The power of supercomputers that simulate weather in the United States has tripled. To these were added two high-performance computing systems developed by General Dynamics to more accurately and quickly predict hurricanes. About this on Tuesday, June 28 was reported by the National Oceanic and Atmospheric Administration (NOAA).
Last year, hurricanes that hit the south and east coasts of the United States cost the country more than 160 lives and $70 billion in damage. Due to climate change, the situation will only get worse, so new facilities are needed to predict the danger.
“This is a big day for NOAA and the state of weather forecasting. Researchers are developing new ensemble-based forecast models at record speed, and now we have the computing power needed to implement many of these substantial advancements to improve weather and climate prediction,” said the director of the National Weather Service.
General Dynamics signed a contract to develop two supercomputers, Dogwood and Cactus, in 2020. They were taken to places in Virginia and Arizona, where they will replace a couple of old systems from IBM.
Each system runs at 12.1 petaflops, or “a quadrillion calculations per second with 26 petabytes of memory.” This means three times as much computing power and twice as much memory as previous systems. Along with other systems, NOAA now owns a computing power of 42 petaflops.
With the updated systems, NOAA will be able to create more plausible models, with better resolution and realistic physics. As a result, weather forecasts need to be more accurate, and storm warnings will be received in advance.
“We are currently developing models that will provide additional time for dangerous weather events and more accurate forecasts of hurricane intensity – both in the ocean and those that hit the land, and we want to have more time before that happens,” said the director of the National Meteorological Service.