OIL & GAS

A new supercomputing system in the UK has achieved an 11 percent reduction in the error rate for its predictions for the northern hemisphere, a spokesperson told Government Computing News on April 27. As we reported on April 24, British scientists also say they have devised a more accurate way to forecast whether the United States is in for a particularly damaging Atlantic hurricane season. The new supercomputing model, developed by climate researchers at University College in London, measures the intensity of the trade winds and the temperature of the water in July to predict whether it is going to be a busy hurricane season and whether those storms will tend to reach land or stay out at sea. The researchers reported that the model correctly predicted the unusually active 2004 hurricane season, when Florida and other Southern states were pummeled with five hurricanes in a row. The UK Met office installed a NEC SX-6 supercomputer in spring 2004, and compares with an average error reduction rate of 3 percent in the other five major weather modelling centers in France, Germany, the US, Canada and Japan. On a global basis, the UK Met Office has achieved a 6 percent reduction in the error rate. The upgrade provided a six-fold increase in processing capacity and enabled the Met Office to use new supercomputing models with a finer resolution to predict the weather. More processing nodes were added on the NEC SX-8. They began to use a new system on April 12. As a result, the Met Office hopes this will enable it to achieve further improvements in its accuracy, but is not making any projections on the future reduction in the error rate. Also, they are working on a model that uses 12km 'boxes' from the atmosphere, that these should be reduced to 4km in the near future, and that there are long term plans to shrink them to 1km. Roger Hunt, chief operating officer at the Met Office, said: "We need to continually improve our forecasts to meet the growing expectations of the public and our other customers. The improvements will continue as we produce more and more detailed forecasts in the future."