Paper on pollution impact from the February 2020 COVID-19 lockdown in China published in Geophysical Research Letters

Figure. Change in pollution over China for February 2020. Observed values (left), expected values if there were no COVID-19 perturbation (center), and their difference (right) are shown for nitrogen dioxide (top) aerosol optical depth (middle), and cloud droplet effective radius (bottom). Shading is such that lighter (left and center) and greener (right) colors align with what would be expected for less pollution.

Our new paper analyzing the effect of China’s strict shutdown measures put in place to stop the spread of COVID-19 on pollution levels and cloud properties was just published in the the American Geophysical Union’s special collection on The COVID-19 Pandemic: Linking Health, Society, and Environment.

The paper is available at:

The University of Washington put out a press release with more information:

I was also recently interviewed by Forbes about this work:

For more information, I wrote about our results and provided links to my “iPosters” virtually presented at the JpGU-AGU Joint Meeting and a special World Meteorological Organization workshop, previously on the blog.

Substantial Cloud Brightening From Shipping in Subtropical Low Clouds published in AGU Advances

Figure. The concentration of cloud droplets as observed by satellites (left) and a statistical estimate of what the concentration of cloud droplets would be without the effect of shipping pollution the southeast Atlantic Ocean (right).

Our new paper analyzing the effect of shipping pollution on cloud properties and the implications for regional and global climate was just published in the first issue of the American Geophysical Union’s new gold open access journal AGU Advances.

The paper is freely available at:

The University of Washington and American Geophysical Union put out press releases with more information: UW / AGU

For further reading, I have also written up a few short blog posts about the study in general, its implications for climate change, and its implications for deliberate climate interventions (marine cloud brightening).

New work on the impacts of the February 2020 COVID-19 shutdown on pollution in China presented at JpGU-AGU and WMO-AGU meetings

I presented a new paper that has just been accepted (pre-print available at at two virtual conferences this summer: the Japanese Geophysical Union (JpGU)-American Geophysical Union (AGU) joint meeting and a special symposium on “Climatological, Meteorological and Environmental factors in the COVID-19 pandemic” sponsored by AGU and the World Meteorological Organization (WMO). Links to the posters and a blog post version of the content are included below.

JpGU-AGU joint meeting poster:

WMO-AGU joint meeting poster:

Limited Regional Aerosol and Cloud Microphysical Changes Despite Unprecedented Decline in Nitrogen Oxide Pollution During the February 2020 COVID-19 Shutdown in China

Michael Diamond & Robert Wood

Department of Atmospheric Sciences, University of Washington, Seattle


In late December 2019, cases of a pneumonia of unknown cause were reported in the city of Wuhan. By January 2020, the pathogen responsible—a novel zoonotic coronavirus—had already spread throughout China. To arrest the spread of COVID-19 (the disease caused by the novel coronavirus), a series of unprecedentedly strict restrictions on travel and other activities were adapted across China, slowing the spread of the epidemic in China even as the disease became a global pandemic. Unsurprisingly, this socio-economic “shutdown” had a catastrophic effect on the Chinese economy. Figure 1 shows the Purchasing Managers’ Index (PMI) for both manufacturing and non-manufacturing sectors as reported by the National Bureau of Statistics of China. The PMI is a survey-based estimate of economic activity, with values above 50% corresponding to growth and below to contraction. February 2020 stands out sharply, featuring a decline in manufacturing PMI deeper than any point during the aftermath of the 2008 financial crisis and the only period of contraction in non-manufacturing PMI since records for that index began in 2007, followed by a rapid recovery.

Figure 1. Purchasing Manager’s Index from January 2005-May 2020


To assess how (and whether) pollution levels changed as a result of this abrupt February 2020 shutdown, we analyze satellite data from the Ozone Monitoring Instrument (OMI) on Aura and the Moderate Resolution Imaging Spectrometer (MODIS) on Aqua, which both take daytime measurements at ~13:30 local as part of NASA’s A-train constellation.

From OMI, we analyze tropospheric column NO2 (screened for cloud fractions below 30%). NO2 is a major component of air pollution and has been linked to health problems in humans.

From MODIS, we analyze aerosol optical depth (AOD) and liquid-phase cloud droplet effective radius (re). Aerosols are particles suspended in the atmosphere and are important for both human health and the climate. Particles with aerodynamic diameters below 2.5 µm (PM2.5) are known to have severe health effects, with some estimates of annual deaths due to outdoor PM2.5 pollution approaching 10 million people a year. At the same time, aerosols can change Earth’s energy balance by absorbing sunlight or reflecting it back to space. Aerosols can also indirectly affect the climate by changing cloud properties. When more aerosol particles are available, liquid clouds can form with a greater number of cloud droplets (for which aerosol particles act as “seeds”). If the total amount of water in the cloud doesn’t change, this leads to the droplets being smaller on average. This effect makes liquid clouds more reflective, which has a cooling effect.

To see if the COVID-19 shutdown in China led to pollution changes that are out of the ordinary, we fit an ordinary least squares linear regression model for each environmental variable using trends before and after a major clean air policy went into effect, a “holiday effect” to account for pollution changes associated with the Chinese Lunar New Year, and an idealized seasonal cycle. The model is fit independently at every 0.25 x 0.25 degree grid box for the natural logarithm of NO2 and at every 1 x 1 degree grid box for AOD and re. Data from January 2005 to December 2019 are used to train the model and then values for 2020 are predicted.

Pollution changes during the February 2020 shutdown

We find a very large and statistically significant decrease in NO2 pollution during February 2020 compared to what would otherwise have been expected (Figure 2, top row). Gray stippling in Figure 2 indicates that absolute differences do not exceed two root-mean-square (RMS) errors. We do not, however, see any similarly consistent results for the aerosol or cloud properties (Figure 2, middle and bottom rows).

Figure 2. Maps of observed and expected environmental variables and their difference for February 2020

To look at the differences between the observed and estimated values in greater historical perspective, we average the observed and estimated ln(NO2) and AOD over eastern China and re values over the East China Sea (boxes in the rightmost row of Figure 2). Results are displayed in Figure 3, with the RMS error in this case calculated using the differences between the spatially-averaged observed and expected values from January 2005 to December 2019. Like the PMI indices, ln(NO2) shows a pronounced and unprecedented decline in February 2020 followed by a rapid recovery. In contrast, AOD and re values during 2020 are not perceptibly distinct from the 2005-2019 record.

Figure 3. Timeseries of regionally-averaged observed and expected values and their difference from January 2005 to May 2020

Factors affecting the pollution changes: Emissions

One explanation for the different NO2 and aerosol responses is that economic sectors which disproportionately emit one or the other pollutant may have been impacted differently by the shutdown. Figure 4 shows changes in passenger transportation, energy generation, and iron and steel production from 2005-2020. (Economic data is compiled from the National Bureau of Statistics of China.) Passenger transportation, in particular, was devastated by the shutdown. In contrast, total January-February power generation was down ~10% (similar to 2008-2009), implying a decrease of ~20% in February alone. Heavy industries like steel production (slightly up) were comparatively unaffected, with the remainder of the economy somewhere in between.

Figure 4. Timeseries of economic indicators from 2005-2020

Anthropogenic emissions (in units of ng/m2/s) of NOx (NOx = NO + NO2), PM2.5, and SO2 (a precursor for sulfate aerosol, which is particularly good at seeding cloud droplets) for the year 2015 from the Emissions Database for Global Atmospheric Research (EDGAR) are combined to create aggregate “transportation” and “industry and power” sectors, with the remainder lumped into an “other” category primarily consisting of agriculture and waste management. Figure 5 shows maps of the contributions of the three pollutants by economic sector. Transportation is a major source of NOx pollution, comparable to the industry and power sectors, whereas the industry and power sectors dominate emissions of PM2.5 and SO2. Summed over the region highlighted in Figure 5, the transportation sector accounts for 26.2% of all NOx emissions but only 4.7% of PM2.5 and 3.6% of SO2 emissions, while the industry and power sectors account for 72.3% of NOx, 92.8% of PM2.5, and 95.1% of SO2 emissions.

Figure 5. Maps of 2015 emissions by economic sector

Factors affecting the pollution changes: Meteorology

Of course, emissions changes may not have been the only difference between February 2020 and previous years. In particular, we know that meteorology can have a major effect on pollution concentrations and the occurence of severe haze events. Figure 6 shows 2-m temperature (T) and specific humidity (qv) and 10-m zonal and meridional winds (vectors; longest ~10 m/s) from the Modern‐Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2). February 2020 was anomalously warm and humid (white to dark gray contours in the bottom row indicate anomalies exceeding -2, -1, 1, and 2 standard deviations). Such conditions contribute to the chemical loss of NO2 but could enhance secondary aerosol formation. Le et al. (2020) and Wang et al. (2020) argue such an effect was responsible for the haze events that occured in Beijing despite the lockdown. This effect would tend to amplify the decrease in NO2 but blunt any decreases in aerosol from emissions changes alone.

Figure 6. February 2005-2019 climatology and February 2020 meteorological anomalies

Factors affecting the pollution changes: Chemistry

As a further complication, meteorological anomalies, emissions changes, and their interactions influenced atmospheric chemistry and therefore pollution concentrations in February 2020.

During the winter, the atmospheric lifetime of NOx (~1 day) over eastern China decreases with decreasing emissions as higher ozone concentrations allow for more loss via reaction with hydrogen oxide radicals (HOx) during the day and via hydrolysis of N2O5 (an important NOx reservoir) within aerosols at night. We may therefore expect decreases in NO2 concentrations to exceed reductions in NOx emissions.

Shi & Brasseur (2020) and Huang et al. (2020) reported increasing ozone levels as NOx emissions fell during the shutdown. The increase in ozone increases OH concentrations and thus the atmospheric oxidizing capacity, which could further reduce NO2 but facilitate secondary aerosol formation. This in combination with the relatively warm and wet February 2020 meteorological conditions offers a compelling explanation for the apparent increase in aerosol surrounding Beijing, although the effect appears to be weaker in other regions, perhaps due to differences in background conditions.

Our results are consistent with either a negligible aerosol change or a moderate emissions-driven decrease that was compensated by increased secondary production.

Summary & Conclusions

Despite unprecedented declines in economic activity and NO2 concentrations during the February 2020 COVID-19 shutdown in China, we find no detectable perturbation in aerosol and related cloud properties. The severe curtailment of passenger transportation (a disproportionate NOx source) but comparatively muted changes in power generation and heavy industry (disproportionate PM2.5 and SO2 sources), along with meteorology and complex chemical interactions, help explain this discrepancy.

Further study of the environmental consequences of COVID-19 is warranted, not least because potential links between long-term and short-term air quality and vulnerability to the disease remain unresolved. There is some evidence that short-term exposure to air pollution increased the case fatality rate of the 2002-2003 Severe Acute Respiratory Syndrome (SARS) outbreak in several Chinese cities, which raises the possibility of feedbacks between containment measures that happen to reduce pollution and population-level resilience. Additionally, dramatically reduced transportation sector emissions without similar changes in other sectors could represent a plausible future emissions mix if widespread electrification of transportation is adopted but other sectors do not adopt similar pollution mitigation measures.


Diamond, M., & Wood, R. (in revision), Limited Regional Aerosol and Cloud Microphysical Changes Despite Unprecedented Decline in Nitrogen Oxide Pollution During the February 2020 COVID-19 Shutdown in China. Geophysical Research Letters. Pre-print available at:

Huang, X., Ding, A., Gao, J., Zheng, B., Zhou, D., Qi, X., et al. (2020). Enhanced secondary pollution offset reduction of primary emissions during COVID-19 lockdown in China. National Science Review.

Le, T., Wang, Y., Liu, L., Yang, J., Yung, Y. L., Li, G., & Seinfeld, J. H. (2020). Unexpected air pollution with marked emission reductions during the COVID-19 outbreak in China. Science, eabb7431.

Shi, X., & Brasseur, G. P. (2020). The Response in Air Quality to the Reduction of Chinese Economic Activities during the COVID-19 Outbreak. Geophysical Research Letters, 47, e2020GL088070.

Wang, P., Chen, K., Zhu, S., Wang, P., & Zhang, H. (2020). Severe air pollution events not avoided by reduced anthropogenic activities during COVID-19 outbreak. Resources, Conservation, & Recycling, 158, 104814.

What can inadvertent cloud changes due to shipping pollution in the southeast Atlantic teach us about deliberate climate interventions?

Figure 1. Estimates of the increase in the concentration of liquid droplets within clouds (red) and how much more reflective those clouds are (blue) due to the effects of shipping pollution in the southeast Atlantic Ocean for different observational time periods (lighter curves have fewer years of observations, gray shading is estimated from all available years). It takes ~5 years of consecutive observations for the signal to become clear.

We care about air pollution not only because of its harmful effects on human health, but also because tiny airborne particles of pollution known as aerosols can affect the climate. Aerosol particles themselves can affect temperatures by absorbing light (think of the eerie orange skies produced by smoke over Seattle in recent summers) and scattering it (reflecting it away — think of how much visibility declines on hazy, high pollution days). Aerosol particles can also affect the climate indirectly through their interactions with clouds.

Clouds are made up of tiny droplets of water, each of which must be “seeded” by an aerosol particle. When the number of aerosol particles increases, a cloud’s water can be spread out over a larger number of small droplets instead of a small number of big droplets. This ends up increasing the reflectivity, or brightness, of the cloud.

Ship tracks, or curvilinear trails of brightened clouds following individual ships, is the quintessential example of these aerosol-cloud interactions in action. Mysteriously, however, past attempts to measure the effects of ship tracks over long time periods and large regions have failed to find substantial effects, despite climate models predicting sizable cooling effects from brighter clouds reflecting more sunlight. This is in part due to how much variation occurs in the clouds naturally, making the signal of the aerosol effects very difficult to detect.

Figure 2. a) Winds (white barbs) blow along the major shipping corridor in the southeast Atlantic (yellow shading). b) The concentration of cloud droplets (shading) is enhanced over the shipping corridor. Black contours show regions of greater than 80% and 90% cloud occurrence — it really is that cloudy over the southeast Atlantic Ocean during spring!

In our study, we took advantage of a unique meteorological situation in the southeast Atlantic (see Figure 2) in which the prevailing winds keep pollution relatively constrained around a major shipping corridor that passes through one of the cloudiest regions on Earth. By analyzing the statistics of cloud properties of nearby, unaffected regions, we predicted what cloud properties would have been in the shipping corridor in the absence of shipping pollution. We then estimated the effect of the shipping pollution by comparing that prediction to what we actually see in the satellite data. Our study is the first to observe a significant cooling effect due to cloud brightening over climate-relevant temporal and spatial scales.

Even before their large-scale effects were confirmed, some scientists have suggested that clouds over the oceans could be deliberately seeded with sea salt particles to produce a cooling effect that would help counteract some of the warming we have experienced due to rising greenhouse gas concentrations in the atmosphere. This proposal, called marine cloud brightening, still requires a lot of research before it could be implemented — or before we decide if it would ever be a good idea to implement it in the first place. Our study contributes to answering two of the open questions about potential field tests of marine cloud brightening: Would they work? And if so, how long would they be required to go on for to see an effect?

As for the first question, our results are fairly positive for proponents of marine cloud brightening: If ships are already inadvertently producing large cooling effects in certain regions, it seems very likely that we could replicate and even enhance these effects deliberately.

On the question of detectability, however, our results are somewhat less optimistic for marine cloud brightening’s prospects. Even in the ideal conditions of the southeast Atlantic, it takes ~5 years of data before the shipping signal of increasing cloud brightness becomes clear (see Figure 1 at the top of this post). Of course, if the test were to aim for a larger effect, it would be more detectable. There will be a real tradeoff in terms of having a light touch with the size of the intervention or the length of time we have to intervene before we can observe how successful the test was.

Below, I summarize what our new study does, and, importantly, does not, teach us about the prospect of deliberate marine cloud brightening.

What the study does tell us about deliberate marine cloud brightening

  • A well-designed test will likely be able to produce a significant cloud brightening effect
    • In the southeast Atlantic, shipping pollution produces a statistically significant, ~2 W/m2 cooling effect during the spring
  • A test may need to be carried out for a substantial period of time to be observable from space
    • In the southeast Atlantic, it took ~5 years of springtime data to clearly detect the size of the cloud brightening effect in satellite observations

What the study does not tell us about deliberate marine cloud brightening

  • Under what conditions should we consider deploying marine cloud brightening?
    • Even if tests of marine cloud brightening are successful (in terms of producing the desired cooling effect), there would be many ethical, socioeconomic, and political questions to answer before it could/should be implemented
    • For instance, if it appeared that a limited deployment of marine cloud brightening could limit warming to 1.5 °C (2.7 °F) instead of 2 °C (3.6 °F), would it be worth the risk?
      • We will need physical scientists, social scientists, those in the humanities, and voices from impacted communities and citizen input to be able to answer a question like this
  • Would it be technologically feasible to produce aerosol particles that are not harmful to human health?
  • What kind of unintended consequences could result from marine cloud brightening?
  • How would testing or implementation of marine cloud brightening be governed?

What can cloud changes due to shipping pollution in the southeast Atlantic tell us about climate change?

Figure. Radiative forcing measured directly (gold) or estimated via the change in total scene reflectivity (black), changes in cloud reflectivity alone (dark purple), and changes in cloud abundance alone (light purple). There is approximately 2 W/m2 of cooling due to pollution-induced cloud brightening in the shipping corridor.

In our new paper in AGU Advances, we are able to attribute increases in cloud brightness to pollution from the shipping industry in the southeast Atlantic Ocean. In that region, we calculate ~2 W/m2 of cooling due to more sunlight being reflected back to space by the clouds rather than being absorbed at the ocean surface. This is a pretty large number — in comparison, the warming (in energy units) that would result from a doubling of atmospheric carbon dioxide concentrations is ~4 W/m2.

But that’s just for the southeast Atlantic. What does this mean for the rest of the world?

By using estimates of the increase in all industrial sulfate pollution from “historical” (1850 to 2015) simulations produced by the latest generation of global climate models, we took the relationships between sulfate pollution and cloud properties from the southeast Atlantic shipping corridor and scaled them worldwide. Our “observationally-informed” global value for cooling due to pollution-induced cloud brightening is ~1 W/m2. This is approximately twice as large as the “best guess” value reported in the most recent Intergovernmental Panel on Climate Change (IPCC) assessment report.

The IPCC estimates that the warming due to increasing greenhouse gases (like carbon dioxide, methane, and nitrous oxide) is ~3 W/m2 today. If our global estimate is correct, then, it means that almost one-third of the warming we could have experienced due to greenhouse gas emissions has been “masked” by the cooling effect of pollution-induced cloud changes.

So far, the world has already warmed ~1 °C (1.8 °F) since the late 1800s. Without the cooling effect of pollution-cloud interactions, it is possible we would have already warmed by 1.5 °C (2.7 °F). The IPCC has found that there would be significant negative effects on society and natural ecosystems if the world were to warm that much, with even more harmful effects being felt at 2 °C (3.6 °F).