Predicting intraurban PM_2.5 concentrations using enhanced machine learning approaches and incorporating human...

Area of Science:

• Environmental Science
• Data Science
• Urban Planning

Background:

• Urban air pollution, particularly fine particulate matter (PM2.5), poses significant risks to human health.
• Traditional statistical and machine learning (ML) models for predicting urban PM2.5 often rely on linear correlations and limited variables.
• Existing models frequently overlook dynamic, human-related factors such as building occupancy and real-time mobility.

Purpose of the Study:

• To develop and evaluate an enhanced ML approach for predicting urban ambient PM2.5 concentrations.
• To investigate the impact of incorporating dynamic human-related factors, including mobility and building occupancy, into air quality prediction models.
• To identify the most effective ML algorithm for modeling intraurban PM2.5 variations.

Main Methods:

• Hybridized cascade and Principal Component Analysis (PCA) methods to reduce data dimensionality and capture nonlinear relationships.
• Tested nine state-of-the-art ML algorithms, including Gaussian-kernel support vector regression (SVR).
• Utilized hourly PM2.5 concentration data from Chicago, IL, incorporating meteorological, land use, co-pollutant, mobility, and building occupancy data.

Main Results:

• Gaussian-kernel SVR demonstrated superior performance, achieving 118% higher accuracy than traditional Multiple Linear Regression (MLR).
• The enhanced ML approach with SVR improved model performance by up to 18.4% for yearlong and 98.7% for month-long datasets.
• Including building occupancy patterns enhanced model performance by 4% to 37%.

Conclusions:

• The proposed enhanced ML approach significantly improves the accuracy of urban PM2.5 prediction models.
• Dynamic human-related factors, especially building occupancy, are crucial for accurate intraurban air quality modeling.
• This innovative approach offers a more effective alternative to conventional methods for monitoring and managing urban air quality.