User Uploaded Weather Info Imptoves Radar College App
December 15, 2020
A.I. model shows promise to generate faster, more accurate weather forecasts
Today's weather forecasts come up from some of the most powerful computers on Globe. The huge machines churn through millions of calculations to solve equations to predict temperature, wind, rainfall and other weather events. A forecast's combined demand for speed and accuracy taxes even the nigh mod computers.
The future could have a radically different approach. A collaboration betwixt the University of Washington and Microsoft Enquiry shows how artificial intelligence can analyze by atmospheric condition patterns to predict future events, much more efficiently and potentially someday more accurately than today's engineering.
The newly developed global weather model bases its predictions on the past 40 years of weather data, rather than on detailed physics calculations. The simple, data-based A.I. model can simulate a year'southward weather around the world much more quickly and almost as well as traditional weather models, past taking similar repeated steps from ane forecast to the next, according to a paper published this summer in the Journal of Advances in Modeling Earth Systems.
"Machine learning is essentially doing a glorified version of pattern recognition," said lead author Jonathan Weyn, who did the inquiry equally function of his UW doctorate in atmospheric sciences. "It sees a typical pattern, recognizes how information technology usually evolves and decides what to do based on the examples it has seen in the past forty years of data."
On the left is the new paper's "Deep Learning Weather condition Prediction" forecast. The middle is the actual weather condition for the 2017-18 year, and at right is the boilerplate weather for that 24-hour interval. Weyn et al./ Journal of Advances in Modeling Earth Systems
Although the new model is, unsurprisingly, less accurate than today'due south top traditional forecasting models, the current A.I. design uses about 7,000 times less computing ability to create forecasts for the aforementioned number of points on the globe. Less computational work means faster results.
That speedup would allow the forecasting centers to quickly run many models with slightly different starting weather, a technique chosen "ensemble forecasting" that lets conditions predictions encompass the range of possible expected outcomes for a conditions event – for instance, where a hurricane might strike.
"At that place'south then much more efficiency in this approach; that'due south what's so important nearly it," said author Dale Durran, a UW professor of atmospheric sciences. "The promise is that it could allow us to deal with predictability issues by having a model that's fast enough to run very large ensembles."
Co-author Rich Caruana at Microsoft Inquiry had initially approached the UW group to advise a project using artificial intelligence to brand weather predictions based on historical data without relying on physical laws. Weyn was taking a UW informatics course in machine learning and decided to tackle the projection.
"Later training on past weather condition information, the A.I. algorithm is capable of coming up with relationships between dissimilar variables that physics equations just can't exercise," Weyn said. "We can afford to utilize a lot fewer variables and therefore brand a model that's much faster."
To merge successful A.I. techniques with weather forecasting, the team mapped six faces of a cube onto planet World, then flattened out the cube's vi faces, like in an architectural paper model. The authors treated the polar faces differently because of their unique function in the weather as one manner to improve the forecast's accuracy.
First the authors carve up the planet's surface into a filigree with a six-sided cube (top left) and and so flatten out the six sides into a 2-D shape, like in a paper model (bottom left). This new technique let the authors employ standard machine learning techniques, developed for 2-D images, for weather forecasting.Weyn et al./ Journal of Advances in Modeling Globe Systems
The authors then tested their model by predicting the global height of the 500 hectopascal pressure, a standard variable in weather forecasting, every 12 hours for a total year. A recent paper, which included Weyn as a co-author, introduced WeatherBench as a benchmark test for information-driven atmospheric condition forecasts. On that forecasting examination, developed for three-mean solar day forecasts, this new model is one of the top performers.
The data-driven model would need more item before it could begin to compete with existing operational forecasts, the authors say, but the idea shows promise as an culling approach to generating weather forecasts, especially with a growing amount of previous forecasts and weather condition observations.
Weyn is now a data scientist with Microsoft'due south weather condition and finance division. This research was funded by the U.S. Role of Naval Research and a Department of Defense force graduate fellowship.
For more data, contact Durran at drdee@uw.edu or Weyn at jweyn@uw.edu.
Tag(s): College of the Environment • Dale Durran • Department of Atmospheric Sciences • atmospheric condition
Source: https://www.washington.edu/news/2020/12/15/a-i-model-shows-promise-to-generate-faster-more-accurate-weather-forecasts/
Belum ada Komentar untuk "User Uploaded Weather Info Imptoves Radar College App"
Posting Komentar