Table 1 Comparison of performance metrics across study areas and pollutants.
Study | Methodology | Data sources | Key innovation | Research gap |
|---|---|---|---|---|
Chen et al12 | CNN-LSTM + attention | Single satellite + ground | Neighborhood selection | Limited data fusion |
Ahmad et al20 | BiGRU-1DCNN | Ground stations | Multi-station analysis | Single pollutant, single region |
Nguyen et al17 | Hybrid DL + QPSO | Ground + meteorological | Quantum-inspired optimization | High computational cost |
Ahmad et al21 | RNN-BiGRU | Time series | Novel imputation | Single modality |
Xia et al15 | Multi-modal DL | Satellite + time-series | Beijing/Tianjin fusion | Region-specific, no uncertainty |
Mahmood et al23. | WaveNet-XGBoost | Ground stations | Ensemble learning | Limited spatial coverage |
Kumar et al22 | Hybrid time series | Ground + meteorological | Spatio-temporal analysis | Linear assumptions |
Duan et al9 | ARIMA-CNN-LSTM | Time series | Dung beetle optimization | Extensive tuning needed |
