Modeling of atmospheric particulate matters via artificial intelligence methods

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Artificial Intelligence in Environmental Monitoring

Background:

• Increasing atmospheric pollutants pose significant environmental and health risks.
• Particulate matter (PM), including PM10 and PM2.5, is a major air pollutant formed through natural and anthropogenic sources or atmospheric reactions.
• Accurate prediction of PM levels is crucial for effective air quality management.

Purpose of the Study:

• To predict particulate matter (PM10 and PM2.5) concentrations in an industrial zone.
• To evaluate the performance of various artificial intelligence methods for air pollutant prediction.
• To identify the most effective AI model for forecasting PM levels.

Main Methods:

• Air pollutant and meteorological data were collected and evaluated.
• PM10 and PM2.5 components were modeled using R software.
• Artificial intelligence methods including ANFIS, SVR, CART, RF, KNN, and ELM were employed for prediction.

Main Results:

• The adaptive neuro-fuzzy inference system (ANFIS) model demonstrated the highest accuracy in predicting PM10 (R²=0.95) and PM2.5 (R²=0.97).
• ANFIS outperformed other tested AI methods, including SVR, CART, RF, KNN, and ELM.
• Performance metrics like RMSE and MAE further validated ANFIS's superior predictive capability.

Conclusions:

• The ANFIS model is highly effective for predicting air pollutant concentrations, specifically PM10 and PM2.5.
• AI-driven models offer a promising approach for real-time air quality monitoring and forecasting.
• Findings support the integration of ANFIS in environmental management strategies for industrial zones.