Environmental Engineering Urban Environmental Intelligence Framework for Air Quality Prediction Using Multi-Source Mobility and Climate Data

Abstract

Rising urban mobility, intensive anthropogenic activity, and limitations in conventional monitoring systems capable of delivering continuous spatial coverage have turned urban air pollution into a pressing environmental concern. Existing air-quality prediction approaches tend to depend on single-source observations or are built primarily for large metropolitan areas, which limits how well they apply to secondary cities with localized pollution dynamics. To address this, the study introduces an Urban Environmental Intelligence Framework that brings together multi-source air quality, climate, spatial, temporal, and air-quality information for urban air-quality prediction. Evaluation of the framework was conducted using hourly observations of pollutant concentrations (PM2.5, NO2, and CO), mobility indicators (traffic volume, vehicle count, average speed, and congestion index), meteorological variables, and urban spatial attributes, collected from Lhokseumawe, Indonesia. A multi-source XGBoost model was benchmarked against Single-Source XGBoost, Random Forest, and LSTM models using a chronological data partition and multiple evaluation metrics, including MAE, RMSE, MAPE, and R². The proposed framework achieved the strongest predictive performance across all target variables evaluated, attaining an R² of 0.912 and an RMSE of 5.84 µg/m³ for PM2.5 prediction. Ablation analysis confirmed that temporal and mobility information contributed most substantially to prediction accuracy, while climate and spatial variables provided complementary contextual value. Feature interpretation further revealed that traffic intensity, historical pollutant levels, and meteorological conditions were the dominant drivers of urban pollution outcomes. Overall, the proposed framework offers an effective environmental approach for supporting short-term air-quality forecasting, pollution hotspot identification, and evidence-based urban environmental management in secondary cities