India, May 3 -- In the last two articles, we discussed the basics of climate model building and the problems therein. We saw that one had to be cautious while applying the outputs of various climate models and their simulations. In this article, we will take a harder look at the gaps between climate models and reality: a task that is straightforward in weather prediction models, but tricky in climate models. We will also discuss how best to apply the results from climate model simulations to serious public policy questions, spanning from mitigation measures in industry and energy to adaptation measures in irrigation, flood control, and agriculture.
A Big Chasm? We saw in the earlier articles that climate models are run by 'hindcasting', i.e., by simulating the past. The model's results are then compared with recorded observations of temperature, rainfall, etc., and a close match validates the model. It tells us that the scientific basis and the parameterisation of the model are sound. Climate models can also be run with and without a rise in greenhouse gas emissions, which tells us the effect of greenhouse gases on temperature and other variables. Finally, climate models also help us distinguish natural variability from human-induced variability. But the problem with 'hindcasting', or simulating the past climate, is that models can be compared with only one climate that unfolded in the past. There is no way to analyse and compare with other possible ways in which the past climate would have unfolded. Another problem is that while assessing a climate model, the same data is used which was available to the model builder: what David Stainforth has called the 'in-sample' problem. As Stainforth says, the best we can do is be aware of the limitations of the model while using it.
However, we still have a problem: even if a model is robust in terms of yielding valid results of past simulations, we still cannot say for sure that the model will predict the future climate with the same robustness. The reasons for this are that there can be multiple pathways to the future climate. Further, since climate change is a reality, we are reasonably certain that the climate will behave very differently in the future: not only are we unsure of the nature of future climate change, we are even more uncertain about its rate. Finally, there is no way to test climate prediction decades from now: note that this is very different from 10-day weather forecasts, which can be tested easily, since observations are available in a short time.
In the Coupled Model Intercomparison Program (CMIP), many simulations of the past have been run and compared with actual observations. This has been done at both the global and regional levels. Even the global annual average temperature, or the average annual temperatures at regional levels, have seen a wide divergence between actual observations and those that have come from past simulations. It appears that climate models do not really give us accurate measures of the actual climate system. It follows that climate models, in their current form, may not help much with climate prediction. They may confirm broad trends about global warming as a result of greenhouse gas emissions, or that rainfall across the world will become more erratic in its frequency and intensity-but these are broadly what science would also tell us. When it comes to predicting precise climate events, such as average annual temperatures in a particular city or the intensity of rainfall in a coastal town, climate models may have their constraints.
There is no easy answer to this conundrum: some say that the models need to become more sophisticated, with higher resolution and greater complexity. But again, there is no consensus on where to stop. However, all is not lost: climate models can still be useful in helping answer public policy questions. We will discuss more of this in the coming articles. Conclusion Climate models have become increasingly complex over the years, with the availability of more computing power. However, climate models remain a work in progress. They have limitations in predicting the climate, and even 'hindcasting' poses its problems. Nevertheless, climate models can be useful in understanding the climate system and also provide pointers to policy issues from the many simulations of the future.
Published by HT Digital Content Services with permission from Millennium Post.