conference_footprint

compute the CO2 footprint of an academic conference
git clone https://a3nm.net/git/conference_footprint/
Log | Files | Refs | LICENSE

README.md (6705B)


      1 This file explains how the carbon footprint of Highlights'24 was computed.
      2 
      3 ## Data collection
      4 
      5 We collected information about the travel plans of participants using a
      6 web form (originally hosted at [this
      7 URL](https://framaforms.org/highlights-and-jewels-of-automata-theory-2024-1715936947)).
      8 Filling in the form was part of the registration process.
      9 
     10 One participant registered as attending on-site and then as attending online, so
     11 we counted this participant as attending online. We then eliminated online
     12 participants. We arrive at 160 on-site participants.
     13 
     14 We then eliminated local participants, estimating a CO2 footprint of 0 for them.
     15 We arrive at 135 on-site non-local participants.
     16 
     17 We sanitized the data by hand as follows:
     18 
     19 - when participants indicated multiple possible places, we selected the first
     20 - when participants did not specify a place, we use their affiliation location
     21 - when participants did not select a means of transportation, we assumed that
     22   trips of >400km were done by plane (which covered all cases with missing
     23   information)
     24 - we manually fixed some typos in locations and disambiguated some locations to
     25   ensure a correct geocoding output
     26 
     27 The registration form asks about "Other scientific activities during your stay
     28 (including HCRW)", giving people to indicate the option "Yes, I am extending my
     29 trip for other scientific reasons.". The form also asks participants whether
     30 they will participate to HCRW. (Not all participants who ticked the second box
     31 also ticked the first.) We propagate this information about extended stays (both
     32 fields) in the data that we generate and release, but we do not take it into
     33 account in the computation.
     34 
     35 From the data, we then use the Geonames service to transform the location
     36 indicated by participants, by extracting to locations.txt the locations and
     37 geocoding them using geocode.py to the file locations_with_latlon.txt.
     38 
     39 We then have the file locations_with_latlon.txt giving all locations preceded by
     40 their latitude-longitude in the format, e.g.,:
     41 
     42   44.84044 -0.5805 Bordeaux 
     43 
     44 And we have the file
     45 highlights_and_jewels_of_automata_theory_2024_onsite_nonlocal_manualclean.csv
     46 containing lines of the following form for each onsite nonlocal participant
     47 (numbered from 0, and tab-separated):
     48 
     49 - fields 0 and 1 are irrelevant
     50 - fields 2 and 3 give first and last name (only used for debugging)
     51 - field 4 is irrelevant
     52 - field 5 gives university (only used for debugging)
     53 - fields 6 say "I'm coming to Bordeaux"
     54 - field 7 gives participant type (unused)
     55 - field 8 says "External Participant"
     56 - field 9 gives the origin place (text)
     57 - field 10 gives the origin mode among "Plane", "Train", "Bus/Coach"
     58 - fields 12 and 13 give the same information for the destination place
     59 - fields 14 and 15 are the information of the two boxes about extended stays
     60   (propagated in the files but not used in the computation)
     61 
     62 (These files are not versioned because they can be considered private
     63 information.)
     64 
     65 We run: 
     66 
     67   ./generate_trips.py 44.84044 -0.5805 0.2
     68 
     69 Where the arguments are the latitude and longitude of Bordeaux, and 0.2 is the
     70 noise to add. This generates a file trips_anonymized.csv containing, for each
     71 trip leg, the mode ("plane", "train", "bus/coach"), the distance (in km,
     72 rounded, with noise), and the information about extended stays. A file
     73 trips.csv is also produced for debugging (with the data without noise and with
     74 personal information). A file map.geojson is also produced with the map of
     75 participants and transportation modes and private information (to be used as an
     76 image only).
     77 
     78 The file trips_anonymized.csv can then be fed to co2.py which computes the
     79 carbon footprint (see below). This gives (from the anonymized data):
     80 
     81 total CO2e emissions (tons): 41.159883
     82 for mode train: CO2e emissions (tons): 5.264101
     83 for mode plane: CO2e emissions (tons): 35.871842
     84 for mode bus/coach: CO2e emissions (tons): 0.023940
     85 for distances <2000 km, plane is used for 68/243 trips
     86 for distances >=2000 km, plane is used for 22/27 trips
     87 flights of over 2000 km account for 18149946.000000 CO2e emissions (tons) i.e. 44.096204 percent of total for 22/270 total legs
     88 distance by plane: 201579
     89 
     90 Hence, the total CO2 footprint is 41 tons CO2e (it is the same with the
     91 non-anonymized file). Around 87% of emissions are due to plane travel, and 44%
     92 of the emissions are due to 8% of the transportation legs, namely,
     93 the plane trips of over 2000 km. (Note that most trips of over 2000km are done
     94 by plane, but not all.)
     95 
     96 The average footprint per onsite non-local participant (135) is around 
     97 307 kgCO2e. The average footprint per onsite participant (160) is around 
     98 260 kgCO2e. (These figures are computed from the non-anonymized data.)
     99 
    100 ### Carbon footprint
    101 
    102 Like in 2022, we compute the CO2 fotprint following the
    103 [labos1point5](https://labos1point5.org/ges-1point5) data, which is adapted from
    104 the French agency [Ademe](https://www.ademe.fr/). We use the values from 2022
    105 without updating them to ensure that the methodology is comparable.
    106 
    107 - For train, we count **37 gCO2e/pkm** (international train). This is pessimistic in France, very
    108   pessimistic for TGV, but similar to the 41 gCO2e/pm for national (UK) rail
    109   given by [Our World in
    110   Data](https://ourworldindata.org/travel-carbon-footprint).
    111 - Plane is counted following
    112     [labos1point5](https://labos1point5.org/ges-1point5), including the effect
    113     of contrails:
    114   - 258 gCO2e/pkm for less than 1000km
    115   - 187 gCO2e/pkm between 1001km and 3500km
    116   - 152 gCO2e/pkm above 3500km. This value is consistent to the 150 gCO2e/pkm
    117       value for long-haul flight given by [Our World in
    118       Data](https://ourworldindata.org/travel-carbon-footprint) (also including
    119       contrails)
    120 - For bus/coach, we count 28 gCO2e/pkm as the coach value given by [Our World in
    121     Data](https://ourworldindata.org/travel-carbon-footprint) as there is no
    122     value in labos1point5.
    123 
    124 ## Trends relative to 2022
    125 
    126 We now compare the footprint relative to 2022. (In 2023, there was no
    127 computation of the footprint.)
    128 
    129 In 2022, there were 173 registered onsite participants, 127 registered onsite
    130 nonlocal participants, and a total of 42 tons of CO2e. Relative te 2022, and
    131 with the same methodology:
    132 
    133 - The total CO2 footprint of Highlights'24 is essentially the same as in 2022
    134 - the CO2 footprint per registered onsite participant has evolved from 240
    135   kgCO2e to 260 kgCO2e, i.e., a 8% increase
    136 - the CO2 footprint per registered onsite nonlocal participant has evolved from
    137   330 kgCO2 to 260 kgCO2e, a 22% decrease
    138 
    139 In an nutshell, the total emissions are about the same, but Highlights'2024 has
    140 slightly less onsite participants but slightly more onsite nonlocal
    141 participants.
    142