Urban Greenness and Summer Land Surface Temperature: Evidence from a 1-km Grid Panel of Berlin, Budapest, and Vienna

Urban greening is commonly promoted as a heat-mitigation strategy, but most of the remote-sensing evidence is cross-sectional and thus makes the vegetation image associated with persistent differences in urban form and location. In this study, we build a 2012-2021 summer panel of 1,818 one-kilometre grids across Berlin, Budapest, and Vienna, with 18,156 grid-year observations. Using grid fixed effects, city-by-year fixed effects, ERA5 weather controls, and grid-clustered standard errors, we estimate the annual summer LST to be related to the Normalized Difference Vegetation Index (NDVI) with a 0.1 increase in NDVI (SE = 0.033) and a 0.835 °C reduction in the summer LST (SE = 0.033). The estimation is stable when boundary-area weights are used, interior grids are limited, balanced panels are used, winsorization is used, and Driscoll-Kraay inference is performed. The city-specific estimates are -0.708 °C in Berlin, -0.930 °C in Budapest, and -0.739 °C in Vienna. The MODIS data show that a higher NDVI is associated with richer evapotranspiration and latent heat flux, as expected from an evaporative cooling channel, but the missing values over an otherwise impervious surface will not be used in the formal mediation case. The cooling is weaker in grids with high 2020 built-up fractions. In contrast, the lagged NDVI is not significant, and the estimates should be considered to be robust within-grid contemporaneous associations rather than causal effects. The results indicate that greening can be achieved from a spatial perspective and support the use of policy-based quasi-experimental designs and better time-varying measures of the green space structure.