key: cord-0531014-8gfibnv2 authors: Liu, Zhu; Deng, Zhu; Ciais, Philippe; Lei, Ruixue; Davis, Steven J.; Feng, Sha; Zheng, Bo; Cui, Duo; Dou, Xinyu; He, Pan; Zhu, Biqing; Lu, Chenxi; Ke, Piyu; Sun, Taochun; Wang, Yuan; Yue, Xu; Wang, Yilong; Lei, Yadong; Zhou, Hao; Cai, Zhaonan; Wu, Yuhui; Guo, Runtao; Han, Tingxuan; Xue, Jinjun; Boucher, Olivier; Chevallier, Frederic; Boucher, Eulalie; Wei, Yimin; Zhang, Qiang; Guan, Dabo; Gong, Peng; Kammen, Daniel M.; He, Kebin; Schellnhuber, Hans Joachim title: COVID-19 causes record decline in global CO2 emissions date: 2020-04-28 journal: nan DOI: nan sha: e7a821216e5187358350e37331cf9838043a0c92 doc_id: 531014 cord_uid: 8gfibnv2 The unprecedented cessation of human activities during the COVID-19 pandemic has affected global energy use and CO2 emissions from fossil fuel use and cement production. Here we show that the decrease in global fossil CO2 emissions during the first quarter of 2020 was of 5.8% (542 Mt CO2 with a 20% 1-{sigma} uncertainty). Unlike other emerging estimates, ours show the temporal dynamics of emissions based on actual emissions data from power generation (for 29 countries) and industry (for 73 countries), on near real time activity data for road transportation (for 132 countries), aviation and maritime transportation, and on heating degree days for commercial and residential sectors emissions (for 206 countries). These dynamic estimates cover all of the human induced CO2 emissions from fossil fuel combustion and cement production. The largest share of COVID-related decreases in emissions are due to decreases in industry (157.9 Mt CO2, -7.1% compared to 2019), followed by road transportation (145.7 Mt CO2, -8.3%), power generation (131.6 Mt CO2, -3.8%), residential (47.8 Mt CO2, -3.6%), fishing and maritime transport (35.5Mt CO2, -13.3%) and aviation (33.4 Mt CO2, -8.0%). Regionally, decreases in emissions from China were the largest and earliest (-10.3%), followed by Europe (EU-27&UK) (-4.3%) and the U.S. (-4.2%). Relative decreases of regional CO2 emissions are consistent with regional nitrogen oxides concentrations observed by satellites and ground-based networks. Despite the unprecedented decreases in CO2 emissions and comparable decreases in economic activities, we monitored decreases in the carbon intensity (Emission per unit of GDP) in China (3.5%), the U.S. (4.5%) and Europe (5.4%) over the first quarter, suggesting that carbon-intensive activities have been disproportionally impacted. COVID-19 has caused hundreds of thousands of deaths worldwide since December of 2019, together with large-scale ongoing reductions in human activities and profound effects on different national economies. Industrial production and energy consumption in some countries were reported to decline by up to 30% in just a few weeks 2,3 as lockdowns were imposed to protect public health. Fossil fuel and cement CO2 emissions are directly linked to human activities. Initial estimates of emissions changes based on a limited sample of power plants and indirect satellite observations of atmospheric pollutants 4, 5 have suggested that we may be witnessing the largest drop of emissions since the end of the Second World War. However, quantitative analyses of the pandemic's effects on energy use and CO2 emissions for all key sectors are still lacking 1, 6 . A major limitation is that detailed inventories of energy and fuel use that have historically been used to assess CO2 emissions are only available with a lag of one or two years [7] [8] [9] [10] [11] [12] . Here, we compile and analyze high-frequency and up-to-date electricity-related emission data, emission estimates by the industry, along with online data-streams of human activity that are closely related to emissions in order to provide a near real time assessment of the impacts of the COVID-19 pandemic on country-and sector-specific CO2 emissions during the first quarter of 2020 and to enable a monitoring of how CO2 emissions will change in the following period. Details of our data sources and analytic method are provided in the Methods section. In summary, we calculate regional CO2 emissions between January and April 2020 and compare them to the same period in 2019, drawing on hourly datasets of electricity power production and CO2 emissions in 29 countries (including the substantial variations in carbon intensity associated with electricity production), three different indexes of daily vehicle traffic / mobility in 416 cities worldwide, monthly production data for cement, steel and other energy intensive industrial products in 73 countries, daily maritime and aircraft transportation activity data, as well as proxies for the residential and the commercial building emissions (see Extended Data Table 1 for data sources). Together, these data cover almost all fossil and industry sources of global CO2 emissions. It is important to note that during the winter, CO2 emissions from the power sector, as well as from residential and commercial buildings are highly sensitive to temperature (i.e. heating degree days-HDD 7 ). The first months of 2020 were exceptionally warm across much of the northern hemisphere, meaning that 2020 CO2 emissions would have been somewhat lower than the same period in 2019 even without the disruption in economic activities and energy production caused by COVID-19 and related lockdowns. Fig. 2a) , and relatively small decreases in residential emissions (-47.8 Mt CO2, 9% of the total; yellow in Fig. 2a ). The rest of the reduction comes from aviation, ships emissions. Our estimates of power sector emissions rely on near real time hourly data (except for China and Japan where only monthly data are available, SI Figure 1 ) of electricity production and the carbon content of the electricity mix. Thus, we are able to resolve the effects of weather-driven variations of renewable electricity supply (See Methods for temperature adjustment) as well as the increases in natural gas relative to coal for power generation in 2020 in the U.S. that has been caused by very low oil prices. Figure 2b shows that in the first quarter of 2020, global CO2 emissions from the power sector declined by -3.8% (-131.6 MtCO2), with a large decline in Industry emissions from steel, chemicals and other manufactured products from fossil fuel combustion and the cement production process represent on average 29% of the global CO2 emissions, with a much larger share of national emissions in developing countries (39% in China and 33% in India) 13 . We collected data separately for steel, chemicals (based on 8 chemical products) and 26 other industrial products, as well as for cement production in China, and data for steel and cement production in India, for a better attribution of industrial CO2 emissions changes (SI Data Table 4 .1-4.5). Only emissions from direct fuel consumption and chemical process emissions by the industry sector were considered, electricity related emissions for the industry being already counted in the power generation sector. In the first quarter of 2020, Table 4 .1-4.5). In China, emissions from steel production (41.6% of national industrial emissions from fuel combustion) increased by +1.4% and +5.0% in January and February, and then dropped by -1.7% in March, compared to 2019. Despite the COVID-19 pandemics, emissions from the steel industry thus remained 1.4% larger in the first quarter of 2020. Emissions from ground transportation (SI Figure S4 respectively (SI Figure S4 ). Emissions from global aviation decreased by 8% during Q1 (-33.4MtCO2), among which those from international aviation decreased by -19.8MtCO2 (SI Figure S5 other 19 ship classes including fishing, show a reduction of -11.5%. We estimated that emissions from commercial and residential buildings declined by 3.6% in the first quarter of 2020 compared to the same period in 2019. This sector is the one for which emission changes have the largest uncertainty, as there is no actual fuel use data worldwide. The emissions from fuel use (oil and gas) in commercial and residential buildings was estimated using population-weighted heating degree days by the ERA5 15 reanalysis of 2 meters air temperature for 206 countries. We found that the global heating demand in Q1 declined by -5.0% compared to 2019, owing to the abnormal warm northern-hemisphere winter 16 , resulting in the decreased of emissions. Note that the effect of increased time spent in households on emissions was not accounted for. Our estimates of decreases in fossil and industry CO2 emissions (See Methods and SI Table S9 ) are consistent with observed changes in nitrogen dioxide (NO2) emissions, which are also mainly produced by fossil fuel combustion. Tropospheric NO2 column concentration data from satellites 17, 18 , and surface NO2 concentrations from air quality stations show a decrease (See Table 2 ) consistent with the reduction of fossil carbon fuels emissions presented above. Overall, NO2 decreased over China in the first quarter of 2020 is consistent with our calculated NO2 emission declines based on near real time activity and emission data (See Methods and SI Table S9 ),) . Over the U.K., France, Germany, and Italy, NO2 decreased by a similar amount than in the U.S.. Over India, NO2 showed a weaker decline, also consistent with satellite data. Overall, NO2 declines in January and February over China are the largest declines since the OMI data become available in 2004 (SI Figure S10 ). The consistent results from both ground based and satellite monitoring systems confirm the significant decline of the NO2 concentrations due to COVID-19(See Methods). Based on the OMI satellite data, Over the U.K., France, Germany, and Italy, NO2 decreased by a similar amount than in the U.S.. Over India, NO2 showed a weaker decline, also consistent with satellite data. It is still unclear to what extent annual CO2 emissions will be continue to be affected by the COVID-19 pandemic, which will depend on the efficacy and stringency of public health policies and the recovery of economies and human activities around the world. The IMF predicts that the global annual economic output (GDP) will decrease by -3.0% in 2020, which is worse than the financial crisis in 2008 19 , and yet this projection was based on the assumption that the COVID-19 epidemics will fade globally in the second half of this year. Based on near real time activity and emission data, we estimate a decrease of 5.8% (542 MtCO2) of global CO2 emissions in Q1 of 2020, the largest global quarterly emission decrease ever recorded, larger than during the 2009 economic crisis. This signal reflects unprecedented impacts of the pandemics on the global economy. Nevertheless, given negative impacts on the carbon intensive industry sector such as cement production, we inferred improvements of the ratio of the emission intensity (CO2 emissions per unit of GDP) in China (3.5%), US (4.5%) and Europe (1.8%), although such improvement are the consequence of highest-ever cost payed for the reduction of 1t of CO2between 1k and 10k US$ per ton, it still suggesting a unique opportunity for green investments and low carbon development in the years to come, for which global concerted efforts will urgently be needed. The ability to monitor trends in emissions in near real time that we demonstrate here will be invaluable in adaptively managing the transition. The CO 2 emissions and sectoral structure in 2018 for countries and regions are extracted from EDGAR V4 3.2 10 , and the emissions are scaled to the year 2019 based on the growth rates from Liu et al. 20 and the Global Carbon Budget 2019 21 . For countries with no current estimates of emission growth rates in 2019 such as Russia, Japan and Brazil, we assume their growth rates of emissions were 0.5% based on the emission growth rates of rest of world 21 . Given the large uncertainty of CO 2 emission in China 22,23 , we calculated CO 2 emissions based on the methodology developed 24 previously: , , reflect the regions, sectors and fuel types respectively. In our calculation, covers XX countries that representing 70% of global total emissions. covers four sectors that are power generation, industry, transportation and household consumption, covers three primary fossil fuel types which are coal, oil and natural gas. Emission factors can be further separated into the net heating values for each fuel "v", the energy obtained per unit of fuel (TJ per t fuel), the carbon content "c" (t C TJ-1 fuel) and the oxidization rate "o", which is the fraction (in %) of fuel oxidized during combustion and emitted to the atmosphere. We assumed that the emission factors and the structure remain unchanged for each country in 2020 when comparing with 2019. Thus, the rate of change of the emission is calculated based solely on the change of the energy consumption data in Q1 of 2020 compared to Q1 of 2019. Based on the assumption of sectoral carbon intensity and energy structure remain unchanged from 2018, the EDGAR sectors were aggregated into four main sectors, including power sector, transport sector, industry sector and residential sector. We used actual near-real time emissions data for the power sector, industry emission data (including cement), and scaled CO 2 emissions to 2020 by using proxy data for other sectors. We include in the estimate of global emissions those from domestic and international aviation and shipping. For China, only monthly electricity generation statistics are available on the National Bureau of Statistics in China (http://www.stats.gov.cn/tjsj/). Then the electricity generation data were disaggregated to daily emission data by the daily coal consumption of six main power company (https://www.wind.com.cn/). For Japan, we collect the daily electricity generation data from 10 electricity companies in Japan (Hokkaido Electric Power, Tohoku Electric Power Network, Tokyo Electric Power Company, Chubu Electric Power Grid, Hokuriku Electric Power Transmission & Distribution Company, Kansai Electric Power, Chugoku Electric Power Company, Shikoku Electric Power Company, Kyushu Electric Power and Okinawa Electric Power Company). For US, we used daily total electricity generation data of 48 states from Energy Information Administration's (EIA) Hourly Electric Grid Monitor (https://www.eia.gov/beta/electricity/gridmonitor/). For EU countries and UK, electricity generation data every 15 minutes by production types are collected from ENTSO-E Transparent platform (https://transparency.entsoe.eu/dashboard/show) and near real time variations of the carbon content of electricity was taken from Electricity Map (https://www.electricitymap.org/). Due to the poor data quality or missing data, Croatia, Cyprus, Ireland, Luxembourg and Malta are excluded from the calculation. The hourly data of the other 23 EU countries (including Austria, Belgium, Bulgaria, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Italy, Latvia, Lithuania, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden) and United Kingdom are aggregated into daily total electricity generation data. For India, daily total electricity generation data are updated by Power System Operation Corporation Limited (https://posoco.in/reports/daily-reports/). For Russia, daily electricity generation data are collected from United Power System of Russia (http://www.soups.ru/index.php). For Brazil, daily electricity generation data are downloaded from the Operator of the National Electricity System (http://www.ons.org.br/Paginas/). All the daily electricity generation data are further aggregated into monthly data to calculate the monthly growth rates. We use the linear regression between electricity consumption and daily temperature in the winter of 2018 and 2019, and correct temperature effect to the electricity generation in 2020. For the monthly emissions in 2019, we use the monthly electricity generation for countries to allocate the emissions in power sector for each month. Then we calculate the monthly emissions in 2020 based on the monthly growth rates compared to the same periods in 2019. For China, the industrial sector is divided into four industries including steel industry, cement industry, chemical industry, and other industry based on the structure of industrial emissions calculated by IEA 25 data. For the steel industry, we collect the global monthly crude steel production data from World Steel Association (https://www.worldsteel.org/). For cement industry, China's monthly cement production data are available on the National Bureau of Statistics. We estimate the growth rate of chemical production in China chemical industry by calculating the average cumulative growth rates of sulfuric acid, caustic soda, soda ash, ethylene, chemical fertilizer, chemical pesticide, primary plastic and synthetic rubber. For other industry, we estimate the change in China other industry by calculating the average cumulative growth rates of 26 industrial products (crude iron ore, phosphate ore, salt, feed, refined edible vegetable oil, fresh and frozen meat, milk products, liquor, soft drinks, wine, beer, tobaccos, yarn, cloth, silk and woven fabric, machine-made paper and paperboards, newsprint, plain glass, ten kinds of nonferrous metals, refined copper, lead, zinc, electrolyzed aluminum, industrial boilers, metal smelting equipment, and cement equipment). Based on the emission distribution of these four industries in industrial sector in China, we finally get the annual growth rate of industrial sector in China. For US and Russia, we use the cumulative Industrial Production Index (IPI) from Federal Reserve Board (https://www.federalreserve.gov) and Federal State Statistics Service (https://eng.gks.ru) respectively, to estimate the growth rates of the emissions in these countries. For India, the industrial sector is divided into steel industry, cement industry, and other industry. Emissions from steel industry and cement industry account for 60% of the total emissions in industrial sector of India. The steel production in India is extracted from World Steel Association. India's monthly cement production data could be found from the Office of the Economic Advisor, but the last observation was on February. We assume that the cement production in India has dropped 40% in March. For other industry, we assume the growth rate is the average of growth rates of steel production and cement production. For Europe (EU27 & UK), Japan and Brazil, we use the cumulative Industrial Production Index (IPI) to estimate the growth rates of emissions in these countries or regions, collected from Eurostat (https://ec.europa.eu/eurostat/home), Ministry of Economy, Trade and Industry (https://www.meti.go.jp), and Brazilian Institute of Geography and Statistics (https://www.ibge.gov.br/en/institutional/the-ibge.htm) respectively. However, the last observation in EU27 and UK was in January 2020 and those in Japan, Russia and Brazil were in February 2020. To estimate the current growth in March 2020, for EU27 and UK, we assume that the monthly growth rate in February 2020 equals to that of the same month in 2019. The monthly growth rate in March 2020 in EU27 and UK together has fell 5%, based on the assumption of the average growth rate of industrial sector in China (-8.1%) and US (-2.2%) because of the implement of locking down policies later than China but earlier than US. For Japan and Brazil, as looser travel restriction in these countries, we assume the monthly growth rates in March were -3% compared to February. The monthly industrial emissions are allocated to daily emissions by electricity data. For the emissions from chemical process during cement production, we only had access to data from China and India which nevertheless account for 63% of the global cement production and CO 2 emissions. We collected monthly cement production in China and India and calculated emissions from cement production by multiplying emission factors of cement production process as: = × The mobility trends from the Google COVID-19 Community Mobility Reports (https://www.google.com/covid19/mobility/) are mainly used in this study. The mobility changes are calculated based on the visits and visit duration in six places categories, including grocery & pharmacy, parks, transit stations, retail & recreation, workplace, and residential. The daily mobility changes are compared to the median value during the first five weeks in 2020 for the corresponding day of the week as the baseline value. Except for residential, we consider that the mobility changes of visits in the other five places representing the total mobility, so we calculate the daily average mobility change of five of the place categories to represent as the changes of the total traffic volume in these countries, based on the assumption the mobility in the same period in 2019 and January in 2020 is at the same level of the baseline value. The average mobility of EU countries (except for Cyprus due to the lack of data) and UK is used to estimate the emission growth of Europe. In China, we use the change of the Baidu Migration Scale Index to characteristic the change of transport sector, due to the lack of data in Google Community Mobility Reports in China. This index is aggregated from migration flows within the nation based on the positioning requests on Baidu Map services. The migration scale index is reported since Jan. 12 th in 2019 and Jan. 1 st in 2020, 9 days before the start of the Spring Festival travel rush every year. To fill the data gap between Jan. 1 st 2019 to Jan. 11 th , 2019, we use the average value of migration index between Jan. 12 th , 2019 and Jan. 20 th , 2019. Then we aggregate the daily index into monthly total and calculate the monthly changes in 2020 compared to 2019. In Russia, because the lack of data and the earliest lockdown policy implementing on Mar. 30 th in Moscow, we assumed that the transport emissions remain unchanged in the first quarter in 2020. In addition, we collected TomTom congestion global level data from TomTom website (https://www.tomtom.com/en_gb/traffic-index/). The congestion level represents the extra time spent on a trip, in percentage, compared to uncongested condition. TomTom congestion level data are available for 416 cities across 57 countries at a temporal resolution of 10 to 15 minutes. We compared the time series of TomTom congestion level in the first quarter of 2019 and 2020. The emission changes were first calculated for individual cities, and then weighted by city emissions to aggregate to national changes. Although both the International Civil Aviation Organization (ICAO) and the International Air Transport Association (IATA) publish yearly statistics of aircraft operations, CO2 emissions from commercial aviation are usually reconstructed from bottom up emission inventories based on the knowledge of individual flights 26 . The International Council on Clean Transportation (ICCT) 27 published that CO 2 emissions of commercial aviation from international passenger movements totaled 444 Mt in 2018 and implied annual compound growth rate of total emissions from commercial flights, 5.7%, during the past five years from 2013 to 2018 1 . Sub-national CO 2 emissions from international passenger operations based on the location of flights departing airports are provided by ICCT 28 . It is known that total CO 2 emissions from all commercial operations include passenger movement, belly freight, and dedicated freight. ICCT published that passenger transport accounted for 81% of total emissions in 2018, so we use this proportion to estimate the international aviation emissions from commercial flights in 2018. Due to the lack of emissions data after 2018, we used a compound growth rate of commercial aviation emissions, 5.7%, to estimate the international aviation emissions from commercial flights in 2019. Based on the assumption that the number of international commercial flights and the number of total commercial flights including domestic and international flights have the same distribution over time, we collect the number of domestic commercial flights for each week in the first quarter in 2019 and the same week in 2020 from aviation worldwide Limited-OAG (https://www.oag.com/about-oag) to estimate international aviation emissions from commercial flights for each week during this period. Daily tracking statistics about commercial flights and total flights in 2020 are provided by Flightradar24 (https://www.flightradar24.com/51.5,-0.12/8), to estimate all aviation emissions in the first quarter in 2020. As processing such data will require some time, we mix different sources of information to gain a more rapid knowledge into the time evolution of CO2 emissions. For this we use daily statistics from AirNav/RadarBox published daily on their website (2019 and 2020). As these statistics are expressed as number of daily flights grouped by destinations, they are not sufficient to assess CO2 emissions and need to be combined with previous CO2 emission inventories for commercial aviation. Specifically we use the B1 2025 QUANTIFY emission inventory at monthly and 1°x1° resolution 28 26 for year 2018. This includes all commercial traffic (passenger and freight). We take this distribution to be also representative of year 2019. We then define 12 main regions of air traffic upon which we project the AirNav/RadarBox categories. For instance, the US box corresponds to a mix of the "US <-> US" and "US <-> NON US" categories; the North Atlantic flight corridor corresponds to the "US <-> Europe" category, etc. This mapping and the weight attributed to the different categories are somewhat arbitrary but this has limited impact on the results as all categories of traffic decrease sharply albeit with somewhat different timing. The 2019 emission inventory is then scaled from the 2019 inventory using weighted ratios of daily 2020 to monthly-mean daily 2019 flights for each corresponding category. A global daily CO 2 emission is then estimated from the spatially-distributed daily 2020 emission map (SI Figure S5 ). The Third International Maritime Organization greenhouse gases study published global shipping CO 2 emissions from 2007-2012 29 . We completed these data by global shipping CO 2 emissions during the period of 2013-2015 from the ICCT report 30 . In addition, IEA released international shipping CO 2 emissions from 2000 to 2018 (https://www.iea.org/reports/tracking-transport-2019/internationalshipping). International shipping CO 2 emissions released by IEA were on average 23% lower than those published by IMO. We used this ratio to modify international shipping CO 2 emissions from IEA. According to the Third IMO GHG Study, CO 2 emissions from international shipping accounts for 87% of global shipping emissions, domestic and fishing accounts for 9% and 4%, respectively. We estimated global CO 2 shipping emissions from 2016-2018 with the modified IEA's international emissions and the ratio between international shipping and global shipping emissions. And we extrapolated emissions from 2007-2018 to estimate emissions in 2019. We obtained emissions for the first quarter of 2019 based on the assumption that monthly variation is flat in shipping CO 2 emissions. In addition, we assume that the change in shipping emissions is linearly related to the change in ship's volume. We collected the change in container ships (https://www.straitstimes.com/business/economy/coronavirus-global-containershipments-set-to-fall-30-in-next-few-months), bulk carriers ship (http://www.eworldship.com/html/2020/bulk_market_0408/158450.html), Oil Tankers (http://www.eworldship.com/html/2020/ship_market_observation_0415/158673.html) and Other 19 ship class (http://www.msivt.com/news/archive/2020). The calculation of emissions was performed in three steps: 1) Calculation of population-weighted heating degree days for each country and for each day based on the ERA5 reanalysis of 2-meters air temperature, 2) Using the EDGAR estimates of 2018 residential emissions as the baseline. For each country, the residential emissions were split into two parts, i.e., cooking emissions and heating emissions, according to the EDGAR guidelines. The emissions from cooking were assumed to remain stable, while the emissions from heating were assumed to depend on and vary by the heating demand. 3) Based on the change of population-weighted heating degree days in each country, we scaled the EDGAR 2018 residential emissions to 2019 and 2020. Since the index of heating degree days are daily values, we can get daily emissions update for the residential sources globally. Note that the effect of increased time spent in households on residential buildings and decreased time in commercial and public buildings was not accounted for, since we did not have fuel consumption data for urban areas and building types. Our estimates of residential emissions changes are consistent with those obtained from the City of Paris, based on individual electricity use (https://data.enedis.fr/) and population surveys (Y. Françoise pers. comm.). For power sector, given the daily statistics of China is not available, we use the daily coal consumption of six main power company to re-allocate the total amount of monthly emissions from power sector to get the daily emissions of power sector in China. For the emissions from power sector in rest of world, we use the average growth rates of emissions from power sector in the quarter of 2020 of India, US, Europe (EU27 & UK), Russia, Japan and Brazil to estimate the growth rate of power emissions in the rest of world. For industrial sector, the statistical data are often lack of real-time and dynamic tracking updates, so we assume the industrial emissions have dropped -5% in Europe (EU27 & UK) and -3% in Japan and Russia in March. In addition, we use the daily electricity generation on national level to disaggregate the monthly industrial emissions into daily industrial emissions. For transport sector, because Google Mobility Reports only publish detailed mobility trends since February 15th, 2020, we assume the emissions are equal to the baseline value in all countries (except for China) before February 14th, 2020. In China, due to the lack of Baidu Migration Scale Index before January 11th, 2020, we use the average value of migration index between January 12th, 2019 and January 20th, 2019, to fill the data gap between January 1st 2019 to January 11th, 2019. We will keep updating our estimates based on fuller and more reliable data sources. Uncertainty of using proxy data to assess changes during 2020 is within the uncertainty of the statistical data usually used to assess monthly emissions. Previous estimates suggested an uncertainty of ±20%(1σ) for fossil fuel CO2 emission in emerging economies like China and ±5% for the global average 31-33 .. As a conservative uncertainty of using monthly, weekly or hourly proxy statistics of activity data (power generation, industrial output, road traffic) we assumed the same emission uncertainty when using these proxy data than when using monthly statistics in China to assess emissions 20 . This is a conservative estimation given that monthly statistics from China have a large uncertainty among those from other large emitters. Uncertainty from monthly statistics was derived from 10000 Monte Carlo simulations to estimate a 68% confidence interval (1-sigma) for China as in ref. 20 . We calculated the 68% prediction interval of linear regression models between emissions estimated from monthly statistics and official emissions obtained from annual statistics at the end of each year, to deduce the one-sigma uncertainty involved when using monthly data to represent the whole year's change (SI Figure S66) . The squared correlation coefficients are within the range of 0.88 (e.g., coal production) and 0.98 (e.g., energy import and export data), which represent that only using the monthly data can explain 88% to 98% of the whole year's variation 20 , while the remaining variation not covered yet reflect the uncertainty caused by the frequent revisions of China's statistical data after they are first published. To validate the response of the atmosphere, including CO 2 concentration and air quality, to the decreased fossil fuel burning and transportation, we collected NO 2 , aerosol optical depth (AOD) and column-averaged dry air mole fraction of CO 2 (XCO 2 ) data from satellites (NO 2 from OMI, AOD from MODIS and XCO 2 from GOSAT) and surface daily average nitrogen dioxide (NO 2 , μg/m 3 ), carbon monoxide (CO, μg/m 3 ) from 1600 air quality monitoring sites (China and US, SI Figure S7 ) in to investigate the impact of COVID-19 on air quality and atmospheric CO 2 . Surface air quality data in China was collected from the daily report by Ministry of Ecology and Environment of China (http://www.mee.gov.cn/). Measurements of daily average nitrogen dioxide (NO2, μg/m3), carbon monoxide (CO, μg/m3), and particulate matter smaller than 2.5 μm (PM2.5, μg/m3) from 1580 sites used to estimate pollution changes between the first quarters of 2019 and 2020. Surface air quality data in U.S. is downloaded from the Air Quality System (AQS) operated by the U.S. Environmental Protection Agency (https://www.epa.gov/aqs). Measurements of daily maximum 1-hour NO2 (ppb), daily maximum 8-hour CO (ppm), and daily average PM2.5 (μg/m3) from 983 sites are used. For March 2020, data availability is limited in U.S. with 20 sites for NO2, 31 for CO, and 309 for PM2.5. Sites with missing data for NO2/CO (PM2.5) at over 20 (5) days in any months will be excluded. We obtained monthly NO 2 data from the Ozone Monitoring Instrument (OMI) provided by Tropospheric Emission Monitoring Internet Service (TEMIS), which has with a spatial resolution of 0.125° x 0.125° and a temporal coverage from October 2004 to March 2020. We only included the data from January 2013 to March 2020 in the work (SI Figure S8 ). For AOD, we chose daily Level 2 MOD 04 data from MODIS 34 and then calculated the monthly averaged AOD f from January 2013 to March 2020. Only "good" and "very good" data (in AOD_550_Dark_Target_Deep_Blue_Combined_QA_Flag 2 and 3) were kept in the calculation. At last, we calculated the monthly XCO 2 data with a resolution of 2.5° x 2.5° from the Greenhouse Gases Observing Satellite "IBUKI" (GOSAT). Because of the delay in the data processing at National Institute for Environmental Studies (NIES), we used a bias-uncorrected version V02.81 for the period of January 2013 to March 2020. With the consideration of the focus on an abnormal event due to COVID-19, the bias-uncorrected data is proper for this study. All of the monthly averaged data were re-gridded to 1° x 1°. We focused on four emitting regions, China, USA, EU4 (UK, France, Germany, and Italy), and Indian, and then calculated the country level monthly averaged NO 2 , AOD, and XCO 2 values. Surface air pollution in China was significantly reduced during the epidemic period (SI Figure 7) . A deep reduction of NO2 by 31.7% was observed on January 24th 2020, one day after the lockdown for many provinces (Si Figure S7b ). The reduction rates were 13.7% for PM2.5 and 16.5% for CO on the same day. A clear rebound (U shape) could be found for all pollution after the spring festival (February 5th) in 2019. However, such recovery was missing in 2020 due to the lockdown, leading to a decreasing trend all through the first quarter. On average, pollution concentrations decreased by 23.0% for NO2, 15.4% for PM2.5, and 12.5% for CO during January-March 2020 relative to the same period in 2019. Pollution level in U.S. was also reduced by the epidemic but with smaller magnitude compared to that in China. Surface PM2.5 decreased in all first three months in 2020 relative to 2019 with the largest reduction of 20.6% in March. NO2 also exhibited large reductions of 9.0% on March 2020 compared to 2019, however, such reduction seemed affected by the limited site numbers (only 20). For example, one site in Salt Lake, Utah reported >200 ppb (normally <40) NO2 during March 20-23, 2020. Such episodes were likely caused by fires but weakened the reduction rate of NO2 after Middle March (SI Figure 7e) . Changes of CO were also limited in U.S., with opposite signs in January and March. Such tendencies could also be biased due to the limited site numbers (only 31). The observed tropospheric nitrogen dioxide (NO2) column concentration data from satellite observation 35 and surface air quality data from ground monitoring networks have exhibited a decrease (SI Table S9 ) consistent with reduction of fossil carbon fuels emissions. In China, January, February, and March 2020 decreased by -32.27%, -34.22%, -4.53% respectively compared to 2019. Overall, NO2 decreased over China by -23.67% in the first quarter of 2020 compared to 2019. In the US, the decrease of NO 2 first started in Feb and continued to decrease at least until March 2020. Compared to the same period of the year in 2019, NO 2 over the US decreased by -23.08% and -14.32% in February and March 2020, respectively (SI Table S9 ). For the UK, France, Germany, and Italy, we observe similar NO 2 decreases than over the USA India had weaker decline in NO 2 than other regions. The decline rate of NO 2 (-25.73%) based on atmospheric observations can be used to check the consistency of the decrease of NO 2 emission from the inventory, and given the NO2 is mainly contributed by fossil fuel combustion with life time short than one day, the temporal change of NO2 emission can could verify the decrease of the fossil fuel combustion and the associated CO 2 emissions. For China where the most significant decrease of tropospheric NO 2 column concentration observed, the inventorybased estimates 36,37 of power generation (-6.8%), transportation (-37.2%) and industry(-8.1%) are adopted with result of weight mean -23.94% % NO2 emission in first quarter of 2020 when comparing with 2019. These three sectors together account for 96% of China's total NO2 emissions. The -23.42% decline of the NO 2 emissions from our bottom-up inventory is consistent with the satellite observed -26% decrease of column NO 2 , and with the -23% decrease of near surface concentrations at the 1680 ground-based stations. For US, the inventory-based estimates of power generation (-4.9%), transportation (-2.7%) and industry(-2.2%) are adopted with result of -2.57% NO2 emission in first quarter of 2020 when comparing with 2019, slightly smaller than -4.76% tropospheric NO 2 column concentration, but difference with the site observation data ( -8.98% in March and +0.34%for first quarter), which may be affected by site numbers (only 20 sites in US). We calculated 1° x 1° monthly mean of NO2, AOD, and XCO 2 from OMI, MODIS, and GOSAT, respectively (SI Table S9 ). For satellite observations, the overall uncertainty of tropospheric NO2 columns monthly mean is 10% 38 . The uncertainty of AOD is approximately 0.03+0.20τ M , where τ M is AOD at 550 nm 39 . In other words, the uncertainties in percentage in low AOD regions (US and EU4) is higher in high AOD regions (China and India). The standard deviations of XCO 2 monthly mean over land are about 0.5-1.5 ppm 40 . Here we conservatively considered uncertainty of monthly XCO 2 as 1.5 ppm. To estimate the uncertainty of changes of 2020 compared to 2019 from January to March, we input above uncertainties of monthly means and run Monte Carlo simulations of 10000 trials to calculate the 68% confidence intervals (i.e., one sigma range) which are shown in Table 1 . All data generated or analyzed during this study are included in this article (and its Data descriptor paper and supplementary information files). Global Energy Review 2020: The impacts of the COVID-19 crisis on global energy demand and CO2 emissions The global macroeconomic impacts of COVID-19: Seven scenarios Coronavirus temporarily reduced China's CO2 Analysis: Coronavirus set to cause largest ever annual fall in CO2 emissions The Great Lockdown: Worst Economic Downturn Since the Great Depression High resolution temporal profiles in the Emissions Database for Global Atmospheric Research CO2 Emissions from Fuel Combustion EDGAR v4. 3.2 Global Atlas of the three major Greenhouse Gas Emissions for the period 1970-2012 Trends in global CO2 and total greenhouse gas emissions Trends in global CO2 and total greenhouse gas emissions ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate European Centre for Medium-Range Weather Forecasts. The boreal winter season 19/20 was by far the warmest winter season ever The version 3 OMI NO2 standard product S5P TROPOMI NO2 slant column retrieval: method, stability, uncertainties and comparisons with OMI World Economic Prospects Monthly New dynamics of energy use and CO2 emissions in China Global carbon budget 2019 The gigatonne gap in China's carbon dioxide inventories China's uncertain CO 2 emissions Reduced carbon emission estimates from fossil fuel combustion and cement production in China The International Council on Clean Transportation Flying into the future: aviation emissions scenarios to 2050 Third IMO Greenhouse Gas Study Greenhouse gas emissions from global shipping Aerosols and their relation to global climate and climate sensitivity Quantifying the uncertainties of a bottom-up emission inventory of anthropogenic atmospheric pollutants in China Gridded uncertainty in fossil fuel carbon dioxide emission maps, a CDIAC example MODIS Atmosphere L2 Aerosol Product: NASA MODIS Adaptive Processing System An improved tropospheric NO2 column retrieval algorithm for the Ozone Monitoring Instrument Effect of changing NOx lifetime on the seasonality and long-term trends of satellite-observed tropospheric NO2 columns over China Trends in China's anthropogenic emissions since 2010 as the consequence of clean air actions Improving algorithms and uncertainty estimates for satellite NO2 retrievals: results from the quality assurance for the essential climate variables (QA4ECV) project Validation and uncertainty estimates for MODIS Collection 6 "Deep Blue" aerosol data Global mapping of greenhouse gases retrieved from GOSAT Level 2 products by using a kriging method Authors acknowledge Paul O. Wennberg for insightful comments to improving this paper. The code generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Authors declare no competing interests.