Satellite Image-Based Estimation of Urban Air Pollution in Tehran Under Climate Change Using Digital Image Processing for Air Quality Monitoring

Abstract

Introduction: Urban air pollution is a major global health threat, exacerbated by climate change through altered temperature inversions, boundary layer dynamics, and pollutant dispersion. Ground based monitoring networks are costly and spatially limited. This study assesses the feasibility of estimating the Air Quality Index (AQI) from Google Earth satellite imagery of Tehran using three digital image processing methods in MATLAB. Materials and Methods: Seven cloud free satellite images acquired between July 2022 and April 2024 (AQI range: 64-172) were analyzed. Three image derived features were correlated with AQI: (1) pixel intensity standard deviation from normalized grayscale histograms, (2) Pearson cross correlation coefficients relative to a clean day reference, and (3) mean edge pixel density from Canny edge detection. Linear regression models were trained on five images and validated on two withheld test images. Results and Discussion: Pixel standard deviation showed a strong inverse relationship with AQI (R² = 0.82; regression slope: 1160.98 ± 312.4, intercept: 361.71 ± 75.3), with test errors of 1.25% and 29.66%. Canny edge detection mean density yielded R² = 0.80 (slope: 925.41 ± 268.7, intercept: 319.35 ± 64.9) with balanced errors of 10.11% and 12.81%, indicating greater seasonal robustness. Cross correlation confirmed that pollution distorts structural image similarity, but showed limited predictive consistency across seasons due to illumination variability. Conclusion: The results suggest that freely available optical satellite imagery may encode air quality relevant information accessible through simple image operations, without specialized sensors or atmospheric correction.