Estimating PM2.5 concentrations in Northeastern China with full spatiotemporal coverage, 2005-2016
Xia Meng1, Cong Liu1, Lina Zhang1
1School of Public Health, Fudan University, Shanghai, China.
Remote Sensing of Environment
|September 22, 2021
Summary
This study developed a novel gap-filling method to accurately predict historical fine particulate matter (PM2.5) in China, even with missing satellite data. The approach significantly improved air quality predictions, revealing pollution trends and hotspots.
Area of Science:
- Environmental Science
- Atmospheric Science
- Public Health
Background:
- Accurate long-term prediction of fine particulate matter (PM2.5) is crucial for understanding its health impacts and evaluating air pollution control policies in China.
- Satellite-retrieved aerosol optical depth (AOD) is valuable for high-resolution PM2.5 monitoring, but high missing rates in regions like Northeastern China (NEC), especially during winter, hinder accurate predictions.
- Addressing AOD data gaps is essential for reliable historical PM2.5 assessment and policy evaluation.
Purpose of the Study:
- To develop and validate a robust gap-filling approach for predicting historical ground-level PM2.5 concentrations in Northeastern China (NEC) using satellite AOD and other relevant data.
- To assess the accuracy and reliability of the developed model in filling gaps caused by missing AOD data, particularly during winter months.
- To analyze the spatiotemporal trends and identify pollution hotspots of PM2.5 in NEC from 2005 to 2016.
Main Methods:
- Utilized random forest algorithms to integrate satellite AOD, meteorological data, land use parameters, population density, and visibility for PM2.5 prediction.
- Developed a gap-filling model combining predictors with and without AOD to achieve full spatial coverage at a 1-km resolution.
- Employed daily and monthly level validation, including leave-one-year-out cross-validation, to assess prediction accuracy (R², RMSE).
Main Results:
- The full-coverage PM2.5 prediction model achieved a daily R² of 0.81 (RMSE: 18.5 μg/m³) and a monthly R² of 0.65 (RMSE: 16.3 μg/m³).
- Gap-filling significantly reduced prediction errors on days with missing AOD, decreasing relative error from 28% to 2.5% in winter.
- Predicted PM2.5 levels in NEC showed an increase from 2005, peaking between 2013-2015, followed by a decline in 2016, with a south-to-north gradient and identified pollution hotspots.
Conclusions:
- The developed gap-filling approach effectively overcomes challenges posed by missing AOD data, enabling reliable long-term historical PM2.5 prediction in high-latitude regions.
- The model enhances the accuracy of PM2.5 estimates, providing valuable data for health research and air pollution policy assessment.
- The study highlights the importance of integrating multiple data sources for comprehensive air quality monitoring and analysis, especially in data-scarce areas.
