Nowcasting and forecasting urban air pollution with meteorological and wind profile data: A framework for Smart City Digital Twins

This research outlines an integrated approach for nowcasting and short-term forecasting of urban air pollution considering the meteorological and wind profile data. The system aims to improve the management of urban air quality and the monitoring of the built environment's durability in the context of smart city digital twins. The Series-cOre Fused Time Series forecaster (SOFTS) model is the main predictive element of the framework. Its efficacy and robustness have been thoroughly tested using real-world data from Lamezia Terme, Italy, under fully and partially reconstructed data conditions. For benchmarking, the framework utilizes two conventional machine learning methods, Extreme Gradient Boosting (XGBoost), and Long Short Term Memory (LSTM) networks, for comparative

analysis. The framework demonstrates high accuracy and resilience, according to numerical results. During real-time nowcasting in Lamezia, the SOFTS model consistently outperformed the XGBoost benchmark and had coefficients of determination above 0.91 and weighted mean absolute percentage errors below 21% for several target pollutants. For short-term forecasting up to six hours, the model maintained predictive capability for particulate matter with R2 values above 0.70 at the six-hour mark. Validation from the Cabauw site in the Netherlands further substantiates the generalizability of the framework.

Message: Hi FE, I need help with “Nowcasting and forecasting urban air pollution with meteorological and wind profile data: A framework for Smart City Digital Twins” in “Environmental Engineering”