Proposing and investigating PCAMARS as a novel model for NO₂ interpolation

Area of Science:

• Environmental Science
• Spatial Analysis
• Epidemiology

Background:

• Accurate air pollution measurement, particularly nitrogen dioxide (NO2), is crucial for epidemiological studies and urban air quality management.
• Existing interpolation methods struggle to precisely determine pollutant concentrations in unmonitored urban locations.
• There is a need for advanced spatial predictive models to enhance the accuracy of air pollution mapping.

Purpose of the Study:

• To propose, develop, and test a novel spatial predictive model for NO2 concentration.
• To integrate diverse spatial and meteorological data with air quality measurements for improved prediction accuracy.
• To evaluate the model's performance against established interpolation techniques.

Main Methods:

• Development of a spatial predictive model combining Multivariate Adaptive Regression Splines (MARS) and Principle Component Analysis (PCA).
• Inclusion of spatial data (population, road networks, points of interest) and meteorological data (temperature, pressure, wind speed, humidity) as independent variables.
• Validation against reference interpolation methods: Inverse Distance Weighting, Thin Plate Splines, Kriging, Co-Kriging, and MARS.

Main Results:

• The proposed MARS-PCA model demonstrated superior accuracy in predicting NO2 concentrations compared to reference methods.
• Integration of spatial and meteorological data significantly enhanced the model's predictive capabilities.
• Consistent performance was observed over a 12-month interpolation period.

Conclusions:

• The developed MARS-PCA model offers a more accurate approach for interpolating NO2 concentrations in urban environments.
• This method provides a valuable tool for precise air pollution assessment and effective urban air quality management.
• The findings support the use of integrated data sources and advanced modeling techniques for environmental monitoring.