Green Algorithms: Making computational science more environmentally sustainable
16 May 2024
A research team has developed an online calculator to help scientists easily estimate the carbon footprint of a computation.
Overview
Computational science is an essential part of modern discoveries, but it comes with significant environmental impacts.
The Green Algorithms project promotes more environmentally sustainable computational science by building tools for carbon footprint estimation, raising awareness about the environmental impacts of algorithms, and driving change in the community.
The Challenge
Scientific research and development improves the human condition and has the potential to tackle climate change, yet the computational tasks can also impact on the environment. The global carbon footprint of data centres is an estimated ~100 mega tonnes of CO2e, equivalent to the entire US commercial aviation.
In August 2023, a UK Research and Innovation (UKRI) report highlighted the need for action to reach the target of net-zero digital research infrastructures by 2040.
Scientists were becoming aware that the environmental effects of computing should be taken into account and mitigated where possible. However, there was no methodology or tool broadly available to scientists to quantify the carbon footprint of their research activities.
The Solution
Dr Loïc Lannelongue, Research Associate, University of Cambridge and Prof Michael Inouye, University of Cambridge and Baker Heart and Diabetes Research Institute, developed an online calculator to help scientists easily estimate the carbon footprint of a computation. This is based on a number of factors, including the hardware requirements of the tool, the runtime, and the location of the data centre. They then developed GAHPC, a version of the calculator tailored to high-performance computing (HPC) infrastructures, which they are now looking to build a dashboard for.
The tools use CO2 equivalent (CO2e) as an indicator of carbon impact and contextualise emissions calculations by providing ‘tree-months’: the amount of CO2 a mature tree is able to absorb in one month.
These open-source and open-access tools, developed with input from stakeholders in different areas (environmental Life-Cycle Assessments, HPC, bioinformatics, and sociology), are used to inform the design, selection, and implementation of algorithms, as well as which analyses to run.
In collaboration with HDR UK, EMBL’s European Bioinformatics Institute (EMBL-EBI), and others, the team developed the GREENER principles to make environmental sustainability a core element of research. They have also released information on ways to avoid unnecessary CO2 emissions, carbon benchmarks, and a landmark publication to increase awareness of green computation.
Dr Loïc Lannelongue, Research Associate, University of Cambridge said:
“It all started with the Australian bush fires in January 2020, which made us reflect on the climate emergency and the carbon footprint of our own work. Building the calculator, we were able to see that some bioinformatics analyses, for example, can reach dozens of tonnes of CO2e, when the IPCC target is just two tonnes per year, per person. We needed to make it easy for scientists to track their energy usage and consider ‘do I really need to run that extra test?’”
The Impact
The calculator has approximately 200 users per week and to date has had 24,000 overall users from around the world, making it easier for them to track their energy usage and implement broader reporting of carbon footprints. The Green Algorithms methodology has been adopted in a number of fields, such as Nextflow (a pipeline tool widely used in bioinformatics), and rolled out in individual institutions such as EMBL-EBI, for carbon footprint monitoring.
The team has engaged with public policy makers, by contributing to Royal Society reports. The Green Algorithms tool has also become an integral part of the decision process for a €40m funding call by the French Department for the Environment dedicated to ‘Frugal AI for sustainable development in local communities’. AI-based project proposals must include an environmental impact statement quantifying the carbon footprint for computing usage and the French Ministry has therefore made it compulsory for applicants to use the Green Algorithms tool as part of their applications.
The next step is to add new functionalities and they aim to review the effectiveness of the suite of tools and how they have changed behaviours.
Find out more at Green Algorithms