The role of urban trees in reducing land surface temperatures in European cities

The study addresses how effectively urban trees reduce land surface temperatures (LST) across diverse European climates and how this cooling compares with treeless urban green spaces. Prior analyses of surface urban heat islands (SUHI) often compare cities with their rural surroundings and do not disentangle vegetation types or focus on limited regions, leaving a gap in understanding the climate-dependent performance of urban trees versus grass-dominated green spaces. The purpose is to quantify and compare LST differences among urban trees, treeless urban green spaces, and urban fabric across European cities and seasons, including during hot extremes, to inform urban heat mitigation strategies.

Previous work has shown that urban vegetation can reduce urban heat, but the magnitude varies by climate and city. SUHI-based approaches, which contrast urban and rural LST, make it difficult to isolate effects of specific vegetation types, an important distinction evidenced in studies of land-use/land-cover (LULC) change. Many urban-focused studies either do not distinguish tree canopies from treeless green spaces or are limited to specific regions, missing continental-scale climatic gradients. Mechanistically, trees influence urban climate through shading, transpiration (evapotranspiration, ET), and albedo. Shading can strongly reduce daytime LST and near-surface air temperature, with larger effects over asphalt than grass and influenced by canopy morphology. However, how these processes translate into regional patterns in cooling across Europe, and how trees compare to treeless green spaces under average and extreme conditions, has remained underexplored.

Study domain and data: 293 European cities (subsampled from Copernicus Urban Atlas coverage across the EU and selected additional countries including Turkey) were analyzed. Land-cover and urban tree data were derived from the Copernicus Urban Atlas and an associated urban tree layer capturing contiguous forest or urban tree patches ≥500 m² and ≥10 m width. Elevation and aspect came from EU-DEM v1.0. LST observations were obtained primarily from Landsat (Landsat 5/7/8; 2006–2018) with 90 m resampled resolution and ~16-day revisit around 10:15 a.m. local solar time, complemented by ASTER (coarser temporal coverage, used for comparison/validation). In total, a large high-resolution LST dataset (over 120,000 Landsat scenes) under cloud-free conditions was compiled. For broader biophysical context, MODIS products were used to estimate monthly albedo (MCD43A3) and ET proxies (MYD13A2-derived; aggregated 2006–2018), at 500 m–1 km resolution, with multiple linear regression used to attribute city-level albedo and rural ET to LULC fractions (ET estimated for rural forests and pastures; ET within urban classes often unavailable). Modeling approach: For each city and each available LST observation, a Generalized Additive Model (GAM) was calibrated with LST as the response and predictors including fractions of LULC classes (e.g., continuous urban fabric, urban trees, treeless urban green spaces), topography (elevation, aspect), spatial coordinates (as tensor-product smooths), and other covariates. From each fitted GAM, counterfactual predictions were generated for hypothetical 100% coverage by specific LULC types to estimate LST differences (ΔT) between categories, notably urban trees vs continuous urban fabric (UT−UF), treeless urban green spaces vs urban fabric (UGS−UF), and urban trees vs treeless green spaces (UT−UGS). Analogous contrasts were computed for rural forests and pastures versus urban fabric to contextualize urban-rural differences. Background air temperature conditions were characterized using E-OBS gridded air temperature data to identify summertime medians (JJA) and the hottest (extreme) observations per city. Seasonal analyses were performed for spring (MAM), summer (JJA), and autumn (SON). Regional aggregations (e.g., Scandinavia, British Isles, Alps/Mid-Europe, Eastern Europe, Mediterranean, Iberian Peninsula, Turkey) were used to summarize spatial patterns. Inter-city relationships between ΔT and biophysical drivers were analyzed by correlating ΔT with rural ET; the role of albedo was also examined. Model diagnostics indicated reasonable explanatory power (average R² ~0.70 at tested sites).

- Urban trees are substantially cooler than continuous urban fabric in summer across most European cities. Average LST differences (UT−UF) are about 8–12 K in Central Europe and 0–4 K in Southern Europe (Mediterranean, Iberian Peninsula, Turkey). - During hot extremes, cooling by urban trees often decreases in Southern and Southeastern Europe (Mediterranean, Iberian Peninsula, Turkey), while in Scandinavia, the British Isles, and parts of Mid-Europe/Alps, cooling is similar to or greater than median summertime values, with the strongest cooling shifting northward during extremes. - Seasonal patterns: In Southern Europe and Turkey, spring cooling (MAM) is comparable to or exceeds summer cooling (JJA), whereas elsewhere in Europe, summer cooling is highest. Autumn (SON) shows the least cooling everywhere. - Treeless urban green spaces cool less than urban trees. Their LST reductions are typically about 2–4 times smaller than those provided by trees. In parts of Southern Europe and Turkey, treeless green spaces and rural pastures can be as warm as or warmer than urban fabric, indicating minimal or negative cooling benefits. - Rural forest vs urban comparisons mirror urban tree patterns but with nuances: in Central Europe, urban tree LSTs are warmer than rural forests; in Turkey, rural forests cool more than urban trees. An east–west gradient is observed in rural forest vs urban fabric differences, with smaller absolute differences in Eastern Europe. - Biophysical drivers: Inter-city variation in ΔT for trees vs urban fabric correlates strongly with rural forest ET, and ΔT for green spaces vs urban fabric correlates with rural pasture ET. Albedo differences explain little of the inter-city variance (R² < 0.1).

Findings indicate that the cooling benefits of urban trees are climate-dependent, with ET playing a dominant role in spatial patterns of LST reduction. In drier Southern regions and during hot extremes, elevated vapor pressure deficit coupled with limited soil moisture constrains stomatal conductance and reduces transpiration, diminishing tree-driven cooling. In more mesic Northern and Central regions, sufficient soil moisture permits higher ET during heat, maintaining or enhancing cooling. While albedo contributes to local temperature differences, its role in explaining inter-city variance in ΔT is minor relative to ET. Vegetation type matters: trees, with higher leaf area, deeper roots, and greater roughness, generally provide greater cooling than grass-dominated green spaces. However, LST-based metrics capture surface heating and ET effects but not the direct shading benefits of trees, which can be especially important for human thermal comfort in sunny, dry climates. The results underscore that heat mitigation strategies must be tailored to climate context: in drying regions, reliance on vegetation cooling alone may be insufficient during extremes; complementary measures (e.g., higher-albedo materials) and potentially irrigation (subject to water constraints) may be needed. The clear divergence between average summertime cooling and cooling during extremes cautions against simple temporal averaging when evaluating adaptation benefits.

This study provides a continental-scale, observation-based assessment of how urban trees reduce LST relative to urban fabric and treeless green spaces across 293 European cities. Urban trees deliver substantial summertime cooling, especially in Central and Northern Europe, and generally 2–4 times greater LST reduction than treeless urban green spaces. Cooling effectiveness varies by climate and season and can diminish during hot extremes in Southern Europe, while remaining robust or increasing in Northern regions. Results highlight the importance of vegetation type, background climate, and water availability for urban heat mitigation. Future research should: (1) integrate shading and wind effects with ET to assess overall thermal comfort and air temperature impacts; (2) improve temporal resolution and include extreme events explicitly; (3) incorporate detailed urban morphology and canopy structure; (4) refine urban tree datasets to capture street trees and small patches; and (5) link LST differences to near-surface air temperature and human-relevant metrics to guide policy and design.

- Reliance on satellite-derived LST with limited temporal resolution and morning overpass (~10:15 a.m.) introduces uncertainty, especially for hot extremes; only cloud-free scenes are used. - LST is a proxy and does not directly represent near-surface air temperature or human thermal comfort; shading benefits are not captured by LST contrasts. - Spatial configuration and urban morphology (e.g., street canyons, building heights) may confound temperature differences; effects only partially controlled via model predictors. - Urban tree dataset excludes isolated/street trees and narrow strips, potentially underrepresenting tree cooling in some settings. - MODIS albedo and ET products are coarse (500 m–1 km), limiting precision for small urban patches; ET not available for urban classes; some negative ET predictions occurred and were discarded. - E-OBS air temperature data are unavailable for some Turkish cities; background temperature characterization may thus vary. - GAMs fitted per observation cannot fully eliminate spatial autocorrelation or biases; model structure assumes average effects rather than within-city interactions. - Regional aggregation masks city-specific heterogeneity; results for individual cities and absolute ΔT should be interpreted cautiously.