Sensitivity Analysis for Predicting Sub-Micron Aerosol Concentrations Based on Meteorological Parameters.

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Air Quality Monitoring

Background:

• Sub-micron aerosols, measured as particle number concentration (PN), are critical air pollutants with significant health impacts.
• Data scarcity in PN measurements at air quality stations necessitates robust modeling approaches.
• Developing accurate PN models is crucial for comprehensive air quality assessment.

Purpose of the Study:

• To present a novel PN modeling framework utilizing sensitivity analysis.
• To evaluate the performance of artificial neural networks (ANNs) for PN concentration prediction.
• To identify optimal meteorological parameters as descriptors for PN modeling.

Main Methods:

• A year-long aerosol measurement campaign in Amman, Jordan, was used for model development and testing.
• Feed-forward neural networks (FFNN) and time-delay neural networks (TDNN) were employed as modeling tools.
• Sensitivity analysis was conducted on various combinations of meteorological parameters (temperature, humidity, pressure, wind speed) as model inputs.

Main Results:

• FFNN achieved high accuracy (R² = 0.77) for daily averaged PN data.
• TDNN demonstrated good performance (R² = 0.66) for hourly averaged PN data.
• Temperature, relative humidity, pressure, and wind speed were identified as key meteorological descriptors.

Conclusions:

• The developed PN models successfully capture diurnal patterns and provide satisfactory estimations.
• These models offer a viable solution for estimating PN concentration when direct measurements are unavailable or incomplete.
• The framework enhances air quality monitoring capabilities by leveraging meteorological data for PN prediction.