Enter the maze

A tornado in Texas: weather, climate and chaos

Andy Russell tells us about the computer science in climate and weather

a tornado touches down in a field

Let’s start with a paradox: of the two greatest and best-known computer experiments in the science of weather and climate, one’s about how you can’t predict the future and the other’s a prediction. There’s actually no paradox here and I can show why not by describing both of these experiments. I know them really well as it was studying their smart maths and understanding the importance of the science early in my education that made me want to work on weather and climate. The first showed that weather was fundamentally unpredictable beyond a few days and the second has shown, beyond much doubt, that human activity will have a worrying impact on our future weather.

Beautiful accidents

You’d never have predicted the unpredictability result. It was a beautiful accident made by one man. In the early 1960s Edward Lorenz – a distinguished American weather scientist who sadly died this year – was working on a simplified model of the atmosphere on an unreliable computer. His model required input values with six decimal places each, but when repeating one of his runs he tried to save a little time by shortening the six decimal places to three. It made a seemingly insignificant difference of less than 0.1% to his numbers. However, the change in the result was much, much bigger than the change in the initial values. After first thinking that this was a fault with the computer, Lorenz eventually ruled this out and published one of the most important results in what would become chaos theory.

Chaos is often summarised by the analogy that the flap of a butterfly's wings in Boston could set off a tornado in Texas. This catches the essence of chaos: in a chaotic system a small change in one thing has the power to later produce a very big change in the results. The choice of a butterfly in the analogy was not an arbitrary one – if all the results from a loop of Lorenz's model are plotted out in 3 dimensions then they look remarkably like a butterfly! Chaos is also found in other places, like the maths of how populations grow, the ways that molecules vibrate and the movement of satellites in the solar system.

Experiments around the world

The second important computer experiment I want to look at is the input of climate scientists to the Intergovernmental Panel on Climate Change, or IPCC for short. The IPCC produced their first major report on the state of the climate in 1990 and have written another 3 increasingly large ones since then. The biggest issue the IPCC reports look into is how we think our climate will change in the future. In the climate predictions section of the latest report, a collection of identical climate experiments were run by different institutions around the world using their different climate models. Some of the most powerful computers in the world were used to run the most advanced climate models (up to millions of lines of code) in order to make these predictions. Running so many experiments allows the IPCC to make an overall prediction as well as a calculation of uncertainty or error. Despite not being very sexy science, the description of uncertainty is a big deal. It’s important for scientists to know how accurate their predictions are likely to be.

Not for the faint of heart

I think that it’s very brave of these scientists to step up to the challenge and make predictions of such importance, predictions that will be tested within our lifetime. However, we mustn’t forget that these are models and not the real world. There’s a subtle but important point here: they produce real forecasts for a simplified system not simplified predictions for the real system. Nevertheless, they still combine our best understanding of the climate system and shouldn’t be taken lightly. Even so, many people who want us to ignore climate change have professionally and personally attacked some of these scientists and we should be thankful that most of them persevere in their science despite these attacks.

Climate v weather

So, back to that first paradox. Given that we know weather is chaotic, are the IPCC results good enough for governments to make big policy decisions based on them? The answer is yes but to justify this I must explain why climate is predictable over long time periods whereas weather is not. To do this we can use an analogy of a car driving on a motorway. Looking on from the side of the road, it would be very difficult to predict what a single car's top speed would be during its journey – the motorist could, for example, be a ‘Sunday driver’ or a ‘boy racer’. However, it would be much easier to predict the average top speed of all the cars that were on the motorway on any day – it would be around 70mph. Similarly, predicting the weather for a certain day in the year 2050 would be very difficult but predicting how much, on average, the greenhouse effect will increase our temperatures between now and then would not be so difficult.

Best behaviour

The IPCC has done a great job, not only by giving us predictions for the future of our climate, but also by being clear on how accurate it’s likely to be. Right now climate scientists want to acquire more and more powerful computers to improve the predictions even further. This is an important direction to move in but, for the time being, there’s enough evidence for governments and individuals to start improving our environmental behaviour without knowing the exact weather we have in store. Because, moving back to our analogy, failing to reduce the greenhouse gases in our atmosphere would be like removing the motorway speed limit.

Andy Russell is a NOISEmaker, and campaigns to let teenagers know the joys of a science career. Find out more at www.noisemakers.org.uk.