The effects of unexpected extreme weather can be devastating. Extreme cold – like that brought by the polar vortex – kills upwards of 500 Americans every year. Add in extreme heat, and it’s closer to 1,000 per year. With extreme weather – and in particular, dangerously hot days – increasing under the effects of a warming world, these temperature shifts can be a very serious threat.
But now, thanks to remarkable improvements in accurate and timely forecasting, we can see extreme weather events coming. The crushingly cold weather that swept across the U.S. was widely (and accurately) forecast weeks in advance, allowing residents in affected areas to better prepare and survive.
The scale of our advances in forecasting ability was highlighted in a recent paper written by Richard B. Alley, Kerry A. Emanuel, and Fuqing Zhang and published in Science. According to the authors, a modern 5-day forecast is as accurate as 1980’s 1-day forecast – and now, 9- to 10-day forecasts are accurate enough to be useful.
These forecasting improvements have also helped our ability to prepare for hurricanes. The authors note that “72-hour predictions of hurricane tracks are more accurate than 24-hour forecasts were 40 years ago[.]” In fact, even as hurricane-prone coastal populations have boomed, death tolls from extreme weather have dramatically fallen.
Forecasts are improving thanks primarily to leaps in modern technology. Satellites now closely monitor the land and atmosphere and feed that data into models that better represent our atmospheric and physical systems – models that we are able to run faster and faster as supercomputing systems improve. Just recently, IBM announced a new global weather forecasting system powered by 84 AC922 nodes (each equipped with four GPUs) that processes 10 terabytes of data per day and is capable of forecasting thunderstorms anywhere on the globe.
The authors anticipate that these improvements will continue for some time, before eventually slowing down near a theoretical 2-week constraint on accurate forecasting – but even this constraint, they say, will soften with time as we become better able to model the effects of initial conditions. As currently disparate datasets (such as ocean state) begin to interlink more heavily, models will improve even further.
“Computation is essential in everything discussed here,” the authors emphasize. “Progress will involve larger ensembles of model runs at higher resolution leading to improved probabilistic forecasts, including those of hazardous weather. This can be realized if governments maintain a steady schedule of investment in high-speed computing, recognizing the strong evidence that such investments will be repaid many times over in savings to the economy.”
About the paper
The paper discussed in this article, “Advances in weather prediction”, was authored by Richard B. Alley, Kerry A. Emanuel, and Fuqing Zhang. It can be found as a publicly-available publication of Science and appears in the Jan. 25, 2019 issue.