New report details AI infrastructure for Earth system predictability

Extreme weather

Credit: CC0 Public Domain

The use of artificial intelligence (AI) to help gather, understand and analyze information has the potential to revolutionize our ability to observe, understand and predict processes in Earth’s systems.

Researchers and scientists are working together to implement AI and modeling techniques such as machine learning (ML) to advance the earth and environmental sciences. In particular, a team of scientists and experts aims to integrate state-of-the-art technology into the workings of earth system models, observations and theories, as well as provide computational capabilities that can provide speed, accuracy and more accurate information, quick decision-making.

In a joint effort between the Bureau of Biological and Environmental Research (BER) of the US Department of Energy (DOE) and the DOE Advanced Scientific Research Program, as well as community experts, the Artificial Workshop for Earth Prediction (AI4ESP) Held. From October to December 2021. The five-week virtual workshop explores challenges and infrastructure developments that will integrate the integration of technological capabilities and human activities in the field and in the laboratory with computer resources. BER developed the process as a “model-experiment” or ModEx.

AI4ESP Director Nicki Hickmon said: “Effective improvements to Earth systems forecasting require radical advances across the ModEx environment. To work together towards an understanding of the progress needed. ” Deputy Director of Operations for the DOE Atmospheric Radiation Measurement Office of the Scientific Utility at DOE Argonne National Laboratory.

According to a newly released report summarizing the AI4ESP workshop, the event brought together more than 700 participants from both the private and public sectors, with representatives from the Earth and Science, Environment, Computers and AI. Together, about 100 experts designed the workshop based on 156 white papers provided by 640 authors from 112 institutions around the world.

The information is compressed into 17 topics related to the combined water cycle and extreme weather phenomena in that cycle. The experts discussed nine key points related to the forecasting of the Earth system, including sessions related to the hydrological sciences of watersheds and coastal dynamics. Atmosphere, land, oceans and ice; And climate variability and extreme. Throughout the session, participants explored the potential of AI to unlock scientific discoveries using tools such as neural networks, machine learning, information, knowledge, AI architecture, and collaborative design.

In each session, researchers identified challenges that support the need for the AI ​​technology revolution and the infrastructure that can be applied to manage complex tasks in the field of environmental sciences.

Charu Varadharajan, a research scientist at the DOE Lawrence Berkeley National Laboratory, said: “Applicable under future climate regimes.” AI and data application domain. “This workshop is unique in that it discusses how AI can improve observation and theoretical models that incorporate DOE ModEx methods.”

“Workshops and reports have allowed us to develop 2-, 5- and 10-year goals for the development of an integrated framework for each focal point. Add AI4ESP mission Varadharajan added.

Experts have compiled a comprehensive list of opportunities that AI research and development can help address some of the biggest challenges facing Earth science. These challenges include managing and analyzing large data sets to increase the ability to observe and predict serious events and promote the integration of human activities into theories and models.

“One of the most exciting opportunities for modeling is the development of new hybrid models that include process-based and ML-based modules,” said Forrest Hoffman, head of the Earth Science computing team at DOE Oak Ridge National Laboratory. “These model frameworks enable the input of less-understood process data, which can improve accuracy and often lead to computational performance for earth system models that allow simulations. Do and further analysis is done within the given resource limits. ”

Workshop participants also identified a number of priorities to address computational challenges, including advances in both AI and ML, data management algorithms, and more. The results of those priorities can help to develop technology infrastructure that is more efficient, accurate, strategic and convenient and more accessible across resources.

There is also a need for program and cultural change to support a more cohesive mission across scientific and governmental agencies, as well as a trained workforce that can successfully integrate technology into their research and humanitarian activities. Experts have identified solutions that will include AI-specific research centers for environmental sciences, a framework that enables shared services across communities, and ongoing ongoing training and support missions.

Participants in the 2021 AI4ESP Workshop continue to discuss community calculation activities, including those from the American Geological Survey and the American Meteorological Association. Looking forward to more seminars and meetings in the future, collaboration, participation and further framework development will further the AI4ESP mission.

Additional information:
Nicki Hickmon et al, Artificial Intelligence for Earth System Predictability (AI4ESP) Workshop Report, (2022). DOI: 10.2172 / 1888810

Provided by Argonne National Laboratory

Excerpt: Detailed new report on AI infrastructure for Earth system forecasting (January 24, 2023) taken from January 24, 2023 from .html

This document is subject to copyright. Except for fair settlement for the purpose of private study or research, no part may be reproduced without written permission. Content is provided for informational purposes only.


Also Read :  Congress reauthorizes DOD innovation grants with new China safeguards

Leave a Reply

Your email address will not be published.