Spatio-Temporal Dynamics of Urban Thermal-Pollution Interactions in Delhi and Its Surrounding Areas: A Remote Sensing and GAM Modelling Approach.

Abstract

Urban thermal dynamics, influenced by air pollution and meteorological variability, are critical for assessing urban resilience in rapidly urbanizing megacities. LST serves as a vital indicator for understanding these interactions under changing climatic and environmental conditions. This study investigates the long-term spatio-temporal variations of LST and criteria air pollutants, their correlations with meteorological parameters and dominant factors influencing LST variability using the Generalized Additive Model (GAM) modelling. MODIS-derived LST and spectral indices, ERA5 meteorological data and pollutant data from the CPCB were retrieved. Results reveal that Premonsoon daytime and nighttime LST varied between 33.88-36.53 °C and 19.57-21.72 °C, respectively, during 2014-2023. The UTFVI analysis found that the urban core falls within the "worse" and "worst" thermal stress categories (> 0.020) at night. Sen's slope found that SO2 shows a persistent increasing trend across all seasons, with a positive slope in winter (4.58 μg/m3 per year). Air temperature positively correlates with pre-monsoon daytime LST (r = 0.86), while wind speed negatively correlates with monsoon nighttime LST (r = -0.95). The GAM model demonstrates predictive performance for both daytime and nighttime LST (R2 > 0.93). SMI dominates daytime LST in all seasons (ΔR2 ≈ 0.004-0.018) whereas nighttime LST is mainly influenced by SO2 (ΔR2 = 0.024) in monsoon and EVI in pre- and post-monsoon (ΔR2 = 0.014 and 0.020). The pollutant variables contribute similarly (ΔR2 ≈ 0.005-0.006) in winter nights. This study reveals the urban thermal-pollution-meteorological dynamics, providing valuable insights for climate-responsive urban planning, heat mitigation strategies and improved air-quality management.