Trend towards virtual and hybrid conferences may be an effective climate change mitigation strategy

Since 2020, the COVID-19 pandemic has urged event holders to shift conferences online. Virtual and hybrid conferences are greener alternatives to in-person conferences, yet their environmental sustainability has not been fully assessed. Considering food, accommodation, preparation, execution, information and communication technology, and transportation, here we report comparative life cycle assessment results of in-person, virtual, and hybrid conferences and consider carbon footprint trade-offs between in-person participation and hybrid conferences. We find that transitioning from in-person to virtual conferencing can substantially reduce the carbon footprint by 94% and energy use by 90%. For the sake of maintaining more than 50% of in-person participation, carefully selected hubs for hybrid conferences have the potential to slash carbon footprint and energy use by two-thirds. Furthermore, switching the dietary type of future conferences to plant-based diets and improving energy efficiencies of the information and communication technology sector can further reduce the carbon footprint of virtual conferences.

-Concerning the impact from food and heating/electricity at the conference venue vs. participants staying at home. I was wondering whether the estimations assume that at home footprint from food and also from electricity/heating is the same as at the conference venue/conference hotel. In my view this would not make much sense since conference catering or eating at a restaurant probably produces much more food waste, than cooking at home. The same is probably true for heating/electricity: at a large conference venue/hotel I assume the footprint to be higher than at home. I did not find any information on this issue in the paper.
-Concerning the travel induced footprints the authors assume that participants take the train in Europe if the travel distance is below 600 km. I have two questions here: 1) is the travel distance for the train travel calculated by Google Distance Matrix API actually based on the railway network? Or is it based on a street network? Since in many European countries the railway network differs a lot from the street network in terms of density and directness (e.g. in France the high-speed train network connects all major cities with Paris in straight lines, whereas in Germany the train network follows the routes historically from one city to the next, which results often in less direct connections than using the street-network). 2) Assuming a certain threshold for traveling by car/bus/train or by airplane makes sense, yet I doubt that the travel distance itself is the right factor here. I suppose that most conference participants decide primarily on the basis of the travel time. Thus in cases where a good high-speed train network exists such as in France (e.g. travelling from Marseille to Paris is 775 km by car takes about 8h, by train less than 4h; flying is probably no real option here). Thus, using not a fixed km-threshold, but a threshold for the time accepted for getting to the conference venue might be a better option. See Jäckle (2019, 2021) for a similar approach.
-Different organizations report different emission factors for trains, cars and airplanes depending on various assumptions (European Environment Agency, 2021; UK Department for Business, Energy & Industrial Strategy, 2021; Umweltbundesamt, 2021). E.g. on the electricity mix used to power the trains, the average passenger-load factor, whether radiative forcing is included or not for flying etc. This naturally leads to deviating results in the estimation of the overall carbon footprint. The authors did not state -at least I did not find anything on this issue in the paper or the supplement -which assumptions they make, i.e. which emission factors they use for their estimates. Instead of presenting only point estimates it could also make sense to present a minimum-to-maximum range. This might particularly be the case for the travel induced carbon footprint since it clearly -as the results in the paper show -make up by far the highest percentage of emissions for the whole conference attendance. Variations in the emission factors are therefore very relevant for the overall estimates.
-Since air-travel is by far the biggest component of the carbon footprint of conferences, it makes sense to be as exact as possible here. I was wondering whether the distances used to calculate the carbon emissions from flying are solely the great circle distances between the airports or whether the fact that many airplanes do not fly the shortest route but more inefficient detours, or have stop overs has been included in the estimations. It has been shown that the actual distances aircraft fly are between 6 and 10% longer than the great circle routes between the departure and the destination airports (Kettunen et al., 2005). Furthermore, is there a difference in the estimation between long-haul and short-haul flights? Since long-haul flights reach higher altitudes where CO2 exerts more harmful effects it may makes sense to make a distinction. Does the analysis account for the different electricity mix in the participants' countries for those joining virtually?
Minor point: • Fig 2c is difficult to read. The colors/shading of the legend should be enlarged. best regards, Sebastian Jäckle -European Environment Agency (2021) CO2-emission intensity from electricity generation. Available at: https://www.eea.europa.eu/data-and-maps/daviz/sds/co2-emission-intensity-from-electricitygeneration-2/@@view. Dear authors, the paper is well-structured and covers an important gap in scientific literature on the GWP of virtual and hybrid conferences. However, before to be published some adjustments are still needed. Introduction, I would extend the introduction or added a paragraph on the current state of the art on LCA and Carbon Footprint of conferences, to clearly highlight the innovative aspects of the current studies. You cited some studies with 4-13 references, I would add at least a table and a chapter indicating the main content in the previous papers. The paper has lots of information, but it is not always clear reading it which are the data you have used and there are some Life cycle inventory data missing in the main manuscript. I know that the most are reported in the supplementary materials but it doesn´t help the transparency of the study. Please report the main data of life cycle inventory in the main manuscript: e.g. the consumption of each transport system considered, the FAO data on Food, and so on. The sequence of the manuscript is also not conformed to the an LCA study in according to the ISO 14040: Methodology with all 4 phases of LCA should be reported after the introduction and state of the art to allow the reader to understand the environmental impacts obtained. It is very important for an LCA study to be transparent in its assumptions made and ensuring transparency and reproducibility. In page 7 Fig. 2 Comparison between the 1-hub in-person conference and the virtual conference. a, Carbon footprint associated with transportation for individual participants, indicating that the unit carbon footprint for trips primarily by plane tends to be smaller than that for driving. I do not see this result in the figure and it is not coherent with the results of other studies, please clarify it.

Thank you for your comments. During the virtual conferences, both the electricity used for accommodation and the information and communication technology (ICT) service contribute to the electricity consumption. The electricity used for accommodation is estimated based on the average daily residential electricity consumption (U.S. Energy Information Administration, 2020);
The electricity use for the ICT services involved with the video-conferencing is quantified following existing literature (Pärssinen et al., 2018) and is detailed in the Method section of the original manuscript. In terms of the electricity consumption for watching recordings after the virtual talk, the electricity consumption can be different from attending the real-time virtual talk. Both  "Other than the food preparation, electricity consumption for accommodation during the virtual conference and the video-conferencing contribute considerably to most impact categories." On page 19 of the revised manuscript: "However, diminishing synchronous virtual participation may add difficulty to communication and collaboration for virtual conferences that do not provide sufficient asynchronous attendance options, such as the recording proceedings 42 . To create more opportunities for exchange and follow-up discussion, virtual conference organizers are encouraged to provide contact information, recording proceedings, and electronic documents submitted by the presenters to all participants. These asynchronous attendance practices could support virtual conferences to improve equity, diversity, inclusivity, networking, early career research promotion, and career development 18,21-23 . On the other hand, post-conference activities, such as downloading and playing asynchronous recordings, sending follow-up emails, and searching for materials, can result in environmental impacts. The relationship between daily virtual participation and the amount of post-conference activities is unknown. Therefore, we cannot expand our system boundary to account for the variation in the environmental impacts of these consequential activities. Future work could further explore the consequential environmental impacts of the rapidly expanding video-conferencing industry." On page 20 of the revised manuscript: "Post-conference activities of participants, such as downloading and playing asynchronous recordings, sending follow-up emails, and searching for materials, are out of the scope of this attributional LCA." On page 25 of the revised manuscript: "Information and communication technology To estimate the environmental impacts of virtual conferences, we incorporate an environmental impact assessment framework of Internet services 53 . The system boundary of the ICT stage consists of infrastructure, network, and server related to video-conferencing. Specifically, four analysis layers are examined, including the energy consumption of the infrastructure, the shares of access network traffic and shares of IP protocols delivering the investigated Internet services, shares of traffic classes representing the end-user activities, and shares of the investigated Internet services in each traffic class. As participants of conferences usually use search engines or other office software simultaneously on the computer, mobile devices are excluded from the system boundary. Energy consumption of infrastructure, including router, Internet access equipment, computer, and data center equipment (cooling systems, lighting, and power supplies), are considered following previous studies 39,54 . Additionally, the production and distribution of router, Internet access equipment, and computer are included in the system boundary, while the construction of data center infrastructure is excluded, because of the unavailable LCIA data and dominance of the operational phase to the overall environmental impacts 55 . The energy intensity of the network for the video traffic class (EInetwork) is calculated as follows: where λfixed line, λIPv4, λTCP, λHTTP, λvideo represent the share of fixed line traffic in the total IP traffic, the share of IPv4 based traffic in the total IP traffic, the share of transmission control protocol (TCP) based traffic in the total IP traffic, the share of hypertext transfer protocol (HTTP) based traffic in the TCP based traffic, the share of video traffic class in the HTTP based traffic, respectively. K indicates the set of network components, including packet switched core, fixed line customer premises equipment, operator data center, office networks, and Internet core. Ek denotes the energy consumption of each component of the network. DTvideo represents the fixed line traffic for the video traffic class. As the best available data of the network's energy consumption and data traffic is from 2016, an annual electricity efficiency improvement of 10% is considered following the setting of the expected scenario in previous work 43 . The energy intensity of the server (EIserver) is estimated from the best available data of a 2015 Sweden study on the ICT sector 56 , and extrapolated to the value for 2020 with an annual electricity efficiency improvement of 10% 43 . The calculation is as follows: where Eserver represents the total energy consumption of servers, and DTserver represents the total data traffic of the server. Data traffic of the virtual conference is computed following the survey of a recent study, which recognized 80% of participants attending the conference each day with a daily online duration of 5.5 hours 5 . The energy consumption related to the network and server for the virtual conference (VEnetwork and VEserver) is computed following equation (3) and (4).
(4) where DTvirtual is the data traffic of the virtual conference calculated by multiplying the bandwidth of the video-conferencing with the total amount of online time for all participants. Downstream and upstream bandwidth of a Zoom group video calling for 720 high-definition videos is obtained from Zoom as 1.8 and 2.6 megabits per second (Mbps), respectively 57 . Following the 2020 projection with expected improvement in energy efficiency from a previous study 54 , the energy consumption of data center can be broken down into four equipment categories, namely infrastructure (33%), network (3%), storage (11%), and servers (53%). The material and energy inventories of the ICT stage are summarized in Table 6." Table 6 Material and energy consumption of the information and communication technology for virtual-conferencing 5,39,43,53,54,56,57 .  6,13 . Half of them focused exclusively on round-trip transportation 4, 5,9,11,12,29 while the rest considered life cycle stages of preparation, execution, catering, accommodation, and transportation [6][7][8]13,26 . However, due to differences in assumptions associated with in-person conferences (e.g., duration, size, and locations of the conference, geographical distribution of participants, transportation mode, system boundary, and selection of characterization factors), the carbon footprint ranges from 92 to 3540 kg CO2 eq. per capita. All of these studies identified transportation as the environmental hotspot. The conference site and geographical distribution of participants determine the transportation distance and mode for participants. From those who reported the average transportation distance, the average round-trip transportation distance varies from 1980 km 13 to 9564 km 9 . However, the average distance does not illustrate the complete picture of the participant transportation, and it was found that the 10-20% of participants with the most polluting trips contribute to a substantial portion (20-70%) of the total transportation-induced emissions 4-7,12 . These values depend on the distribution of participants, which reveals whether the conference is more localized or more internationalized. As shown in Supplementary Table S1, most participants are from the region where the conferences are held. The conference location is also important in determining the transportation profile, in which a conference location with better train connection to other major cities is capable of allowing more participants to transport by train and thus has more potential of reducing carbon footprint while a conference located in the southern hemisphere usually perform much worse in terms of carbon footprint compared to the northern hemisphere 11 S1 Comparison of per capita carbon footprint results from previous literature and this study. By taking the whole life cycle of a virtual conference into consideration, the carbon footprint of the virtual conference in this study is substantially higher than those in other studies 1-4 . The carbon footprint of 1-hub in-person conferences is within the range of values reported by existing studies. Only a few studies investigated the carbon footprint of multi-hub inperson conferences. The carbon footprint of 1-hub and 2-hub in-person conferences from Ewijk and Hoekman 3 is higher than that from our study because the participants of that conference are more geographically distributed than those in this study. Due to the same reason, the carbon footprint of the 3-hub in-person conference from Ewijk and Hoekman 3 is significantly reduced and becomes lower than our result.

13
Thank you for your comments. We agree that short-haul flights have higher greenhouse gas emissions than those of medium-and long-haul flights on a per passenger-kilometer basis. Indeed, we took this factor into consideration and adopted the characteristic factors of air transportation that decrease over distance, as shown in Supplementary Fig. S2 of the original Supplementary Information.
In terms of the hybrid conferences that involve no flying, several "maximum virtual participation" hybrid scenarios, as summarized in Table 1, involve no flying due to the 600-km threshold for rail transport and 500-km threshold for road trips. Fig. 5a  e62668, (2020). Achakulvisut, T. et al. Improving on legacy conferences by moving online. eLife 9, e57892, (2020). Actions: On page 19 of the revised manuscript: "However, diminishing synchronous virtual participation may add difficulty to communication and collaboration for virtual conferences that do not provide sufficient asynchronous attendance options, such as the recording proceedings 42 . To create more opportunities for exchange and follow-up discussion, virtual conference organizers are encouraged to provide contact information, recording proceedings, and electronic documents submitted by the presenters to all participants. These asynchronous attendance practices could support virtual conferences to improve equity, diversity, inclusivity, networking, early career research promotion, and career development 18,21-23 . On the other hand, post-conference activities, such as downloading and playing asynchronous recordings, sending follow-up emails, and searching for materials, can result in environmental impacts. The relationship between daily virtual participation and the amount of post-conference activities is unknown. Therefore, we cannot expand our system boundary to account for the variation in the environmental impacts of these consequential activities. Future work could further explore the consequential environmental impacts of the rapidly expanding video-conferencing industry. Dietary type as well as the electricity and food consumption rate are the next most sensitive variable for a virtual conference."

Reviewer 2
The authors are most grateful to the editor for the helpful comments.

Reviewer's comment (1)
The paper includes in my view two major innovations compared to earlier studies: 1) it shows to what extent hybrid conferences at multiple hubs with a spatially optimized distribution of participants helps to reduce the environmental impact of conferences, and 2) it estimates not only the carbon footprint but applying a complete LCA approach the impact of conferences (virtual/hybrid and in person) on a number of outcomes (e.g. water depletion, or marine ecotoxicity). While the first innovation is made very clear, the second innovation is less prominent in the paper. To some extent I am also not sure whether focusing solely on the carbon footprint as the major outcome variable would be better in order not to overstretch the paper. All in all, I think this is a very important article which helps to base the decision on how to move forward with the conference business on a stable empirical base. It therefore makes a significant contribution to the literature and fits well into the scope of Nature Communications. In the final section the authors could give more clear recommendations for conference organizers how to organize these events in an environmentally friendly manner. From a methodological point of view, I have to admit that I am definitely not an expert in LCA. The following questions should be considered accordingly against this background. Answer: Thank you for your comments. In this study, we account for full-spectral impact categories because some of the investigated life cycle stages have been reported contributing to many environmental issues, such as food consumption to land use, water use, and pollution of aquatic and terrestrial ecosystems (Springmann et al., 2018) Supplementary  Fig. S13-S14." On page 18 of the revised manuscript: "Therefore, from the environmental perspective, it is beneficial to hold a hybrid conference and decide the location of the hubs using the registration information or survey responses. The hub locations can be sub-optimal because differences may exist between the pre-conference information and the real attendance. Adding hubs and increasing virtual participation levels tend to provide more environmental benefits, but this benefit becomes less prominent as the number of hubs and virtual participation level are high enough. It is therefore important for conference organizers to consider the trade-off between organizational challenges and environmental sustainability." On page 20 of the revised manuscript: "For virtual conference organizers, advocating energy saving from heating and other residential electricity use (e.g., air conditioning, lighting, electronics, and appliances), food waste reduction, ovo vegetarian diet can be effective practices to improve sustainability."  . Although clearly not as sophisticated in the methodological LCA approach as this paper, the results are nevertheless very similar. It also shows that hybrid solutions can -particularly if those participants who would have to fly in from far away join online -and all others accept longer travel times by bus/train instead of flying can significantly reduce the emissions from conferences. Answer: Thank you for your comments. We have discussed the methodology of this study in the Introduction section of the revised manuscript, following the reviewer's suggestions. Actions: On page 4 of the revised manuscript: "A recent study on the carbon footprint of virtual, in-person, and hybrid conferences accounted for the video-conferencing-related emissions, transportation, execution, catering, and accommodation 26 . However, it considered a single conference hub for both in-person and hybrid conferences and thus neglected the geographical effects of hub selection and participant assignment." References: 26 Jäckle  figure 3,  However, due to differences in assumptions associated with in-person conferences (e.g., duration, size, and locations of the conference, geographical distribution of participants, transportation mode, system boundary, and selection of characterization factors), the carbon footprint ranges from 92 to 3540 kg CO2 eq. per capita. All of these studies identified transportation as the environmental hotspot. The conference site and geographical distribution of participants determine the transportation distance and mode for participants. From those who reported the average transportation distance, the average round-trip transportation distance varies from 1980 km 13 to 9564 km 9 . However, the average distance does not illustrate the complete picture of the participant transportation, and it was found that the 10-20% of participants with the most polluting trips contribute to a substantial portion (20-70%) of the total transportation-induced emissions 4-7,12 . These values depend on the distribution of participants, which reveals whether the conference is more localized or more internationalized. As shown in Supplementary Table S1, most participants are from the region where the conferences are held. The conference location is also important in determining the transportation profile, in which a conference location with better train connection to other major cities is capable of allowing more participants to transport by train and thus has more potential of reducing carbon footprint while a conference located in the southern hemisphere usually perform much worse in terms of carbon footprint compared to the northern hemisphere 11 Fig. S1 Comparison of per capita carbon footprint results from previous literature and this study. By taking the whole life cycle of a virtual conference into consideration, the carbon footprint of the virtual conference in this study is substantially higher than those in other studies 1-4 . The carbon footprint of 1-hub in-person conferences is within the range of values reported by existing studies. Only a few studies investigated the carbon footprint of multi-hub inperson conferences. The carbon footprint of 1-hub and 2-hub in-person conferences from Ewijk and Hoekman 3 is higher than that from our study because their participants are more distributed than ours. Due to the same reason, the carbon footprint of the 3-hub in-person conference from Ewijk and Hoekman 3 is significantly reduced and becomes lower than our result.

Reviewer's comment (4)
-Concerning the impact from food and heating/electricity at the conference venue vs. participants staying at home. I was wondering whether the estimations assume that at home footprint from food and also from electricity/heating is the same as at the conference venue/conference hotel. In my view this would not make much sense since conference catering or eating at a restaurant probably produces much more food waste, than cooking at home. The same is probably true for heating/electricity: at a large conference venue/hotel I assume the footprint to be higher than at home. I did not find any information on this issue in the paper. Answer: Thank you for your comments. The amount of electricity consumption at home and at the hotel is different while food and heating are assumed to be the same. Details were presented in the original manuscript as follows. To address the variation in resource consumption, we have added a sensitivity analysis in the revised manuscript, following the reviewer's suggestions. On page 21 of the original manuscript: "The accommodation stage only considers utilities and waste treatment for staying at a hotel or guest home. Specifically, electricity consumption for a guest home is estimated from the U.S. Energy Information Administration's State Energy Data System 36 , electricity consumption for hotel, thermal energy and water consumption for both guest home and hotel are estimated from previous studies (Supplementary Table S6 Table 4. The maximum and minimum total food consumption rate in Table 4 is considered as the best case and worst case. b, Air transportation distance. Great-circle distance is considered as the base case, and 110% of the great-circle distance is considered as the worst case 21 ." On page 22 of the revised Supplementary Information:  "  Supplementary Fig. S19 Sensitivity analysis on heat and electricity consumption rate at guest/private home and hotel. Virtual and in-person scenarios with 1 to 6 hubs for an average participant are assessed. a, Heat consumption rate. b, Electricity consumption rate. The base-, worst-, best-case value of heat and electricity consumption rate at the hotel is obtained from Filimonau et al. 22 ; The base-, worst-, best-case value of heat and electricity consumption rate at the hotel is obtained from the U.S. Energy Information Administration 23 Platform, 2021). Therefore, we use the travel distance as a threshold in this study to determine the transportation mode of participants. Following the reviewer's suggestions, we have clarified the mechanism of travel distance calculation and the reason for using a distance-based threshold rather than a time-based threshold in the revised manuscript. conducted a sensitivity analysis to address the variations in these emission factors, based on the Ecoinvent database and references suggested by the reviewer.

References:
Tuchschmid, M., ecoinvent database version 3.7.1. transport, passenger train, DE, Allocation, cutoff by classification, (2020) (Accessed 04/25/2021). ecoinvent database version 3.7.1. market for transport, passenger car, medium size, petrol, EURO 5, GLO, Allocation, cut-off by classification, (2020) (Accessed 04/25/2021). On page 9 of the original manuscript: "The characterization factors for air transportation in Ecoinvent are classified based on four distance categories (i.e., very short-haul, short-haul, medium-haul, and long-haul flights). To be discriminative on the air transportation distances, we adopted the characterization factors from a comprehensive LCA study on air transportation 27  The base-, best-and worst-case characterization factors for the carbon footprint of rail transportation and driving are from the Ecoinvent database (Supplementary Table S4) and the Network for Transport Measures 17,20 . Specifically, the base-, best-and worst-case characterization factors for rail transportation are chosen among the country-specific processes for passenger train or high-speed passenger train, while the selection is conducted among passenger cars with different size and fuel for driving." On page 25 of the revised Supplementary Information: " Supplementary Table S4 Comparison of characterization factors for the carbon footprint of air transportation, rail transport, and driving. The base-case characterization factor for the carbon footprint of air transportation is obtained from Cox et al. 14 ; the best-case and worst-case characterization factors are obtained from the UK Department for Business, Energy & Industrial Strategy (International flights for an average economy-class passenger and long-haul flights for an average first-class passenger, respectively) 19 . The base-, best-and worst-case characterization factors for the carbon footprint of rail transportation and driving are from the Ecoinvent database and the Network for Transport Measures 17,20 . Specifically, the base-, best-and worst-case characterization factors for rail transportation are chosen among the country-specific processes for passenger train or high-speed passenger train, while the selection is conducted among passenger cars with different sizes and fuel for driving. Rest-of-World 7.66E-02 kg CO2 eq./pkm Worst transport, passenger train, high-speed, DE 17 Germany 6.10E-02 kg CO2 eq./pkm transport, passenger train, high-speed, FR 17 France 2.00E-02 kg CO2 eq./pkm transport, passenger train, high-speed, IT 17 Italy 4.88E-02 kg CO2 eq./pkm transport, passenger train, high-speed, RoW 17 Rest-of-World 7.14E-02 kg CO2 eq./pkm High spend train with green electricity 20 -5.70E-04 kg CO2 eq./pkm Best High speed train 20 EU 3.76E-02 kg CO2 eq./pkm Inter city train with green electricity 20 -7.10E-04 kg CO2 eq./pkm Inter city train 20 EU 4.63E-02 kg CO2 eq./pkm Regional train with green electricity 20 -8.00E-04 kg CO2 eq./pkm Regional train 20 EU 5.25E-02 kg CO2 eq./pkm Driving market for transport, passenger car, small size, petrol, EURO 5, GLO 17 Global 2.71E-01 kg CO2 eq./pkm market for transport, passenger car, small size, diesel, EURO 5,GLO 17 Global 2.34E-01 kg CO2 eq./pkm market for transport, passenger car, medium size, petrol, EURO 5,GLO 17 Global 3.40E-01 kg CO2 eq./pkm Base market for transport, passenger car, medium size, diesel, EURO 5,GLO 17 Global 3.09E-01 kg CO2 eq./pkm market for transport, passenger car, large size, petrol, EURO 5,GLO 17 Global 4.09E-01 kg CO2 eq./pkm Worst market for transport, passenger car, large size, diesel, EURO 5,GLO 17 Global 3.86E-01 kg CO2 eq./pkm market for transport, passenger car, electric, GLO 17 Global 2.26E-01 kg CO2 eq./pkm Best  Previous LCA studies mainly focused on quantifying the carbon footprint of in-person conferences, while only two of them presented results for other impact categories using the life cycle impact assessment (LCIA) method, including CML2001, USEtox, Eco-Indicator 99, and UBP 97 6,13 . Half of them focused exclusively on round-trip transportation 4, 5,9,11,12,29 while the other half considered life cycle stages of preparation, execution, catering, accommodation, and transportation [6][7][8]13,26 . However, due to differences in assumptions associated with in-person conferences (e.g., duration, size, and locations of the conference, geographical distribution of participants, transportation mode, system boundary, and selection of characterization factors), the carbon footprint ranges from 92 to 3540 kg CO2 eq. per capita. All of these studies identified transportation as the environmental hotspot. The conference site and geographical distribution of participants determine the transportation distance and mode for participants. From those who reported the average transportation distance, the average round-trip transportation distance varies from 1980 km 13 to 9564 km 9 . However, the average distance does not illustrate the complete picture of the participant transportation, and it was found that the 10-20% of participants with the most polluting trips contribute to a substantial portion (20-70%) of the total transportation-induced emissions 4-7,12 . These values depend on the distribution of participants, which reveals whether the conference is more localized or more internationalized. As shown in Supplementary Table S1, most participants are from the region where the conferences are held. The conference location is also important in determining the transportation profile, in which a conference location with better train connection to other major cities is capable of allowing more participants to transport by train and thus has more potential of reducing carbon footprint while a conference located in the southern hemisphere usually perform much worse in terms of carbon footprint compared to the northern hemisphere 11,12 . Bossdorf et al. suggested food and accommodation accounted for 18% and 13% of the total carbon footprint, respectively 8 . On the other hand, owing to the exclusive vegetarian menu and the much higher transportation emissions, Astudillo and Azarijafari reported that food and accommodation only accounted for 1% and 2% of the total carbon footprint 7 .
Recent studies compared the carbon footprint of in-person and virtual conferences 4,5,26 , which ranges from 0 to 5.87 kg CO2 eq. per capita. Among which, Ewijk and Hoekman assumed carbon neutrality for the virtual conference 4 ; Jäckle computed the carbon footprint from the electricity needed for devices and servers 26 ; Burtscher et al. considered emissions related to network, laptop, and Zoom-server 5 . Several studies considered multi-site conferences, yet the choices of locations are arbitrary 4,13 . Stroud and Feeley chose to optimize the conference location by minimizing the carbon footprint while restricting the potential locations to participants' origins 9 . Astudillo and Azarijafari considered the geometric median of all participants as the optimal conference location 7 . As discussed above, none of the existing studies has explicitly explored to what extent virtual conferences and multi-hub hybrid conferences with spatially optimized conference hubs and participant assignments can reduce the environmental impact of in-person conferences."

Reviewer's comment (2)
The paper has lots of information, but it is not always clear