A huge BBC-backed grid computing project to predict future climate change has hit a “major error”, forcing the experiment to be restarted after two months.
The BBC Climate Change Experiment, climateprediction.net, has tapped the spare computing power of PCs owned by 200,000 volunteers who have downloaded special software. The application is a version of the Met Office’s climate model. It processes data automatically when participants’ PCs are online, sending it to researchers at Oxford University.
But the project’s principal investigator, Myles Allen, told participants that the date of the simulation had been reset to 1920 after it emerged that the model had shown the earth warming up faster than the real world did during the 20th century.
The error occurred because one of the model’s input files did not increase the amount of sulphate pollution in the atmosphere correctly, distorting the picture of global warming.
Allen told participants, “The problem was a single entry in a file header, which meant that the model started reading from the wrong point in the file. Because the data and the dates in the file were OK, the problem was far from obvious.”
Computer scientists would “have strong views about a file format that allows model dates and input-file dates to get out of synch in this way”, he admitted, noting that everyone at the Met Office “would agree”.
But he added, “These climate models are some of the most complicated pieces of software in the world, having evolved over many years, and the Met Office model was designed to be the world’s best, not to be an easy piece of software to run on a PC.”