Daily TROPOMI tropospheric NO₂ vertical column density (TVCD) Level-3 data at 0.022° x 0.022° resolution (~2.45 km) were aggregated to urban polygons from the Global Urban Polygons and Points Dataset (GUPPD v1). The analysis was restricted to urban areas with estimated 2020 population >100,000. For each city and day, NO₂ TVCD was calculated as an area-weighted mean of all grid cells intersecting the polygon, provided that at least 80% of the polygon area was covered by valid satellite retrievals. Meteorological normalization was performed separately for each city with the AirGAM framework, a generalized additive model (GAM) approach implemented in R using `mgcv`. The response variable was satellite-observed NO₂ TVCD, modeled with a log-link as a function of ERA5-Land meteorology and time covariates. Meteorological predictors included wind speed, wind direction, 2 m air temperature, relative humidity (derived from dew point), boundary layer height, and surface solar radiation, sampled around the Sentinel-5P overpass (~13:30 local time) and aggregated as area-weighted city means. Additional covariates included day of week, day of year, and total elapsed time. The displayed trend metric is the mean annual meteorology-adjusted relative change for 2019-2024, derived from the fitted AirGAM trend term. Annual growth ratios were computed from successive yearly values of the fitted trend curve, and their geometric mean was converted to percent change per year. Uncertainty is propagated from the 95% confidence bounds of the fitted trend term. Trends are classified as significant only when the 95% confidence interval implies a consistently increasing or decreasing annual change over the analysis period. Popup mini-plots show the fitted meteorology-adjusted trend term for 2019-2024 together with its 95% confidence bounds. For more detailed information about methodology and results, see our paper at URL.