Overheating is becoming an increasingly significant issue within the UK housing stock, driven by rising external temperatures, heat-retentive building design, and the limited prevalence of active cooling systems. Research has historically focused on heating-season comfort, leaving summertime thermal performance comparatively underexplored despite increasing exposure to warmer conditions (Velashjerdi Farahani et al. 2021). Numerous monitoring campaigns and modelling studies have documented internal temperatures exceeding recommended thresholds during summer, even under present-day conditions (Pathan et al. 2017; Jones et al. 2016; Ji et al. 2014; Cherian et al. 2025). This is further supported by empirical analyses of UK dwellings, which demonstrate that overheating can occur frequently in naturally ventilated homes and is not always effectively mitigated by occupant behaviour alone (Rodrigues et al. 2015). In response, recent policy and guidance documents, including Building Regulation Part O (HMG 2021) and CIBSE Technical Memorandum 59 (CIBSE 2017) have been introduced to provide a consistent framework for the assessment and mitigation of overheating risk in residential buildings. Key contributors to overheating include high insulation levels (Rodrigues and Gillott 2013), airtightness, large glazing areas, internal heat gains and occupant behaviours.

As building standards have progressively improved, recent dwellings generally exhibit higher internal temperatures. Research indicates that average indoor temperatures can be approximately 1.5 °C warmer in post-1990 homes compared with those constructed before 1919 (Beizaee et al. 2013). Furthermore, reducing external wall U-values from 0.9 W to 0.3 W/m²K has been shown to raise indoor temperatures by 15-17% across multiple climate scenarios (Elsharkawy and Zahiri 2020). While building age provides a broad categorisation of the housing stock, it is complemented by variation in dwelling typology. Common types include terraced, semi-detached, detached and bungalow, along with converted flats and purpose-built low/ high-rise flats (Piddington et al. 2020). Terraced and semi-detached homes comprise the largest proportion of the housing stock at 27.4% and 25.0%, respectively, followed by flats (20.9%) and detached dwellings (17.9%), while bungalows account for the smallest share at 8.8% (Piddington et al. 2020). Building typology strongly influences heat retention, solar exposure, and ventilation potential (Beizaee et al. 2013; Mulville et al. 2025). Evidence indicates that flats tend to exhibit the highest summertime temperatures, largely due to the greater solar gains experienced by upper-storey units and in some cases, lower levels of thermal mass (Lomas 2021). Smaller dwellings, such as mid-terrace houses and bungalows, also appear to be at elevated risk (potentially because internal and occupancy gains are distributed over a smaller floor area) resulting in higher heat gains per unit area (kW/m²) (Bugenings and Kamari 2022). Literature further suggests that mid-terrace and semi-detached homes demonstrate broadly comparable thermal behaviour, with end-terrace properties tending to run slightly warmer and detached dwellings generally cooler (Beizaee et al. 2013). Overheating in UK dwellings has been strongly linked to adverse health outcomes, particularly for elderly and vulnerable populations. For example, a London-based study estimated approximately 274 heat-related deaths over a 55-day period during which outdoor temperatures exceeded 24.8 °C (Taylor et al. 2015).

The UK exhibits considerable climatic diversity, with southern regions generally warmer and subject to more intense heatwaves than northern areas. UKCP18 projections indicate significant warming across all regions, but the magnitude and temporal pattern of change vary geographically especially when considering summer maximums (Murphy et al. 2018). Several studies (Toledo et al. 2025; Lomas et al. 2021; Morgan et al. 2015) have shown that overheating risk varies geographically, with severity consistently higher in southern England (Porritt et al. 2012). As a result, many overheating assessments rely on London weather data as a default, a practice that may lead to systematic under- or overestimation of overheating risk in other regions.

Weather files are fundamental inputs to dynamic thermal simulations, and their construction introduces significant variability into overheating outcomes. Conventional weather datasets such as Typical Meteorological Years (TMY) and design-day files remain widely used for building energy modelling, including design, compliance, operation and retrofit assessment (Chan et al. 2006). In a changing climate, this approach is increasingly inadequate, as buildings are sensitive to both future warming trends and short-duration extreme events that drive peak thermal stress and cooling demand (Yan et al. 2025; Hosseini et al. 2021). Future weather assessment is grounded in climate scenario development. Early work relied on the Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios, which defined four scenario families and shaped impact research for over a decade. As global socioeconomic and technological pathways diversified, these scenarios became insufficient, prompting the development of Representative Concentration Pathways and their later integration with Shared Socioeconomic Pathways (Zeng et al. 2025). Despite this evolution, many building-performance studies continue to rely on RCP-based datasets due to their availability and embedded use in modelling tools.

Future weather files for buildings are generated using either dynamical or statistical downscaling. Dynamical downscaling applies regional climate models at finer spatial resolution, including topography and urban effects, but remains computationally intensive and limited in availability (Roberts et al. 2018; Feser et al. 2011). Statistical downscaling offers a more accessible alternative by linking large-scale climate outputs to local weather variables, though it assumes the stability of historical climate relationships under future conditions (Giorgi 2008; Ekström et al. 2015). Within this category, commonly used approaches include morphing techniques that modify baseline weather files using climate change signals (Leith and Chandler 2010, Belcher et al. 2005), regression-based methods and stochastic weather generators designed to explore variability and extremes (Wilks and Wilby 1999).

Tools such as CCWorldWeatherGen exemplify the morphing approach and are widely adopted in building simulation studies due to their simplicity and compatibility with standard modelling workflows (Shen et al. 2025). However, CCWorldWeatherGen relies largely on SRES-based projections and applies mean monthly adjustments to historical weather, which limits its ability to represent future extremes, heatwave persistence, and sub-daily variability (Duan et al. 2025). In contrast, PROMETHEUS is based on a stochastic weather generator framework that incorporates probabilistic representations of temperature, solar radiation and persistence of extreme events. This enables PROMETHEUS to produce future weather series that better reflect changes in heatwave frequency, duration and intensity (Eames et al. 2010, UoE). Peer-reviewed comparisons indicate that such methodological differences can lead to substantial divergence in predicted indoor temperatures and cooling demand, particularly for residential buildings under future climates (Liu et al. 2023; Nik et al. 2016).

The methodological distinction between morphing and stochastic generation is particularly important for overheating analysis. Overheating is commonly assessed using threshold-based criteria, in which compliance depends on the number or proportion of occupied hours above a defined temperature. Where indoor conditions remain close to that threshold, relatively small changes in external temperature, nocturnal cooling or heatwave duration can produce disproportionately large changes in exceedance hours. Weather-file methodology may therefore affect not only the magnitude of predicted overheating, but also whether a dwelling passes or fails the selected criterion. A morphing file that primarily represents changes in mean conditions may produce a different risk classification from a probabilistic high-percentile file that incorporates a longer or more intense heat event. This makes weather-file choice a source of methodological uncertainty rather than merely a difference in input format.

Within the European Union, the United Kingdom has the highest percent of buildings built pre 1945 with the fewest built post 2000s (Piddington et al. 2020). Analysis of the types of dwellings within the UK literature commonly points at 6 typically identifiable points Pre 1919s, Interwar (1919 to 1944), Post war (1945-1964), Industrial (1965-1980), modern (1980-1990) and post 2002 each with their own construction configuration used:

Despite substantial progress in understanding overheating risk in UK dwellings, several key gaps remain:

Together, these gaps create a fragmented evidence base in which overheating risk is difficult to compare across studies and highly sensitive to modelling choices.

CIBSE Technical Memorandum 59 (TM59) provides a standardised methodology for assessing overheating risk in residential buildings using building energy simulation tools (CIBSE 2017). In this study, TM59 profiles (summarised in Table 2) were implemented within a detailed simulation framework to capture the key drivers of overheating. The framework accounted for variations in occupancy patterns and behavioural use to enable realistic assessments of both daytime and nighttime thermal performance. Two occupants were assigned to each double bedroom, while three occupants were assumed for the combined living room and kitchen areas, in accordance with the occupancy schedules and internal gain profiles in Table 2. Appliance and equipment loads were defined using room-specific profiles aligned with TM59 activity assumptions. Window operation was represented using a deterministic binary control that was applied consistently to all simulations. Windows were enabled to open when indoor temperature exceeded 24 °C and were closed when outdoor temperature exceeded 26°C or during the defined winter period. This standardised control was selected to prevent differences in assumed occupant behaviour from confounding comparisons between regions, construction periods and exposure configurations. Space heating was provided by a gas boiler with a setpoint temperature of 18°C, a seasonal efficiency of 0.92 and a corresponding seasonal coefficient of performance of 0.882 kW/kW.

Overheating assessments are traditionally based on the number of hours during which indoor temperatures exceed defined comfort thresholds by a specified margin. In accordance with CIBSE Guide A, a space was regarded as ‘overheating’ when indoor temperatures exceeding 28 °C in living rooms and 26 °C in bedrooms during peak summer conditions (CIBSE 2017; Chartered Institution of Building Services Engineers 2006). Thermal comfort was evaluated using two compliance criteria. Criterion A applies to living rooms, kitchens and bedrooms, requiring that indoor temperatures exceed the relevant comfort threshold by no more than 1 °C for less than 3% of occupied hours between 1 May- 30 September. Criterion B applied to bedrooms only, limiting temperatures above 26 °C between 22:00 - 07:00 to no more than 1% of total annual hours (CIBSE 2017).

The analysis is based on a structured scenario matrix that combines regional climatic variation, dwelling form and age, weather-file methodology and future scenarios (as seen in Table 3). Three regions are considered within the UK: South, Midlands and North with Brighton (R1), Birmingham (R2) and Aberdeen (R3). For each region, three representative dwelling archetypes were modelled: a mid-terrace house (B1), a semi-detached house (B2) and end terraced house (B3). In addition, a further set of permutations was considered based on building age, comprising the pre-1919 (A1), 1945-1964 (A2), 1965-1980 (A3), and post-1980 construction periods (A4). Future climate projections were derived using two weather-file methodologies: CCWorldWeatherGen (W1) and PROMETHEUS (W2-W5), applied uniformly across all regions and dwelling typologies. The PROMETHEUS framework includes four projections: a medium emissions 50th percentile scenario (W2), a medium emissions 90th percentile scenario (W3), a high emissions 50th percentile scenario (W4), and a high emissions 90th percentile scenario (W5). Climate change is represented using three time periods: the baseline current-period (T1), and future time slices for the 2050s (T2) and 2080s (T3). Each scenario is uniquely defined by the combination of region (R), building type (B), building age (A), weather method (W) and time slice (T). For example, scenario R1-B2-A2-W1 corresponds to a semi-detached 1945-1964 dwelling in Brighton simulated with CCWorldWeatherGen weather files. In total, the full scenario matrix comprises 540 distinct simulations.

Figure 2 compares overheating in south and north-facing bedrooms across dwelling typologies (B1: mid-terraced, B2: semi-detached and B3: end-terraced). A consistent ordering was observed, with B3 generally recording the highest overheating, followed by B2 and B1. However, the magnitude of the difference varied considerably according to the weather scenario and future time slice. Under the medium scenario, the separation between dwelling configurations was relatively small and was frequently below one percentage point. However, this separation becomes more pronounced under elevated percentile scenarios and in the later time slices. By 2080, under the medium 90th percentile scenario, overheating increases progressively from mid-terraced to end-terraced dwellings, with B3 consistently recording the highest exceedance. This divergence widens further under the high 90th percentile scenario, where overheating approaches 30% in B3, exceeding corresponding B1 values by 3.2 percentage points. While the 2050 results exhibit the same ordering, absolute magnitudes are lower, indicating that both climatic progression and percentile severity amplify typological differences.

A similar hierarchy is evident in north-facing bedrooms, although the magnitude of differences is slightly reduced under the medium scenario. Under the medium 90th percentile projection in 2080, overheating increases substantially across all typologies, with end-terraced dwellings again exhibiting the highest values. This pattern intensifies under the high 90th percentile scenario, where overheating in B3 approaches 30%, clearly exceeding mid-terraced values. The persistence of this hierarchy in north-facing rooms, where direct solar gains are limited, indicates that typological differences are not solely attributable to orientation-driven solar exposure. Rather, the increased external envelope area associated with end-terraced dwellings likely contributes to greater sensitivity to elevated ambient temperatures and external heat transfer. Across both room orientations, a non-linear escalation in overheating is evident between scenarios.

Across all regions and rooms, a consistent age-related trend emerged: overheating increased progressively from A1 (pre-1919) to A4 (1980-1990). In both 2050 and 2080 projections, newer dwellings (A3 and A4) recorded higher overheating percentages than older stock (A1 and A2). This pattern was most pronounced in south-facing bedrooms. As seen in Figure 3, in Brighton (R1) south bedroom overheating in 2080 increased from 13.9% (A1) to 15.5% (A4). Birmingham (R2) showed a similar progression, from 8.1% (A1) to 10.5% (A4). Even in Aberdeen (R3), where absolute overheating values remained below 1%, the relative increase with construction age persisted (0.4% in A1 rising to 0.9% in A4), however this is within margin of error. North-facing bedrooms and living rooms followed the same trajectory, though with smaller absolute magnitudes. The consistency of this progression across regions indicates that construction age exerts a systematic influence on overheating risk. The only notable deviation from the general age-related trend was observed in the north-facing bedroom and in some instances, the living room, where A1 recorded slightly higher overheating than A2 and A3. This can be attributed to the differing wall constructions. A1 comprises 300 mm solid brick walls, whereas A2 and A3 utilise cavity constructions with lighter internal blocks and improved insulation levels. The solid masonry in A1 provides greater thermal mass, resulting in higher heat storage capacity and a slower diurnal thermal response. In north-facing rooms, where direct solar gains are limited, internal gains and heat transfer from adjacent spaces become the dominant contributors to overheating. Given that bedrooms are assessed against a 26 °C threshold and are particularly sensitive to elevated nocturnal temperatures, this delayed heat release can increase cumulative exceedance hours. By contrast, the lighter cavity constructions in A2 and A3 are likely to cool more rapidly once external temperatures decline, thereby reducing night-time overheating duration.

The increasing overheating from A1-A4 corresponds directly with improvements in thermal fabric performance. Wall U-values reduce from 1.80 W/m²K (A1) to 0.29 W/m²K (A4) and roof U-values from 2.45 W/m²K to 0.19 W/m²K over the same period. While these improvements reduce winter heat losses, they also increase heat retention during summer. Airtightness values vary only marginally between age bands (7.9-8.7 m³/h·m² @50 Pa), suggesting that the dominant driver of age-related variation is insulation level rather than infiltration. The results therefore reinforce the established trade-off between winter efficiency and summer resilience in progressively better-insulated dwellings. Although the age-related trend was consistent, its magnitude varied by region. The increase from A1-A4 was greatest in Brighton, moderate in Birmingham and smallest in Aberdeen. This indicates that warmer climates amplify the overheating penalty associated with improved envelope performance. In cooler climates, while the relative pattern remains, the absolute impact is significantly dampened due to lower external temperature stress.

Although the three regions are situated within the same national context, the UK’s geographical extent results in distinct climate profiles, with measurable variation in solar exposure, ambient temperature profiles, and diurnal ranges from north to south. Although the analysis of individual cities does not constitute a stock-level assessment of housing across the country, it provides valuable insight into how buildings perform under different regional climatic conditions and whether overheating represents a significant concern in each location. This study therefore evaluated multiple future weather files alongside variations in dwelling age bands and typologies to determine whether overheating constitutes a systemic regional issue on average, rather than being driven by a particular building type or construction period within the domestic stock. The objective was to isolate climatic influence while controlling for archetype-specific characteristics, thereby avoiding bias toward any single segment of the housing landscape. When comparing regions R1 and R2 against R3, a consistent upward trend in overheating frequency is observed across all weather files (as seen in Figure 4). The largest reductions are seen in R3, where reductions in overheating exceed 20% relative to R1. R2 overheating values remain broadly comparable to R1, with only minor deviations across the assessed building spaces. By 2080, the average reduction associated with R2 is approximately 4%.

Regional comparison across the three case study locations confirmed the expected climatic gradient, with overheating risk decreasing progressively towards the north. The southern location (Brighton) consistently exhibited the highest overheating frequencies, while Aberdeen recorded the lowest values across all dwelling typologies and age bands. More notably, under 50th percentile projections, including both medium and high emissions scenarios (W2 and W4) as well as the CCWorldWeatherGen datasets (W1), the northern location did not exceed CIBSE overheating criteria, even when projected into 2080, suggesting under median climate projections future overheating risk in northern regions remains limited. Overheating in the north emerged primarily under the 90th percentile weather files, which represent near-extreme or heatwave conditions. However, even under these more severe projections, exceedance above comfort thresholds was limited, with operative temperatures surpassing the comfort limit by approximately 0.6% under medium emissions and 4% under high emissions (as seen in Figure 5). This indicates that while extreme events introduce some risk, the magnitude of thermal discomfort in northern regions remains comparatively constrained relative to southern counterparts. The north-facing bedroom exhibited patterns consistent with south facing bedroom, though operating at lower overheating percentage points. Consequently, under median climate projections to 2080, north-facing bedrooms are unlikely to present significant overheating risk, with vulnerability emerging primarily under near-worst-case climate scenarios. As illustrated in Figure 5, overheating risk in the Brighton and Birmingham is substantial and warrants concern. Under the medium emissions 50th percentile projections, CIBSE overheating criteria are exceeded as early as 2050, with the magnitude of exceedance increasing progressively towards 2080. The 90th percentile weather files present a more severe outlook. Under the medium emissions pathway, overheating reaches up to 30%, while the high emissions scenario produces values approaching 45%.

Future weather files inherently carry substantial uncertainty due to their reliance on emissions scenarios, climate model projections and downscaling methodologies. The speculative nature of long-term climate forecasting introduces variability that must be acknowledged when interpreting simulation outputs. Consequently, comparative evaluation of different future weather generators is essential to assess robustness in overheating projections. The W5 high emissions scenario at the 90th percentile produced the most severe overheating outcomes across the dataset. By 2080, overheating approached 30% (as seen in Figure 6), representing a substantial exceedance of comfort thresholds. The medium emissions scenario at the 90th percentile also demonstrated markedly elevated risk, with overheating values reaching approximately 20%. Importantly, exceedance of recommended overheating limits was not confined to extreme percentile files. Across all weather datasets, including lower percentile projections, every dwelling typology and age band surpassed the acceptable thresholds defined under CIBSE guidance. Comparable patterns were also observed within the north-facing bedrooms and living rooms, the latter of which was assessed using the higher 28°C overheating criterion (as seen in Figure 7).

Across the full matrix of model permutations, including dwelling age bands, typologies and UK regions, the CCWorldWeatherGen datasets demonstrated strong alignment with the Prometheus medium (50th percentile) scenarios. This similarity was particularly evident in overheating metrics derived from dynamic thermal simulation. When aggregating results across all spaces, regions and configurations, the mean difference in overheating frequency between the two datasets ranged ±0.5 percentage point, indicating negligible divergence due to it being within margin of error. Regional analysis revealed that discrepancies were smallest in northern locations. However, this convergence should be interpreted in context, as absolute overheating frequencies in northern regions were substantially lower than those observed in southern climates. The high-emissions (50th percentile) scenario produces similar overall overheating trends but consistently yields higher overheating than CCWorldWeatherGen. On average, overheating is approximately 3.1% greater under the high-emissions dataset. This divergence is primarily driven by results in the Midlands, where the discrepancy becomes more pronounced. In south-facing rooms, overheating frequencies are nearly 7% higher compared to those generated using CCWorldWeatherGen. North-facing rooms also exhibit elevated overheating under the high-emissions scenario, with increases of approximately 6.5% in the Birmingham and 4.1% in Brighton. The 90th percentile weather files, which are interpreted as near-extreme or heatwave conditions within both medium and high emissions pathways, produced substantially elevated overheating across all building typologies and regions. These datasets amplify peak dry-bulb temperatures, reduce nocturnal cooling potential and increase the persistence of consecutive hot days, thereby compounding internal heat accumulation in highly insulated dwellings. Under the medium emissions pathway, overheating exceeded those derived from CCWorldWeatherGen by more than 17% by 2080. The divergence was even more pronounced under the high emissions scenario, where overheating increased by up to 31%. Although 90th percentile files are not typically used to represent “expected” annual performance and therefore may overstate overheating risk in probabilistic terms, they serve an important analytical function. Specifically, they provide insight into system resilience under extreme climatic stress and enable evaluation of passive survivability limits. From a risk-based design perspective, these scenarios are valuable for stress-testing retrofit strategies and identifying thresholds beyond which mechanical cooling becomes unavoidable.