First Chinese ultraviolet-visible hyperspectral satellite instrument implicating global air quality during the COVID-19 pandemic in early 2020

The COVID-19 pandemic led to widespread lockdowns that reduced anthropogenic emissions. Prior studies using TROPOMI, OMI, in situ monitoring, and chemical transport models documented declines in NO2 and changes in HCHO, highlighting satellite remote sensing advantages in spatial coverage and consistency. The Environmental Trace Gases Monitoring Instrument (EMI) aboard GaoFen-5 is China’s first UV–visible hyperspectral spectrometer for air pollutants, but its poorer spectral quality had hindered its use for pandemic-related air quality analyses. This study aims to overcome EMI data-quality limitations via algorithmic optimizations and use EMI-retrieved tropospheric vertical column densities (TVCDs) of NO2, SO2, and HCHO to quantify air quality changes associated with lockdown measures across countries and cities, and to explore links between air quality changes and economic activity during early 2020.

The paper reviews satellite-based assessments of pandemic-related air quality changes, noting strong NO2 reductions seen by high-resolution instruments such as TROPOMI and OMI, and complementary evidence from in situ monitoring and chemical transport modeling. It highlights prior retrieval frameworks (e.g., QA4ECV for NO2) and methodological elements like DOAS/BOAS fitting, stratosphere–troposphere separation, and AMF calculations used by OMI/TROPOMI, which inform EMI retrievals. Previous global studies (e.g., Bauwens et al., Sun et al.) reported significant NO2 and HCHO changes during early 2020. The authors also reference known SO2 trends due to desulfurization policies, with persistent hotspots in India seen in satellite data, and earlier findings of limited SO2 changes during initial lockdowns in China and India.

Instrument and data: EMI on GaoFen-5 (launched May 2018) is a push-broom UV–visible spectrometer (240–710 nm) with nadir resolution ~12 × 13 km2. EMI shares overpass time with TROPOMI. Retrieval overview: For NO2, a three-step process is used: (1) retrieve total slant column density (SCD) via DOAS, fitting cross-sections for O3, NO2, O4, H2O, liquid water, and Ring effect (per QA4ECV recommendations); (2) separate stratospheric NO2 using the STRatospheric Estimation Algorithm from Mainz; (3) convert tropospheric SCD to TVCD using air mass factors (AMFs): VCD = SCD/AMF. For HCHO, differential SCD (DSCD) is first retrieved by BOAS in a selected fitting window including cross-sections for O3 at two temperatures, NO2, O4, and BrO. DSCDs are converted to SCDs by reference sector correction (daily Earth radiance over the remote Pacific). AMFs convert SCD to VCD; as HCHO is concentrated in the lower troposphere, VCD approximates TVCD. Cloud information and surface albedo from TROPOMI are used in AMF calculations, and GEOS-Chem provides a priori NO2 and HCHO profiles. For SO2, an optimal estimation (OE) algorithm minimizes a cost function balancing measurement–simulation mismatch and deviation from a priori, with constraints from measurement and a priori covariance. VLIDORT simulates radiance with Rayleigh scattering and O3 absorption; effects of other gases are introduced via Lambert–Beer law with GEOS-Chem a priori profiles. Cross-sections include O3 at four temperatures, BrO, and HCHO. Algorithmic enhancements for EMI: (1) On-orbit spectral calibration to compute daily ISRFs (FWHM) and wavelength shifts, addressing larger temporal/spatial ISRF variability and wavelength drift than TROPOMI due to lack of preflight ISRFs and on-orbit degradation. (2) Adaptive iterative retrieval to handle EMI’s low SNR (~one-third of TROPOMI). The algorithm selects fitting settings yielding minimum uncertainty across candidate windows, polynomial orders, and interfering species. Selected wavelength windows: NO2 (420–470 nm), HCHO (326.5–356 nm), SO2 (310.5–320 nm). (3) Use of simulated solar irradiance in place of infrequent/degraded on-orbit irradiance (EMI provides solar spectrum roughly every 6 months and shows diffuser issues), removing cross-track stripes and reducing residuals. Additional improvements for SO2 include measurement error correction and pixel merging because the strongest SO2 absorption (300–330 nm) lies near the edge of EMI spectra, where noise is high and traditional settings overestimate SO2 in China/India. Validation and uncertainties: EMI monthly NO2 and HCHO TVCDs show good consistency with operational TROPOMI products across multiple cities in Q1 2019–2020, though absolute values differ and EMI HCHO random errors are larger. SO2 comparisons with TROPOMI operational products are avoided due to known systemic deviations. Systematic uncertainty for NO2 TVCD combines SCD retrieval (<3%), AMF (15–40%), and stratospheric separation (<10%), yielding ~18% in summer and ~42% in winter. HCHO VCD systematic errors combine SCD (~17%), AMF (15–51%), and reference sector correction (10–40%), totaling ~25% in clean and ~67% in polluted regions; AMF dominates. SO2 systematic error inferred from remote Pacific background is ~0.29 DU. Random retrieval error at city scale is approximated as average pixel error divided by sqrt of the number of pixels over the city, with spatial sampling the primary driver. Meteorology adjustment: For SO2 in India, 2019 TVCDs were deweathered to 2020 meteorological conditions using GEOS-Chem to isolate emission-driven changes.

- Algorithmic advances (on-orbit ISRF/wavelength calibration, adaptive fitting settings, simulated irradiance, and additional corrections) enable reliable EMI retrievals of NO2, HCHO, and improved SO2, despite EMI’s initially poor spectral quality and low SNR. - Global NO2 reduction: In March 2020, global average NO2 TVCD from EMI was lower than March 2019 by 1.3 × 10^14 molecules cm^-2 (~20%), with strongest decreases over eastern China, western Europe, and eastern North America. - City-level NO2 captured lockdown timing: Abrupt NO2 drops aligned with intervention timing—January 2020 in Chinese cities (e.g., Wuhan), February in Seoul, and March in Tokyo and many European/American cities. In Wuhan, January 2020 NO2 TVCD decreased by 56% vs January 2019 (51% after meteorology adjustment), with random errors <1% of monthly TVCDs. - SO2 showed limited initial change: March 2020 SO2 TVCDs exhibited a slight global increase vs March 2019. In India’s Chhattisgarh–Odisha power-plant region, March 2020 SO2 exceeded both March 2019 and deweathered March 2019 levels, indicating continued power-sector emissions during early 2020. Prior studies reported larger SO2 reductions only starting April–May 2020 under stricter lockdowns. - HCHO changes: Global average HCHO TVCD decreased by 1.5 × 10^15 molecules cm^-2 (~21%) in March 2020 vs March 2019, with regional heterogeneity. Marked HCHO decreases in Wuhan, Shanghai, Guangzhou, and Seoul imply predominantly anthropogenic VOC sources there, whereas little change in Beijing and New Delhi suggests a stronger natural (biogenic) VOC contribution. - EMI–TROPOMI consistency: Monthly NO2 and HCHO TVCDs from EMI agree well in temporal variability with TROPOMI across 13 global cities in Q1 2019–2020. - Economic link: Comparing relative NO2 changes with GDP indicates the pandemic affected the secondary industry more in China, while primary and tertiary industries were more impacted in Korea and across Europe and America.

The study demonstrates that, after targeted calibration and retrieval optimizations, EMI can provide scientifically robust NO2, HCHO, and improved SO2 TVCDs suitable for diagnosing rapid emission changes. The pronounced and temporally aligned NO2 decreases reflect reductions in combustion-related activities due to lockdown measures, validating NO2 TVCD as an effective proxy for anthropogenic NOx emissions across diverse regions. The limited initial response of SO2, especially over power-generation hotspots in India, suggests that sectors like electricity production and residential heating were less immediately affected than transportation and some industrial processes. Contrasting HCHO responses—large decreases in several Chinese and Korean cities versus muted changes in Beijing and New Delhi—indicate varying VOC source contributions, with anthropogenic dominance in some megacities and stronger biogenic influence in others. The observed relationships between NO2 reductions and GDP components provide insight into sector-specific economic disruptions: stronger impacts on secondary industry in China, versus primary/tertiary sectors in Korea and Western countries. Together, these findings address the research goal of using a Chinese satellite instrument to capture pandemic-driven air quality changes and to relate these to policy measures and economic activity, expanding the global observing capability beyond European/U.S. instruments.

This work delivers the first comprehensive use of China’s EMI UV–visible hyperspectral instrument to assess global air quality changes during early 2020. Through on-orbit spectral calibration, adaptive retrieval settings, and auxiliary strategies (simulated irradiance, improved SO2 handling), the authors retrieved reliable TVCDs of NO2, HCHO, and improved SO2. EMI captured sharp NO2 declines that correspond to city-specific lockdown timings worldwide, revealed heterogeneous HCHO responses indicating differing VOC source contributions, and showed limited early SO2 changes over power-sector regions. The study also links air quality changes to economic sectors across regions. These advances establish EMI as a valuable complement to existing satellites for air quality monitoring. Future work could refine AMF inputs (e.g., dynamic surface albedo and cloud/aerosol characterizations), improve SO2 sensitivity in noisy spectral regions, expand validation with ground-based networks, and extend analyses to later pandemic phases and additional species (e.g., aerosols, ozone).

- Instrument constraints: EMI has lower SNR (~1/3 of TROPOMI), strong spatial/temporal variability in ISRF FWHM, and infrequent/abnormal irradiance measurements, necessitating simulated irradiance and complex calibration. - Retrieval uncertainties: AMF-related uncertainties dominate systematic errors for NO2 and HCHO; HCHO uncertainties are large in polluted regions (~67%). SO2 retrievals are noisy due to spectral band-edge issues, with a systematic retrieval error ~0.29 DU inferred from the remote Pacific. - Product comparisons: SO2 was not compared with TROPOMI operational products due to known systemic deviations, limiting cross-sensor evaluation. - Model dependence: Stratosphere–troposphere separation (NO2), AMFs, and deweathering rely on GEOS-Chem simulations and TROPOMI-derived cloud/albedo inputs, introducing dependence on model and ancillary data quality. - Data completeness: The paper focuses on early 2020; longer-term and seasonal dynamics and later lockdown phases are less explored within the presented excerpts.