Conferences are the heart of the PL research community. The best PL research is published at conferences, following a rigorous peer review process on par or better than the process of high-quality journals. Conferences are also where science gets done. As the respective community gathers to learn about the latest results, its members also network and interact, developing collaborations or carrying on projects that could produce the next breakthroughs. At conferences, students and young professors can rub elbows with luminaries, and researchers can develop problems and exchange ideas with practitioners.
One drawback of conferences (compared to other forms of research publication) is their cost. The monetary cost is obvious: at least one of a paper’s authors must pay to attend the conference, a cost that includes registration, travel, and hotel. Another cost that is less often considered is the environmental cost. In particular, I’m thinking of the impact that travel to/from the conference has on global warming. Most conference attendees travel great distances, and so travel by airplane. But air travel is particularly bad for global warming. So I wondered: what is the cost of conference travel, in terms of carbon footprint?
To get some idea, I decided to estimate the carbon footprint of the PLDI’16 program committee (PC) meeting, held just before and at the same venue as POPL’16. The result directly sheds some light on the carbon footprint of in-person PC meetings, and by treating it as a sample of the PL community overall, sheds light on the carbon footprint of PL conferences. In this blog post I present the results of my analysis and conclude with thoughts about possible actions to mitigate environmental cost.
I asked the PLDI’16 program committee to tell me the itinerary they followed, door-to-door, to get to the PC meeting in St. Petersburg, FL, USA, which was held at the same hotel as POPL’16. The meeting was held over two days, 18-19 January 2016. The PC comprises 35 members, all of whom attended the meeting but one, due to a personal emergency; I used the itinerary he would have followed if not for the last-minute cancellation. Two student helpers also attended. About two thirds of the PC responded to me with their exact itinerary; for those who didn’t respond I used one that seemed sensible, based on their affiliation.
To compute the carbon footprint of travel, I used the calculator at carbonfootprint.com. The site states: “The online calculators on this web site follow the methodology outlined in Defra’s Voluntary Reporting Guidelines, and uses the most up to date emission factors (currently using 2015 Government Conversion Factors for Company Reporting Methodology).” 1 (Defra is a the UK Department of Environment Food and Rural Affairs.) For air travel, this calculation essentially estimates the fuel consumed for the distance traveled, with a boost for each takeoff. This consumption is then divided amongst the expected number of passengers on the plane. 2 For my calculation I assumed everyone flew economy class; the calculator attributes a higher amount of the carbon to first-class passengers because they occupy a greater percentage of the plane’s space. I enabled the option to apply a radiative forcing factor into the calculation. Doing so multiplies the carbon emissions to better reflect the fact that emissions from airplanes at high altitude have an increased effect on global warming. As a result, what is reported is metric tons of “CO2 equivalent,” or CO2e. For ground transportation, the site has options for train, tram, taxi, and personal auto travel (with the footprint divided among the expected greater number of people on a train than in a car).
The estimated total was 44.53 metric tons of CO2e for the PC meeting. Of this, 43.73 was due to air travel, and the rest was due to transport to/from the airport (whether by auto, train, etc.). Distances traveled over ground were usually short, and thus usually either 0.01 or 0.02 tons per individual. In general, commercial air travel is more carbon friendly than driving a personal automobile once you exceed a certain threshold. For example, my round-trip flight from DCA to TPA was 0.36 tons of CO2e, but if I’d driven the 1914 miles (alone) instead, it would have been 0.64 tons. The overall average per attendee was 1.20 tons. The average for North American attendees was 0.77 tons; for Europeans it was 2.32 tons; and for the lone Australian was 4.34 tons. The least CO2e was due to me, at 0.38 tons, as it turned out I was the most proximate to the PC meeting location.
As points of comparison:
- The average footprint for US residents is 20.40 metric tons per year (says carbonfootprint.com).
- The average for residents of industrial nations is about 11 metric tons per year (says carbonfootprint.com).
- I drive a 2002 Saturn SL2 about 10,000 miles per year. The carbon footprint for doing so is 3.31 tons of CO2e.
- I flew about 25,000 miles last year; tallying each trip totaled to 9.04 tons.
Scaling to a conference
What if this was a full conference, and not an in-person PC meeting? If we assume that the PC’s geographic distribution and travel habits are a good sample for the whole of the attendees to a conference, then we can scale up the result to the total number of attendees.
POPL’16 in particular had 461 attendees. Then we can use 461/37 = 12.46 as a multiplicative factor, which yields 555 tons. This probably underreports the total. For the PLDI’16 PC, about 75% of the attendees were from North America, 20% from Europe and the Middle East, and 3% from Asia and Oceania. For the POPL’16 conference, 57% were from North America, 35% from Europe and the Middle East, and 7% from Asia and Oceania. The greater representation from the latter two would account for greater distances traveled than for the PC. A rough redistribution of estimated emissions nets about 701 tons instead.
Accounting the carbon
An important question is how to account for this cost, along with the other costs and benefits of conference travel. Doing so is particularly important for climate change researchers and policymakers, whose own conference travel contributes to the condition they are warning us about; by one estimate, the 2015 Paris climate talks accounted for 21,000 tons of CO2e. I believe it is also important for any scientific body to justify the environmental impact of its activities. Here are several ideas (there are others online too).
One idea is to hold the conference as usual, but then afterwards buy carbon offsets based on the estimated carbon footprint. There are many ways to pay to offset this carbon emission, e.g., via planting trees. carbonfootprint.com offered a variety of options. For example, their Global Portfolio VC funds a range of projects that aim for carbon reduction, at the cost of $8.99 per ton. For 701 tons (the estimated footprint of POPL’16), that would be a cost of $6340.36.
Carbon-aware conference location
One could also take into account carbon footprint when choosing the conference location. As of now, I don’t believe carbon footprint is part of the process. My casual observation is that the main factors for conference choice are pre-arrangements with an eye toward serving the existing community (“twice in the USA, then once in Europe or elsewhere”), attempts to expand reach (e.g., by hosting the conference in China, India, etc.), attempts to diversify the location (don’t return to the same place twice in 10 years), and attempts to minimize monetary cost (e.g., costs of hotel rooms, conference facilities, etc.). Locations that are geographically central and well-connected by air would minimize carbon cost. For PL conferences, having a conference in the northeast USA would probably minimize carbon cost. The challenge is to figure out how much this matters in the overall deliberation about where to hold the conference.
Another idea is to reduce the number of meetings by colocating venues that are normally distinct. The easiest thing would be to colocate conferences that have a similar calendar and are likely to have repeat attendees. For PL conferences, the most obvious thing would be to attempt to co-locate ICFP and SPLASH/OOPSLA, since both happen in the Fall. But doing so will not help much if there are few repeat attendees. 3
A more drastic idea is to shift the calendar of conferences in order to co-locate them. As an extreme, consider the following: We have one large PL event that colocates the four major SIGPLAN conferences: PLDI, POPL, ICFP, and OOPSLA. Submission deadlines happen as usual during the year, and all papers accepted by a certain calendar date are slated to appear at the combined event. We could even imagine generalizing this idea so that instead of the standard submission deadlines, we allow rolling submissions, like a journal-first publishing process. Then papers can be submitted at any time during the year to the appropriate “conference,” and all of those accepted by the (standing) program committee at a certain date appear at the next mega-PL conference.
Having one big conference is standard in other communities and might even offer advantages over PL’s current approach of having many conferences throughout the year. For example, it would be convenient to consolidate trips. It would encourage more of the community to get together at once. My impression is that today, few people attend many conferences, and so there is a good chance that some of the people you’d like to meet won’t be at the conference you are attending. One mega-conference would reduce that risk.
Account for in-person PC meetings
While the carbon cost of an in-person PC meeting pales in comparison to that of entire conference, we might also want to carefully account for their cost. The above strategies apply: buy carbon offsets; hold the meeting at a geographically central location; and/or co-locate the event with another event of common interest (like PLDI’16’s PC meeting did). We might even consider dropping in-person PC meetings altogether, as some conferences have (e.g., ACM Computer and Communication Security, which is a large security conference). There are clear benefits to in-person meetings, so it would take some thought to understand how much the environmental cost counteracts them.
Conferences are a central part of the research landscape in computer science, and in programming languages especially. While their benefits are numerous and important, they do impose an environmental cost due air travel (as well as other costs). As the world grapples the threat of global warming, it expects the scientific community to set an example. As such we should at the least consider how our activities could be adjusted to reduce their environmental impact. This post presents some ideas in this direction in terms of better understanding and possibly reducing the carbon footprint of conferences.
- I also tried using the flight calculator at offsetters.ca. It reported higher emissions for flights than the one at carbonfootprint. For example, carbonfootprint reported round trip from Tel Aviv to Tampa Bay via JFK as 2.98 tons but offsetters reported 3.61 tons. I stuck with carbonfootprint because it specifically stated using 2015 guidelines, and because I suspect a more conservative estimate is the right one. ↩
- It turns out that the carbon cost per passenger of air travel has been improving over time, as planes are more fuel efficient, and tend to be more full. ↩
- As it turns out, there was little overlap in 2015. OOPSLA’15 had 533 registrants, ICFP’15 had 388, but only 15 people attended both conferences. ↩