12 July 2023

Top science news headlines | Phys.org


Two new AI-based weather-forecasting systems challenging the status quo

by Bob Yirka , Phys.org

Two new AI-based weather-forecasting systems challenging the status-quo
Case study of a precipitation event starting at 23:50 UTC on 25 March 2021, with a tornado outbreak across several states of Alabama, Georgia and Tennessee. NowcastNet provides the only results that have forecast skills on high-intensity precipitation and show the sharp structures of several supercells for the 3-h horizon. a, Geographic context for the predictions. b, A single prediction at T + 1 h, T + 2 h and T + 3 h lead times for different models. c, CSI neighborhood at thresholds 16 mm h−1 and 32 mm h−1. d, PSD at different wavelengths. Images are zoomed in 768 km × 768 km to highlight local details. Precipitation data obtained from the MRMS26 dataset and maps produced with cartopy and Natural Earth. Credit: Nature (2023). DOI: 10.1038/s41586-023-06184-4
Two teams of AI engineers have independently designed, built and tested two new AI-based weather-forecasting systems that have been found to be as accurate as existing systems.
The first team, made up of engineers at Huawei Cloud, in China, built a system called Pangu-Weather. It was designed to predict weather a week in advance. The second team, with engineers from Tsinghua University, in China, working with one colleague from the China Meteorological Administration and another from the University of California, Berkeley, has built a system called NowcastNet. It was designed to predict precipitation levels for the upcoming six hours.
Both teams have published papers in the journal Nature describing their systems and how well they have done during testing trials. Imme Ebert-Uphoff and Kyle Hilburn with the Cooperative Institute for Research in the Atmosphere at Colorado State University, have published a News & Views piece in the same journal issue outlining the challengers of building AI weather-predicting systems and the work done by the teams on these two new efforts.
Currently, the most reliable form of weather-forecasting comes courtesy of numerical models that accept current weather data and apply math and physics formulas to make predictions about upcoming weather. Such systems are considered to be quite reliable, at least for major metropolitan areas—but they are CPU intensive, taking hours to calculate results.
But because weather forecasting is so important to farmers and for providing warning of dangerous storms, scientists continue to look for ways to improve prediction results. One promising area of research is AI—where instead of running formulas, systems are given historic weather data and use it to make predictions about the future.
In the first effort, the team behind Pangu-Weather, trained their system on 39 years of weather data and then asked it to make predictions based on current weather patterns. They found that it was as accurate at doing so as existing systems, and did its work in just a fraction of the time.
But it does suffer from one major drawback—it does not make any predictions about precipitation amounts. Instead, it returns estimates of temperature, , air pressure and other weather-related data. Humans are then left to make prediction estimates based on what the system shows them.
NowcastNet, on the other hand, does just the opposite—it tries to give  about the amounts of precipitation a given area will receive in the upcoming six hours. It does so using both  and physical rules. It too proved to be as accurate as conventional systems and also returned results much faster than conventional systems.
Both teams note that their systems are currently still at the test-of-principle stage but also suggest that their results thus far hint at the possibility of AI-based  predicting soon becoming the standard approach.

More information: Yuchen Zhang et al, Skilful nowcasting of extreme precipitation with NowcastNet, Nature (2023). DOI: 10.1038/s41586-023-06184-4

Kaifeng Bi et al, Accurate medium-range global weather forecasting with 3D neural networks, Nature (2023). DOI: 10.1038/s41586-023-06185-3

Imme Ebert-Uphoff et al, The outlook for AI weather prediction, Nature (2023). DOI: 10.1038/d41586-023-02084-9

Journal information: Nature 

More news stories

New biodegradable plastics are compostable in your backyard

We use plastics in almost every aspect of our lives. These materials are cheap to make and incredibly stable. The problem comes when we're done using something plastic—it can persist in the environment for years. Over time, ...

POLYMERS

JUL 10, 2023

0

0

The rise and fall of the Roman empire preserved in pollen

Sediments at the bottom of the ocean can offer a window into the past, indicating environmental conditions not just from the sea but washed in from terrestrial runoff, as well as preserving the flora and fauna of the time. ...

EARTH SCIENCES

JUL 10, 2023

0

59

Bioengineers explore why skin gets 'leathery'

Received wisdom says that staying out in the sun too long can make your skin tougher over time. Think about the "leathery" complexions of farmers, road crews and others who work long hours outdoors, or someone who spends ...

CELL & MICROBIOLOGY

JUL 10, 2023

0

24



Planetary Sciences

Scientists use NASA MESSENGER mission data to measure chromium on Mercury

The origin of Mercury, the closest planet to the sun, is mysterious in many ways. It has a metallic core, like Earth, but its core makes up a much larger fraction of its volume—85% compared to 15% for Earth.

JUL 10, 2023

0

1

Developing a human malaria-on-a-chip disease model

In a new report published on Scientific Reports, Michael J. Rupar, and a research team at Hesperos Inc., Florida, U.S., developed a functional, multi-organ, serum-free system to culture P. falciparum—a protozoan that predominantly ...

CELL & MICROBIOLOGY

feature

JUL 7, 2023

0

45

New insight into how plant cells divide

Every time a stem cell divides, one daughter cell remains a stem cell while the other takes off on its own developmental journey. But both daughter cells require specific and different cellular materials to fulfill their ...

PLANTS & ANIMALS

JUL 6, 2023

0

158

How one of nature's most fundamental molecules forms

Life runs on ribosomes. Every cell on Earth needs ribosomes to translate genetic information into all the proteins needed for the organism to function—and to in turn make more ribosomes. But scientists still lack a clear ...

CELL & MICROBIOLOGY

JUL 6, 2023

0

336

Why does matter exist? Roundness of electrons may hold clues

In the first moments of our universe, countless numbers of protons, neutrons and electrons formed alongside their antimatter counterparts. As the universe expanded and cooled, almost all these matter and antimatter particles ...

GENERAL PHYSICS

JUL 6, 2023

31

1882

at

No comments:

Post a Comment

Subscribe to: Post Comments (Atom)