Spatial modeling of PM2.5 concentrations with a multifactoral radial basis function neural network

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Data Science

Background:

• Accurate fine particulate matter (PM2.5) monitoring is crucial for public health.
• Sparse stationary monitoring networks limit comprehensive spatial and temporal characterization of PM2.5 concentrations.

Purpose of the Study:

• To introduce and evaluate a radial basis function (RBF) neural network model for estimating PM2.5 concentrations.
• To assess the utility of meteorological and land-related factors in improving PM2.5 estimation accuracy with sparse data.
• To enhance air quality assessment for epidemiological and environmental research.

Main Methods:

• Developed a radial basis function (RBF) neural network model.
• Utilized meteorological and land-related factors as covariates for PM2.5 estimation.
• Evaluated model performance using statistical indices like mean square error and correlation coefficient in Texas, USA.

Main Results:

• RBF models incorporating meteorological and/or land-related factors significantly improved PM2.5 concentration estimation accuracy.
• Combined meteorological and land-related factors yielded the best performance compared to individual factor models.
• The RBF neural network demonstrated effective PM2.5 estimation with sparse monitoring data.

Conclusions:

• Meteorological and land-related factors are valuable predictors of PM2.5 concentration variability.
• RBF neural networks offer a viable solution for accurate PM2.5 estimation in data-scarce regions.
• Improved PM2.5 estimations support better local air pollution characterization and exposure reduction strategies.