NCCS Supports Monitoring the Global Decline of Air
Pollution During the COVID-19 Pandemic


NASA Goddard Space Flight Center and university scientists researched the decline and subsequent recovery of nitrogen dioxide (NO2) related to COVID-19 stay-at-home orders around the globe. Researchers sought to understand the impact of the lockdown on NO2—a key air pollutant formed during the combustion of fossil fuels—so they combined a near-real-time computer simulation with surface observations using substantial NASA Center for Climate Simulation (NCCS) resources.

Study results are being presented Thursday, December 10 at the virtual 2020 American Geophysical Union (AGU) Fall Meeting by Mathew Evans, professor of Atmospheric Chemistry Modeling, Department of Chemistry, University of York, in session A098-03 - Global Impact of COVID-19 Restrictions on the Atmospheric Concentrations of Nitrogen Dioxide and Ozone (meeting registration required).

This snapshot of May 1, 2020 shows the response of nitrogen dioxide (NO2) to COVID-19 restrictions in Europe. The hemispheres mark the locations of daily NO2 measurements, colored according to their departure from a baseline model and additionally marked as “no lockdown” (a grey dot) or “lockdown” (a green dot). The line plot shows the 7-day running average for Wuhan, China (yellow); Madrid, Spain (blue); and New York, U.S. (green). Visualization by Trent Schindler, NASA’s Scientific Visualization Studio. View animation.

The air quality simulation used NASA Goddard’s global GEOS Composition Forecast (GEOS-CF) model at 25-kilometer resolution. GEOS-CF runs on 3,416 cores of the NCCS Discover supercomputer and produces approximately 300 gigabytes of output per day, stored on the NCCS DataPortal.

The researchers computed a NO2 baseline by combining terabytes of GEOS-CF model output with millions of surface NO2 observations from 4,778 air quality monitoring sites in 46 countries. This baseline represents the “business as usual” scenario, which is what NO2 concentrations would be in a world without COVID-19 lockdowns. For this analysis, they used an XGBoost machine learning algorithm on Discover and the NCCS ADAPT Science Cloud.


Impact: High-resolution NASA computer simulations can explain the confounding impact of meteorology on air pollution, as well as provide a business-as-usual baseline of nitrogen dioxide that would have been expected had COVID-19 not occurred.


“NASA supercomputers offer the computational power needed to simulate air pollution at high spatial resolution,” said study leader Christoph Keller, a Universities Space Research Association (USRA) scientist in NASA Goddard’s Global Modeling and Assimilation Office. “This power is essential to capture the fine-scale features of air pollution. Without such global high-resolution simulations, we would have much less confidence in our observation-model comparisons.”

The team’s comparisons of surface observations against their model-simulated baseline concentrations show that surface NO2 declined sharply after the COVID-19 outbreak. Moreover, NO2 reductions correlate with timing and intensity of COVID-19 restrictions, ranging from 60% in severely affected cities (e.g., Wuhan, Milan, and New York) to little change (e.g., Rio de Janeiro and Taipei). On average, NO2 concentrations over the studies locations were 18% lower than the baseline between February and August 2020.

Estimated change of surface NO2 relative to a business-as-usual baseline at six cities from January 1, 2018 through August 1, 2020. Blue-shaded areas indicate lower observed NO2 concentrations relative to baseline; red-shaded areas indicate higher observed NO2 concentrations than expected. Gray-shaded areas indicate time periods with official COVID-19 lockdowns. Image by Christoph Keller, NASA Goddard Space Flight Center/USRA.

For their study, the researchers also calculated country-level changes in NO2. During the second quarter of 2020 (April to June), such changes range from +5% (in Brazil) to more than -40% (e.g., Ecuador, India, the United Kingdom, and Spain). As presented at last month’s SC20 conference, “these NO2 declines correlate well with reported year-over-year changes in quarterly Gross Domestic Product (GDP), highlighting the close relationship between NO2 concentration and economic activity.”

Comparison of estimated NO2 change (x-axis) vs. change in Gross Domestic Product (GDP, y-axis) during the first two quarters of 2020. Blue circles show values for first quarter 2020 (Q1, January–March); red circles show values for second quarter 2020 (Q2, April–June). Only shown are countries significantly impacted by COVID-19 during Q1 or Q2 and having available GDP data. Circle size corresponds to economy size. Image by Christoph Keller, NASA Goddard Space Flight Center/USRA.

Related Links


Jarrett Cohen, NASA Goddard Space Flight Center