This study presents a machine learning approach called S-MESH (Satellite and ML-based Estimation of Surface air quality at High resolution) for estimating daily surface NO₂ concentrations over Europe at 1 km resolution using TROPOMI observations and XGBoost models. The model uses a combination of TROPOMI NO₂ tropospheric column density, VIIRS night light radiance, MODIS NDVI, and meteorological parameters such as planetary boundary layer height, wind velocity, and temperature. The model achieves a mean absolute error of 7.77 μg/m³, a median bias of 0.6 μg/m³, and a Spearman rank correlation of 0.66. The model performs best at NO₂ concentrations of 10–40 μg/m³ with a bias of less than 40%. The SHAP analysis highlights TROPOMI NO₂ column density as the main source of information for deriving surface NO₂ concentrations, indicating its significant potential for such studies. The SHAP values also indicate the importance of anthropogenic emission proxy inputs such as VIIRS night lights in complementing TROPOMI NO₂ values for deriving higher resolution and detailed spatial patterns of NO₂ variations. The model is computationally efficient and can provide daily estimates of surface NO₂ concentrations over Europe. The model's performance is influenced by NO₂ concentration levels, with the most reliable predictions at concentrations of 10–40 μg/m³. The model's spatial and temporal error analyses indicate its robustness across the study area, with better prediction accuracy during winter months and associated higher NO₂ concentrations. The model's performance is evaluated across different spatial and temporal dimensions, showing good agreement with station observations and demonstrating the potential of satellites for surface NO₂ estimations. The study also highlights the capability of ML models in mapping non-linear, complex relationships, making them a good candidate for further exploring the potential of deriving surface NO₂ concentrations from satellite measurements. The study uses various datasets representing factors that influence NO₂ such as the spatiotemporal distribution of NO₂, location and pattern of anthropogenic activities, meteorological conditions, solar radiation, land use and change, and topography. The study period ranges from 2019 to 2021 with high resolution daily surface NO₂ maps derived for all three years. The model's performance is evaluated across different spatial and temporal dimensions, showing good agreement with station observations and demonstrating the potential of satellites for surface NO₂ estimations. The study also highlights the capability of ML models in mapping non-linear, complex relationships, making them a good candidate for further exploring the potential of deriving surface NO₂ concentrations from satellite measurements.This study presents a machine learning approach called S-MESH (Satellite and ML-based Estimation of Surface air quality at High resolution) for estimating daily surface NO₂ concentrations over Europe at 1 km resolution using TROPOMI observations and XGBoost models. The model uses a combination of TROPOMI NO₂ tropospheric column density, VIIRS night light radiance, MODIS NDVI, and meteorological parameters such as planetary boundary layer height, wind velocity, and temperature. The model achieves a mean absolute error of 7.77 μg/m³, a median bias of 0.6 μg/m³, and a Spearman rank correlation of 0.66. The model performs best at NO₂ concentrations of 10–40 μg/m³ with a bias of less than 40%. The SHAP analysis highlights TROPOMI NO₂ column density as the main source of information for deriving surface NO₂ concentrations, indicating its significant potential for such studies. The SHAP values also indicate the importance of anthropogenic emission proxy inputs such as VIIRS night lights in complementing TROPOMI NO₂ values for deriving higher resolution and detailed spatial patterns of NO₂ variations. The model is computationally efficient and can provide daily estimates of surface NO₂ concentrations over Europe. The model's performance is influenced by NO₂ concentration levels, with the most reliable predictions at concentrations of 10–40 μg/m³. The model's spatial and temporal error analyses indicate its robustness across the study area, with better prediction accuracy during winter months and associated higher NO₂ concentrations. The model's performance is evaluated across different spatial and temporal dimensions, showing good agreement with station observations and demonstrating the potential of satellites for surface NO₂ estimations. The study also highlights the capability of ML models in mapping non-linear, complex relationships, making them a good candidate for further exploring the potential of deriving surface NO₂ concentrations from satellite measurements. The study uses various datasets representing factors that influence NO₂ such as the spatiotemporal distribution of NO₂, location and pattern of anthropogenic activities, meteorological conditions, solar radiation, land use and change, and topography. The study period ranges from 2019 to 2021 with high resolution daily surface NO₂ maps derived for all three years. The model's performance is evaluated across different spatial and temporal dimensions, showing good agreement with station observations and demonstrating the potential of satellites for surface NO₂ estimations. The study also highlights the capability of ML models in mapping non-linear, complex relationships, making them a good candidate for further exploring the potential of deriving surface NO₂ concentrations from satellite measurements.