Near-real-time monitoring of global CO2 emissions reveals the effects of the COVID-19 pandemic

The study addresses how the COVID-19 pandemic altered human activities, energy consumption, and associated CO2 emissions, resolving uncertainties stemming from the lack of timely national inventories that typically lag by 1–2 years. Prior estimates based on limited power plant data, satellite proxies of pollutants, and agency demand projections suggested significant emission drops but lacked daily, sector-specific, and country-level detail. The authors aim to provide near-real-time, daily, sectoral, and country-resolved CO2 emission estimates to precisely quantify the magnitude, timing, and sectoral drivers of changes due to COVID-19 and to separate these effects from seasonal, weekly, and holiday variations. Using the Carbon Monitor framework, they construct daily emissions from Jan 1, 2019 to Jun 30, 2020 to evaluate global and regional dynamics before and during the pandemic.

Existing literature and reports indicated notable but variably estimated declines in emissions early in 2020. The International Energy Agency projected about a -5% decline in global CO2 emissions for Jan–Apr 2020 versus 2019 based on monthly energy demand data. Le Quéré et al. estimated a 17% reduction in daily CO2 emissions in April 2020 compared to the 2019 daily mean, scaling declines to confinement indices. Initial satellite observations of atmospheric pollutants and limited power plant data also suggested large decreases. However, these approaches lacked comprehensive daily, sectoral, and country-specific resolution and did not uniformly include process emissions (e.g., cement). The present study builds on and extends these efforts by providing globally consistent, daily, sector-specific estimates using multiple near-real-time activity datasets.

Baseline and scope: The study estimates daily, sector-specific CO2 emissions for countries/regions from Jan 1, 2019 to Jun 30, 2020 (with some sectors to Jul/Aug 2020), covering fossil fuel combustion and industry process emissions (including cement). Baseline 2019 national totals and sectoral structures are based on EDGAR v5.0 (2018) scaled to 2019 using growth rates (Liu et al., Global Carbon Project). China’s totals were computed separately due to known inventory uncertainties, using China Energy Statistical Yearbook and Statistical Communiqué growth rates. General emissions formulation: Emissions were computed as the product of activity data and emission factors, with sectoral and fuel structure assumed unchanged between 2019 and 2020. Daily 2020 emissions changes were derived from near-real-time activity data relative to 2019. Sectors and activity datasets: - Power generation: Near-real-time hourly/daily electricity data used to estimate thermal generation changes. Data sources include: China (daily thermal generation; coal consumption proxies for late May–June), India (POSOCO daily by fuel), U.S. (EIA Hourly Electric Grid Monitor), EU27 & UK (ENTSO-E), Russia (SO-UPS), Japan (OCCTO), Brazil (ONS). For other countries (collectively 26% of global power emissions), changes were inferred using lockdown timing (from compiled sources) and average change rates observed in countries with closures. A temperature correction was applied where power generation correlated with temperature (R2>0.5) to attribute winter differences (Jan–Mar) to weather vs. COVID-19. - Industry and cement production: For China, four sub-sectors (steel, cement, chemicals, other industries) used monthly production indices as activity proxies (with process emissions for cement). For U.S., EU27 & UK, India, Russia, Japan, Brazil, monthly/cumulative Industrial Production Index (national sources/Eurostat) was used; June values forecast via Trading Economics where needed. Daily allocation of industry totals used daily thermal production patterns. Rest-of-world changes followed analogous approaches. - Ground transportation: Used TomTom congestion index data for 416 cities in 57 countries (hourly to 15-min resolution). A sigmoid function relating congestion (X) to vehicle counts (Q) was calibrated using Paris traffic counts and applied to other cities to infer relative daily activity; city results aggregated to national using EDGAR road transport weights (2010 spatial distribution). Countries without TomTom coverage adopted mean regional/global patterns. Uncertainty from prediction intervals of the Q–X regression. - Aviation: FlightRadar24 data on daily flights and distances were used to compute emissions via a constant CO2-per-km factor calibrated to ICCT’s 2018 estimate (scaled +3% to 2019). Great-circle distances were summed per flight; domestic vs international classified by origin/destination. Global totals include international flights; country allocations include domestic only. Assumes similar fleet mix and completeness between 2019 and 2020. - International shipping: Daily 2019 emissions assumed flat intra-annually; changes in 2020 linked linearly to estimated -25% change in shipped volume for H1 2020 based on industry reports. - Residential and commercial buildings (direct fuel use): Daily emissions inferred from population-weighted heating degree days (HDD) by country (ERA5 2-m air temperature) with EDGAR 2018 baseline split into cooking (assumed stable) and heating (HDD-dependent). Case studies of daily natural gas data (France, Italy, Belgium, Spain) after temperature normalization suggested minimal intrinsic change due to lockdowns; this assumption was extended globally. Global heating demand in the first seven months of 2020 was 2.1% lower than in 2019 due to an anomalously warm winter. Attribution and validation: - Temperature attribution in power: For Jan–Mar 2020, removing temperature effects suggested ~85% of observed power-sector reductions were attributable to COVID-19 impacts, with the remainder due to warmer weather. - Atmospheric verification: Changes in tropospheric NO2 columns (OMI/TROPOMI) and surface NO2/CO/PM2.5 over China, U.S., EU4, and India were analyzed (2013–2020 context). Observed NO2 declines were consistent with inventory-based reductions by sector and timing of lockdowns. Uncertainty: Sector-specific uncertainties followed IPCC 2006 guidelines: power (±14%), industry (notably 20% for China’s share; Monte Carlo for monthly-to-annual inference), ground transport (prediction interval of congestion–counts regression, with unquantified additional structural uncertainty), aviation (difference between distance-based and flight-count methods), shipping (13% per IMO), residential (2σ ~40%). 2019 projection uncertainty ~2.2%. Total uncertainties propagated using IPCC error propagation for sums and products. Data and code: Daily emissions and sectoral time series are available at Carbon Monitor (carbonmonitor.org/.cn); satellite and air quality datasets from OMI, MODIS, GOSAT, and national agencies. Code available upon request from corresponding author.

- Global decline: CO2 emissions fell by 8.8% (−1551 Mt CO2) in the first half of 2020 relative to H1 2019, the largest first-half-year drop on record, exceeding declines during past economic crises and World War II. Mean daily emissions Jan–Jun 2020 were 88.4 Mt CO2/day vs 98.2 Mt CO2/day in 2019 (−10%). The largest monthly global drop occurred in April 2020 (−16.9% vs 2019). - Recovery dynamics: Emissions began to recover in late April–May as restrictions eased, with China rebounding fastest; by June, power-sector emissions were only 1.1% lower year-on-year (vs −9.7% in April). Mobility-related emissions remained more depressed: ground transport was −13.0% in July 2020 vs July 2019 (after −38.6% in April and −32.6% in May). - Country/region declines in H1 2020 (Mt CO2, %): U.S. −338.3 (−13.3%); EU27 & UK −205.7 (−12.7%); India −205.2 (−15.4%); China −187.2 (−3.7%); Japan −43.1 (−7.5%); Russia −40.5 (−5.3%); Brazil −25.9 (−12.0%). Additional examples: Germany −54.0 (−15.1%); France −21.5 (−14.2%); UK −26.8 (−15.0%); Italy −22.9 (−13.7%); Spain −23.1 (−18.8%). Timing of declines aligned with national lockdowns, with China’s earliest and sharpest initial drop and rapid rebound (Feb −18.4%, Mar −9.2%, Apr +0.6%, May +5.4% vs 2019 months). - Sectoral contributions to global H1 2020 decrease: Ground transportation −613.3 Mt (40%); power −341.4 Mt (22%); industry −263.5 Mt (17%); aviation (domestic + international) −200.8 Mt (13%); international shipping −89.1 Mt (6%); residential/commercial fuel use −42.5 Mt (3%). - Sectoral percentage changes: Power −5.0% globally (China −1.4%; U.S. −7.6%; India −12.7%; EU27 & UK −19.3%); Industry −5.5% globally (China −2.1%; U.S. −9.1%; EU27 & UK −14.1%; India −22.1%); Ground transportation −18.6% H1 (−17.8% through July); Domestic aviation −35.8%; International aviation −52.4% (July 2020 still −72% vs July 2019); International shipping about −25% H1; Residential/commercial direct fuel use −2.2% (largely temperature-driven). - Attribution: Warm early-2020 temperatures reduced heating demand; after correcting for temperature, ~85% of power-sector reductions in Jan–Mar 2020 were attributable to COVID-19 impacts. - Atmospheric consistency: Satellite tropospheric NO2 and ground-based air quality data showed large declines consistent with inferred fossil fuel emission reductions (e.g., China NO2 −20.2% Jan–May 2020 vs 2019; substantial decreases over U.S. and EU4; weaker over India).

Near-real-time, daily, sector-specific emissions data reveal that the magnitude and timing of CO2 emission reductions in early 2020 were closely synchronized with the implementation of COVID-19 lockdowns and mobility restrictions across countries. The analysis separates holiday/seasonal effects and temperature anomalies from COVID-19 impacts and demonstrates that transportation (especially ground and aviation) and power/industry activity declines drove most of the global reduction, while residential direct fuel use changed little after accounting for weather. Rapid rebounds, particularly in China, indicate that without structural changes, emission declines linked to activity suppression are transient. The persistent deficits in the U.S., Brazil, and India by late June mirror ongoing pandemic intensity. These findings underscore that achieving climate targets requires transformational shifts in energy systems, transportation decarbonization, and efficiency improvements rather than relying on reduced activity. The daily estimates also provide a basis for monitoring recovery pathways, assessing changes in carbon intensity (emissions per GDP), and informing adaptive policy to prevent a carbon-intensive rebound and to leverage opportunities for a green recovery.

The study delivers the first global, daily, sector- and country-resolved estimates of CO2 emissions spanning pre- and mid-pandemic periods, quantifying an unprecedented 8.8% (−1551 Mt) global decline in H1 2020 driven primarily by ground transport, power, industry, and aviation. It demonstrates strong temporal alignment with lockdowns, rapid partial rebounds as restrictions eased, and cross-validation with atmospheric observations. The methodology and data infrastructure (Carbon Monitor) enable near-real-time tracking of emissions to inform timely policy responses and to evaluate recovery trajectories and structural changes in carbon intensity. Future work should: refine sectoral proxies (e.g., inter-/intra-city transport differences), integrate broader real-fuel-use data for residential/commercial buildings, improve aviation and shipping activity representation, extend coverage beyond mid-2020, and couple with economic indicators to robustly assess decoupling and policy effectiveness.

- Activity proxy assumptions: Ground transport relies on a Paris-calibrated congestion–traffic relationship applied globally and mainly reflects intra-city conditions; inter-city changes may be underrepresented. Residential/commercial estimates assume fuel use changes are driven primarily by temperature (HDD), extrapolating from limited European gas data; intrinsic behavioral shifts may be missed. Shipping assumes flat intra-annual seasonality in 2019 and linear linkage to shipment volume changes. - Data completeness and scaling: Aviation emissions use a constant CO2-per-km factor and FlightRadar24 coverage with calibration to ICCT 2018; fleet mix or coverage changes could bias results. Some 2019 baselines are projected (±2.2% uncertainty). Lockdown timings for some countries are compiled from secondary sources. - Temperature confounding: Warm early-2020 conditions reduced heating demand; while corrections were applied for power where possible, residual attribution uncertainties remain. - Uncertainty quantification: While sector-specific uncertainties are provided (e.g., power ±14%, residential 2σ ~40%), some structural uncertainties (e.g., applying the congestion model globally) are acknowledged but not fully quantified due to data limitations. - Temporal coverage: Sectoral series extend to different end dates (e.g., ground transport and aviation to July/August 2020), and some industrial indices required forecasting for June in certain regions.