Spatially and temporally coherent reconstruction of tropospheric NO₂ over China combining OMI and GOME-2B measurements

Abstract

Tropospheric NO2 columns retrieved from ozone monitoring instrument (OMI) are widely used, even though there is a significant loss of spatial coverage due to multiple factors. This work introduces a framework for reconstructing gaps in the OMI NO2 data over China by using machine learning and an adaptive weighted temporal fitting method with NO2 measurements from Global Ozone Monitoring Experiment–2B, and surface measurements. The reconstructed NO2 has four important characteristics. First, there is improved spatial and temporal coherence on a day-to-day basis, allowing new scientific findings to be made. Second, the amount of data doubled, with 40% more data available. Third, the results are reliable overall, with a good agreement with Multi-AXis Differential Optical Absorption Spectroscopy measurements (R: 0.75–0.85). Finally, the mean of reconstructed NO2 vertical columns during 2015 and 2018 is consistent with the original data in the spatial distribution, while the standard deviation decreases in most places over Mainland China. This novel finding is expected to contribute to both air quality and climate studies.

National Land Use Regression Model for NO2 Using Street View Imagery and Satellite Observations

“…33 Those results suggest that images may capture similar features as traditional GIS predictors while also allowing for the inclusion of higherresolution (i.e., street-level) features that traditional GIS variables lack. 37 Another notable trend in recent air quality modeling studies is the adoption of advanced machine learning approaches to improve model performance, including random forest, 39−41 gradient boosting, 42,43 artificial neural network, 44−46 and hybrid algorithms. 47,48 Our previous study successfully developed single-city LUR models solely using street view images.…”

Ground-level gaseous pollutants across China: daily seamless mapping and long-term spatiotemporal variations

“…Satellite remote sensing data used here include the daily seamless tropospheric NO2 products (0.25° × 0.25°) generated by first combining OMI/Aura and Global Ozone Monitoring Experiment-2B retrievals (He et al, 2020), and then gap-filling using CAMS tropospheric NO2 simulations via machine learning (Wei et al, 2022b), and MODIS monthly NDVI (0.05° × 0.05°), LandScan TM annual population (POP, 1 km) (Bright et al, 2000), and the SRTM digital elevation model (DEM, 90 m). ERA5-Land (0.1° × 0.1°) and ERA5 global reanalysis (0.25° × 0.25°) provided hourly meteorological fields (Muñoz-Sabater et al, 2021;Hersbach et al, 2020) surface mass concentrations were also included from the MERRA-2 and GEOS-FP global reanalysis every 1 and 3 hours at horizontal resolutions of 0.625° × 0.5° and 0.3125° × 0.25°, respectively.…”

Combing GOME-2B and OMI Satellite Data to Estimate Near-Surface NO2 of Mainland China

“…At present, large-scale monitoring research has been realized based on the huge advantages of satellite remote sensing [21], [25]. The retrieval of troposphere NO2 vertical column densities (VCDs) from satellites sensors such as Ozone Monitoring Instrument (OMI) [26], [27] and Global Ozone Monitoring Experiment (GOME) [28]- [30] have been widely used in the mapping of the spatiotemporal distribution and change analysis of tropospheric NO 2 [31]- [34], while higher quality upward looking surface observations such as MAX-DOAS have been used to establish such conditions in specific locations [35], [36]. The relationship between surface and column observations have been constructed to estimate the NS-NO 2 concentration with the support of chemical transport models [37]- [40], geographically and temporally weighted regression [41]- [43], and machine learning methods [13], [20], [32], [35], [44], [45].…”

“…Most related studies [32], [37], [42], [46] have employed the NO2 column data from OMI, since it is the first sensor to provide daily global coverage of NO 2 at a relatively high resolution [47]. However, due to there being only a single pass per day, data loss near the swath edges, cloud cover, and other issues, these results generally do not reflect the actual NS-NO 2 concentration [35]. This study aims to obtain a more accurate measure of NS-NO 2 by combining GOME-2B and OMI satellite measurements in a machine learning framework, taking advantage of the individual strengths of both sensors.…”