Cheap and noisy chips could improve climate predictions
17 Apr 2009
As scientists start to fill out the picture of a future globe dramatically changed by catastrophic global warming, the use of climate models is increasingly important when forecasting the risks faced by various regions.
Tim Palmer and researchers at the European Centre for Medium-Range Weather Forecasts say that running simulations on cheap computer chips that produce results tainted with random noise could improve those models.
Many climate processes – such as cloud formation or the movement of air currents – are too complex to simulate exactly. Models approximate these processes over a particular region of the planet, dividing it into grids typically around 100 kilometres across.
Different models, however, often fail to precisely match up because they vary in how their approximations are built. While researchers are striving to make the models more realistic, they are limited by the processing power of the supercomputers that run climate models, Palmer says. “That determines how fine of a grid we can solve the equations on, because of the computing cost,” he says.
Try, try again
Adding a degree of randomness to a particular model and running it multiple times could provide a cheaper way to increase realism, Palmer and colleagues argue, as it could be a “poor man’s surrogate for high-resolution models”.
If multiple, slightly different runs of a model come up with the same answer, it provides a hint of the strength of a prediction, according to the team. The technique has already been shown to work for weather forecasting over periods of a few weeks.
“The time is now right to integrate this into climate models,” Palmer says.
Just generating randomness to feed such models can eat up a lot of computing power. A way around this could be to use cheap hardware – low-cost computer chips that generate output with some random noise due to the way electrons bounce through them. Essentially, those chips produce the necessary randomness for free.
“It’s very speculative,” Palmer says. “But if it can be made to work, it would make much more efficient use of power.” The idea of adding randomness into the models is “very interesting and might be helpful for some cases”, says Reto Knutti of the Swiss Federal Institute of Technology in Zurich, “but in my view it will not solve all problems.”
Randomness will work best for well-understood processes such as the movement of air currents, Knutti says. “But for a case where we don’t understand the process in the first place” – such as how climate change will affect plant growth – “it may not help”.
Journal reference: A pre-print of a paper on the proposal is available on the arXiv: 0812.1074