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

The global events industry has grown rapidly, driving significant economic activity but also generating substantial greenhouse gas emissions. Prior life cycle assessments (LCAs) indicate per-participant footprints up to 3000 kg CO2e, implying the industry’s annual emissions are comparable in magnitude to those of large emitting nations. With the COVID-19 pandemic prompting a shift to virtual formats, a key question has emerged: should future conferences return to fully in-person, remain virtual, or adopt hybrid models, and what are the environmental trade-offs? While objections to virtual formats include digital fatigue and reduced informal interactions, proponents highlight improved accessibility and sustainability. Existing studies largely focus on travel emissions and often omit information and communication technology (ICT) and other life cycle stages. There remains limited quantitative understanding of environmental impacts across virtual, in-person, and hybrid conferences, especially regarding the role of hub selection and participant assignment. This study aims to quantify life cycle environmental impacts (carbon footprint, cumulative energy demand, and 17 ReCiPe midpoint indicators) across these modes and to analyze trade-offs between in-person interaction and carbon footprint using optimized hub selections and participant assignments.

The authors compare their work with prior conference LCA studies, noting earlier analyses primarily quantified carbon footprints of in-person events, with few covering broader impact categories (CML2001, USEtox, Eco-Indicator 99, UBP). Many studies focused only on transportation; others added preparation, execution, catering, and accommodation. Reported per-capita footprints varied widely (92–3540 kg CO2e), driven by assumptions (conference duration, size, location, participant distribution, transport modes, system boundaries, characterization factors). Transportation consistently emerged as the hotspot, with a small fraction of long-distance trips contributing disproportionately (10–20% of participants responsible for 20–70% of transport emissions). Conference location affects transport profiles (e.g., better train connections reduce emissions; southern hemisphere locations often fare worse). Food and accommodation contributions varied (e.g., 18% and 13% in one study vs 1% and 2% in another with vegetarian menus and high travel emissions). For virtual conferences, some studies assumed zero emissions or included limited ICT sources (network, laptops, servers), with per-capita footprints reported up to 5.87 kg CO2e. Multi-site conference analyses existed but often used arbitrary or simplified hub choices (e.g., participants’ origins or geometric median). No prior work comprehensively assessed hybrid conferences with optimized, geographically informed hub selection and participant assignment while including ICT and non-travel life cycle stages, leaving a gap this study addresses.

The study integrates life cycle assessment (LCA) with spatial optimization using data from the 2020 American Center for Life Cycle Assessment virtual conference (536 participants; geographic data available for 383; impacts estimated per average participant). Following ISO 14040, four phases were applied: goal and scope, life cycle inventory (LCI), life cycle impact assessment (LCIA), and interpretation. Scope and functional unit: One average participant; life cycle stages include food preparation, accommodation, preparation, execution, ICT, and transportation. Post-conference activities (e.g., downloading recordings, follow-up emails) are excluded. Impacts reported as carbon footprint (GHG), cumulative energy demand (CED), and 17 ReCiPe midpoint indicators (hierarchist perspective): agricultural land occupation, fossil depletion, freshwater ecotoxicity/eutrophication, human toxicity, ionizing radiation, marine ecotoxicity/eutrophication, metal depletion, natural land transformation, ozone depletion, particulate matter formation, photochemical oxidant formation, terrestrial acidification/ecotoxicity, urban land occupation, water depletion. Recycled content approach used for end-of-life modeling. Food and accommodation LCI: Country-level per-capita food supply from FAOSTAT maps to representative foods (fruits, vegetables, grains, dairy, protein foods, legumes, oils, animal fats, sugars), including distribution/preparation losses and utilities for preparation (per capita per day: 4.0 kWh electricity, 6.0 MJ heat, 28 kg water, 0.028 m3 wastewater). Accommodation utilities for hotel and guest home include electricity, thermal energy, water, wastewater; MSW composition and management from US EPA; fast/slow-moving consumer goods excluded. End-of-life uses recycled content approach. Preparation and execution LCI: Preparation includes activities of permanent/local committees, secretariat, participants (printing, computer usage), website maintenance, booklets, souvenirs (jute bags). Printing includes paper/toner production, printer electricity, and waste disposal. Execution includes venue electricity, water, and waste treatment. Material and energy inventories compiled from literature and databases. ICT LCI: Environmental impact framework for Internet services encompasses infrastructure, network, and servers specific to video-conferencing. Four layers considered: infrastructure energy, traffic shares (fixed line, IPv4, TCP, HTTP), traffic classes (video), and service shares. Mobile devices excluded; includes production/distribution of routers, access equipment, laptops; data center construction excluded due to limited LCIA data and dominance of operations. Network energy intensity computed from component shares and energy (packet-switched core, fixed line CPE, operator data center, office networks, Internet core), adjusted with 10% annual energy efficiency improvement to 2020. Server energy intensity derived from a Sweden ICT sector study (2015) extrapolated to 2020 with 10% efficiency improvement. Virtual conference data traffic based on 80% daily participation and 5.5 h/day online time; Zoom bandwidth 1.8 Mbps downstream and 2.6 Mbps upstream for 720p. Data center energy breakdown assumed: infrastructure 33%, network 3%, storage 11%, servers 53%. Transportation LCI: Participant origins geocoded; ground distances via Google Distance Matrix API; air distances as great-circle. Mode selection thresholds: rail (>600 km one-way in Europe) or car (>500 km one-way outside Europe) triggers air travel as primary mode; air trips include ground segments origin-to-airport and airport-to-hub. Rail distances based on public transit routes; nearest airports drawn from top 60 global and top 60 US busiest airports; otherwise, nearest practical airports via Google Flights. LCIA: Characterization factors primarily from Ecoinvent v3.7.1; air transport characterization refined using distance-dependent factors from Cox et al. (2018), fit to 2020 polynomial functions; additional factors for specific processes (e.g., egg, juice) from literature when missing. Spatial optimization: Facility location model selects hubs from alternative locations (top 30 global and top 15 US busiest airports) to minimize total participant travel distance, with constraints on assignments, minimum participants per hub (β=20), and number of hubs N (1–6). Hybrid scenarios: (1) Maximum Travel Distance (MTD) scenarios limit one-way travel distance (1000, 3000, 5000, 10,000 km) while prioritizing in-person attendance; distances exceeding limit penalized via big-M. (2) Maximum Virtual Participation (MVP) scenarios require minimum in-person share (i.e., cap virtual participation at 10%, 30%, 50%, 70%). Results represent upper bounds for realistic hybrid conferences since they are optimized hypotheticals. Sensitivity analyses: Examined parameter uncertainty for both in-person and virtual scenarios, including ICT energy efficiency trajectories, dietary types, air transport characterization factors, and travel distances; assessed impacts of virtual participation rates and postulated practices (e.g., asynchronous options) qualitatively for system-boundary exclusions.

- Switching from in-person to virtual conferencing reduces carbon footprint by 94% and CED by 90% per participant. - Baseline virtual conference per-participant impacts: 46 kg CO2e and 767 MJ equivalent; dominated by food preparation, accommodation electricity, and ICT. Food-related impacts dominate agricultural land occupation, terrestrial ecotoxicity, and water depletion. ICT is a hotspot (notably metal depletion due to device production). - In-person (1-hub) per-participant footprint reported at 840 kg CO2e (consistent with literature) and comparable to an average U.S. citizen’s monthly footprint (2018). For all participants, transportation emissions total 280 t CO2e, with 50% arising from round-trip distances >10,000 km. The most polluting 10–20% of trips contribute disproportionately to total emissions. - Transportation dynamics: For a single passenger, driving emissions equal air emissions at ~500 km; beyond this, per-km air emissions tend to be lower than car. Air travel dominates multiple midpoint categories (fossil depletion, marine eutrophication, natural land transformation, ozone depletion, photochemical oxidant formation) due to kerosene production and flight operations (LTO and CCD phases). Car travel dominates freshwater and marine ecotoxicity, metal depletion, and terrestrial acidification due to vehicle and fuel life cycles. - Multi-hub in-person scenarios: Adding hubs reduces air travel distances and cuts carbon footprint and CED roughly by half compared to 1-hub, but even with 6 hubs, footprints remain 739% (carbon) and 637% (CED) higher than virtual. Increasing hubs can increase driving distances, making some impact categories (freshwater/marine ecotoxicity, terrestrial acidification) worse at higher hub counts due to car travel. - Hybrid trade-offs (MVP scenarios): With optimized spatial hubs and participant assignments, capping virtual participation at 50% reduces in-person conference carbon footprint by about two-thirds across hub counts; at 70% virtual, air travel is eliminated, yielding >80% reduction. - Hybrid trade-offs (MTD scenarios): Prioritizing in-person attendance while limiting maximum one-way travel distance performs poorly if the limit is large. With limits of 3000, 5000, or 10,000 km, virtual shares remain low (<4% multi-hub; <13% 1-hub), and carbon footprints can exceed in-person scenarios (up to 1799 kg CO2e per capita for 1-hub with 10,000 km limit). Setting maximum travel distance below ~3000 km and/or using more hubs yields substantial footprint reductions. Effectiveness depends on participant geospatial distribution and hub count. - Sensitivity: In-person results are highly sensitive to air transport characterization factors and distances; more hubs reduce variability. Virtual impacts are sensitive to daily participation, dietary type (ovo-vegetarian yields lower footprint than vegan due to egg vs soybean differences), residential energy use, ICT energy intensity (network improvements to 2030 could reduce per-participant footprint by ~2.9–5.9%). - Grid mix matters: Benefits of virtual conferences diminish when many participants are in regions with carbon-intensive electricity mixes; increasing renewable penetration enhances virtual performance.

The study addresses whether virtual and hybrid conferences can mitigate climate impacts relative to in-person formats by quantifying full life cycle impacts and exploring spatially optimized hybrid designs. Results demonstrate that virtual conferences dramatically reduce carbon and energy burdens primarily by avoiding air travel, although food, residential electricity, and ICT become dominant hotspots. Multi-hub in-person configurations reduce air travel distance but cannot match virtual performance; increasing hubs eventually raises car travel impacts in some categories. Hybrid conferences provide a practical balance: with optimized hub placement and participant assignment, organizers can maintain substantial in-person interaction while achieving large emissions reductions. MVP scenarios produce a clear trade-off curve (Pareto front) between in-person share and carbon footprint, highlighting feasible configurations (e.g., ≤50% virtual yields ~two-thirds reduction). MTD scenarios show that without stringent distance limits and adequate hub numbers, prioritizing in-person attendance undermines mitigation benefits. The findings emphasize the importance of spatial optimization, diet choices, and ICT efficiency improvements, and they highlight regional electricity mix as a key factor affecting virtual conference benefits. Sensitivity analyses reinforce the need to minimize stopovers and long-haul flights for in-person attendees and to provide asynchronous virtual participation options to manage participation-related uncertainties while acknowledging potential unaccounted post-conference ICT use.

This work provides a holistic life cycle comparison of in-person, virtual, and hybrid conferences, integrating transportation, food, accommodation, preparation, execution, and ICT. Transitioning to virtual conferences can reduce per-participant carbon footprints by 94% and energy use by 90%. Strategically designed hybrid conferences with optimized hubs and participant assignments can cut impacts by roughly two-thirds while retaining substantial in-person engagement (<50% virtual). The analysis identifies environmental hotspots (air travel for in-person; food, residential electricity, and ICT for virtual) and offers actionable levers: minimize long-haul flights and stopovers, select geographically optimal hubs, encourage plant-forward diets (e.g., ovo-vegetarian), improve residential energy efficiency, and advance ICT energy performance. Diminishing returns appear with many hubs or very high virtual shares, suggesting organizers should balance logistical complexity with environmental gains. Future research should: (1) quantify consequential and post-conference ICT-related impacts (e.g., streaming recordings, follow-up communications); (2) refine regionalized modeling of electricity and food systems for globally distributed participants; (3) incorporate dynamic attendance behaviors and uncertainties in participation; (4) evaluate broader social outcomes (equity, accessibility, networking) alongside environmental metrics; and (5) extend to other event types and participant distributions to generalize hub selection and participation strategies.

- System boundary excludes post-conference activities (e.g., downloading/streaming recordings, follow-up emails, material searches), potentially underestimating virtual impacts. - Participant geolocation available for 383 of 536 attendees; results are based on this subset and may not fully represent all participants. - Hybrid and in-person hub scenarios are hypothetical and optimized to minimize travel distance using busiest airports as candidate hubs; real-world constraints (venue availability, costs, scheduling) and deviations from registration data may reduce achievable benefits. - Air transport characterization relies on distance-dependent factors and polynomial fits to 2020 projections; results are sensitive to these choices and to actual travel distances and stopovers. - ICT modeling excludes data center construction impacts due to limited LCIA data and assumes efficiency improvement rates; device use assumptions (e.g., laptops only, excluding mobile devices) and bandwidth estimates may vary. - Electricity grid mixes are regionalized but limited by available datasets; participants residing in regions with carbon-intensive grids can alter outcomes. - Dietary modeling uses representative foods from FAOSTAT and literature; actual conference diets vary and influence results. - Mode choice thresholds (500 km car, 600 km rail) and airport selection pools introduce discretization that may not capture individual behaviors.