-term climate; at best, they put a lower bound on the uncertainty. He also warns against the large amounts of parameterisations that climate models have to resort to: they are just approximations and are not based on fundamental physics. As a result, different climate models use different approaches to parameterisation.

The Intergovernmental Panel on Climate Change uses such models, as well as the many Coupled Model Intercomparison Projects (CMIP), to get some idea about policy in various socio-economic areas. However, at the end of the day, climate models are just that: models representing reality, but not reality itself. They are not exact, but they do provide a basis to take forward research.

Conclusion

With a rise in computing power, many sophisticated and comprehensive climate models with greater resolution are being built. While these models do take forward research on the complex subject of climate science, we are still left with the basic question: what to do with the output of such models when applying them to real-world economic policies and steps to mitigate and adapt to climate change. We will discuss this in the coming articles.