The hidden pollution cost of online meetings

It is often assumed that online meetings are better for the environment than physical meetings. That is not always the case.

  • A one-hour audio call consumes about 36 MB of data per person.
  • A one-hour standard-definition video call consumes about 270 MB per person.
  • A one-hour high-definition video call consumes about 540 MB per person.
  • A one-hour ultra-high-definition video call consumes about 1.3 GB per person.

Assuming an average of one one-hour meeting a day involving two people, 250 days a year, then:

  • The audio-only calls would emit 0.08 kg of CO2.
  • The standard-definition video calls would emit 0.6 kg of CO2.
  • The high-definition video calls would emit 1.1 kg of CO2.
  • The ultra-high-definition calls would emit 2.8 kg of CO2.

An average tree can absorb about 10 kg of CO2 per year. Here’s the equivalent number of people calling that would be required in order for it to be necessary to plant one tree in order to offset the pollution:

  • 270 people for audio only
  • 36 for standard-definition video
  • 18 for high-definition video
  • 7 people for ultra-high-definition video.

The average CO2 emissions from new passenger cars registered in the European Union in 2018 was 0.1204 kg of CO2 per kilometer. Thus, on a yearly basis:

  • The audio-only calls would be the equivalent of driving 0.7 of a km.
  • The standard definition video calls would be the equivalent of driving 5 km.
  • The high-definition video calls would be the equivalent of driving 9 km.
  • The ultra-high-definition calls would be the equivalent of driving 23 km.

It’s much better to have an online meeting than to drive to a meeting, it would seem. Well, it’s not that simple. The above estimates relate to the streaming costs for online meetings. There are also processing costs. In many organizations meetings are saved and stored and sometimes watched later by others. There are costs relating to the devices used for the meetings.

Based on initial calculations, we estimate that streaming may represent no more than 5% of the total costs. So, for meetings using high-definition video, for example, based on our scenario above, it could be the equivalent of driving 375 km per year. This still a quite reasonable figure, but is a lot higher than the streaming-only costs of 9 km per year.

What happens if people drive to the office and go on conference calls with other people in that very office? What happens if far more meetings now occur online than were held offline? What happens if far more people attend online meetings?

These are the things that I’ve noticed. When working with larger organizations, I’ve regularly been part of meetings with anywhere from 20 to 100 people; far more people than would have attended if the meetings had been held in a physical space.

Used wisely, digital can be better for the environment. However, time and time again, I have found that digital behaves as an accelerant, as a duplicator, as a copier, as a reproducer, as an encourager of wasteful behavior.  

Digital is not green. What is so much better than digital meetings are no meetings, fewer meetings, meetings that take a lot less time, meetings that are useful.

GB per hour for audio calls, Zoom

GB per hour for video calls, Taraspan

Electricity Intensity of Internet Data Transmission, J. Aslan, K. Mayers, J. Koomey, C. France, Journal of Industrial Ecology, 2015

Greenhouse gas reporting: conversion factors 2018, GOV.UK

What is the carbon footprint of streaming video on Netflix? George Kamiya, International Energy Agency in Paris.

Cisco digital forecasts 2017 – 2022, Cisco

7 thoughts on “The hidden pollution cost of online meetings

  1. Alex Hull

    This is a very compelling article and some of the figures are quite surprising, but what is the source of the calculations? There must also be other key assumptions made to get these figures.

    If we intend to create change, it must come with credibility and scientifically-backed evidence, otherwise it is just as easily dismissed by those who may find these facts inconvenient. There’s never been so much misinformation on the web as there is now, and without something to ground the article in reality, it’s difficult to understand what may be real versus what is sensationalism targeted to earn ad revenue. I’m not suggesting that’s what this is but just observing how easy it would be to ignore this information without further context.

    Reply
    1. Gerry McGovern Post author

      Thanks for pointing this lack of sources out, Alex. You’re totally right. I have now added sources to the article.

      Reply
  2. Em Ission

    These numbers are totally misinformed. The real numbers are around 411 times lower than the ones presented here, depending on use cases. The real consumption of an hour ultra-high definition video call (1,3 GB/h) are more like 0,006 kg CO2. Which means that you can call 17 hours of ultra-high definition for each km of car driven. Below are the numbers and sources.

    According to a study named “Electricity Intensity of Internet Data Transmission” (Journal of industrial ecology) published in 2017 the estimated power consumption per gigabyte of traffic in 2015 was 0.06 kWh/GB. This has decreased by half approximately every 2 years between 2000-2015 for developing countries, which would give an estimate of around 0.06/2^(5/2)= 0.01 kWh/GB in 2020. (Source: https://doi.org/10.1111/jiec.12630)

    Let’s say you use two phones to do an online call, and they use around 5W each (the reality is probably more like 2-3W). That would give us 10W = 0.01 kW. Running an hour long call is therefore here 0.01 kWh.

    Lets then assume the call is in the European Union area where the average 2016 numbers according to the European environment agency is 295,8 gCO2/kWh. (Source: https://www.eea.europa.eu/ds_resolveuid/0EKWQC5FRZ)

    The calculations would then be (based on an excel-document, so formatting are not ideal):
    Electricity intencity of internet data transmission projected 2020 0,01 kWh/GB

    One hour ultra-high definition 1,3 GB/h energy consumption 0,013 kWh
    Two phones (5W each) 0,01 kWh
    European environment agency electricity CO2 emission intensity 295,8 gCO2/kWh
    CO2 emissions in grams 6,8034 gCO2
    CO2 emissions in kg 0,0068034 kgCO2

    One-hour ultra high definition video call article 2,8 kg CO2
    Real consumption 0,0068034 kg CO2
    Difference 411,5589264x

    CO2 emissions per km from new passenger cars according to article 0,1204 kg CO2
    Hours of ultra-high definition call for each km car driven 17,69703384

    Reply
    1. Gerry McGovern Post author

      1) Our calculations include transmission costs which yours do not.

      2) We took the same study by Komey et al and projected to 0.015 for our
      2019 calculations, so that’s similar to yours.
      However, the IEA are still using a 0.058 figure, so that’s the last measured point; we’re assuming log decrease is continuing.
      I have queried this with Jonathan Koomey as I wondered how valid it can be to project a log relationship, but it’s the best we have to go on so far.

      3) Our assumption is on corporate use which is mainly desktop PCs or laptops.
      These use much more power than phones but use very efficient data transmission networks.
      Mobiles are very energy efficient as a device, but on the other hand they are much more expensive on transmission costs.
      See the slide “Data Transmission Networks” from this 2017 paper by George Kamiya of the IEA which shows the relative balance of fixed and mobile transmission in terms of both volume and efficiency:
      https://www.ieta.org/resources/COP%2023/Side-Event-Presentations/IETA%20-%20Digitalization%20-%20George%20Kamiya.pdf

      4) We also agree on the CO2.kWh figure; we use 280g from a 2018 gov.uk source.

      5) We count the transmission costs per participant. Our figure of 1.35 GB per participant per hour comes from the
      bandwidth recommendation of Lifesize quoted in
      https://www.taraspan.com/blog/bandwidth-requirement-for-hd-video-conferencing/

      Reply
      1. Em Ission

        First of all thank you for your reply. I might have been overstating the 411x numbers. Could you also present your calculations here? I will try to answer your points below as good as I can.

        You also added some sources to the article that I read through. I was surprised to see that the article “What is the carbon footprint of streaming video on Netflix? George Kamiya, International Energy Agency in Paris.” presented a carbon footprint of 0.028-0.057kgCO2 equivalents of watching a 30 minute show on Netflix. Kamiya also compares this to driving around 200 metres in a conventional car. Much less than the 23 km presented in your article. That is with almost double the carbon intensity of power generation with 478 gCO2/kWh and a larger total file size of 1,9 GB. So, using the same basis their calculations should be higher than yours. Also Jonathan Koomey as I mention in the answer to your second point have a much lower estimate of less than 20 grams CO2 of half an hour of Netflix.

        1) It is true I did not include the transmission cost of mobile networks, but the article I referred to and that you also now added as sources has a scope of all the IP core network and access networks. The only processes not included are home/on-site networking equipment and user devices. In a wired or WIFI environment of a business this should have an minimal impact. Most likely in the single digit wattage. Especially if you consider that they are turned on and idle already, the delta wattage change from the introduced traffic should be minimal. Let us say this is at a bad scenario around 5W. That is only a 14 % increase ((0,005/(0,026+0,01))*100) = 13,8888) (21.74% if you do not add the double data transfer rate that you corrected in point 5). Still there is a large difference. Where is this coming from?

        According to an article in the peer reviewed sustainability journal, the mobile network energy cost of mobile data Finland in 2017 was 0,30 kWh/GB, and they projected it to be 0.1 kWh/GB.
        (see page 5 at https://doi.org/10.3390/su10072494).

        This also corresponds to a large electricity usage of communication technology research done in 2015 called “On Global Electricity Usage of Communication Technology: Trends to 2030”. They are presenting both 2010 numbers and projected 2020 numbers in a table. First 2010 numbers are wired – 0.50 kWh/GB, Wi-Fi – 0.36 kWh/GB, WAN (mobile network) – 6-15 kWh/GB and data centers – 0.13-0.14 kWh/GB. Remembering back to cell phone data transfer prices in 2010 there should be no surprise that energy costs were quite high. Now the projections to 2020 are vastly different. Especially in the mobile data range. Here they are presented as wired – 0.11-0.28 kWh/GB, Wi-Fi – 0.07-0.17 kWh/GB, WAN (mobile network) – 0.047-1.04 kWh/GB and data centers – 0.027-0.085 kWh/GB. I will take the lower numbers with a grain of salt as the research are conducted by a telecom provider, but the projected range are perfectly within the expectations from the peer reviewed study from Finland quoted above.
        (see table 2 on page 142 at https://doi.org/10.3390/challe6010117).

        Now the business web call would almost always be on the low wattage corporate wired of Wi-Fi network, but lets do an worst case scenario where the two participants are transmitting an ultra-high quality call over mobile networks for an hour. This would introduce 2*0,1 kWh/GB extra energy consumed and increase my estimates by around 9 times ((internet 2x 1,3 GB + two phones 10W + mobile network transmission (0,1 kWh/GB*2*1,3GB = 0,26 kWh) * 295,8gCO2/kWh) = (0,026 + 0,01 + 0,26) * 295,8) = 87 grams CO2).

        This new estimate of 87 grams in a less than ideal situation is still far away from your 2800 grams estimate. Again, I would like to see your calculations.

        2) You mentioned that you queried with Jonathan Koomey, but he seems to be disagreeing with your numbers quite much. I checked his website, and one of the recent blogposts he is interviewed by BBC to answer the question of the environmental footprint of watching half an hour of Netflix. Many news outlets reported CO2 footprints of 1,6 kg, where he calculated 80x less than this and concluded on lower than 20 grams. Below is the quote and link:

        “In October 2019, many news outlets (including Phys.org) reported that watching half an hour of Netflix would emit the same amount of carbon dioxide (1.6 kg) as driving four miles. This appears to be yet another amazing “factoid” about information technology’s environmental footprint that has little relationship to reality.

        I dug into the calculations, at the prompting of the BBC, and figured out the real story. Half an hour of Netflix emits less than 20 grams of carbon dioxide, probably much less. The BBC interviewed me last week and did a nice story about it.”
        (https://www.koomey.com/post/190502951713)

        3) Mobile network transmission costs is answered in the first point. The data presented here dated 2017 are first of all not explained in any other way than a graph. Secondly they seem to come from an old study not only using 2010 numbers, but using questionable methods of power consumption divided by data numbers. Resulting in 0.08 kWh/GB for fixed broadband, 2,9 kWh/GB for average 3G and 37 kWh/GB for average 2G (https://doi.org/10.1111/jiec.12145). If a telecommunications operator had those power consumption levels today for their mobile data network, they would have been out of business a long time ago. Those numbers are simply too high to be economically feasible in a 2020 mobile data demand.

        4) Great, that is very similar numbers. My numbers should be higher though. Do you have any feedback on where they differ?

        5) Yes, that is true for some services, so I added double data transfer rates for my two person example above. An WebRTC service with a SFU mixing architecture will give the total downlink to all participants, and mesh both download and upload. MCU on the other hand the server remixes the video feed and each participants network costs are minimized. If you see the statistics of a zoom call for example they have a mixed approach to this. They are not using WebRTC, but they scale the data transfer according to who is talking and how the screen are viewed if it is gallery mode or active speakers. You linked to their bandwidth requirements in your sources, and this never gets above 3 Mbps both ways even in a large group call. Highest bandwidth an hour with zoom is therefore ((3/8)*60*60)*1000 = 1,35 GB/h. Exactly around our highest estimates. But this rarely happens in a real world scenario, and you even have to go into the setting of your account to enable it as the default maximum is 720p video for new accounts.

        Lastly one interesting comparison if we turn the problem upside down. You mentioned that you used the estimate of 280 gCO2/kWh in your calculations. If we reverse engineer those calculations you are then suggesting that the ultra-high definition call are using (2800gCO2/280gCO2/kWh = 10kWh). Lets assume that this electricity would be paid by household consumers, the average electricity price in EUR of the EU-28 was 0.2159 per kWh. This is around 0.245404 USD in currency price. This means that the cost of the 10 kWh projected energy demand of the 2,6 GB call is 2,4540 USD in energy cost for an average home consumer. Now we see the market bandwidth cost by an Amazon EC2 on-demand instance in Ireland and transfers the equivalent 2,6 GB at the bandwidth price of 0.02 USD per GB to a total of 0,052 USD for the transfer. This means that the bandwidth charges Amazon projects to the customer is 47 times lower than the estimated energy cost you are presenting. Now the datacenters have a much lower energy bill, but not that much lower. Also this example is the pure energy price, and not taken into account the cost of infrastructure and network investments required to actually provide the bandwidth needed to transfer 2,6 GB out from the datacenter. Which should make the comparison even worse than the pure energy difference. This might be a simple comparison with possible flawed estimates, but it proves my point. Who pays for the energy consumption of 10kWh that you are suggesting? The datacenter? The ISP? The business? The end user? Again comparing the energy demand to a tesla model 3 long range edition the 10 kW suggested is enough to drive a Tesla model 3 long range edition 65 km (10kWh / 0,154kWh/km = 64,9km).

        As we all know money talks, and the 10kWh estimate is simply way out of proportions.

        I’ve had an interesting journey trying to find realistic numbers to the environmental impact of online meetings, and I probably used too much time to write these comments. A small comment on a blog hidden in the whole wide web of the internet are not much of an impact. I guess my surprise to see those large numbers caught my attention and got me on this journey to find out if this was real or not. I do not doubt your intentions in presenting these numbers, but I believe that they are involuntarily misinforming people of the real situation. Rather than people facilitating good online meeting solutions that help the environmental impact, such estimates feed the misinformed notion that fossil fuel transportation are somehow an environmental impact equivalent to online meetings.

        Reply
        1. Gerry McGovern Post author

          Thank you for this extremely interesting and well researched reply. I now realize where the misunderstanding lies. Our figures are for a year of conferencing (250 days). If we look at an individual conference call with two people, I think you will see that our figure is more conservative than Kamiya. I think the fault lies with us, though, in that we weren’t clear enough that we were dealing with a year’s worth of data. I’m going to update the article to make it clearer.

          Thank you again. The work and analysis you are doing is extremely important.

          Reply
          1. Em Ission

            Ah. I see. No wonder my estimates where hundreds of times lower than yours when those were consisting of 250 meetings in a year. Then our figures converge to similar estimations. My bad. Maybe it would be good to add a one-meeting estimate in the article for clarity?

            I also did a calculation based on our misunderstanding where the energy demand of 7.69 kWh/GB multiplied by the total yearly internet traffic of 3 zettabytes actually totaled the entire world energy demand of 23 petawatts, but I did not publish it. When this number is divided by 250, it makes more sense.

            Well. I’ve learned a lot in this process, so thank you for waking my interest! Even though less meetings are the best option, this shows how much more efficient online meetings are compared to physical meetings when there is travel involved. 23km/250times is 92 meters driving equivalents each meeting. About the same distance you would walk to get your coffees in the office or to the cantina throughout the day. While it is not zero, that is still a very low energy footprint.

            Thank you also for using your time to read through these comments and approve them to be published. I did not catch the yearly aggregation in your original version, so then your article is on point for yearly aggregates. Looking forward to explore more in your book “world wide waste” scheduled to be published next month. At least if it is provided in a digital form 😉

            I guess there is nothing more to say then. I left my email in this comment if that is of interest.

Leave a Reply

Your email address will not be published. Required fields are marked *