It was one of those gatherings. Someone had given a short speech about their worthy new initiative and then herded us in to a large room to enjoy warm Sauvignon Blanc. I have no idea how I got onto the mailing list for these events and I've even less of a clue how to get off it. Over the wine, I met a retired physicist who was a fellow of the Royal Society. He was rather deaf and, when I said I couldn't hear very well either, he demanded to know how we were supposed to communicate. But it turned out that he did have something interesting to say; something that I've since verified with various other sources. It was about climate change and feedback loops.
We all know that the world's average temperature hasn't changed much for ten or fifteen years. As a result, warnings about dangerous global warming appear to have been awry. Static temperatures for the last decade don't disprove the predictions of pessimistic climate models, but they do make them less likely. Admittedly, the problems with predicting climate are manifold. As well as increases in carbon dioxide, you have to factor in the ocean's capacity to absorb heat, the formation of clouds and the importance of other greenhouse gases. But the more factors you build into your model, the higher the uncertainties attached to it. Worse, climate information is very noisy. That's to say, there is a lot of natural variation that has nothing to do with long-term trends. You want your model to fit the long-term signal and not all the natural variation. As I've learnt from Nate Silver's excellent book The Signal and the Noise, there is a tendency to "over fit" models so that they accidentally reflect past noise. That makes them less good at forecasting the future.
This seems to have happened to climate models. There was relatively rapid warming in the 1980s and 1990s. In fact, there was more warming than you'd expect from rising carbon dioxide levels alone. To reflect the data, the models built in positive feedback, whereby increasing temperatures gave rise to knock-on effects that make the world even hotter. These models necessarily expected this rapid warming to continue. But it looks likely that the big increases in temperatures in the 1980s and 1990s were noise rather than signal. The cooler 2000s were just a return to a more modest trend. In that case, the models had been over fitted to the data.
Of course, warming is happening and it is caused by humanity belching carbon dioxide into the atmosphere. The question is how much affect all this greenhouse gas has. As my physicist friend explained and other studies confirm, if you strip out all the complicated feedback loops from the models, they become more accurate. We've known for a long time that if you double the carbon dioxide in the atmosphere, temperatures increase by about 1.5 degrees centigrade. And that is exactly what the data, noise and all, suggests has happened over the last century and a half. As so often in science, it turns out our best model is the simplest one. William of Ockham would have been proud (or he should, if he'd invented Ockham's razor).
All this is good news. As Matt Ridley has been saying for some time, scary increases in temperature look unlikely this century. Environmental activists who want to return us to the Neolithic won't be happy. Neither will renewable energy subsidy junkies. But the rest of us should breathe a sigh of relief. Slower increases in temperature mean we have more time to adapt. Replacing coal with gas is the most sensible option. Natural gas produces about half the carbon per unit of energy as dirty coal. We don't have to bankrupt ourselves with inefficient solar power, ugly windmills or expensive nuclear plants. In fifty years or so, we will need an alternative to gas. Until then, global warming shouldn't be a crisis. We will only make it one if we increase the cost of energy for no good reason.