Fine particulate matter (PM₂.₅) poses severe public health and environmental risks in Nepal’s Tarai and Dun Valley regions, where ground-based air quality monitoring is spatially sparse and temporally inconsistent. This study integrates ground-based observations, satellite-derived Aerosol Optical Depth (AOD) from MODIS MCD19A2, TROPOMI trace gases (CO, NO₂, SO₂), and ERA5 meteorological reanalysis (temperature, relative humidity, wind components) to characterize the spatiotemporal variability of PM₂.₅ and to reconstruct long-term trends from 2000 to 2023. Daily PM₂.₅ measurements from six monitoring stations across southern Nepal were analyzed alongside collocated satellite and reanalysis data. Correlation analysis revealed strong positive associations between PM₂.₅ and both AOD (r = 0.59) and CO (r = 0.62), while temperature (r = −0.38) and relative humidity (r = −0.46) showed moderate negative correlations. A Random Forest model incorporating AOD, CO, temperature, relative humidity, and wind components achieved robust predictive performance (R2 = 0.65, RMSE = 22.6 µg/m3), substantially outperforming multiple linear regression (R2 = 0.54). The strong contribution of CO performance dropped to R2 = 0.62 when excluded, highlights the dominance of combustion sources (biomass burning, vehicular emissions, forest fires) in driving PM₂.₅ pollution. Long-term reconstruction using AOD and meteorological variables (excluding CO due to limited historical data) revealed a distinct east–west gradient. Eastern stations (Jhumka, Bharatpur, Hetauda) show statistically significant increasing trends (up to + 0.41 µg/m3/year, p < 0.01), while western stations (Dhangadhi, Bhimdatta, Dang) exhibit stable or slightly declining trends. Seasonal analysis showed the highest concentrations during winter and pre-monsoon (January–April, > 60 µg/m3) and substantial reductions during monsoon months due to rainfall-driven washout. The results underscore the importance of integrating satellite and ground-based data with machine learning to assess historical air quality, identify pollution hotspots, and inform evidence-based mitigation strategies in data-sparse regions. This framework provides a robust basis for air quality management and public health planning in southern Nepal and across the Indo-Gangetic Plain, where transboundary cooperation is essential to reverse worsening pollution trends.

Integrating satellite AOD, trace gases, and meteorological reanalysis to estimate PM₂.₅ and its long-term trends in southern Nepal

期刊论文
发表日期：

2026年6月19日

查看全文

Govinda Prasad Lamichhane,
Nabina Maharjan,
Niroj Timalsina

摘要

Fine particulate matter (PM₂.₅) poses severe public health and environmental risks in Nepal’s Tarai and Dun Valley regions, where ground-based air quality monitoring is spatially sparse and temporally inconsistent. This study integrates ground-based observations, satellite-derived Aerosol Optical Depth (AOD) from MODIS MCD19A2, TROPOMI trace gases (CO, NO₂, SO₂), and ERA5 meteorological reanalysis (temperature, relative humidity, wind components) to characterize the spatiotemporal variability of PM₂.₅ and to reconstruct long-term trends from 2000 to 2023. Daily PM₂.₅ measurements from six monitoring stations across southern Nepal were analyzed alongside collocated satellite and reanalysis data. Correlation analysis revealed strong positive associations between PM₂.₅ and both AOD (r = 0.59) and CO (r = 0.62), while temperature (r = −0.38) and relative humidity (r = −0.46) showed moderate negative correlations. A Random Forest model incorporating AOD, CO, temperature, relative humidity, and wind components achieved robust predictive performance (R² = 0.65, RMSE = 22.6 µg/m³), substantially outperforming multiple linear regression (R² = 0.54). The strong contribution of CO performance dropped to R² = 0.62 when excluded, highlights the dominance of combustion sources (biomass burning, vehicular emissions, forest fires) in driving PM₂.₅ pollution. Long-term reconstruction using AOD and meteorological variables (excluding CO due to limited historical data) revealed a distinct east–west gradient. Eastern stations (Jhumka, Bharatpur, Hetauda) show statistically significant increasing trends (up to + 0.41 µg/m³/year, p < 0.01), while western stations (Dhangadhi, Bhimdatta, Dang) exhibit stable or slightly declining trends. Seasonal analysis showed the highest concentrations during winter and pre-monsoon (January–April, > 60 µg/m³) and substantial reductions during monsoon months due to rainfall-driven washout. The results underscore the importance of integrating satellite and ground-based data with machine learning to assess historical air quality, identify pollution hotspots, and inform evidence-based mitigation strategies in data-sparse regions. This framework provides a robust basis for air quality management and public health planning in southern Nepal and across the Indo-Gangetic Plain, where transboundary cooperation is essential to reverse worsening pollution trends.