12

The carbon footprint, in carbon dioxide equivalent, of sending an email message, has been estimated at 0.3 gCO2e for a spam email, 4 gCO2e for a "proper email", and 50 gCO2e for an "email with long and tiresome attachment that you have to read". These figures have been repeated in the media many times, even recently in 2018, in 2017, etc., and implied to mean that it is climate-conscious to avoid sending too many emails. Indeed, with these figures, an average 121 emails per day would amount to 176 kgCO2e/year, which is certainly not negligible. My question is whether these figures are accurate (in 2020).

The figures seem to come from the 2010 book How Bad Are Bananas? The carbon footprint of everything by Mike Berners-Lee. I found them in the book, but no info about how they were computed, except a mention of research by McAfee without a specific source. I think it is probably this 2009 report, which does have the 0.3 gCO2e figure for a spam email, and a figure of 131 kgCO2/year in email-related emissions for the average business email user. But there is little information there about how the statistics were computed.

My reason to doubt this is that watching Netflix was more recently estimated (in 2020) to have a 28-47 gCO2e footprint for half an hour. Even taking the lower value, this would mean that a "proper email" would have the same emissions as 4 minutes of watching Netflix, and a spam would correspond to 20 seconds of Netflix. This is weird to me, as Netflix does seem to require much more bandwidth, computation power on the server and client, storage, etc. One explanation for why the email estimate seems high may be that the carbon footprint of network transfer has steadily decreased since 2009-2010: see this study Figure 3 for instance.

So I am wondering whether these estimations have been updated or refuted since 2009-2010. Looking online, I haven't been able to find anything that revisited or contradicted these statistics.

a3nm
  • 301
  • 2
  • 8
  • 4
    Does the claim perhaps include the computer + screen time spent *reading* the email? An hour of reading emails on a nice 32" 4k monitor vs. switching the computer off and working in the garden? – gerrit Jun 04 '20 at 11:57
  • 1
    The wording appears to suggest that the computation includes the human effort required to "digest" the email. – Daniel R Hicks Jun 04 '20 at 12:26
  • 1
    Yeah, but reading a spam takes <10s, reading a "proper email" certainly takes <4 min on average, so the comparison to Netflix still applies. – a3nm Jun 04 '20 at 13:00
  • I feel sure this is a duplicate, but I am not finding the question it is reminding me of :-( – Oddthinking Jun 04 '20 at 14:46
  • 2
    These kinds of numbers are not useful at all. How much higher is your carbon footprint if instead of spending 30 minutes watching Netflix you hop in your old clunker and go to the store to buy some milk? Numbers in the aggregate, where all of the initial conditions are known, are far more useful (say, comparing the carbon savings of a wind farm with the carbon spent building it). – Robert Harvey Jun 04 '20 at 14:47
  • I suppose a very significant portion of the footprint for long emails is the monitor - both of the writer while composing the mail and the reader while reading. White backgrounds are quite power-hungry. – Polygnome Jun 04 '20 at 18:07
  • E-mail returned to sender - insufficient voltage. – Robert Columbia Jun 04 '20 at 18:32
  • 8
    @Polygnome: for most display types, those that filter colored pixels from a backlight, background color doesn't matter. – dandavis Jun 04 '20 at 19:51
  • 1
    @Polygnome only if you use an OLED or proper LED display, and those are expensive. On phones, OLEDs are becoming more popular than LCDs, but most computers still use backlit screens, where the colors displayed make absolutely no difference in power consumption. – TheWanderer Jun 05 '20 at 12:59
  • 7
    IMO it's a **nonsense statistic**. My computer runs all day, and it makes no difference to the carbon cost what the length of the email is. Similarly the network server runs all the time. It's just an average divided by another average. Similarly in my UK doctor's surgery a notice informs that it costs the NHS £xx every time I consult a doctor, which again is an average of total costs divided by consultations, taking no account of the knock-on cost of me *not* visiting a doctor. It costs the surgery no less to run, whether or not I visit (although if many visit they need more staff). – Weather Vane Jun 05 '20 at 21:43
  • 2
    ...there is also the point that some electronic services run off the back of other services, so there is **no** additional carbon cost, and the whole discussion is no more than a "space filler" in a "news" spot, like "did you know that only 15% of the hairs in a rabbit's ear are actually pink/grey/whatever?" You would also have to consider the carbon footprint of alternatives such as a letter, a fax, driving there with the message, and so on. – Weather Vane Jun 05 '20 at 21:46
  • 3
    @WeatherVane There cannot be no additional carbon cost. The transmission of an email necessitates that a lot of devices worldwide change a couple of bits in their internal state and that requires a certain growth of entropy, the resulting heat of which can only come from various electric power supplies. The *theoretical* physical minimum is perhaps tiny, but of course actual and typical (in particular: fast) devices do not even get close to such a minimum – Hagen von Eitzen Jun 05 '20 at 21:54
  • 2
    @HagenvonEitzen but the devices aren't doing *nothing* otherwise. They are continually monitoring and making requests, and change *different* bits when an email is sent. There isn't a "zero activity" state when they are running. And as I mentioned, what is the *alternative* cost of sending a message by another method? Is it more, or less? Email is very cheap, as an 'extra' cost on a system already in place, and not expensive. – Weather Vane Jun 05 '20 at 21:57
  • 3
    @WeatherVane, there isn't a *zero* activity state when they're running, but there is a *very low* activity state. In the past decade or so, a great deal of effort has gone into designing devices so that the "waiting for something to happen" state consumes very little power. – Mark Jun 05 '20 at 22:13
  • 1
    +1 @Mark. Routing an email consumes some CPU, and a comptuer that's routing traffic consumes more than a computer that does nothing. So there is a non-zero footprint (although it may be extremely small). – a3nm Jun 06 '20 at 11:27
  • 1
    @a3nm are you sure it is not a *negative* footprint? Assuming a message or document has to be sent by *some* means, does an email have a bigger carbon cost than a letter, or a courier? This is what I mean by a "nonsense statistic." On its own, it is meaningless. – Weather Vane Jun 06 '20 at 13:29
  • @WeatherVane: Uh, well, yes, sure, but carbon footprints are usually computed relative to not doing the task in question; not by assuming that the task must be done anyway. When we talk about the carbon footprint of a trip, it's not based on assuming the trip must be done and comparing to the option with the least footprint... – a3nm Jun 06 '20 at 13:31
  • So to be meaningful, you should also quote the carbon cost of the alternatives. If you don't send any messages, don't eat any food, don't buy anything, the cost of living is very little. Otherwise it's just a piece of trivia, obtained by dividing one number by another. – Weather Vane Jun 06 '20 at 13:32
  • @WeatherVane. Sure. But my question isn't about the carbon footprint of the alternatives. The question is about the carbon footprint of sending an email, and the widely-reported statistic that I quoted. – a3nm Jun 06 '20 at 15:48

1 Answers1

8

The carbon footprint of sending an email has been estimated by the Shift Project in this document to be between 0.5 and 2 Watt.hour (for a 1 MB email sent in 3 minutes), let's take 1 Watt.hour. The US average emissions are estimated here to be 0.429 kgs CO2e per kWh. This would give an estimate of about 0.5 g CO2e for sending a 1 MB email (so a rather large email), so two orders of magnitude less than the figures I am asking about in my question.

The Shift Project estimate, unlike the estimates I am asking about, is sourced, in particular to the 2015 paper by Andrae and Edler, ''On Global Electricity Usage of Communication Technology: Trends to 2030'', available here, which estimates 0.1-0.2 TWh per exabyte (p10), so 0.1-0.2 Wh per megabyte — the corresponding Shift Project values in the "Network impact" cell of their spreadsheet seems somewhat higher but of the same order of magnitude (they also include the cost of the electricity used to operate the device). This estimate is in line with the 2017 paper by Aslan, Mayers, Koomey and France, Electricity Intensity of Internet Data Transmission, available here, which estimates 0.1 kWh.hour for 1 GB.

Long story short, I think the figures in my question are indeed overestimated by about two orders of magnitude.

a3nm
  • 301
  • 2
  • 8
  • 2
    Pretty sure the average email is closer to 10 kB than 1 MB too (even when most of it useless non-standard headers and wrapped in HTML). – pipe Nov 11 '20 at 23:23
  • There are also more recent figures here (from 2020): https://carbonliteracy.com/the-carbon-cost-of-an-email/ Somewhat surprisingly, the largest impact may be in the time that it takes to read/write the email (as you are using your computer during that time). – a3nm May 12 '23 at 19:54