County-Level Assessment of Building Stock Thermal Resilience During Heat Waves and Power Outages

Abstract

This paper demonstrates how building stock models, such as ResStock in the US, can be useful for resilience testing. We examined the thermal resilience of residential buildings in Illinois during simulated extreme weather events and power outages using ResStock models. Our results reported significant vulnerability, with the median dry bulb temperature (DBT) peaking at 32.5°C within 41 hours during the heatwave with power outage, and the 85 th percentile DBT reaching 38.4°C in 42 hours. The Heat Index (HI) exceeds the danger level, with a median of 36.14 °C within 21 hours on the first day indicating extreme caution for 11 consecutive hours, and the 85 th percentile DBT reaching 51.3 °C in the same timeframe indicating danger level for 11 hours. Analysis of building characteristics highlights the crucial role of building design, emphasizing the importance of thermal regulation in building infrastructure, particularly in those with finished attics or concrete masonry units (CMU). A geospatial vulnerability assessment identified the most thermally vulnerable and resilient counties and further investigated these counties' socioeconomic status based on the federal poverty level (FPL). We found that low-income households are often in buildings with less thermal resilience and are at a greater risk during extreme temperature events.

Full text

PLEA 2024 WROCŁAW
(Re)thinking Resilience
County-Level Assessment of Building Stock Thermal
Resilience During Heat Waves and Power Outages
MOHAMED A. BELYAMANI1 KRITIKA KHARBANDA2 NAN MA1,* HOLLY SAMUELSON2
1 Worcester Polytechnic Institute, Worcester, United States of America
2 Harvard Graduate School of Design, Cambridge, United States of America
ABSTRACT: This paper demonstrates how building stock models, such as ResStock in the US, can be useful for
resilience testing. We examined the thermal resilience of residential buildings in Illinois during simulated extreme
weather events and power outages using ResStock models. Our results reported significant vulnerability, with the
median dry bulb temperature (DBT) peaking at 32.5°C within 41 hours during the heatwave with power outage,
and the 85th percentile DBT reaching 38.4°C in 42 hours. The Heat Index (HI) exceeds the danger level, with a
median of 36.14 °C within 21 hours on the first day indicating extreme caution for 11 consecutive hours, and the
85th percentile DBT reaching 51.3 °C in the same timeframe indicating danger level for 11 hours. Analysis of
building characteristics highlights the crucial role of building design, emphasizing the importance of thermal
regulation in building infrastructure, particularly in those with finished attics or concrete masonry units (CMU). A
geospatial vulnerability assessment identified the most thermally vulnerable and resilient counties and further
investigated these counties’ socioeconomic status based on the federal poverty level (FPL). We found that low-
income households are often in buildings with less thermal resilience and are at a greater risk during extreme
temperature events.
KEYWORDS: Passive Survivability, Heatwave Events, Residential Energy Modelling, Indoor Heat Vulnerability,
Geospatial Vulnerability Assessment
1. INTRODUCTION
1.1. Background
The impacts of climate change have been felt
across the globe, exposing populaons to more
frequent and more severe weather events [1-2]. As
global temperatures rise due to the accumulaon of
greenhouse gases in the atmosphere, various regions
are experiencing changes in weather paerns and an
increase in extreme weather phenomena. Between
2020 and 2022, the U.S. experienced 60 separate
weather and climate disasters, each exceeding a billion
dollars in damages [3]. Notable occurrences among
these included unprecedented heatwaves and
wildres in California, Oregon, and Washington during
the fall of 2020 [4], a historic winter storm/cold wave
event focused on the Deep South and Texas in
February 2021 [5], and a drought and heatwave in the
Western and Southern Plains states in the summer of
2022 [6]. Extreme weather occurrences, along with
power outages, present a specic threat to energy
infrastructure and fundamental services reliant on
electricity. This situaon jeopardizes the eecve
funconing of buildings in maintaining safe indoor
environments. However, buildings should play a
fundamental role in enhancing resilience against
climate change events such as heatwaves to safeguard
the safety and well-being of occupants. Therefore, a
research queson arises: which segments of our
residenal building stock are most at risk of
overheang with the increasing frequency and
severity of extreme weather events and power
outages? Thus, the objecve of this paper is to 1)
invesgate the thermal resilience of exisng
residenal building infrastructure, with a focus on the
ecacy of homes in migang heatwave challenges
with power outages and 2) create a replicable
framework for such analysis.
1.2. Previous studies
Previous studies on thermal resilience emphasize
its role in providing human comfort and safety during
extreme climate events. Sheng et al. [7] delved into an
assisted living facility’s response to heatwaves and
power outages, emphasizing the signicance of
passive envelope strategies and natural venlaon.
White et al. [8] explored resilient, sustainable design
approaches, demonstrang how passive building
techniques bolster resistance to outages. Sengupta et
al. [9] explored the resiliency and passive survivability
of an oce building, sustainable design approaches,
demonstrang that implementaon of cooling
systems acve or passive as well as sun blinds are
important to combat overheang risks during
heatwaves. Another Sengupta et al.’s [10] study
focused on overheang risks in educaonal buildings
during heatwaves and outages, shedding light on the
vital role of resilient cooling strategies.
Prior work primarily explored thermal resilience in
a specic building. However, our research aims to

extend this scope by geo-spaally evaluang the
thermal resilience of a diverse set of residenal
buildings. This expansion allows us to conduct a
comprehensive analysis of the thermal resilience of
exisng residenal infrastructure in the face of
heatwave challenges and power outages, with the
ulmate goal of establishing a replicable methodology
for such invesgaons.
2. METHODS
In this study, we invesgate thermal resilience in
the US housing stock using ResStock [11]. Built on the
OpenStudio / EnergyPlus building energy simulaon
engine, the ResStock database has an extremely rich
documentaon of US residenal building
characteriscs across various geographical resoluons
ranging from naonal to county level. ranging from
naonal to county level. Detailed informaon on the
characteriscs of the housing stock and how to access
their metadata is provided in Ref [11]. To our
knowledge, this is the rst published study using the
ResStock database for resilience tesng. We randomly
selected a 20% sample of ResStock buildings (n=4,374)
in Illinois to create a meaningful subset of housing and
achieve a county-level assessment of building thermal
resilience. We focused on Illinois for two primary
reasons: 1) Diversity: e.g., dense urban, suburban, and
rural areas with mulple climate zones. 2) Notable
changes in weather paerns: an increasing frequency
of extreme weather events in recent years [12,14].
2.1 Methodological framework for building thermal
resilience
To assess thermal resilience, we selected July
2012’s peak heat period in Chicago as a representave
extreme heat scenario. Two metrics such as Dry Bulb
Temperature (DBT) and Heat Index (HI) [15] are used
to measure occupant heat exposure and vulnerability.
As illustrated in Figure 1, our methodological
framework involves the following four-phase process.
The rst phase gathers representave samples from
the 2023 ResStock database, with each building model
having an XML le and a CSV le for annual schedules
(i.e., window, HVAC operaons). For Illinois’ two
climate zones (4A and 5A), we coupled each building
model with appropriate weather les, using Chicago
and Springeld as reference cies for zones 5A and 4A,
respecvely. The second phase modies XML les to
simulate power outages and window opening
schedules during heatwaves which includes disabling
HVAC systems, then seng higher thresholds for the
natural venlaon temperature setpoints, maximum
air exchange rates, and outdoor humidity raos (to
enable window opening to avoid indoor overheang).
The third phase focuses on conguring the simulaon
plaorm using the OpenStudio-HPXML workow
framework. This framework integrates the OpenStudio
soware suite with the Home Performance XML
(HPXML) data standard, which is especially useful for
large-scale computaon as it simplies managing and
automang mulple building simulaons and data
handling. Along with this framework, a batch
simulaon can be launched for a representave
sample of buildings (in our study, n=4,374). The nal
phase involves extracng and analyzing the me series
data to assess thermal resilience based on the
simulaon outputs.
Our methodological framework, built in Python
3.7, automates the majority of the simulaon process,
oering a pipeline for future research in assessing
building thermal resilience. Our developed framework
is applicable to buildings that rely on HVAC for cooling
during heatwaves. The framework incorporates a pre-
outage (normal operaon), phase of three days,
followed by a simulated three-day power outage,
resulng in a total evaluaon period of six days.
3. RESULTS AND DISCUSSION
3.1 Building thermal resilience analysis during
heatwaves and power outage
Figure 2 presents our simulated results of the
hourly DBT before and during the blackouts. On the
rst day of the outage, the median DBT value reached
31.5°C within an 18-hour period from the outage
onset. The 85th percenle DBT peaked at 37.4°C in the
same meframe. The highest peak in the three-day
outage scenario occurred on the second day, with the
median DBT hing 32.5°C at 41 hours post-outage,
and the 85th percenle of residenal buildings
reached a maximum of 38.4°C at 42 hours. These
results demonstrate the risk for signicantly elevated
indoor temperatures in residences without
Figure 1: Methodological framework for building thermal resilience simulation.

intervenons, posing risks of discomfort, as well as
health and safety concerns. These ndings are
consistent with prior research [7], where
temperatures reached 30°C within a 20-hour period.
However, our study reveals an oscillang paern in
line with outdoor temperature uctuaons, diering
from previous study.
In examining HI performance, Figure 3 shows the
HI variaon during a three-day power outage. On the
rst day, the median HI hits 36.14 °C in a 21-hour
period from the start of the outage, indicang extreme
cauon for 11 consecuve hours. The 85th percenle
HI peaked at 51.35 °C in the same meframe reaching
the danger zone and persisng for 11 hours. By the
second day, the maximum HI across all simulated
buildings in Illinois reached a median of 35.5 °C at 47
hours post-outage and persisted for 9 consecuve
hours. The 85th percenle HI peaked at 52 °C at 46
hours post-outage. This temperature, indicave of the
danger zone, persisted for 11 hours at the state-level.
These observaons in the simulated data set highlight
a potenal need for beer preparedness in the
majority of Illinois buildings to sustain safe condions
during simultaneous power outages and heatwaves,
posing a signicant threat to occupants’ safety. Our
ndings align with those in Ref [5], which also
idened an extreme cauon level within an 8-hour
period, indicang similar building performance.
Notably, even short periods of power outages, such as
a one-day blackout, can lead to reaching the extreme
cauon level within the rst day. This highlights the
urgent need for migaon and adaptaon strategies
to promote resilient building pracces in Illinois. Here,
our results are reported with the 85th percenle
instead of the 95th percenle to migate the inuence
of potenal outliers and extreme values.
3.2 Thermal resilience profiles across different
climate zones
The comparison of simulated building resilience
results across the two dierent Illinois climate zones,
as shown in Figure 4, demonstrates the variaon in
DBT, Relave Humidity (RH), and the HI. There is a
marked dierence between the overall state
condions and climate zone 4A, where 4A consistently
shows higher temperatures and median thermal
performance for both DBT and HI compared to the
state average. These ndings prompt further
invesgaon into how dierent building
characteriscs contribute to resilience during power
outages and heatwaves.
Figure 4: Boxplot of indoor DBT, RH, and HI for the 4,374
residences across two Illinois climate zones during the 2012
heatwaves on the three-day power outage.
3.3 Influence of building characteristics
In general, buildings are designed based on a group
of xed assumpons and condions in the design or
renovaon phases. However, the actual performance
of buildings during occupancy oen diverges from
these architect-intended inial condions. In this
Figure 2: Hourly DBT distribution among the 4,374 residences and outdoor air temperature during the 2012 heatwave.
Figure 3: Hourly HI distribution among the 4,374 residences and outdoor air temperature during the 2012 heatwave.

secon, we invesgate a range of building
characteriscs to determine which are correlated with
overheang, parcularly by analyzing mean and
maximum DBT temperatures. Examined design
variables include ac types, wall insulaon levels, and
air changes per hour (ACH).
The ridgeline plots in Figure 5 show the probability
distribuon of the mean and maximum temperatures
across dierent ac types during heatwaves and
blackouts. As can be seen, residences featuring
nished acs or cathedral ceilings demonstrate a
noteworthy reducon in mean temperature compared
to other types with an average mean of 27 °C. It also
shows less variability in temperature, suggesng a
more regulated and consistent indoor climate. This
paern may highlight the crical role of ac
construcon in moderang indoor thermal
environments, poinng to the potenal benets of
strategic ac design for improved thermal resilience.
Figure 5: Ridgeline plot illustrating the distribution of mean
and maximum temperatures across various attic types on the
first day of heatwaves without power.
Figure 6 reveals significant variations in the thermal
resilience of different wall types. Concrete masonry
units (CMU) with a 6-inch hollow and uninsulated
structure are correlated with better thermal
performance than brick and wood stud walls, possibly
due to CMU’s high thermal mass and efficient thermal
behavior [16]. CMU-walled dwellings are typically
correlated with stable indoor temperatures, averaging
around 27°C and peaking at 32°C. Brick walls are
correlated with slightly less resilience than CMU, but
their peak temperature probability is substantially
higher, reaching the extreme caution level as defined
by the HI. Wood stud walls are correlated with the
least thermal resilience, with a 10-15% likelihood that
peak temperatures could exceed 42°C, entering the
danger level per HI standards.
The ACH in a building is commonly used as an
indicator of air inltraon. Evaluang the degree of
inltraon helps in assessing the potenal for
migang indoor heat by exchanging it with outdoor
air or prevenng the loss of cooler indoor air. Figure 1
illustrates the ACH values and their corresponding
average and peak temperatures as computed by the
HI. Notably, airght dwellings (1 ACH and 2 ACH) show
reduced temperature variability and lowest mean
temperature proles. However, there is sll a notable
probability of these dwellings reaching peak
temperatures in the extreme cauon zone as dened
by the HI. This suggests that while airghtness
contributes to resilience, it could also lead to
overheang concerns during certain periods.
Conversely, buildings with higher ACH (e.g., 3 ACH, 5
ACH and higher) oen experience wider and more
elevated temperature ranges. As can be seen from
Figure 7, on average there is a 20-30% likelihood of
these dwellings reaching the extreme cauon level.
We are not suggesng here that the ACH should be
indiscriminately increased or decreased to improve
resilience. Instead, we aim to highlight what could be
“good” versus “poor” design based on overheang risk
assessment.
Figure 7: Ridgeline plot illustrating the distribution of mean
and maximum temperatures across ACH variations on the
first day of heatwaves without power.
3.4 Geospatial vulnerability and resilience
assessment
Exploring thermal resilience at the county level is
critical due to the distinct climatic and architectural
variations that exist within different regions. Such a
granular approach allows us to tailor resilience
strategies more effectively, addressing specific local
challenges and enhancing the overall safety and
comfort of residents in varying geographic and climatic
conditions. We employed a geographical information
Figure 6: Ridgeline plot illustrating the distribution of mean
and maximum temperatures across various wall insulation
levels on the first day of heatwaves without power.

system (GIS) approach to visually convey the regional
disparities in thermal vulnerability and resilience.
Figure 8 presents the results, which averaged the peak
HI of each residential building in a county, aggregating
the data for all residential buildings in that county on
the first day of a heatwave coincided with a power
outage. The color gradient represents the severity of
heat exposure measured by the HI. This county-level
analysis reveals significant spatial disparities in
thermal resilience. Counties in blue display relatively
lower HI values, which suggests that residences in
these areas maintained cooler conditions and their
dwellings are indicative of better thermal resilience.
The red-shaded counties, on the other hand, show
higher HI values, signifying areas where the indoor
conditions are likely more stressful and potentially
hazardous, thus highlighting a greater need for
effective cooling solutions and thermal design
improvements in these counties. This breakdown can
guide the state resource allocation for enhancing
thermal safety.
Figure 8: County level spatial distributions of HI.
Addional analysis was conducted for the most
resilient and vulnerable counes in Illinois, outlined in
white, in Figure 8. The most vulnerable counes, i.e.
those with the highest HI are Pia County with a mean
HI of 47.7°C, and mean DBT of 35.6°C, and Mason
County with a mean HI of 49.5°C, and a mean DBT of
35.5 °C. The most resilient counes in terms of HI are
Woodford County, with a HI of 25.5°C and a mean DBT
of 24.7°C, and Marshall County, with a HI temperature
of 25.8 °C and a mean DBT of 25.5°C. We further
correlated these counes with their socioeconomic
status according to the federal poverty level (FPL). The
FPL, an economic measure, uses a percentage to
compare household income against the poverty
threshold. Lower FPL percentages indicate incomes
closer to the poverty line. Figure 9 shows that
households with annual incomes below 100% of the
FPL experience a wide range of HI values, suggesng
these buildings are more prone to inadequate thermal
control, elevang health risks during heatwaves and
power outages. A similar paern was observed in
households with 150-200%, 200-300%, and 300-400%
of the FPL. This suggests a correlaon where lower-
income households, which are oen in buildings with
less investment in thermal resilience measures, are at
a greater risk during extreme temperature events. Our
ndings emphasize the need for targeted
intervenons in building design and energy assistance
programs to protect vulnerable populaons from
extreme heat.
Figure 9: Heat disparities and socioeconomic status in the
most resilient and vulnerable counties.
These findings not only illuminate the varying degrees
of vulnerability across Illinois counties but also provide
essential context for enhancing preparedness and
resilience in the face of extreme temperatures and
power outages. This analysis also allows us to delve
deeper into understanding the building characteristics
associated with vulnerability to overheating. This type
of analysis can equip decision-makers with the
knowledge needed to implement targeted strategies
for improving the overall resilience of buildings and
communities in the region.
4. LIMITATIONS AND FUTURE WORK
In this study, we used the ResStock energy models,
which do not represent individual exisng buildings
but are based on building stock stascs and, in
aggregate, have been validated to match measured
energy data [17]. These models have not been
validated in terms of indoor thermal condions, and,
therefore, their accuracy for these outcomes is
unknown. Furthermore, we only used a 20% sample
of Illinois residenal buildings from the ResStock
database for thermal resilience evaluaon due to
computaonal cost constraints. For more
comprehensive analysis, future research could expand
the dataset size, possibly using the full database to
explore wider paerns and characteriscs for building
resilience evaluaon across construcon years and
varying degrees of retrong. This would provide a
more detailed understanding of the factors that
contribute to thermal resilience in residenal

buildings. In addion, naonwide thermal resilience
evaluaon also would benet from idenfying the
vulnerable state and/or the most vulnerable counes
across the naon, which is possible with this dataset
and high-performance compung clusters.
Future research on building thermal resilience
could incorporate more demographic data. Naonal
surveys like the U.S. Census can supplement ResStock
models by providing detailed demographic
informaon. This would allow users to quanfy
resilience measures through compung metrics like
physiologically equivalent temperature (PET) and
perceived temperature (PE). Addionally, our study
assumed that occupants would open windows when
outdoor temperatures are lower than indoor
temperatures, but this may not always be true without
empirical evidence or measurements of window-
opening behavior during power outages. Therefore,
there is a need to collect more data on such behaviors.
5. CONCLUSION
Our study developed a methodological framework
which can invesgate the vulnerability of residenal
buildings to extreme temperatures during power
outages, a consequence of climate change-induced
weather events. Through the use of ResStock models,
our research demonstrates a means to understand the
risks posed to indoor thermal condions and delves
into the factors inuencing resilience. Building
characteriscs such as ac type, wall material,
insulaon, and inltraon are correlated with an
infrastructure’s ability to withstand extreme
condions. Our analysis shows that the HI reached
crical levels indicang extreme cauon and danger
zones during blackouts, with the 85th percenle HI
peaking at 52°C. This highlights the acute threat to
occupant safety during simultaneous power outages
and heatwaves. Moreover, our geospaal vulnerability
assessment reveals regional disparies in thermal
resilience across Illinois counes, correlang
socioeconomic status with overheang vulnerability.
This approach equips stakeholders with knowledge to
develop targeted strategies for enhancing the overall
resilience of residenal infrastructure in the face of
escalang climate challenges. Our research
contributes not only to the understanding of building
thermal dynamics but also oers a replicable
methodology for future studies, guiding eorts to
create thermally safe and climate-resilient
communies.
REFERENCES
1. Anderson, G., & Bell, M. (2011). Heat waves in the United
States: mortality risk during heat waves and effect
modification by heat wave characteristics in 43 U.S.
communities. Environmental Health Perspectives, 119(2),
210–218. DOI: 10.1289/ehp.1002313
2. Hondula, D., Balling, R., & Vanos, J. E. (2015). Rising
temperatures, human health, and the role of adaptation.
Current Climate Change Reports, 1, 144–154. DOI:
10.1007/s40641-015-0016-4
3. Smith, A. B. (2020). U.S. Billion-dollar Weather and Climate
Disasters, 1980 - present (NCEI Accession 0209268),
[Online], DOI: 10.25921/STKW-7W73 [05 June 2023].
4. National Oceanic and Atmospheric Administration (2021).
2020 U.S. billion-dollar weather and climate disasters in
historical context, [Online], Available:
https://climate.gov/disasters2020 [05 June 2023].
5. National Oceanic and Atmospheric Administration (2022).
2021 U.S. billion-dollar weather and climate disasters in
historical context, [Online], Available
https://climate.gov/news-features/blogs/beyond-
data/2021-us-billion-dollar-weather-and-climate-disasters-
historical [05 June 2023].
6. National Oceanic and Atmospheric Administration (2023).
2022 U.S. billion-dollar weather and climate disasters in
historical context, [Online], Available:
https://climate.gov/news-features/blogs/beyond
data/2021-us-billion-dollar-weather-and-climate-disasters-
historical [05 June 2023].
7. Sheng, M., Reiner M., Sun K., Hong T. (2023). Assessing
thermal resilience of an assisted living facility during heat
waves and cold snaps with power outages. Building and
Environment, 230, 110001. DOI:
10.1016/j.buildenv.2023.110001.
8. White LM (2020). ASSESSING RESILIENCY AND PASSIVE
SURVIVABILITY IN MULTIFAMILY BUILDINGS. ASHRAE Top
Conference Proceedings. p. 144–155.
9. Sengupta, A., Deleu, J., Lucidarme, B., Breesch, H., &
Steeman, M. (2023). Assessing Thermal Resilience To
Overheating In An Office Building.
10. Sengupta, A., Breesch, H., Al Assaad, D., & Steeman, M.
(2023). Evaluation of thermal resilience to overheating for an
educational building in future heatwave scenarios.
International Journal of Ventilation, 22 (4): p. 366–376. DOI:
10.1080/14733315.2023.2218424 10. Sengupta A.
11. ResStock, [Online], Available: https://resstock.nrel.gov
[10 June 2023].
12. CLIMATE CHANGE IN ILLINOIS, [Online], Available:
https://stateclimatologist.web.illinois.edu/climate-change-
in-illinois [10 June 2023].
13. Chen K, Newman AJ, Huang M, Coon C, Darrow LA,
Strickland MJ, et al (2022). Estimating Heat-Related
Exposures and Urban Heat Island Impacts: A Case Study for
the 2012 Chicago Heatwave. GeoHealth, 6(1),
e2021GH000535. DOI: 10.1029/2021GH000535.
14. Chicago, IL Temperature Records, [Online], Available:
https://www.weather.gov/lot/Chicago_Temperature_Recor
ds [10 June 2023].
15. NOAA. Heat Index, [Online], Available:
https://www.noaa.gov/jetstream/global/heat-index [15
June 2023].
16. Ben-Alon L, Rempel AR. Thermal comfort and passive
survivability in earthen buildings (2023). Building and
Environment, 230, 110339.
17. Wilson E, Christensen C, Horowitz S, Horsey H (2016). A
High Granularity Approach to Modeling Energy Consumption
and Savings Potential in the U.S. Residential Building Stock.
No. NREL/CP-5500-83077. National Renewable Energy Lab.