Newswise – A collaboration with researchers at the University of California, Los Angeles.
Being able to create accurate weather models to predict weather is essential for every aspect of the US economy from aviation to shipping. To date, weather models have been based primarily on equations related to thermodynamics and fluid dynamics in the atmosphere. These models are computationally quite expensive and typically run on large supercomputers.
Researchers from private sector companies like Nvidia and Google have started developing large-scale artificial intelligence ,Aye) models for weather forecasting, known as foundation models. Recently, scientists at the U.S. Department of Energy (DOE) Argonne National Laboratory, in collaboration with researchers Aditya Grover and Tung Nguyen at the University of California, Los Angeles, have begun investigating this alternative type of model. This model can in some cases produce even more accurate forecasts than existing numerical weather prediction models at a fraction of the computational cost.
Some of these models outperform existing models in their predictive ability out to seven days, giving scientists additional information about the weather.
Foundation models are built on the use of “tokens”, which are small pieces of information. Aye Algorithms are used to learn the physics that drive weather. There are many base models used for natural language processing, which means handling words and phrases. For these larger language models, these tokens are words or fragments of language that the model predicts in sequence. For this new weather forecast model, the tokens are instead pictures – patches of charts showing things like humidity, temperature and wind speed at different levels of the atmosphere.
“Instead of being interested in a text sequence, you’re looking at spatio-temporal data, represented in images,” said Argonne computer scientist Sandeep Madireddy. ”When using these patches of images in models, you have some notion about their relative positions and how they interact because of the tokens.”
Argonne atmospheric scientist Rao Kotamarthi said the scientific team could use significantly lower-resolution data and still make accurate predictions. “The principle of weather forecasting over the years has been to achieve higher resolution for better forecasts. That’s because you’re able to solve the physics more precisely, but of course it comes at a larger computational cost,” he said. “But now we’re finding that even at coarser resolution with the method we’re using we’re actually able to get results comparable to existing high-resolution models.”
While reliable near-term weather forecasting appears to be an achievable goal in the near term AyeTrying to use the same approach for climate modeling, which involves analyzing weather over time, presents an additional challenge. “Theoretically, the Foundation Model can also be used for climate modelling. However, there are greater incentives for the private sector to adopt new approaches to weather forecasting compared to climate modelling,” Kotamrathi said. ”Work on basic models for climate modeling will continue within the scope of national laboratories and universities dedicated to finding solutions in the interest of the general public.”
Troy Arcomano, an environmental scientist at Argonne, said one reason climate modeling is so difficult is that the climate is changing in real time. “With climate, we have gone from a largely stable state to a non-stable state. This means that all of our climate data is changing over time due to additional carbon in the atmosphere. That carbon is also changing the Earth’s energy budget,” he said. “It’s complicated to figure out numerically and we’re still looking for ways to use Aye,
The launch of Argonne’s new exascale supercomputer, Aurora, will help researchers train at much larger scales. Aye-based model that will operate at very high resolutions. “We really need an exascale machine to be able to capture a fine-grained model AyeKotamarthi said.
The research was funded by Argonne’s Leadership-Directed Research and Development program and the model was run on Polaris, a supercomputer at the Argonne Leadership Computing Facility, a DOE Office of Science User Facility.
A paper based on the study received the best paper award in the workshop.tackling climate change machine learning, The workshop was held in conjunction with the International Conference on Learning Representations 2024 in Vienna, Austria on May 10.
Argonne Leadership Computing Facility Provides supercomputing capabilities to the scientific and engineering community to advance fundamental discovery and understanding in a variety of disciplines. Supported by the U.S. Department of Energy (DOE) Office of Science, Advanced Scientific Computing Research (ASCR) program, the ALCF is one of two DOE Leadership Computing facilities in the country dedicated to open science.
Argonne National Laboratory Seeks solutions to national problems in science and technology by conducting leading basic and applied research in almost every scientific discipline. Managed by Argonne UChicago Argonne, LLC For US Department of Energy Office of Science.
US Department of Energy Office of Science is the largest supporter of basic research in the physical sciences in the United States and is working to address some of the most pressing challenges of our time. Visit for more information https://energygy.gov/sscience,
(tagstotranslate)Newswise
