New one artificial intelligence (AI) -Drive weather can change the forecast system, researchers predicted
Researchers have dubbed the Edwark weather that generates a rapid forecast season tens of times compared to traditional forecast systems using a fraction of computing power on Thursday (March 20) in the journal on Thursday (20 March). Nature,
“The weather forecast system on which we all rely on have been developed for decades, but only in 18 months, we are able to make something that is competitive with the best of these systems, using the tenth part of the data on the desktop computer,” Richard turnerAn engineer at Cambridge University in United Kingdom, said statement,
Current weather forecasts are generated by inputing data in complex physics models, a multi-step process that requires several hours on a dedicated Supercomputer,
Aardvark weather enhances the process of this demand: machine learning model uses raw data from satellites, weather stations, ships and weather balloons to create its predictions without relying on atmospheric models. Satellite data is particularly important for model predictions, the team said.
Connected: Google creates an AI model that can predict future weather destruction
Researchers claimed that this new approach could provide large benefits in terms of cost, speed and accuracy of weather forecasts. Instead of the requirement of a supercomputer and a dedicated team, the Aardvark season can create a forecast on the desktop computer in a few minutes.
Weather forecasting pipeline replaced with AI
The team compared Aardvark’s performance to the existing forecast systems that generate global predictions. Aardvark improves US Nationals using only 8% observation data. Global forecast system (GFS) was comparable to the system and forecasts made by the United States Meteorological Service.
However, Aardvark’s spatial resolution is somewhat lower than the current forecast systems, which may make its initial predictions less relevant for hyper-local weather forecast. The Aardvark season is operated on a 1.5-degree resolution, which means that each box in each box includes 1.5 degrees of latitude and 1.5 degrees longitude. For comparison, GFS uses a 0.25-degree grid.
However, researchers also said that because AI learns from the figures fed to it, it can be sewn to predict the weather in specific aranas – such as temperature for African agriculture or air speed for renewable energy in Europe. Aardvark may include high-resolution regional data, where they are present, to refine local forecasts.
“What can these results achieve only Aardvark, it is the beginning,” study Kothor Anna allenIn the statement of the University of Cambridge. “This end-to-end learning approach can easily be applied to other weather forecasting problems, for example storms, forest fire and tornado. Beyond the weather, its applications are expanded to the widespread earth system, including air quality, ocean dynamics and the prediction of marine ice, for the forecast.”
Researchers stated that Aardvark can also support forecast centers in the world regions, with a lack of resources to refine global forecasts in high-resolution regional predictions.
“Aardvark’s success is not only about speed, it is about access,” Scott hosingAn AI researcher at the Allen Turing Institute in the UK said in the statement. “By moving the weather prediction from supercomputers to desktop computers, we can democratic the forecast, allowing these powerful technologies to provide developing countries and data-circular regions worldwide.”
