The September issue of Nature Astronomy focuses on an unusual topic for a specialist astronomy/planetary science journal: climate change. The issue features a number of articles that extensively quantify the greenhouse gas emissions of astronomers as they perform the typical duties of our field: doing research, taking observations, running simulations, travelling to conferences and meetings, and so on. Combined they show that astronomers contribute several times more to global warming (a manifestation of climate change) than the average global citizen, and that’s just the equivalent emission generated through work-related activities — any greenhouse gases generated through personal activities must be added on top. That may sound alarming, but astronomy is in line with many professions that are predominantly practised in developed countries, and there are certainly many that have a greater impact on the environment: a job connected to the aviation industry, for instance, is responsible for generating something like 18 times the global average. The purpose of quantifying emissions and their sources is to target the most critical areas for reductions.
The topical issue contains six articles, listed here:
There is also an Editorial (‘The climate issue’) that provides a brief summary of the messages presented in the articles. However, the subject of climate change is an important one and the articles provide some fascinating, and often ground-breaking, quantitative insight into how we contribute to climate change… hence this Community post, which can dig into some of the findings in more detail. I encourage you to read the individual articles for the full information, and important caveats, but I hope to offer some synthesis here.
Many of the articles state that in order to start combatting climate change, and reducing our greenhouse gas emissions, we need to know how much we are contributing. Quantification is key in order to identify the areas that are most damaging. The authors generally express impact in terms of ‘equivalent tonnes of CO2’, tCO2e. For some context, a flight from the US to Australia, for instance, generates around 3 tCO2e. In the two broad-view articles of the collection, Stevens et al. divide the CO2 contributions of the Australian astronomical community into several areas, and Jahnke et al. do the same for a European institute: MPIA in Germany. The other articles discuss specific areas in detail. Let’s discuss the main ones:
Electricity for supercomputing
It transpires that the principal activity that generates CO2 emissions for astronomers is not travelling to conferences (something that our community has been focusing on in light of the transition to virtual conferencing imposed upon us by the COVID-19 pandemic). The energy required for supercomputing produces substantial amounts of CO2 emissions and is therefore damaging to the environment. Stevens et al. consider the three main supercomputers used by Australian astronomers within Australia, as well as those overseas, by extrapolating the data provided by one supercomputing centre. Australian astronomers use 400 million CPU core-hours per year in total, requiring more than 13 million kWh of electricity. Each Australian astronomer then generates something in the region of 14-28 tCO2e/yr through supercomputing usage. In comparison, at MPIA the astronomers use 3.4 million kWh of electricity on supercomputing, generating ~4.2 tCO2e/yr each. The difference in per capita emission totals mostly reflects Australia’s usage of fossil fuels compared to Germany, since German astronomers do slightly more supercomputing than Australians. Supercomputing contributes close to 60% of an Australian astronomer’s emission budget, whereas in Germany it contributes less than a quarter. In order to reduce emissions, the authors suggest moving to renewable sources of energy (solar), or carbon offsetting in the short term.
Simon Portegies Zwart has been aware of the expense of supercomputing, and has an additional suggestion. In his article, he demonstrates that particularly intensive computations can be as expensive in terms of CO2 production as a rocket launch if they are programmed inefficiently. Running a stellar evolution code or a population synthesis code in a popular language such as Python can cost more than a short plane trip. Python programmers should not despair, however, but optimise their way of programming: use an appropriate language on an appropriate machine to get your result in a reasonable time at a reasonable carbon cost. Sometimes this will mean resorting to C++ or FORTRAN. Run parallelized code on an energy-efficient CPU, or even better, GPU. Consider whether you even need to run that test or parameter set.
Astronomers — like others in academia — often travel a considerable amount in a given year, to attend conferences, meetings, give colloquia and other talks, interview for jobs, observe in visitor mode, install hardware, and probably many other reasons. Flights lead the emissions inventory of MPIA, responsible for 8.5 tCO2e per researcher. In Australia, too, flights rank highly, contributing 6.1 tCO2e per astronomer. At the CFHT, a telescope on Mauna Kea, flights for observatory staff are responsible for 4.2 tCO2e per employee. Although fuel efficiency and aircraft design have improved in recent years, it is unlikely that further performance upgrades to aeroplanes will substantially reduce the emissions associated with flying. The alternatives include changing the prevalent mode of transport (e.g., switching from planes to trains, something that would be challenging for Australia or the US) or cutting back on flying, and say, preferring virtual conferencing to in-person conferences.
Leo Burtscher and colleagues have analysed some of the differences between the face-to-face European Astronomical Society meeting in 2019 and the virtual meeting in 2020 that was put in place in response to the COVID-19 pandemic. The virtual meeting saw 40% more attendees than the face-to-face one, possibly the result of removing some of the barriers (e.g., geographical, financial) to participation. The CO2 emissions of the entire 2020 virtual meeting were estimated to be well below 1 tCO2e. The previous year’s face-to-face meeting in eastern France was responsible for 1,855 tCO2e, comparable to the emissions of MPIA for ~8 months on average. Thus virtual conferencing has a clear environmental advantage, as well as for inclusion. However, face-to-face meetings have their advantages too, and the authors of the article suggest that hybrid meetings, where participants gather in regional/continental hubs to join a conference that links hubs together online may offer advantages from both formats, but be less environmentally expensive. Other suggestions involve scheduling conferences in a clustered way, attaching smaller satellite meetings to larger annual society meetings, or indeed, having virtual conferences with pre-recorded talks but live discussion sessions.
Observatories differ widely in design and construction, and their energy efficiency will to some degree depend on what era they were constructed in. Stevens et al. analyse the electricity consumption of the Australian Telescope National Facility (which includes the Australian Telescope Compact Array, the Parkes Observatory, the Mopra Radio Telescope, and the Murchison Radio-astronomy Observatory, which hosts the Murchison Widefield Array and the Australian Square Kilometer Array Pathfinder). These facilities span six decades of history, from the 1960s to just ~10 years ago, and their power sources depend on their locations: ATCA, Parkes and Mopra in New South Wales rely on mains power, whereas the more remote telescopes of the MRO are powered by solar (at the ~15% level) and diesel. Dividing their equivalent emissions amongst Australian astronomers gives 4.8 tCO2e per capita, with this being a lower limit, since some Australian telescopes and telescopes used by Australians abroad did not provide any data for the estimate.
The CFHT, constructed in the late 1970s, is considerably more expensive. For telescope operations (largely electricity), Flagey et al. estimate around 11 tCO2e per employee for 2019 (and note that the employee count here also includes administrative staff whereas the Australian count does not). Hawai’i is heavily dependent on oil (and coal) for its electricity generation, but the contribution from renewable energy sources is increasing year by year, giving hope that this emission total will reduce naturally over time.
The way forward for both the MRO and CFHT suggested by the authors is increasing the dependence on renewable energy sources, particularly solar energy. MRO already has solar panels on site, and could raise their contribution to the ~40% level. CFHT does not have this facility yet, and has limited space for their installation, but could potentially invest US$2M into a solar farm at Hale Pohaku that would pay for itself in ~7 years.
These three areas — supercomputing, air travel and observatories — are the main contributors to the carbon emissions produced by astronomers that have been identified so far. Other sources (such as heating and lighting for buildings, vehicles, food production) contribute in a small way. The articles presented in this issue have taken a small initial step to reducing the environmental impact of our profession: quantification of the sources. The next steps involve addressing the main areas and identifying the best ways to reduce the contributions. In the short term, carbon offsetting might be appropriate (i.e., directly or indirectly funding projects that will reduce the amount of greenhouse gases in the Earth’s atmosphere), along with changing our behaviours around travelling, conferencing and computing: preferring train travel to flying, holding conferences with an online component, being environmentally conscious with our programming. In the longer term, preferring renewable energy sources to fossil fuel sources looks likely to reap huge benefits, given our reliance on electricity for many of our day-to-day activities. As members of institutes and organisations, we can advocate for more environmentally friendly policies and practices.
Climate effects on astronomy
We have seen that astronomers are contributing disproportionately to the hastening of climate change through the nature of our occupations. However, we are far from the worst offenders, and there are ‘easy’ solutions and improvements within our reach. We should bear in mind that not only should we aim for improvements because climate change will harm ecosystems — including our own — on our planet, it will also harm the observational nature of our work: Faustine Cantalloube and colleagues have succinctly outlined the detrimental effects of climate change on astronomy using the VLT at Paranal Observatory as an example. These authors have studied the climate trends at Paranal over many years. There is a hint that the atmospheric water vapour content there will decrease over the coming decades, which may improve observations, but the increase in atmospheric temperature will increase the turbulence in the atmosphere above the telescopes, likely degrading seeing. Observational programmes that require excellent seeing will become harder to carry out, and data will suffer from more quality issues. The VLT dome ambient temperature systems are struggling to match the higher outside temperatures in recent years.
As we stress in the Editorial, any individual actions we take to reduce carbon emissions will pale in comparison to corporate and industrial pollution. Astronomers certainly have an “ethical obligation … that must not be ignored” (Stevens et al.) but we should not internalise environmental guilt; instead, we must call for systemic change and fight against bad practice; we must not let those governments eschewing their responsibilities (primarily China, USA, India) escape from the spotlight; we must vote wisely and spend our money judiciously. Our planet is under threat; if astronomers care as much about habitable planets as the media think we do, we should start with the one under our feet.
Image credit: University illustration / Michael Osadciw.