Urban Park Vegetation Indices and Local Cooling in Krakow: A Linear Mixed Effects Model Analysis of Landsat Land Surface Temperature

Park cooling in Krakow was quantified with archival Landsat imagery from 1990, 2000, 2013, and 2018. A linear mixed effects model with random intercepts for Area and PointID was applied to 33,949 locations and about 136,000 samples. Land surface temperature (LST) was retrieved with Landsat Collection 2 Level‑2 Surface Temperature product. Cooling was defined as the within scene difference in LST between park cores at 0 m and fixed buffers at 150 and 300 m. NDVI and NDMI were included as standardized predictors. Between 1990 and 2018, mean LST rose by about 4°C. Park cores were cooler than the 150 m buffer by about ∼2.1–2.6°C, and cooler than the 300 m buffer by about ∼2.6–3.3°C. The distance decay changed over time. The core to edge contrast peaked around 2000. Relative to 1990, the contrast in 2000 was larger by 0.50°C at 150 m and by 0.67°C at 300 m. It weakened by 2013 and partly recovered in 2018. NDMI showed a stronger association with LST than NDVI (−1.25°C per one standard deviation; p < 0.001 versus NDVI non‑significant (p = 0.846)). Fixed effects explained about 37.5% of variance and the full model about 95.6%. The results show persistent mitigation of surface heat by urban forests and point to the importance of canopy moisture and the protection of near park buffers. The study provides reference values that can be compared across cities and years to guide evaluation of urban heat mitigation.