Elevating hourly PM2.5 forecasting in Istanbul, Türkiye: Leveraging ERA5 reanalysis and genetic algorithms in a comparative machine learning model analysis
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.
View abstract on PubMed
Summary
This summary is machine-generated.Accurate PM2.5 prediction is crucial for urban air quality management. The Nonlinear Autoregressive with Exogenous Inputs (NARX) model, optimized with a genetic algorithm, demonstrated superior performance in predicting PM2.5 concentrations in Istanbul.
Area Of Science
- Environmental Science
- Data Science
- Urban Planning
Background
- Rapid urbanization and industrialization increase air pollution, posing severe health risks.
- Accurate prediction of fine particulate matter (PM2.5) is essential for effective urban air quality management.
- Istanbul, a densely populated city, faces complex air pollution challenges due to diverse emission sources.
Purpose Of The Study
- To assess and compare the predictive accuracy of various machine learning (ML) models for PM2.5 concentrations in Istanbul.
- To investigate the spatial variability of PM2.5 across different regions of Istanbul using high-resolution data.
- To identify key meteorological, land cover, and vegetation variables influencing PM2.5 concentrations.
Main Methods
- Utilized high-resolution ERA5 reanalysis data for grid-based spatial analysis of Istanbul.
- Compared Multiple Linear Regression (MLR), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting (LGB), Random Forest (RF), and Nonlinear Autoregressive with Exogenous Inputs (NARX) models.
- Employed genetic algorithm optimization for the NARX model and Neighborhood Component Analysis (NCA) for variable selection.
Main Results
- The NARX model demonstrated superior performance (R-value: 0.89, RMSE: 5.24 μg/m³, MAE: 2.94 μg/m³), outperforming RF, LGB, XGBoost, and MLR.
- PM2.5 prediction accuracy varied seasonally, with better performance in autumn and winter.
- Highest accuracy was observed in Region-1 (R-value: 0.94), while Region-5 showed the lowest (R-value: 0.75).
Conclusions
- The optimized NARX model is robust for predicting PM2.5 in complex urban environments, even with limited monitoring data.
- The methodology offers potential for global application in similar urban air quality management contexts.
- Advanced data analysis techniques are vital for developing targeted pollution control strategies and public health policies.
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.
