[PM_2.5 Inversion Using Remote Sensing Data in Eastern China Based on Deep Learning].

Area of Science:

• Environmental Science
• Atmospheric Science
• Data Science

Context:

• Particulate Matter (PM2.5) is a critical air pollutant with significant health implications.
• Accurate monitoring of PM2.5 is essential for public health and environmental management.
• Traditional methods for PM2.5 estimation often face limitations in spatial and temporal resolution.

Purpose:

• To develop and validate a novel multi-element joint PM2.5 inversion method utilizing a deep learning model.
• To assess the seasonal effectiveness of the proposed deep learning approach using remote sensing data in Eastern China.
• To compare the performance of the deep learning model against traditional linear and nonlinear models for PM2.5 estimation.

Summary:

• A deep learning model was developed to invert PM2.5 concentrations using Himawari-AOD, meteorological data (temperature, humidity, pressure), and other factors.
• Seasonal experiments in Eastern China (2016-2018) showed strong correlations between PM2.5 and AOD, precipitation, wind speed, and vegetation indices.
• The deep neural network (DNN) model significantly outperformed traditional methods, achieving R² values up to 0.86 in autumn and providing higher resolution, more accurate PM2.5 distribution maps.

Impact:

• The study presents a more accurate and higher-resolution method for estimating PM2.5 concentrations, crucial for air quality assessment.
• The findings support the use of deep learning in environmental monitoring, offering a powerful tool for understanding air pollution dynamics.
• Improved PM2.5 inversion can aid in developing targeted public health interventions and environmental policies.