A neural network-based method for modeling PM 2.5 measurements obtained from the surface particulate matter network.

Area of Science:

• Environmental Science
• Atmospheric Science
• Data Science

Background:

• Air pollution, particularly fine particulate matter (PM2.5), poses significant global health risks, including respiratory and cardiovascular diseases.
• Effective air quality monitoring is essential for public health and environmental protection strategies worldwide.

Purpose of the Study:

• To develop and evaluate artificial neural network models for estimating PM2.5 concentrations.
• To assess the performance of these models using meteorological parameters and aerosol optical depth (AOD) as inputs.

Main Methods:

• Utilized artificial neural networks to train on time-series PM2.5 measurements from the Surface Particulate Matter Network (SPARTAN).
• Incorporated meteorological data and AOD as input features for the neural network models.
• Validated model performance against SPARTAN measurements at a sub-Saharan site in Ilorin.

Main Results:

• The developed artificial neural network models demonstrated high correlation with measured PM2.5 data, with R² values ranging from 0.59 to 0.95.
• The PRB model showed superior performance in estimating both low and high PM2.5 concentrations compared to other models.
• Most models exhibited good performance, with only one model showing a very low correlation (0.0009).

Conclusions:

• Artificial neural networks are effective tools for accurately estimating PM2.5 air quality.
• The study highlights the potential of these models for PM2.5 monitoring in data-scarce regions like sub-Saharan Africa.
• The PRB model offers a promising approach for comprehensive PM2.5 estimation, outperforming models that tend to under or overpredict emissions.