Researchers issue warning over the increasing carbon footprint of computational science – Physics World

Researchers issue warning over the increasing carbon footprint of computational science – Physics World

Time Stamp: July 20, 2023 8:00 AM

Green plug
Calculation costs: the researchers say that everyone involved in computational science must take responsibility for reducing greenhouse gas emissions (courtesy: iStock_alxpin)

The scientific community must act to prevent an exponential growth in the carbon footprint of computational science as the ubiquity of artificial intelligence, algorithms and data science increases. That is the warning from researchers at the University of Cambridge who say that  those involved in computational science – from individuals to institutions – should take responsibility for reducing greenhouse-gas emissions.

There has been increased interest in recent years in the climate impacts of scientific research. Much of this has focused on conferences ­– particularly emissions from attendees’ flights – and the carbon footprint of laboratories.

But there is one aspect of research that is often overlooked: high performance and cloud computing. The researchers argue that while the environmental impact of laboratories and travelling can be easier to determine, the impact of running algorithms is less clear and often underestimated.

Indeed, information and communication technologies (ICT) can have a substantial environmental impact. In 2020, for example, the ICT sector produced between 1.8% and 2.8% of global greenhouse gas emissions — more than the aviation industry (1.9%).

There have been a small number of studies highlighting the environmental impacts of computing in science, particularly in astronomy and astrophysics. For instance, research in 2020 found that the average Australian astronomer produced around 15 tonnes of CO2 equivalents per year just from their supercomputer usage – which was almost four times their annual emissions from flights.

“Science has led to great benefits to society, but this has come with a significant – and not always well understood – impact on the environment” notes Loïc Lannelongue, a mathematician and physicist who works on biomedical data science at the University of Cambridge.

Energy concerns

To tackle this, Lannelongue and colleagues have come up with a set of principles for best practices in environmentally sustainable computational science. The first step, the team say, is for those involved in computational science to take responsibility for reducing greenhouse-gas emissions. This requires transparency and for the carbon footprint of computations to be estimated and monitored.

Such steps could be achieved, the researchers say, through training, more centralised data infrastructures and hardware procurement considering equipment’s lifetime emissions as well as funding bodies requiring estimates of carbon footprints.

Computational scientists have a real opportunity to lead the way in sustainability, but this is going to involve a change in our culture and the way we work

Loïc Lannelongue

According to the authors, reducing the carbon footprint of the electricity used is one of the quickest ways to reduce computational greenhouse-gas emissions. Relocating to a different setting or country is one way to do this. For example, electricity produced in Australia is three orders of magnitude more carbon intensive than electricity in Iceland.

The researchers caution, however, that scientists should not forget about the footprint of data storage, which is often exacerbated by duplication of datasets so multiple groups can have a copy.

The team also argue that education and research are needed to raise awareness and drive innovation. Recent research has shown that different programming languages and coding practices can make to the energy efficiency of algorithms, for instance, highlighting the need for well-trained software engineers.

“Computational scientists have a real opportunity to lead the way in sustainability, but this is going to involve a change in our culture and the way we work,” says Lannelongue.

Time Stamp:

More from Physics World