Heat stress in social housing districts: tree cover–built form interaction

2.1 Selection of the case study

This study focuses on two case studies located in District VI of L’Hospitalet de Llobregat, the second-largest city in the Barcelona metropolitan area. L’Hospitalet is characterised by a Mediterranean climate (Csa: Köppen classification), with an annual mean temperature of 16.1°C and a mean relative humidity of 69%. The city experiences mild winters and warm to hot summers, with monthly mean temperatures ranging from 9.2 to 24.4°C. However, during summer, maximum daily temperatures frequently exceed 28.5°C (Table 1), with an increasing recurrence of heat episodes surpassing 30°C. Furthermore, recent climatological studies have documented a significant increase in the frequency, severity and duration of heatwaves in the region since 1970 (Serra et al. 2024), a trend that is projected to persist and further exacerbate outdoor thermal discomfort in this urban area.

The two selected case studies, Bellvitge (constructed in 1964) and Gornal (constructed in 1972), are contiguous social housing developments separated by a railway infrastructure (Figure 1). These neighbourhoods were originally designed to accommodate the influx of migrants in the 1960s and to relocate residents from informal settlements. Many of the original inhabitants remain in these areas, contributing to a high ageing index, with 26.1% of the district’s population over 64 years old. This demographic characteristic exacerbates climate vulnerability, as older adults are particularly susceptible to heat stress due to pre-existing health conditions and prolonged heat exposure resulting from reduced mobility. Both neighbourhoods have been officially classified as socio-economically vulnerable and included in the Spanish Catalogue of Vulnerable Neighbourhoods (Ministerio de Transportes Movilidad y Agenda Urbana 2023). Additionally, they exhibit high indice de vulnerabilidad al cambio climatico (IVAC) values, a local measure of climate change vulnerability (Area Metropolitana de Barcelona (AMB) 2022).

Figure 1

The focus on districts such as Bellvitge and Gornal is essential due to their significant influence on urban vulnerability and citizen’s wellbeing. These neighbourhoods not only exemplify challenges common to ageing social housing estates throughout Europe but also constitute a significant proportion of the continent’s existing housing stock, thereby presenting major regeneration challenges for architecture and urban design professionals. Bellvitge and Gornal exemplify the core characteristics of the Spain social housing scheme ‘Polígonos de Viviendas’, developed to support low-income segments of the population through welfare-driven policies and mitigate the negative effects of private suburban expansions. Similar housing models emerged concurrently across Europe, including France’s Grands Ensembles, the UK’s housing estates, Italy’s Piani di Edilizia Economica e Popolare, etc. These schemes, which were conceived as modern-city experimental master plans across Europe, are deeply rooted in the architectural culture of Western Europe. However, despite their well-intended origins, many of these places have proved vulnerable when faced with practical realities.

The experimental design approach of these districts manifests in their distinct spatial and physical characteristics, as defined by specific ranges of urban form indicators (Figure 2 and Table 2). Similar to many other cases, Bellvitge and Gornal follow an urban form model based on zoning principles. This model is characterised by a precise pattern of high-rise structures (↑L, number of floors), low building intensity (↓FSI; floor space index), reduced compactness (↓GSI, ground space index) and elevated open space ratios (↑OSR). Another prominent feature of these social housing districts is the oversized and often undefined nature of public spaces in terms of land use and ownership, and substantial green areas with relatively uniform tree density.

Figure 2

While both case studies demonstrate comparable urban metrics (Table 2), they differ significantly in their spatial configuration, building typologies and open space organisation. Bellvitge primarily consists of linear blocks forming elongated street-like outdoor spaces, resulting in clearly defined corridors and structured pedestrian environments. In contrast, Gornal is characterised by a higher concentration of residential towers, which generate large but relatively undefined open spaces.

These morphological distinctions hold particular significance for outdoor thermal comfort assessment, as they influence exposure to direct solar radiation and the availability of shaded areas, providing an interesting starting point to assess the effects of vegetation.

In the cases of Bellvitge and Gornal, both have a high open space ratio (↑OSR), with a value close to 2, meaning that for every 1 m² of built floor area on a site, there are 2 m² of open, unbuilt space. Such large open spaces can reduce heat if properly planned and maintained (e.g. shading devices, trees, grass), but increase radiant heat otherwise (e.g. paved surfaces), underscoring the need for optimised urban design from a climate-centric perspective.

Bellvitge and Gornal exhibit a relatively similar presence of green areas (green space index = 0.1–0.08). The tree density in Gornal (83 trees/ha) is 24% higher than in Bellvitge (67 trees/ha), but Bellvitge has a higher pervious-to-impervious surface ratio, which may influence its microclimatic behaviour. The study area (delineated by the white square in Figure 1) contains 2261 trees across public open spaces—according to the latest municipal inventory—representing 91 distinct species. Table 3 summarises the predominant ones and their key characteristics.

The selection of Bellvitge and Gornal as case studies is particularly relevant due to their contrasting urban morphologies and shared vulnerability to heat stress. Their characteristics allow for an in-depth exploration of how urban form and tree cover influence outdoor thermal conditions in social housing districts across the Mediterranean region.

2.2 Model development

UMEP uses its own meteorological data file format. In this case, the Barcelona typical meteorological year (TMY) EnergyPlus Weather (EPW) file was used to generate the appropriate format. To conduct the simulations, several types of raster files are necessary as input to calculate MRT:

• Digital elevation model (DEM): this raster file contains terrain elevation. The DEM for each neighbourhood was downloaded from the Barcelona Metropolitan Area (AMB) database (AMB 2024).

• Digital surface model (DSM): this raster file includes terrain elevation, as well as the footprint and height of buildings. It was generated using the UMEP DSM Creator tool, based on a polygonal vector layer with building information from the AMB database and the DEM raster file.

• Canopy digital surface model (CDSM) and trunk digital surface model (TDSM): these raster files provide vegetation-related information. The CDSM and TDSM were created using the UMEP Tree Generator tool, which utilised a point vector layer detailing the location and dimensions of trees (tree type, total height, trunk height and diameter) developed by the authors based on the data from the L’Hospitalet de Llobregat Municipality (Ajuntament de l’Hospitalet de Llobegrat 2024) and Google Earth.

• Land cover types: this raster file categorises land cover types such as grass, water, buildings, paved surfaces, evergreen trees, deciduous trees and bare soil. It was produced using the rasterise with attributes tool in QGIS, utilising a polygon vector layer with the land cover categories and the DEM raster file.

The remaining raster files required for the simulations were generated using UMEP tools. These included the aspect (orientation) and height of the walls and the Sky View Factor raster file, for which two versions were created for each neighbourhood, one including trees and one without.

2.3 Mrt simulation and analysis

Urban heat stress in both neighbourhoods is analysed through simulations of MRT using UMEP version 3.20.9, an open-source QGIS plugin. MRT calculations in UMEP are based on the SOLWEIG model v2022a, which can compute spatial variations of solar (K) and longwave (L) radiation fluxes from six directions in complex 3D environments. SOLWEIG considers the sky, buildings, ground and vegetation components separately (Figure 3).

Figure 3

High-resolution (1 px = 1 m²) MRT maps were generated through pixel-by-pixel simulations of the open spaces under study in two scenarios: with and without trees. The emissivity and albedo values used in the model were as follows. For horizontal surfaces: asphalt = 0.95 and 0.18; grass = 0.94 and 0.16; and bare soil = 0.94 and 0.25. For vertical surfaces (walls), values of 0.90 and 0.50 were applied, respectively, along with a vegetation light transmissivity of 3%. A single emissivity and albedo value was applied uniformly to all vertical surfaces in the model. While SOLWEIG allows these properties to be specified for vertical surfaces, it only permits one global value for each across the entire model. Given this limitation (and considering the high degree of homogeneity in building materials, facade colours and glazing ratios in the study area), the use of averages was considered an appropriate and representative simplification.

The simulations focused on a 500 × 500 m (250,000 m²) study area characterised by a morphologically uniform urban fabric. The simulation period selected for this study was the warmest summer week from the TMY EPW file used. Figure 4 provides a summary of the weather conditions during this week (3–9 August). The results analysis focused on the central hours of the day, from 10:00 to 16:00 local time (coordinated universal time (UTC)+2). Within the study area, a specific 4-ha region of interest (black square in Figure 5) was selected for a more in-depth quantitative analysis of MRT data, utilising zonal statistics, raster layer histograms and the GRASS r.quantile tools of QGIS. Finally, the heat stress category was calculated using the UTCI based on the simulation results.

Figure 4

Figure 5

4.1 Temporal variation of mrt values

The analyses in this paper are based on MRTs during the central hours of the day (10:00–16:00 local time; 08:00–14:00 UTC). While this approach provides a general understanding of the thermal impacts of urban morphology and vegetation, it does not capture their dynamic variations throughout the diurnal cycle. To discuss this aspect, Figure 7 presents hourly MRT distributions (including maximum, minimum and quartile values), which reveal significant temporal variations between both neighbourhoods.

Figure 7

In the treeless scenario, median MRTs remain consistently high throughout the day, peaking around midday (11:00 UTC in Gornal and 12:00 UTC in Bellvitge) in synchronisation with maximum solar radiation. In contrast, the vegetated scenario exhibits consistently lower median MRTs that increase more gradually, with peak temperatures delayed by approximately 2 h. To illustrate the mechanisms underlying these contrasting thermal dynamics, Figure 8 presents the hourly evolution of MRT distribution for the Gornal case.

Figure 8

As shown, the presence of trees introduces a pronounced diurnal variation in MRT due to their dynamic shading effects throughout the day. In the morning, when the sun is low, trees cast long shadows across the urban spaces. These shadows significantly reduce exposure to direct solar radiation, and by extension, prevent surfaces from early overheating, leading to much lower average MRTs compared with the no-tree scenario. As the day progresses and the solar angle increases, the shadows cast by the trees become shorter and shorter until midday, and hence, less effective at covering large surface areas. Consequently, average MRTs rise throughout the day, producing a typical daily temperature curve: low in the morning, peaking in the early afternoon, 1 h after the maximum horizontal irradiance and matching the peak of surface temperatures.

In contrast, the no-tree scenario displays relatively uniform average MRTs throughout the day. Although there is some shading variability due to the hourly movement of shadows cast by tower buildings (08:00–14:00 UTC), surfaces remain exposed to direct sunlight for extended periods. As a result, even shaded areas retain accumulated heat, and MRT remains high. The absence of trees, combined with thermal inertia and limited shading heterogeneity, leads to consistently elevated and homogeneous thermal conditions.

Overall, trees not only reduce peak thermal stress but also modulate the timing of urban heat. During the study period, trees lowered median MRT by 10.7°C (Bellvitge) and 11.5°C (Gornal), with cooling effectiveness varying diurnally. The most substantial changes occur in the early morning (08:00 UTC), where trees reduce median MRT by 19.4°C in Bellvitge and 22.8°C in Gornal. These impacts diminish by early afternoon to 7.9 and 9.2°C, respectively. These results demonstrate a clear diurnal pattern in vegetation benefits, with morning reductions being particularly crucial as they coincide with intense pedestrian activity periods (e.g. commercial establishments typically open at 10:00 local time/08:00 UTC).

The observed 2-h delay in peak MRT under vegetated conditions carries important implications for urban thermal management. By shifting heat accumulation to less sensitive hours, trees not only reduce extreme temperatures but also temporally redistribute thermal stress, offering enhanced comfort during high-usage periods. While averaged values provide useful insights, these findings underscore that temporal resolution is critical for fully understanding mitigation strategies. The identification of morning as the period of maximum cooling efficacy, coupled with the observed phase shift in peak temperatures, suggests that urban greenery’s benefits extend beyond simple temperature reduction to include temporal modulation of thermal extremes. Future work could integrate additional metrics (e.g. hourly extremes or rate of change) to further disentangle these dynamics.

4.2 Critical mrt threshold

Since thermal perception depends on multiple climatic variables beyond MRT, this section examines the relationship between MRT and heat stress levels using the UTCI as an outdoor thermal comfort metric. The analysis employed average simulation conditions (T_air = 28.7°C, relative humidity = 65% and V_air = 3 m/s). Results shown in Figure 9 demonstrate that the observed MRT ranges in both study areas consistently correspond to UTCI values exceeding comfort thresholds, ranging between moderate to very strong heat stress conditions during the study period.

Figure 9

Based on the MRT–UTCI correlation, two critical MRT thresholds were identified at 45 and 70°C. These represent transition points from moderate to strong and very strong heat stress perception (indicated by the dashed lines in Figure 6). Simulations revealed that only shaded areas—whether by trees or buildings—consistently maintained MRTs below the 45°C threshold. In contrast, zones > 70°C were restricted to unshaded areas with high reflected solar radiation gains (from building surfaces). These findings underscore solar radiation as the predominant driver of extreme heat stress conditions and highlight the crucial role of shading in mitigating radiative heat, a key consideration for urban design.

The established MRT thresholds were applied to quantify heat stress levels (moderate, strong and very strong) across open spaces in both case studies (Figure 10). Analysis revealed that in the absence of trees, most of the open space in both neighbourhoods falls within the category of strong heat stress (85.7% for Bellvitge versus 99.9% in Gornal). In contrast, when trees are considered, shaded areas substantially increase in both neighbourhoods, leading to significantly MRT reductions and consequent heat stress mitigation. Thanks to trees, both cases achieve virtually the same percentage of open spaces classified as having moderate heat stress (≈27%), underscoring the compensatory role of vegetation in mitigating heat stress.

Figure 10

The analysis reveals significantly greater thermal comfort improvements from tree coverage in Gornal compared with Bellvitge when using the no-tree scenario as a reference. Although Gornal exhibits higher tree density (83 versus 67 trees/ha), this quantitative difference alone cannot account for the substantial variation in observed benefits: vegetation enhances comfort conditions by 27.3% in Gornal compared with just 12.7% in Bellvitge. Particularly notable is the emergence of ‘cool spots’ in Gornal—areas with MRTs of 35–40°C covering approximately 8% of the total surface area (Table 5)—which are absent in Bellvitge. These divergences can be attributed to several factors: the generally smaller tree size in Bellvitge, and the suboptimal placement of some trees in already shaded building areas.

These findings underscore the critical importance of shadow-related design considerations for maximising vegetation’s cooling potential. While our simulations incorporated individual tree parameters (height, canopy dimensions) for all specimens, software limitations necessitated assuming constant light transmissivity across species. Future studies should assess the sensitivity of results to this factor, accounting for parameter variations due to seasonal dynamics and environmental influences such as water scarcity.

The location-specific nature of the proposed MRT threshold—derived from UTCI heat stress levels and influenced by local meteorological variables (air temperature, humidity, wind speed)—limits its direct transferability to other climatic contexts. However, the methodological framework presented here, combining GIS-based workflows with MRT thresholds, offers transferable value for urban practitioners implementing green infrastructure solutions. As Figure 11 demonstrates, this approach enables the following:

Figure 11

• Hotspot identification through energy simulations of non-vegetated urban models.

• Strategic tree selection by overlaying hotspot maps with urban tree inventories to differentiate critical shade providers (targeting highly irradiated areas) from redundant ones (in building-shaded zones).

• Performance evaluation based on the analysis of vegetated urban models to quantify species-specific cooling effectiveness, enabling microclimatic service rankings.

This systematic process may extend the concept of ‘critical infrastructure’—traditionally applied to non-living structures such as electric or transport networks—to urban trees as essential components for thermal comfort during heat stress. Such classification could inform adaptive management strategies, including prioritised irrigation during water scarcity (increasingly prevalent under climate change), ensuring optimal resource allocation where ecosystem services are most impactful.

Though the interplay between vegetation, building energy consumption and urban microclimate falls beyond the scope of the present study, it is worth noting that energy simulations of existing buildings necessitate careful modelling of urban vegetation, akin to other urban obstructions, in order to yield reliable results. The presence of urban tree cover provides shading to building facades—particularly at lower storey levels—which modifies solar gains through the building envelope throughout the year (Calcerano & Martinelli 2016; Garcia-Nevado et al. 2021; Palme et al. 2021).

With specific reference to the warm season investigated herein, which in the Mediterranean context is projected to increase energy consumption for building cooling due to climate change, urban trees can counteract such increases. Nevertheless, the inverse effects during the cold season must also be considered, depending on the type of tree species present in the urban space (i.e. deciduous or evergreen).

These findings highlight the need for further research examining how tree-mediated heat-mitigation strategies—through both shading and air temperature regulation—influence indoor thermal comfort and building energy consumption patterns.