Spatiotemporal variability of NO<sub>2</sub> and PM<sub>2.5</sub> over Eastern China: observational and model analyses with a novel statistical method

Abstract

Abstract. Eastern China (27–41∘ N, 110–123∘ E) is heavily polluted by nitrogen dioxide (NO2), particulate matter with aerodynamic diameter below 2.5 µm (PM2.5), and other air pollutants. These pollutants vary on a variety of temporal and spatial scales, with many temporal scales that are nonperiodic and nonstationary, challenging proper quantitative characterization and visualization. This study uses a newly compiled EOF–EEMD analysis visualization package to evaluate the spatiotemporal variability of ground-level NO2, PM2.5, and their associations with meteorological processes over Eastern China in fall–winter 2013. Applying the package to observed hourly pollutant data reveals a primary spatial pattern representing Eastern China synchronous variation in time, which is dominated by diurnal variability with a much weaker day-to-day signal. A secondary spatial mode, representing north–south opposing changes in time with no constant period, is characterized by wind-related dilution or a buildup of pollutants from one day to another. We further evaluate simulations of nested GEOS-Chem v9-02 and WRF/CMAQ v5.0.1 in capturing the spatiotemporal variability of pollutants. GEOS-Chem underestimates NO2 by about 17 µg m−3 and PM2.5 by 35 µg m−3 on average over fall–winter 2013. It reproduces the diurnal variability for both pollutants. For the day-to-day variation, GEOS-Chem reproduces the observed north–south contrasting mode for both pollutants but not the Eastern China synchronous mode (especially for NO2). The model errors are due to a first model layer too thick (about 130 m) to capture the near-surface vertical gradient, deficiencies in the nighttime nitrogen chemistry in the first layer, and missing secondary organic aerosols and anthropogenic dust. CMAQ overestimates the diurnal cycle of pollutants due to too-weak boundary layer mixing, especially in the nighttime, and overestimates NO2 by about 30 µg m−3 and PM2.5 by 60 µg m−3. For the day-to-day variability, CMAQ reproduces the observed Eastern China synchronous mode but not the north–south opposing mode of NO2. Both models capture the day-to-day variability of PM2.5 better than that of NO2. These results shed light on model improvement. The EOF–EEMD package is freely available for noncommercial uses.

High-resolution (0.05° × 0.05°) NOx emissions in the Yangtze River Delta inferred from OMI

“…The change in NO 2 VCD from June to August is relatively small, reducing the effect of intraseasonal variability when deriving NO x emissions from summer mean NO 2 VCDs. We screen out the 30 outer pixels with VZA larger than 30 • (crosstrack width larger than 36 km) that greatly smear the spatial gradient of NO 2 , pixels with cloud radiance fraction exceeding 50 %, and pixels with AOD larger than 3 (i.e., when the aerosol data used in the NO 2 retrieval are unreliable and the NO 2 retrieval is subject to an excessive error) (Lin et al, , 2015Liu et al, 2019a). We also exclude data with raw anomaly problems (http://projects.knmi.nl/omi/research/ product/rowanomaly-background.php, last access: 30 January 2018).…”

Response to reviewer#3

“…The gaps between different studies implied potentially large uncertainties in BC bottom-up emission inventories. The uncertainties of BC emission estimates for China were reported at ±484 %, ±208 %, and ±98 % by Streets et al (2003), Zhang et al (2009), and Lu et al (2011), respectively. Due to a lack of sufficient local field tests, emission factors were commonly taken from foreign studies with big variety depending on fuel and combustion condition (Bond et al, 2004;Cao et al, 2006;Lei et al, 2011;Qin and Xie, 2012;Streets et al, 2003Streets et al, , 2001Zhang et al, 2009).…”

Optimal Inversion of Conversion Parameters from Satellite AOD to Ground Aerosol Extinction Coefficient Using Automatic Differentiation

“…As one primary component of the aerosol mass, spatiotemporal variability of the PM 2.5 concentration (µg/m 3 ) is affected by multiple factors, including various anthropogenic and natural emission sources [8][9][10], meteorology likely involved in complex atmospheric chemical processes for generation of secondary PM 2.5 [11,12], local elevation, and terrain [12,13]. Complex atmospheric chemical processes involving these factors and their interactions result in high uncertainty that presents a challenge in the estimation of PM 2.5 , particularly before the launch of the Moderate Resolution Imaging Spectroradiometer (MODIS) in 1999, when no such strong predictors as satellite aerosol optical depth (AOD) were available.…”

“…In this paper, in comparison with the existing methods, a more flexible conversion method is proposed. Considering multiple complex atmospheric and environmental factors (e.g., emission sources [8,10,32], meteorology [11,12], and elevation [12,13]) involved in the aerosol vertical profile, this method introduces a scaling factor (slope) and a shift factor (intercept) for both planetary boundary layer height (PBLH) and relative humidity (RH) to capture the influence of potential other confounders (e.g., atmospheric chemical processes and other meteorological factors) or random factors based on the simplified conversion formula in [15]. For the optimization of multiple parameters in the proposed conversion formula, automatic differentiation (AD) was used in gradient descent to improve learning efficiency.…”