Large discrepancy between observed and modeled wintertime tropospheric NO₂ variabilities due to COVID-19 controls in China

The OMI instrument on the Aura spacecraft has a spatial resolution of 13 km × 24 km (nadir view), which is in a sun-synchronous ascending polar orbit with a local equator crossing time of 13:45 (Levelt et al 2018). OMI provides global coverage with measurements of both direct and atmosphere-backscattered sunlight in the ultraviolet-visible range from 270 to 500 nm; the spectral range 405–465 nm is used to retrieve tropospheric NO₂ columns. The OMI quality assurance for essential climate variables (QA4ECV) retrievals level 2 are used in this work (Boersma et al 2018). The pixel-based OMI NO₂ data are averaged to 0.5° × 0.625° horizontal resolution with daily temporal resolution. Following Jiang et al (2018), and the QA4ECV product user manual (www.qa4ecv.eu/ecv/no2-pre/data), the following filters are applied in our analysis: processing quality flags = 0 to ensure successful processing in the retrieval; surface albedo < 0.3; cloud radiance fraction < 0.5; no edge data (rows 1–5, 56–60); no row anomaly data (rows 27–55).

The GEOS-Chem chemical transport model (www.geos-chem.org, version 12-8-1) is driven by assimilated meteorological data of MERRA-2 with nested 0.5° × 0.625° horizontal resolution. The model includes fully coupled O₃-NO_x -VOC (volatile organic compound)-halogen-aerosol chemistry. It has been widerly used for air quality studies in China (Li et al 2019, Dang et al 2021, Lu et al 2021). The nested domain (E. China) defined in this work is 97.5° E–127.5° E, 17.5° N–47.5° N, with 2 months of spin-up period to remove the influence from initial conditions. The chemical boundary conditions are updated every 3 h from a global simulation with 4° × 5° resolution. Emissions in GEOS-Chem are computed by the Harvard-NASA Emission Component. Global default anthropogenic emissions are from the Community Emissions Data System (Hoesly et al 2018). Regional emissions are replaced by Multiresolution Emission Inventory for China in China, and MIX in other regions of Asia (Li et al 2017). The model configurations in this work are similar with Chen et al (2021), who have demonstrated the capability of this model to simulate surface O₃ concentrations in E. China. We refer the reader to Chen et al (2021) for the details of model configurations. The modeled NO₂ to be compared with satellite measurements (${y_{{\text{trop}}}}$) are smoothed with OMI averaging kernels with the following approach (Huang et al 2014, Qu et al 2020): ${y_{{\text{trop}}}} = {A_{{\text{trop}}}} \cdot {x_{{\text{trop}}}}$, where ${A_{{\text{trop}}}} = A \cdot {\text{AM}}{{\text{F}}_{{\text{total}}}}/{\text{AM}}{{\text{F}}_{{\text{trop}}}}$. Here ${x_{{\text{trop}}}}$ represents the modeled tropospheric NO₂; ${A_{{\text{trop}}}}$ represents the column averaging kernel for tropospheric retrievals; A represents OMI averaging kernel; ${\text{AM}}{{\text{F}}_{{\text{total}}}}$ and ${\text{AM}}{{\text{F}}_{{\text{trop}}}}$ represent total and tropospheric air mass factors, respectively.

We use surface in situ NO₂ concentration data from the China MEE monitoring network (https://quotsoft.net/air/) for the period of 2014–2020. These real-time monitoring stations have the ability to report hourly concentrations of criteria pollutants from over 360 cities in 2020. Concentrations were reported by the MEE in units of µg m⁻³. The MEE NO₂ measurements have been widerly used in recent studies, for example, Liu et al (2018) evaluated modeled surface NO₂ concentrations with MEE NO₂ observations and found underestimation in the emission inventories; Feng et al (2020) estimated the changes in anthropogenic NO_x emissions with MEE NO₂ observations; Shen et al (2021) analyzed the impacts of weather and emission changes on surface NO₂ concentrations provided by MEE stations. To compare with satellite measurements, only surface NO₂ observations (hourly data) matching the OMI overpass time are considered in this work.

Figure 1 shows tropospheric OMI NO₂ columns averaged in the 40–20 d before reference time (25 January 2020, RT) and in the 10–30 d (after the RT) in 2015–2019 and 2020. The RT is set because the Spring Festival (25 January 2020) is a good indication of Chinese economic cycles. The data was further shifted for 2015–2019 to account for the economic cycles due to the Spring Festival. As shown in figure 1, tropospheric OMI NO₂ columns in the 10–30 d (after the RT) in 2015–2019 (figure 1(b)) are dramatically lower than those before the RT (figure 1(a)), caused by the seasonal variability of tropospheric NO₂ and influence of Spring Festival. Tropospheric OMI NO₂ columns in the 40–20 d (before the RT) in 2020 (figure 1(c)) is lower than those in 2015–2019 (figure 1(a)) due to the sustainable reductions of NO_x emissions in China (Zheng et al 2018, Jiang et al 2022). Despite tropospheric OMI NO₂ columns in the 10–30 d (after the RT) in 2020 (figure 1(d)) are lower than those in 2015–2019 (figure 1(b)), it is unclear whether this discrepancy is caused by the decreasing trend of NO_x emissions or COVID-19 controls.

Zoom In Zoom Out Reset image size

To isolate the influences from the decreasing trend of NO_x emissions, figures 2(a)–(c) further show averaged tropospheric OMI NO₂ columns in 2015–2019 and 2020, normalized in the 50–10 d before RT. The shaded areas demonstrate the distributions of OMI NO₂ in 2015–2019. We find a significant decrease of tropospheric OMI NO₂ over E. China in February 2020, driven by COVID-19 controls (figure 2(a)). Figure 3 exhibits the mean tropospheric OMI NO₂ columns over various provinces in E. China in 2015–2019. The high industrialization levels in the North China Plain and Yangtze River Delta have led to strong pollutant emissions in provinces such as Jiangsu, Hebei and Shandong (the locations are shown in figure 1(a)). There are large differences in the responses of OMI NO₂ over different provinces, for example, the perturbations in OMI NO₂ are significant over highly polluted Hebei province (figure 2(b)) but insignificant over medium polluted Hubei province (figure 2(c)). In contrast to satellite measurements, the averaged surface in situ NO₂ observations (figures 2(d)–(f)) revealed significant and comparable declines in NO₂ concentrations over E. China as well as Hebei and Hubei provinces.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

According to figures 2, 3 and figure 4(a) provides a detailed comparison for the observed variabilities in tropospheric NO₂ on 10 February–29 February 2020. We find consistent responses in surface NO₂ concentrations: about 33%–34% reductions over E. China, as well as high, medium and low polluted provinces. By contrast, the perturbations on tropospheric OMI NO₂ columns are smaller than surface NO₂ concentrations. Furthermore, there is larger nonlinearity in the observed tropospheric NO₂ columns: the responses decreased from about 28% (low polluted provinces) to 20% (highly polluted provinces). It is different with Qu et al (2021), in which they found the responses of satellite NO₂ to COVID-19 controls are stronger than those of surface NO₂ over highly polluted areas in the US in March–April 2020. The difference between this work and Qu et al (2021) exhibits possible regional discrepancies in the responses of tropospheric NO₂ to NO_x emissions.

Zoom In Zoom Out Reset image size

The above analysis analyzed the observation-based responses of tropospheric NO₂ to emission changes, we then evaluate the capability of chemical transport models to capture the observed responses. Figure 5(a) shows the modeled tropospheric NO₂ columns in February 2020 over E. China, with similar spatial patterns with OMI NO₂ observations (figure 1(a)). The anthropogenic NO_x and VOC emissions in figure 5(a) are fixed in 2019. Figure 5(b) exhibits the response of modeled tropospheric NO₂ columns to a 30% decline of anthropogenic NO_x emissions in February 2020 to simulate the impacts of COVID-19 controls. The 30% reduction of NO_x emissions resulted in about 30%–40% decreases of tropospheric NO₂ columns (figure 5(b)) over provinces with high NO₂ abundances (in figure 5(a)), and 30%–20% decreases over the rest of E. China. Figure 5(d) further demonstrates the responses of modeled surface NO₂ concentrations to a 30% decline of anthropogenic NO_x emissions in February 2020. The spatial patterns in the responses between tropospheric NO₂ columns (figure 5(b)) and surface NO₂ (figure 5(d)) are similar. However, the reductions of surface NO₂ (figure 5(d)) are weaker than reductions of tropospheric NO₂ columns (figure 5(b)) over provinces with high NO₂ abundances.

Zoom In Zoom Out Reset image size

The simulations with 30% perturbation on NO_x emissions allow us to compare the observed and modeled responses of tropospheric NO₂ to emission changes. To have a more accurate comparison, we sampled modeled surface NO₂ at the locations and times of surface measurements; the modeled column NO₂ are smoothed with OMI averaging kernels (see section 2.2) and then sampled at the locations and times of surface measurements. As shown in figure 4(c), 30% reductions of NO_x emissions lead to about 20%, 30% and 26% decreases of modeled surface NO₂ concentrations, and 36%, 34% and 27% decreases of tropospheric NO₂ columns over highly, medium and low polluted provinces, respectively. The modeled responses of tropospheric column NO₂ are larger than those of surface NO₂.

Following figure 5, we further perturb anthropogenic NO_x and VOC emissions to assess the impacts of nonlinear chemistry and possible changes in atmospheric oxidation capability on modeled tropospheric NO₂. As shown in figure 6(a), both tropospheric column (red) and surface NO₂ (green) demonstrate nearly linear responses to NO_x emission changes as well as insensitive to changes in VOC emissions over E. China. The differences become significant for various provinces: the responses of surface NO₂ are weaker than the 1:1 relationship over highly polluted provinces (figure 6(b)); the responses of column NO₂ are weaker than the 1:1 relationship over lower polluted provinces (figure 6(d)). We find the discrepancies in the modeled responses between column and surface NO₂ are not affected by the selection of NO_x emission perturbations (e.g. 30% in this work): the responses of surface NO₂ are smaller than those of tropospheric column NO₂ over highly polluted provinces (figure 6(b)) within 10%–60% reductions of NO_x emission as well as 10%–40% reductions of VOC emissions.

Zoom In Zoom Out Reset image size

Here we further investigate the impacts of meteorological condition changes on tropospheric NO₂. Figure 7 shows the anomaly of tropospheric OMI NO₂ columns, surface NO₂ measurements, and GEOS-Chem model simulations in 2015–2020. Because the anthropogenic NO_x and VOC emissions are fixed in 2019, the modeled tropospheric NO₂ in figure 7 are driven by changes in meteorological conditions. There are large interannual variabilities in tropospheric NO₂, for example, OMI NO₂ over highly polluted provinces (figure 7(b)) in 2018 is about 18% higher than the 2015–2019 average, and thus, the about 18% decline of OMI NO₂ in 2020 (figure 7(b)) may not represent a significant perturbation due to COVID-19 controls. Furthermore, the interannual variabilities in modeled tropospheric NO₂ columns are much larger than those in surface NO₂ over highly polluted provinces in 2015–2019 (figure 7(b)), indicating stronger effect of meteorological condition changes on free tropospheric NO₂ than surface NO₂. According to figure 7, figure 4(d) provides a detailed comparison for the impacts of meteorological condition changes on modeled tropospheric column and surface NO₂ in 2020. The impacts from meteorological condition changes on surface NO₂ concentrations are generally smaller than 5%, however, the impacts on tropospheric NO₂ columns can be larger than 15% over highly polluted provinces.

Zoom In Zoom Out Reset image size

Figure 4(b) demonstrates the combined influences from 30% reductions of NO_x emissions (figure 4(c)) and meteorological condition changes (figure 4(d)). The modeled responses of surface NO₂ concentrations in figure 4(b) show smaller nonlinearity: about 27%, 34% and 28% over high, medium and low polluted provinces, respectively. By contrast, the responses of modeled tropospheric NO₂ columns in figure 4(b) exhibit larger nonlinearity: about 52%, 38% and 22% over high, medium and low polluted provinces, respectively. The larger reductions of tropospheric NO₂ columns in the simulations over highly polluted provinces are driven by the combined effects from emissions and meteorology: the emission-induced responses over highly polluted provinces are larger than those over low polluted provinces by about 10% (figure 4(c)); the meteorology-induced responses over highly polluted provinces are larger than those over low polluted provinces by about 20% (figure 4(d)).

There are large discrepancies between observed and modeled tropospheric NO₂ variabilities: the observed reductions in tropospheric NO₂ columns are about 40% lower than those in surface NO₂ concentrations over highly polluted provinces (figure 4(a)), but the modeled reductions in tropospheric NO₂ columns are about two times higher than those in surface NO₂ concentrations (figure 4(b)). By contrast, we find good agreement between observations and simulations over low polluted provinces. As shown in figure 8, the modeled NO₂ decreases driven by 30% reductions in NO_x emissions are nearly uniform over low polluted provinces in the lower troposphere (1000–900 hPa), in contrast to the dramatic vertical gradient over highly polluted provinces. The different performances between high and low polluted provinces in figure 4 could thus, be caused by inaccurate simulations of lower tropospheric NO₂ in winter. Furthermore, the observed reductions in the tropospheric NO₂ columns over low polluted provinces are larger than those over highly polluted provinces (figure 4(a)). It indicates a small influence from possible free tropospheric NO₂ backgrounds. As suggested by Qu et al (2021), the NO₂ backgrounds are associated with intercontinental pollution transport or natural emissions (such as soil and lightning NO_x ), which could be weak in E. China in winter.

Zoom In Zoom Out Reset image size