Ground-Level NO₂ Surveillance from Space Across China for High Resolution Using Interpretable Spatiotemporally...

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Artificial Intelligence

Background:

• Ground-level nitrogen dioxide (NO2) is a significant air pollutant impacting environmental quality and public health.
• Satellite data often have gaps, limiting comprehensive analysis of NO2 distribution.

Purpose of the Study:

• To develop an AI model for filling satellite data gaps and deriving high-resolution daily surface NO2 concentrations over mainland China.
• To create a comprehensive dataset for analyzing fine-scale NO2 spatial patterns and temporal variations.

Main Methods:

• Integrated spatiotemporally weighted information into extra-trees and deep forest models.
• Combined surface NO2 measurements, satellite data (TROPOMI, OMI), reanalysis, and model simulations.
• Achieved 1 km daily surface NO2 estimates with 100% spatial coverage for 2019-2020.

Main Results:

• Developed the "ChinaHighNO2" dataset with 49% increased data availability.
• Achieved high accuracy with R-squared values of 0.93 (out-of-sample) and 0.71 (out-of-city).
• Observed significant urban-rural NO2 differences (28% average), holiday effects, and minimal weekday-weekend variations.

Conclusions:

• The "ChinaHighNO2" dataset provides a valuable tool for studying NO2 pollution at fine scales.
• Surface NO2 data better reflect emission changes during events like the COVID-19 pandemic compared to tropospheric columns.