Winter Thermal Resilience of Lightweight and Ground-Coupled Mediumweight Buildings: An Experimental Study During Heating Outages

Abstract

Thermal resilience is critical for building safety in cold climates during heating outages. This study presents full-scale experimental data from two residential buildings in Poland, tested during the winter of 2024–2025 under both typical and extreme outdoor conditions. The buildings—a lightweight timber-frame structure and a mediumweight masonry structure with ground coupling—were exposed to multi-day heating blackouts, and their thermal responses were monitored at a high temporal resolution. Several resilience indicators were used, including the resistance time (RT), degree of disruption (DoD), and hours of safety threshold (HST). Additionally, two time-based metrics—the time to threshold (T_x) and temperature at X-hours (T(t_x))—were introduced to improve classification in long-duration scenarios. The weighted unmet thermal performance (WUMTP) index was also implemented and validated using experimental data. The results show that thermal mass and ground coupling significantly improved passive resilience, enabling the mediumweight building to maintain temperatures above 15 °C for over 60 h without heating. This study provides new empirical evidence of passive survivability in blackout conditions and supports the development of time-sensitive assessment tools for cold climates. The findings may inform future updates to building codes and retrofit guidelines.

1. Introduction

The built environment is becoming increasingly vulnerable to climate-induced disturbances, including prolonged cold spells, polar vortex events, and energy supply disruptions. According to the Intergovernmental Panel on Climate Change [1], such events are expected to intensify throughout the 21st century, posing new challenges to building safety and habitability, especially in heating-dominated regions. While global energy policy has long emphasized decarbonization and energy efficiency [2], recent studies highlight that efficiency alone does not ensure thermal resilience under disruptive conditions [3,4].

Buildings consume a significant share of energy worldwide and are expected to undergo performance shifts due to increasingly variable outdoor conditions. As Hosseini et al. [5] demonstrated through high-resolution simulations, localized microclimatic factors—such as the urban density, vegetation, and thermal mass—can significantly influence indoor temperature decay during winter energy outages. Empirical studies have further confirmed this vulnerability. For instance, Baglivo et al. [6] and D’Agostino et al. [7] reported that buildings compliant with updated national energy regulations may still fail to maintain habitable conditions during extended heating interruptions.

In response, the concept of resilience has emerged as a complementary performance dimension, addressing not only how buildings operate under normal conditions but also how they perform under stress. Originally developed in ecology and materials science [8], the notion of resilience has since been adapted across diverse fields, including engineering, sociology, and climate adaptation. In the context of the built environment, it broadly refers to the ability of a system to absorb, withstand, and recover from a disturbance while maintaining essential functions [9,10].

Two main theoretical frameworks currently shape the resilience discourse: engineering resilience, which emphasizes the speed of recovery and system stability [11], and ecological resilience, which stresses the capacity to adapt, absorb shocks, and reorganize under persistent stress without losing function [10]. Translated into building science, engineering resilience relates to robustness and the recovery of mechanical systems, whereas ecological resilience involves passive survivability, envelope behavior, and occupant adaptability [12,13].

Despite increasing interest, there is no consensus on how to define or operationalize thermal resilience in buildings. Peri et al. [12] found that, while many studies invoke resilience in a general sense, few provide metrics that directly capture dynamic thermal performance during disruptions. Some authors frame resilience in terms of the envelope response and others in terms of temperature restoration or user comfort, while some include human behavior or urban infrastructure as integral components [14,15]. This conceptual fragmentation presents a barrier to standardization and limits comparability across studies.

Against this backdrop, thermal resilience has emerged as a more focused and measurable dimension: the ability of a building to maintain safe and habitable indoor temperatures during heating system failures or extreme weather events [16]. It represents not only a physical response but also a public health concern, as thermal loss can lead to hypothermia, cardiovascular stress, and mortality—especially among vulnerable populations [17,18].

However, most current evaluations remain biased toward overheating scenarios and summer peaks. Winter resilience has received significantly less attention in both research and policy, despite potentially greater risks. Extreme cold conditions introduce unique technical challenges: intensified conductive heat loss, rapid temperature decay in low-mass buildings, and a lack of adaptive occupant responses during night-time hours [3].

The widespread use of typical meteorological year (TMY) datasets in simulation studies exacerbates the problem by smoothing out rare but impactful weather events [19]. Even when extreme cold years (ECY) are modeled, they often omit critical factors such as ground coupling, envelope mass, and thermal preconditioning, all of which are essential in understanding real-world resilience dynamics [5].

Still, there is no consensus on which metrics best reflect operational resilience. For example, DoD accounts for the severity and duration of deviations from comfort but is sensitive to the observation window length and may understate performance in high-mass systems tested over long periods. Only one study [3] has introduced and applied the WUMTP index, albeit exclusively using simulation data. To date, no research has validated WUMTP using measured indoor temperatures or adapted the metric to overheating scenarios [3].

Additionally, the hours of safety threshold (HST), proposed by Wijesuriya et al. [20], quantifies the duration for which a building remains within acceptable temperature bounds. While originally developed for hot climates, it has since been adapted for the assessment of winter scenarios as well.

However, most of these indicators are still primarily applied in simulation environments. As Siu et al. [16] note, empirical validation remains rare, and most existing tools rely on standard meteorological datasets that do not reflect actual winter extremes. Moreover, no international standard currently requires performance assessment under failure conditions, creating a policy blind spot [7,12].

In the European context, while regulations such as the Energy Performance of Buildings [21] promote sustainability and energy savings, they do not explicitly mandate resilience-based performance assessments. Some countries, including Italy and Finland, have introduced more stringent envelope standards [6], yet dynamic resilience modeling is still not required. Rehman et al. [4] show that even well-insulated houses in Nordic regions can reach unsafe indoor temperatures within hours of heating failure.

The World Health Organization [18] recommends maintaining indoor temperatures above 18 °C in winter to reduce the risk of illness and mortality. However, these thresholds are rarely embedded in building codes or certification systems.

Taken together, the literature reveals a clear and persistent gap between the theoretical development of resilience metrics and their integration into real-world design, policy, and evaluation frameworks. While simulation methods offer a valuable starting point, they often lack standardization and robustness when applied to real failure scenarios. At the same time, regulatory frameworks continue to prioritize annual energy performance and carbon reduction targets, frequently overlooking transient risks and critical climatic extremes.

Most notably, the thermal buffering effect of ground coupling, particularly in buildings without insulated floor slabs, remains underexplored. While ground heat exchange has been modeled in some simulations, its empirical impact on resilience under real winter conditions has not been verified. Similarly, metrics such as DoD and WUMTP, although helpful, were developed for short simulation windows and may lose sensitivity in extended blackout scenarios due to penalty saturation or normalization over limited time periods. These limitations call for full-scale experimental studies that assess long-term performance and enable the more robust classification of passive survivability across structural types.

1.1. Research Justification

Thermal resilience has become a critical concern in building science, particularly in the context of climate-induced risks such as prolonged winter cold spells, polar vortexes, and disruptions to energy infrastructure. While many buildings are optimized for steady-state energy efficiency, numerous studies have shown that this does not necessarily translate into adequate performance during extreme or disruptive conditions [3,7,12].

This gap is particularly evident in heating-dominated climates, where loss of thermal services may pose severe health and safety risks. Although several simulation-based methods exist to evaluate thermal resilience, e.g., RT, DoD, and HST, they are often unvalidated against real data and tend to neglect factors such as thermal inertia, thermal history, and substructure behavior.

In particular, the literature lacks experimental investigations into how ground coupling—especially in buildings lacking floor insulation—can affect passive thermal performance during outages. Initial results suggest that preheated soil may act as a thermal reservoir, slowing the cooling process and contributing significantly to resilience. However, its interaction with envelope mass and the building typology remains empirically unverified.

Additionally, existing thermal resilience metrics such as WUMTP and DoD are typically applied in short-duration simulation studies and may fail to capture performance under extended outages. Their sensitivity to the observation window length and the potential for penalty saturation limit their comparability across different building types and durations. These limitations emphasize the need for time-based, physically interpretable indicators that can be validated against empirical data.

This disconnect underscores the urgent need for standardized approaches to resilience assessment, supported by empirically validated indicators and climate-specific thermal safety thresholds, and the stronger incorporation of resilience metrics into energy policy and retrofit practices.

Bridging this gap is essential to ensure that buildings are not only optimized for efficiency under normal conditions but are also capable of providing shelter, continuity, and protection during energy outages and climate-induced stress events. Ultimately, buildings should not be evaluated solely by how they perform on average but by how they protect when it matters most.

1.2. Objectives and Scope of the Study

This study’s setup allowed for an isolated empirical evaluation of how thermal mass and ground contact influence indoor temperature stability during heating interruptions.

Both buildings were exposed to real weather conditions representing two seasonal scenarios: a typical winter period (TDY) and an extreme cold spell (ECY). Controlled heating outages were implemented, and indoor and outdoor temperatures were continuously monitored at a high temporal resolution. The collected data provided a basis for the evaluation of thermal decay, recovery profiles, and resilience under stress.

The analysis applies a combination of established and time-based performance metrics such as RT, DoD, and HST. Additionally, the authors propose a set of new physically grounded indicators, such as T(t_x) and T_x, designed to address the limitations of existing metrics in long-duration blackout conditions. The study also examines the applicability of WUMTP to experimental datasets and benchmarks its output against the measured results.

The focus is limited to the passive thermal response of the building envelope. As such, mechanical system performance, occupant behavior, energy supply reliability, and economic or structural aspects of resilience fall outside the scope of this work. The goal is to generate empirical evidence that can inform the development of standardized resilience indicators and support their future integration into building codes, performance assessments, and retrofit strategies.

1.3. Novelty of the Work

This study offers a novel contribution to the field of thermal resilience by providing full-scale, time-resolved experimental data on the indoor thermal behavior of residential buildings during multi-day heating outages in a winter context. While most prior studies rely on simulation-based assessments using standardized weather data, the present research captures real-time thermal responses under both typical and extreme winter conditions.

A key innovation lies in the controlled side-by-side testing of two buildings with contrasting construction systems: a lightweight timber-frame structure with an insulated foundation (B2) and a mediumweight masonry building without floor insulation (B1). This configuration enables, for the first time, the empirical isolation of ground coupling effects in cold-climate blackout scenarios. It also provides a direct comparison of passive thermal performance in buildings differing in mass and substructure behavior.

In addition to applying established indicators such as RT, DoD and HST, this study demonstrates the practical value of two new categories of resilience metrics: T_x and T(t_x). These indicators are designed to complement existing metrics and better capture performance in long-duration outages.

Furthermore, this study offers the first known empirical application and validation of the full WUMTP model proposed by Homaei and Hamdy [3]. Based on high-resolution indoor temperature data, this validation confirms the model’s overall applicability, while also revealing its limitations in long-duration scenarios due to penalty saturation. This analysis contributes to ongoing discussions about the comparability, scalability, and interpretability of penalty-based vs. time-based performance metrics.

In summary, this research introduces new methodological tools for the evaluation of thermal resilience, validates key simulation frameworks using measured data, and highlights the underexplored role of ground coupling in passive thermal survivability. These contributions offer valuable guidance for performance assessment, retrofit prioritization, and resilience-informed policy development in cold climates.

2. Methodology

2.1. Experimental Investigation

2.1.1. Characteristics of Experimental Buildings

This study was conducted in two full-scale detached residential buildings located at the Science and Technology Park of the University of Zielona Góra, Poland (Cfb climate zone according to the Köppen–Geiger classification) [22]. The buildings, shown in Figure 1, are geometrically identical in layout and orientation but differ in construction type: B1 (left) is a mediumweight masonry structure, while B2 (right) is a lightweight timber-frame structure.

The usable floor area of each building is approximately 123 m², with identical room configurations and main façade orientations. B1’s external walls consist of 24 cm cellular concrete blocks with 20 cm of mineral wool insulation. Its internal load-bearing and partition walls are composed of 24 cm and 8 cm silicate blocks, respectively. In contrast, B2’s walls are timber-framed and filled with 16 cm mineral wool, with an additional 18 cm insulation layer on the outside. Internal walls in B2 consist of wood framing filled with 16 cm and 5 cm mineral wool in structural and partition elements, respectively. A detailed description of the structural assemblies is available in [23].

In 2019, the floor slab in building B1 was modified: the original 30 cm of thermal insulation was removed and replaced with solid concrete, while the foundation wall was insulated with 20 cm of EPS to a depth of 80 cm below ground level. This modification substantially increased B1’s effective thermal mass.

The thermal mass parameter (TMP) was calculated according to ISO 13786:2017 [24], using the formula

$$TMP={\displaystyle \frac{\Sigma {\kappa }_{m }·A}{TFA}}$$

(1)

where κ_m is the areal thermal capacity of each partition, A is its area, and TFA is the total floor area. The resulting TMP was 192 kJ/m²K for B2 (classified as lightweight) and 467 kJ/m²K for B1 (classified as mediumweight).

2.1.2. Measurements and Instrumentation

Two experimental campaigns were conducted during the winter season: 28 December 2024–2 January 2025 and 14–22 February 2025. During these periods, the indoor air temperature and relative humidity were continuously recorded at five measuring points in each building. Outdoor parameters—the air temperature, relative humidity, and global horizontal solar radiation—were measured at a meteorological station installed on the roof of building B1. The experiments were conducted in unoccupied buildings to eliminate the influence of different user preferences and behaviors on the measurement results [25,26].

Indoor temperatures were monitored using P10 sensors placed in representative zones (see Figure 2). The weighted average temperature was calculated using the spatial distribution of room volumes, following the method described by Kuczyński and Staszczuk [23]. Sensors T1–T5 recorded grouped room data, e.g., T1 covered the vestibule, hall, and bathroom, while T4 represented the living/kitchen zone.

A freely programmable PLC controller in each building collected and processed real-time data. Data were stored and visualized using a SCADA system located in the university’s data center and remotely accessible via a secure interface. Measurements were logged at 5 min intervals.

Each building was equipped with four 1 kW electric heaters and one 1.5 kW heater. Mechanical ventilation operated continuously at 0.6 ACH, with heat recovery (efficiency: 75%) and balanced air distribution across zones.

During the December test (December and January), the buildings were not preheated prior to the 5-day stabilization period. Indoor temperatures were then raised and maintained at 21 °C. In February, this setpoint was maintained continuously for over one month prior to the experiment, resulting in warming the ground below the slab, particularly relevant for the B1 building, which lacked floor insulation.

2.1.3. Indicator Definitions and Formulas

To evaluate the thermal resilience of the tested buildings during heating outages, several established and time-resolved performance indicators were applied. This section provides definitions and calculation formulas for each of the key metrics used in the study.

Resistance Time (RT)

RT quantifies the duration during which indoor temperatures remain above a critical threshold (e.g., 15 °C). It is defined as

$$RT=t_{threshold}-t_0$$

(2)

where $t_{threshold}$ is the time when the indoor temperature drops below the selected threshold temperature, and ${ t}_0$ is the time when the heating is turned off.

Intensity of Failure (IoF)

IoF represents the total drop in indoor temperature during the blackout period:

$$IoF = T_{initial} - T_{min}$$

(3)

where $T_{initial}$ is the indoor temperature at the beginning of the outage, and $T_{min}$ is the minimum indoor temperature reached during the outage.

Cooling Speed (CS)

CS describes the average rate of temperature decline from the start of the outage until the lowest temperature is reached:

$$CS = {\displaystyle \frac{T_{initial }-{ T}_{min}}{t_{min} - { t}_0}}$$

(4)

where $T_{initial}$ is the indoor temperature at the beginning of the outage, ${ T}_{min}$ is the minimum indoor temperature during the outage, $t_{min}$ is the time when T_min occurs, and ${ t}_0$ is the time when the heating is turned off.

Recovery Speed (RS)

RS refers to the average rate of indoor temperature increase after heating is restored:

$$RS = {\displaystyle \frac{T_{final }-{ T}_{min}}{t_{final} - t_{min}}}$$

(5)

where $T_{final}$ is the indoor temperature at the end of the recovery period, $T_{min}$ is the minimum indoor temperature during the outage, $t_{min}$ is the time when $T_{min}$ occurs, and $t_{final}$ is the time corresponding to $T_{final }$.

Degree of Disruption (DoD)

DoD integrates the magnitude and duration of thermal discomfort below a defined threshold over the observation period:

$$DoD = {\displaystyle \frac{1}{t_{obs}}} · \Sigma [ max(0, T_{threshold} - T_i) · \mathsf{\Delta }t ]$$

(6)

where $T_i$ is the measured indoor temperature at time step i, $T_{threshold}$ is the selected comfort or habitability threshold (e.g., 18 °C or 15 °C), $\mathsf{\Delta }t$ is the measurement interval (5 min), and $t_{obs}$ is the total duration of the heating outage.

Weighted Unmet Thermal Performance (WUMTP)

WUMTP is a composite metric that penalizes deviations from thermal comfort according to the phase of the event, the severity of the deviation, and the exposure time. It is defined as

$$WUMTP=\sum \left[W_P\left(i\right)·W_H\left(i\right)·W_E\left(i\right)·∆T\left(i\right)·∆t\left(i\right)\right]$$

(7)

where $W_P\left(i\right)$ is the phase weight (e.g., initial, prolonged, recovery), $W_H\left(i\right)$ is the hazard severity weight (e.g., moderate, severe, extreme), $W_E\left(i\right)$ is the exposure duration weight (e.g., short, long), $∆T\left(i\right)$ is the temperature shortfall at time step i (difference between the threshold and $T_i$, if $T_i$ < threshold), and $∆t\left(i\right)$ is the duration of time step i (5 min in this study).

All indicators were computed using high-resolution temperature measurements and normalized to the duration of the heating outage, unless otherwise specified.

3. Results and Analysis

3.1. Cooling Curves—Visual Interpretation

Figure 3 presents the indoor temperature profiles of buildings B1 (mediumweight, black line) and B2 (lightweight, red line) during the experimental heating shutdown conducted in December 2024 and January 2025. The dotted blue line represents the outdoor air temperature (To). The horizontal axis shows the local time (Helsinki) over a multi-day period, while the left and right vertical axes display the indoor and outdoor temperatures, respectively.

Both buildings begin at approximately 21 °C before the heating is turned off. B1 exhibits a slow and consistent decline in temperature, with a long plateau in the 15–16 °C range, while B2 cools rapidly and reaches significantly lower indoor temperatures. The influence of the outdoor conditions, especially daily temperature fluctuations, is clearly visible in the B2 profile, which shows irregular inflections and greater thermal reactivity. The final stage in both profiles shows the sharp recovery of the indoor temperature as heating is restored.

Figure 4 displays the same thermal behavior during the February 2025 experiments, under more extreme weather conditions (ECY scenario). The outdoor temperature fluctuates between −15 °C and 5 °C. Once again, B1 maintains higher internal temperatures with a smoother decay, while B2 exhibits faster and deeper cooling. Notably, the temperature in building B2 drops below 12 °C after approximately 36 h, highlighting a critical loss of habitability.

Thermal oscillations in the outdoor air are mirrored in B2’s internal profile, but they only marginally affect B1. These visual results underscore the importance of thermal inertia and mass in cold-climate thermal resilience.

3.2. Typical Scenario (TDY—December Conditions)

To interpret the results, a set of thermal resilience indicators proposed in [4] was adopted, including the resistance time (RT), intensity of failure (IoF), cooling speed (CS), recovery speed (RS), and degree of disruption (DoD). The reference points used were as follows: preferred setpoint temperature (PST) = 21 °C, resilience threshold (PRT) = 18 °C, and passive habitability threshold (PHT) = 15 °C.

For consistency, the experiments conducted on buildings B1 and B2 were aligned with the typical downscaled year (TDY) and extreme cold year (ECY) weather profiles for Helsinki, as defined by Rehman et al. [4]. Unlike their study, which was simulation-based, the current work is based on experimental measurements that captured the full cooling and recovery cycle. This enabled the direct assessment of real-world resilience indicators.

Table 1. Thermal resilience indicators—TDY (December conditions).

Building	RT [h]	IoF [°C]	CS [°C/h]	RS [°C/h]	DoD [-]	Tmin [°C]
B1	28.5	6.6	0.057	0.634	0.288	14.4
B2	10.3	10.5	0.104	0.427	0.402	10.5
Rehman et al. (old building) [4]	2.0	7.7	0.250	1.910	0.357	—
Rehman et al. (new building) [4]	10.0	5.58	0.188	0.529	0.271	—

summarizes the results obtained in our experiments conducted from 28 December 2024 to 2 January 2025, alongside the simulated data from Rehman et al. [4] for a conventional “old” building and a modern “new” building.

The mediumweight masonry building, B1 (no floor insulation), showed significantly longer temperature maintenance above the resilience threshold (RT = 28.5 h) and a much slower cooling rate (CS = 0.057 °C/h) than the values reported by Rehman et al. [4]. The minimum temperature (14.4 °C) remained above the habitability limit, confirming the strong thermal performance of this envelope. Despite a slower recovery rate (RS = 0.634 °C/h), the DoD (0.288) was lower than that of the “old” simulated building (0.357), indicating higher overall resilience.

The lightweight building, B2, achieved an RT similar to the simulated “new” building (10.3 h vs. 10.0 h), but its other indicators revealed an inferior thermal buffering capacity. The larger temperature drop (IoF = 10.5 °C; Tmin = 10.5 °C) and higher DoD (0.402) indicate a faster loss of comfort and greater vulnerability.

3.3. Extreme Scenario (ECY—February Conditions)

While the TDY scenario represents a typical winter profile, the second experimental set was conducted under more demanding thermal conditions. The February experiments corresponded to the ECY profile for Helsinki, with an average external temperature of –5.5 °C, as defined by Rehman et al. [4].

Table 2. Thermal resilience indicators—ECY (February conditions).

Building	RT [h]	IoF [°C]	CS [°C/h]	RS [°C/h]	DoD [-]	Tmin [°C]
B1	11.0	8.5	0.062	0.210	0.305	13.0
B2	8.0	15.0	0.109	0.335	0.526	6.5
Rehman et al. (old building) [4]	1.0	11.72	0.387	0.977	0.545	—
Rehman et al. (new building) [4]	3.0	10.66	0.334	1.770	0.496	—

shows the results collected between 14 and 22 February 2025, compared with the same reference simulation dataset.

In this scenario, B1 again outperformed the simulated old building, with a more than 10 times longer RT and a significantly slower cooling rate. Although its recovery speed remained modest, the minimum indoor temperature never dropped below 13 °C. The DoD (0.305) was substantially lower than the reference (0.545), confirming its physical resilience under stress.

Conversely, B2 showed typical lightweight sensitivity to deep cold. The RT dropped to 8.0 h, and T_min fell to 6.5 °C. Despite a moderate CS (0.109 °C/h), the limited RS (0.335 °C/h) and high DoD (0.526) indicate a fast loss of habitability and insufficient thermal inertia.

3.4. Validation with HST—Comparison with Wijesuriya et al. [20]

To contextualize our findings, we compared them with the simulation study by Wijesuriya et al. [20], which used only one metric, the hours of safety threshold (HST)—defined as the time until the indoor temperature falls below 15 °C. In their study, simulations were run for a winter storm in Houston with average temperatures of around −1.1 °C. In comparison, our December experiment averaged +0.7 °C, and February averaged −4.5 °C—indicating that our test conditions were at least as demanding, if not more so.

3.5. Summary of Findings

The experimental results clearly demonstrate the superior thermal resilience of the mediumweight structure with no floor insulation (B1). In the December tests, B1 achieved an hours of safety threshold (HST) of 60.4 h, outperforming several retrofit strategies reported in the literature, including PCM-only interventions (44 h) and combined insulation with air sealing (31 h), and approaching the performance level of the most advanced combined PCM + sealing solutions (100 h). The lightweight structure (B2), in contrast, reached an HST of 27.4 h, comparable to moderate retrofit configurations and significantly outperforming the unmodified baseline reported in Wijesuriya et al. [20], which was limited to just 2 h.

Under the more demanding February conditions, with an average outdoor temperature of −4.3 °C, building B1 again maintained high performance, reaching an HST of 61.0 h. B2 also improved in this scenario, with an HST of 31.3 h, yet remained consistently behind B1. These findings confirm the passive thermal stability of the mediumweight structure, even in the absence of supplementary measures such as phase-change materials, preheating protocols, or airtightness enhancements.

4. Discussion of Results

The comparative analysis with Rehman et al. [4] shows that B1 demonstrated exceptional thermal durability in both the TDY and ECY scenarios, with an extended RT, slow CS, and stable RS. Both B1 and B2 had lower CS values than any simulated cases, yet the DoD values were similar or even higher in B2.

This suggests potential methodological limitations in the DoD metric. While the DoD accounts for the cooling depth and duration, it may not scale appropriately with the duration of observation. Rehman et al. [4] used a fixed 31 h window, whereas the present experiments extended to 80–100 h. The DoD may increase disproportionately in longer tests, even if the actual thermal resilience is superior, particularly for buildings with high inertia. This limitation was previously noted by Homaei and Hamdy [3], who emphasized the importance of time-normalized resilience indicators.

Several factors may explain the divergence between the simulated findings of Rehman et al. [4] and the experimental results reported here. These include differences in boundary conditions, observation windows, and modeling assumptions. In particular, the use of a fixed 31 h analysis period in simulations may underestimate the cumulative resilience in high-inertia buildings. Moreover, real-world effects such as preconditioning, ground heat exchange, and thermal bridging are difficult to replicate accurately in simulation environments. These discrepancies highlight the value of empirical data in capturing full-system thermal dynamics and assessing the practical limitations of model-based resilience predictions.

In contrast, the HST approach proposed by Wijesuriya et al. [20] offers a more intuitive classification system, focusing on critical temperature thresholds (e.g., 15 °C). However, its binary nature (safe or unsafe) limits the resolution, especially when comparing nuanced performance aspects across buildings or climates. For operational resilience, such as flexibility under intermittent energy supplies, more detailed indicators like the time to threshold (T₁₈, T₁₅, T₁₂) are more informative and comparable.

The findings suggest that simplified, physically grounded metrics—such as the proposed T_x and T(t_x) indicators—may offer clearer insights into resilience classification. They also enable comparisons across buildings of different masses, typologies, and observation periods without relying on simulation calibration.

An additional dimension requiring attention is the influence of ground heat exchange, particularly in the mediumweight building B1, which lacked floor insulation. From a physical perspective, the absence of insulation at the floor–ground interface facilitates one-dimensional conductive heat transfer between the building and the subsoil. During the heating phase, thermal energy is partially stored in the upper ground layers through downward conduction. When heating is interrupted, the stored energy is gradually released back into the interior, reducing the temperature gradient and slowing the cooling process. This bidirectional exchange is especially relevant in massive constructions with prolonged preconditioning and may explain the relatively stable thermal performance observed in B1 despite the colder ambient conditions.

This mechanism could explain the relatively small difference in the cooling rate between the December and February experiments despite the colder outdoor conditions. The trade-off between energy efficiency and passive resilience, especially in buildings with strong thermal mass, warrants further study.

While most of the resilience literature focuses on the overheating risk, Laouadi et al. [27] emphasize that the indoor environment, through envelope configuration, mass, and design, can amplify or buffer thermal stress more significantly than the outdoor climate alone. Although their review centers on summer conditions in Canada, the underlying principle applies equally to cold-climate scenarios. The present results confirm that, even in the absence of active systems, a properly configured mediumweight structure can delay thermal degradation and maintain habitable temperatures beyond standard emergency response timeframes. Additionally, Laouadi et al. [27] highlight the lack of empirical data on indoor thermal conditions during blackout events, further validating the contribution of the present full-scale experimental results.

Rui et al. [28] also emphasize the limitations of modeling-based approaches that rely on typical meteorological years (TMY), arguing that such datasets underrepresent rare but critical events. Their analysis suggests that resilience assessment should integrate real climate extremes, envelope dynamics, and human behavior. The present findings support this view and provide empirical evidence that high-mass buildings can offer extended thermal protection during blackouts, even under severe winter conditions. Performance of this kind would likely be underestimated in simulation-based studies.

The recent study by Liyanage et al. [29] provides additional confirmation of the severity of blackout conditions in cold climates. In their simulation-based assessment of Canadian homes, code-compliant buildings were shown to reach sub −5 °C indoor temperatures within four days of heating loss. Even with passive measures such as movable insulation, severe conditions could not be entirely avoided. In contrast, the mediumweight building (B1) considered here maintained indoor temperatures above 13 °C for more than 60 h without any dynamic mitigation strategies. This suggests that envelope mass and ground coupling can significantly enhance passive thermal survivability, sometimes more effectively than traditional retrofits. It also highlights the importance of developing resilience indicators that account for not only static efficiency measures but also dynamic thermal behavior during disruptions.

In sum, the comparison with simulated benchmarks and the existing literature underscores the value of empirical data and real-world thermal behavior tracking. The present study complements and extends previous resilience frameworks by incorporating long-duration cold scenarios, high-inertia structures, and passive soil heat buffering, offering new insights for resilience modeling, retrofit design, and cold-climate policy.

Although this study focused on the indoor air temperature as the main resilience indicator, the indoor relative humidity was also monitored. During the experiments, the RH ranged from approximately 30% (before heating shutdown) to 50% (after several days without heating). Within this range, humidity has little influence on thermal perception or comfort in cold environments. Therefore, the air temperature was considered a sufficient proxy in assessing passive thermal resilience.

The findings also have implications for regulatory and policy frameworks. Current building standards prioritize energy efficiency under normal operating conditions but often overlook thermal autonomy during service interruptions. Incorporating resilience-oriented criteria, such as passive survivability thresholds and dynamic temperature decay behavior, into building codes and performance assessment tools could enhance preparedness and safety in heating-dominated climates. Risk-informed regulations that account for thermal resilience would support more robust design strategies, especially in regions facing increasing climate variability and energy supply uncertainty.

5. Comparative Analysis and Development of Thermal Resilience Indicators

5.1. Application of the WUMTP Metric to Experimental Data

To benchmark our results against existing thermal resilience assessment frameworks, we applied the weighted unmet thermal performance (WUMTP) index to our full-scale experimental dataset. WUMTP, proposed by Homaei and Hamdy [3], quantifies building resilience during and after disruptive events by aggregating penalty-weighted thermal deviations based on three dimensions: the event phase, hazard severity, and exposure time.

While this metric has been used extensively in simulation-based studies, it has not yet been widely validated using real-world measurements. Here, we apply the full WUMTP model to empirical data from two residential buildings exposed to multi-day winter blackouts.

In our calculation, we applied the full 12-segment WUMTP formulation, incorporating phase (WP), hazard level (WH), and exposure duration (WE) penalties as proposed by Homaei and Hamdy [3]. The penalty structure distinguishes between three performance zones (acceptable, habitable, uninhabitable) and two exposure time categories (short/easy, long/difficult), resulting in a differentiated response across the temperature trajectory. The analysis covered the first 96 h of the outage, consistent with the original “four-day test framework”, and the results were normalized to the building floor area (123 m²) to yield values in [°C·h/m²].

Table 3. Weighted unmet thermal performance (WUMTP) [°C·h/m²] during the first 96 h of a blackout.

Building	December	February
B1 (mediumweight)	23.64	24.99
B2 (lightweight)	83.82	89.03

presents the calculated WUMTP values for the four test cases.

These results confirm the superior thermal resilience of the mediumweight masonry building (B1), which maintained indoor temperatures above the critical thresholds and accumulated significantly fewer penalty-weighted degree-hours. In contrast, the lightweight structure (B2) experienced rapid and prolonged deviations from comfort conditions, resulting in more than 3.5 times higher WUMTP scores.

Importantly, these values fall within the range reported in simulation-based studies by Homaei and Hamdy [3], where the WUMTP scores for passive designs ranged from approximately 13 to 113 °C·h/m². This close alignment suggests that the penalty scheme is sensitive and transferable to real-world performance and that WUMTP can be reliably applied using monitored temperature data.

Nonetheless, our results also highlight the limitations of WUMTP when used in long-duration experiments. Despite the continued indoor temperature decline in building B2 after 72 h, the WUMTP curve plateaued, due to the saturation of penalty coefficients. This dynamic may underrepresent the extended severity of exposure in lightweight structures. In contrast, our T(t_x) and T_x metrics captured these differences more directly, by resolving the absolute temperature values and crossing times at key thresholds.

While the WUMTP index provides a structured and weighted approach to assessing thermal resilience, our findings suggest that it may lose sensitivity under prolonged blackout conditions—particularly in lightweight buildings, where temperatures continue to decline after penalties saturate. To address these limitations and offer a complementary perspective, we propose a time-based, physically grounded framework that captures the temporal dynamics of indoor temperature decay in a more direct and scalable manner. The following two subsections introduce the proposed indicators: the time to threshold (T_x) and temperature at X hours (T(t_x)).

5.2. Time to Threshold Indicators (T_x)

This metric captures the number of hours from the start of the heating outage until the indoor air temperature drops below a specific critical threshold. The recommended values include

T₁₈—Adaptive comfort threshold;

T₁₅—Passive habitability threshold;

T₁₂—Severe discomfort threshold, where prolonged exposure may impair thermoregulation and increase cardiovascular strain [17];

T₉—Health risk threshold, where temperatures below 9 °C are associated with hypothermia and increased risks of acute health events in elderly or vulnerable occupants [17,18].

The output is a time-based vector, [T₁₈, T₁₅, T₁₂, T₉], which allows for the simple and meaningful comparison of thermal durability across building types, climates, and operational contexts.

5.3. Temperature at X Hours Indicators (T(t_x))

In this approach, the indoor air temperature is measured at fixed time intervals after heating has been disabled; there are typically 12 h, 24 h, 36 h, 48 h, and 60 h.

This generates a temperature vector, [T(t₁₂), T(t₂₄), T(t₃₆), T(t₄₈), T(t₆₀)], which can be plotted as a thermal decay curve or used to construct point-based resilience scores. These values are immediately interpretable by practitioners and policymakers and can be linked to critical exposure thresholds or emergency response timeframes.

5.4. Advantages of the Proposed Method

The proposed time-based resilience framework offers several advantages over existing approaches. First, it is independent of the observation duration, which enables valid comparisons across studies with differing test lengths and monitoring periods. Second, it maintains a high degree of physical transparency—being directly based on temperature and time data, it avoids statistical abstraction and remains closely aligned with occupant-relevant outcomes.

Another key strength of the method is its scalability and replicability. It can be applied in a wide range of contexts, including full-scale field experiments, dynamic building simulations, and post-occupancy evaluations. Its structure also supports flexible classification schemes. For example, buildings can be grouped into standardized resilience classes based on T₁₈ thresholds, such as

Class A (T₁₈ ≥ 24 h);
Class B (12 h ≤ T₁₈ < 24 h);
and Class C (T₁₈ < 12 h),

offering a practical and communicable basis for policy or design guidance. In addition, the framework lends itself well to visual representation, allowing for the development of tools such as resilience maps or performance envelopes. These, in turn, can complement probabilistic modeling and long-term risk assessments under future climate scenarios, helping to bridge the gap between research and actionable insights in thermal safety and climate adaptation.

6. Example Application of the Methodology

To demonstrate the applicability of the proposed metrics,

Table 4. Indoor temperature [°C] after specified time since heating shutdown.

Building	T(t12) [°C]	T(t12) [°C]	T(t12) [°C]	T(t12) [°C]	T(t12) [°C]
B1 (December)	18.4	17.2	16.2	15.6	15.0
B2 (December)	17.2	15.5	13.9	12.8	11.6
B1 (February)	19.0	18.4	17.4	16.3	15.0
B2 (February)	17.6	16.1	14.1	12.7	10.5

presents the recorded indoor temperatures for buildings B1 (mediumweight) and B2 (lightweight) during controlled heating outages in December and February.

The results show a clear difference in thermal performance between the two buildings. The mediumweight masonry structure with no floor insulation (B1) maintained indoor temperatures above the 15 °C habitability threshold for the entire 60 h testing period in both December and February. In contrast, the lightweight timber-frame building (B2) fell below this threshold within the first 24 h and dropped to unsafe levels below 12 °C between 48 and 60 h after heating was disabled. Notably, the contrast in resilience between the two buildings appears to be driven more by ground coupling and differences in the construction type and thermal mass than by variations in outdoor conditions between the two testing periods. To further illustrate the usefulness of the proposed T(t_x) metrics, Figure 5 and Figure 6 present the hourly indoor temperature trajectories of buildings B1 and B2 during the heating outages in February and December, respectively. Horizontal lines in each graph mark the proposed resilience thresholds: 18 °C (adaptive comfort), 15 °C (habitability), 12 °C (severe discomfort), and 9 °C (health risk).

As shown in Figure 5, under December conditions (T_o,avg = +0.7 °C), the lightweight building (B2) experienced a steady and rapid decline in indoor temperature, reaching the 15 °C threshold in approximately 25 h and the 12 °C threshold in about 55 h. In contrast, the mediumweight structure (B1), which had an uninsulated floor, maintained temperatures above 15 °C for approximately 60 h and recorded only a slightly lower temperature over the remaining 57 h of the blackout.

Figure 6 illustrates the hourly indoor temperature trajectories during the February outage (T_o,avg = −4.5 °C). In this case, B2 dropped below 15 °C within approximately 30 h and below 12 °C after 55 h. Starting from the fifth day of the blackout, the night-time temperatures in B2 fell to around 7 °C. B1, on the other hand, remained well above the habitability threshold for the first 60 h, and its temperature never dropped below 13 °C during the entire 7-day heating outage.

Interestingly, the performance gap between the two buildings appears to be even more pronounced in December than in February. This may be attributed not only to the higher thermal mass of B1 but also to the absence of floor insulation, which enabled direct thermal coupling with the ground. As the buildings had not been heated from the beginning of the heating season in October until one week before the December experiment, the soil beneath the uninsulated floor remained relatively cold. Consequently, when the heating was switched off, B1 cooled more rapidly than in the February experiment. In the latter case, the indoor temperatures had been maintained at 21 °C in the intervening period, allowing the ground below the floor to gradually warm up and buffer heat loss more effectively.

These visual results clearly demonstrate the practical value of the T(t_x) framework in identifying resilience classes and guiding passive design decisions. They also underscore the potential of ground-coupled mediumweight construction to passively sustain safe indoor conditions during extended periods without heating.

7. Conclusions

This study evaluated the thermal resilience of two full-scale residential buildings with contrasting construction typologies—mediumweight masonry (B1) and a lightweight timber frame (B2)—under both typical and extreme winter conditions. Indoor temperature decay and recovery were monitored during controlled heating shutdowns. Resilience performance was assessed using standard metrics (RT, DoD, HST) and newly proposed time- and temperature-based indicators (T_x and T(t_x)).

Key findings include the following:

• The mediumweight building (B1) maintained indoor temperatures above the 15 °C habitability threshold for over 60 h, despite the absence of floor insulation or additional passive strategies. This highlights the combined effect of thermal mass and ground coupling.
• The lightweight building (B2), although compliant with modern energy standards, exhibited faster cooling and crossed critical thresholds (below 12 °C) within 55 to 60 h, emphasizing the influence of the construction type on resilience outcomes.
• In comparison with simulation-based assessments, the results indicate that envelope mass and substructure behavior may play a more decisive role in passive survivability than steady-state efficiency ratings.

While the DoD and HST provide valuable insights, these metrics may lack sensitivity when applied to high-inertia structures or long-duration outages. To enhance the applicability, a complementary framework based on the time to threshold and temperature after X hours is proposed.

The results support the use of simplified, physically interpretable metrics in evaluating thermal resilience and emphasize the importance of validating simulation-based approaches with empirical data. The proposed indicators may also contribute to performance classification, retrofit prioritization, and risk-based design practices in heating-dominated regions.

Further research is recommended to expand this framework to additional building typologies, climatic conditions, and occupant behaviors. Interactions between passive and low-energy active systems should also be considered. The observed ground thermal buffering effect in B1 reveals a promising but underexplored aspect of resilience that warrants further investigation through both experimental and modeling approaches. This study also includes the first known empirical application of the full WUMTP model. The calculated values were found to be consistent with those observed in prior simulation-based studies, confirming the metric’s applicability to measured indoor temperature data. However, limitations emerged in extended blackout scenarios, where penalty saturation tended to obscure ongoing performance degradation. These findings reinforce the value of complementary time-based indicators in capturing long-term resilience dynamics.

8. Limitations of the Study

While the findings of this study provide valuable empirical insights into building thermal resilience during winter outages, several limitations must be acknowledged.

First, the sample size and typology were limited to two full-scale experimental buildings, each representing a distinct construction system: a lightweight timber frame and mediumweight masonry. While these cases capture key envelope differences, they do not reflect the full diversity of residential building typologies, especially hybrid or high-performance structures featuring advanced materials or dynamic façade elements.

Second, the climatic context of the experiments was confined to a single cold-climate location under both typical and extreme winter conditions. Although the results are relevant for similar temperate climates, their generalizability to other, warmer climate zones requires further validation.

Third, the buildings were unoccupied throughout the testing period. In real-world scenarios, factors such as occupant behavior, intermittent heating, window opening, and zonal thermal management could significantly influence resilience outcomes. The lack of human interaction may thus limit the realism of the thermal response data.

Fourth, the study focused exclusively on passive resilience, assessing the thermal behavior of buildings without the influence of active backup systems (e.g., battery-supported heat sources or localized electric heaters). As such, the results emphasize passive survivability but do not account for hybrid or adaptive strategies that might be used in emergency conditions.

Finally, while the buildings were maintained at 21 °C prior to each experiment to ensure thermal stabilization, the initial preconditioning period may not have been sufficient—particularly in the mediumweight structure with an uninsulated floor slab. In the December test, the buildings had been unheated before the five-day stabilization period began. Our assumption that five days of heating would establish equilibrium thermal conditions proved insufficient in B1, where the ground-coupled thermal mass likely required a longer period to reach a thermal steady state. As a result, the ground temperatures and their contributions to passive heat exchange may have been underestimated or unequally distributed across the test scenarios. Future studies should more precisely control and measure the thermal history of building–ground systems, especially when evaluating resilience in structures with significant subfloor interactions.