Researchers use Nasa climate data to train AI models
GENEVA: Researchers have copied massive quantities of publicly-available climate and environmental data from Nasa onto a Swiss supercomputer to train artificial intelligence models — and for safekeeping amid US funding cuts. Switzerland’s Federal Institute of...
GENEVA — Researchers in Switzerland have moved a vast collection of publicly available climate and environmental data from Nasa to a Swiss supercomputer, saying the project will help train artificial intelligence systems and preserve the information amid concern over US funding cuts, according to a monitored public feed from Dawn World.
The transfer was announced late last week by Switzerland’s Federal Institute of Technology Zurich, or ETH, which said its researchers had copied about 100 petabytes of Nasa data to the Swiss National Supercomputing Centre in Lugano. Reto Knutti, a professor of climate physics and head of ETH’s Center for Climate Systems Modeling, said the scale of the archive is difficult to grasp. He said the process took about a year and involved roughly six billion files.
By ETH’s estimate, the amount of material is comparable in volume to around 20 million feature-length films. The datasets cover decades of observations and include information on greenhouse gases, clouds, precipitation and ice sheets, giving scientists a wide base of Earth-system data for machine learning projects.
The researchers said the material will be used to train AI models designed to support faster and more reliable forecasting in areas such as weather, climate and natural hazards. The work is part of a broader effort to make large scientific data sets easier to use for advanced computing applications.
ETH professor Thomas Zurbuchen, who previously served as Nasa’s chief scientist and initiated the data transfer, said in a statement that the US has played a central role in Earth observation for decades. He said that without the long-running measurement programmes run by Nasa and the US National Oceanic and Atmospheric Administration, the world would know much less about the planet.
The move also reflects unease among some researchers about the future of US climate science funding. Since returning to office last year, US President Donald Trump has pushed deep cuts to federal climate science and Earth-observation spending, including programmes that feed international monitoring networks, the report said.
Knutti said the team had not encountered any restriction on access so far and emphasized that the data copied to Switzerland were all publicly available. He added that the transfer was driven by practical concerns as much as scientific ones. Going forward, the researchers also plan to copy additional datasets from NOAA.
The report, based on a monitored public feed, described the project as both a research resource and a safeguard for data that scientists view as critical to understanding the planet. Knutti summed up the significance by saying: “Data is the new gold.”
Source: Dawn World - https://www.dawn.com/news/2017159/researchers-use-nasa-climate-data-to-train-ai-models
