Renewal of aging residential buildings for electricity saving and carbon mitigation under climate change

Abstract

Climate change is amplifying Residential Electricity Carbon Emissions (RECE), with aging buildings posing challenges. Due to lack of era-specific assessments of RECE, it still remains unclear whether urban renewal strategies can effectively counteract these climate-driven impacts. Using daily electricity and temperature datasets, we developed temperature-responsive Residential Electricity Carbon footprint (RECF) functions to quantify RECE at building and community’s scale. Aging buildings exhibited 2–4 times higher RECF, while newer buildings contribute >50% of Beijing’s RECE. Higher RECF and compacted conditions led to increased Residential Electricity Carbon Intensity (RECI) in aging communities. Our projections indicated that future climate change will elevate RECE by 6–45% by 2050, with hotspots identified in aging communities inside the 3^rd Ring Road and high-rise clusters beyond the 5^th Ring Road in Beijing. Among the three evaluated renewal strategies, we also found that Near-Zero Energy Buildings (NZEB) could fully offset the climate-induced increase in RECE in the future. Although requiring a higher initial investment, Ultra-Low Energy Buildings (ULEB) could offer substantial long-term carbon reductions and enhanced climate resilience. These findings can provide an integrated perspective on the interplay among climate conditions, building age, and renewal pathways, offering critical insights for policymaking aimed at facilitating low-carbon and climate-resilient transitions in megacities.

Introduction

Consuming substantial energy for thermal regulation, lighting, and other applications, residential buildings represent a critical energy reduction and global decarbonization frontier, accounting for over 35% of electricity consumption and 22% of anthropogenic carbon emissions^1,2,3,4. Climate change, rising hotter summers and extreme weather events, is expected to amplify residential electricity consumption and cascading carbon emissions⁵. In Iran alone, climate-driven increases in residential electricity carbon emissions (RECE) could reach 456 million tons annually⁶, more than three times higher the current level. This escalating RECE trend simultaneously strains household energy affordability and impedes global carbon mitigation goals, demanding integrated strategies that deliver dual household-carbon benefits^{7,8,9,10,11,12}.

Residential building characteristics could critically shape RECE^13,14,15. While energy-efficient “green” residential buildings can reduce energy consumption for heating, ventilation, and air conditioning (HVAC) by up to 16%^16,17, aging residential buildings, particularly pre-1990s stock in China, lack essential thermal-insulation infrastructure and thus are highly vulnerable to temperature extremes. These residential buildings exhibit heavy HVAC dependency to maintain thermal comfort, resulting in disproportionate electricity use and RECE. Future climate change is expected to further intensify this burden^18,19. Beyond environmental and economic strains, aging residential communities face other pressing crises, such as health risks, social segregation, and economic inequalities^20,21,22, further intensifying the urgency of addressing these challenges at a global scale^{22,23,24,25,26,27}. Moreover, these challenges intersect with critical urban planning imperatives^28,29,30,31, underscoring the need for integrated building renewals that simultaneously deliver carbon mitigation through energy-efficient retrofits, enhance climate resilience via hazard-adapted infrastructure, as well as promote social equity by alleviating energy poverty among vulnerable groups and regions^32,33. Therefore, building renewals emerge as proactive, multidimensional interventions that directly promote sustainable urbanization, disaster risk reduction, and distributive justice³⁴.

Rapid expansion of aging residential stocks demands urgent modernization to energy-efficient standards. Renewing aging buildings offers a dual climate-energy benefit by concurrently reducing electricity consumption and mitigating associated carbon emissions^35,36,37,38. Previous studies have confirmed this win-win potential: renewing the 1980s buildings in Illinois could reduce annual RECE by as much as 50%³⁹. Remarkably, aging residential building renewals in the U.S. could achieve up to 91% of per capita RECE reduction⁴⁰. However, a spatially explicit, era-based assessment of RECE for buildings remains lacking, and their responses to future warming are poorly understood. It is also unclear whether urban renewal strategies can offset these climate-driven impacts and how effective they will be under future scenarios. These knowledge gaps hinder the formulation of evidence-based policies for facilitating climate-resilient and low-carbon urban transitions.

As China’s largest megacity, Beijing epitomizes the intertwined challenges of rapid urbanization, climate change, and growing energy demand. The city experienced a 22% surge in RECE between 2007 and 2015, with projections suggesting a further 30% rise in the near term. Critically, its per capita RECE is expected to be 2.3 times higher than the national average⁴¹. The widespread presence of aging residential buildings has exacerbated these pressures, prompting local authorities to prioritize renewal efforts, particularly in climate-vulnerable zones such as Hutongs, urban villages, and aging communities. The heterogeneous age structure of Beijing’s residential buildings results in spatially divergent patterns of RECE, consequently requiring data-driven approaches to prioritize renewal efforts. This highlights high-resolution RECE mapping under varying climate scenarios. Beijing represents an ideal testbed for developing scalable, climate-resilient building renewal strategies, considering its distinct residential building stock, urgent renewal imperative, climate vulnerability, and substantial emissions.

To fill these research gaps, this study compared climate-driven RECE increases with emission reductions achievable through residential building renewal strategies across multiple future climate scenarios. This study advanced the field by developing a modeling framework that integrated daily climate and electricity data to quantify the construction-era-specific Residential Electricity Carbon Footprint (RECF). This approach allowed us to identify age-dependent building vulnerabilities and carbon intensities under various climate scenarios. Leveraging this framework alongside detailed building and climate data, we then mapped the spatial distributions of RECF, RECE, and community-level RECI (community residential electricity carbon intensity) for both current conditions and future scenarios under four Shared Socioeconomic Pathways of SSP126, SSP245, SSP370, and SSP585. Furthermore, by evaluating three renewal pathways—retrofitting to 2000 construction standards (2000CS), ultra-low energy buildings (ULEB), and near-zero energy buildings (NZEB), this study quantified their climate-scenario-dependent mitigation potentials and cost-effectiveness via cost-benefit analysis. Collectively, this study can provide fine-scale evidence of how residential building renewals can simultaneously achieve electricity savings and long-term carbon mitigation; besides, it can also offer a transferable analytical basis for equitable, climate-resilient urban transitions.

Results

Temperature and building construction era strongly influence RECE

The Residential Electricity Carbon Footprint (RECF) demonstrated a robust nonlinear correlation with outdoor air temperature across construction eras, displaying significant seasonal variations (Fig. 1 and Supplementary Fig. 1). During Beijing’s cold winters, RECF remained minimal (0.03–0.05 kg CO₂/m² per day) due to predominant fuel-based heating, while summer months (June-August) showed dramatic spikes in electricity consumption driven by cooling demand. Notably, summer cooling demand increased by 40% of residential electricity consumption, contributing to over one-third of the annual RECE. Higher temperatures had a more pronounced effect, with temperatures above 25 °C triggering exponential RECF growth. Strikingly, the hottest 20% of days contributed to 30% of the annual RECF, while transitional months (April–May, September-October) maintained stable baseline RECF levels that decreased slightly below 20 °C. Summer also exhibited the greatest daily RECF variability, reflecting heightened climate sensitivity during extreme heat events.

Fig. 1: The relationship between RECF and outdoor air temperature for different construction eras.

The four construction eras include pre-1980s, the 1980s, the 1990s, and the post-2000s. a Daily RECF response to outdoor air temperature for different construction eras (results for post-2000 residential buildings are shown; others are in Supplementary Fig. 1; all parameter values in Supplementary Table 3). b Distribution of daily RECF in each month for post-2000s residential buildings. c Monthly RECF for different construction eras. d Distribution of daily RECF at different temperature ranges for residential buildings of different construction eras.

Significant daily and monthly RECF heterogeneities existed across residential building construction eras, with the pre-1980s residential buildings showing 2–4 times higher values than post-2000 buildings. These disparities were amplified during high-temperature days and summer months, with post-2000 buildings reaching daily RECF peaks of 0.31 kg CO₂/m² on extreme hot days (>30 °C) and monthly RECF peaks at 7.1 kg CO₂/m² in July. The pre-1980s residential buildings exhibited 27% higher RECF values during these hot months compared to the post-2000s residential buildings. Such climate vulnerability in aging buildings, characterized by outdated designs and materials that necessitate excessive cooling energy, highlights the critical need for prioritized renewal strategies to mitigate both carbon emissions and energy inequities in Beijing.

Spatial heterogeneity of RECE across Beijing

Beijing’s Residential Electricity Carbon Emissions (RECE) exhibited pronounced spatial heterogeneity strongly linked with residential building construction eras (Fig. 2, Supplementary Fig. 2, and Supplementary Fig. 3). The total annual RECE in Beijing accounted for 22.2 Mt CO₂, with 32% occurring during hot summer months (Fig. 2a). Striking disparities of annual RECE emerged across building construction eras. The pre-1980s buildings represented only 1.6% of the total building stock, but they accounted for 5.6% of annual RECE due to higher RECF values. Conversely, post-2000s buildings comprised 67.5% of the construction area but contributed to 55.1% of annual RECE. This paradox reflects fundamental urban morphology differences. While aging low-rise communities like Baijiatuan Village generated 100 t RECE per building annually, modern high-rises in Taiyanggongyuan Community exceeded 300 t RECE per building. The high-rise residential building clusters inside the 5^th Ring Road collectively emitting up to 5000 t CO₂ of RECE annually, due to their larger construction areas (Fig. 2b). These patterns underscore how building height and density modulate emission impacts despite efficiency advantages in newer buildings.

Fig. 2: Seasonal variation of RECE and their proportion across different building heights and construction eras within different ring roads in Beijing.

a Daily average RECE for each month under different seasons in Beijing. b RECE proportion across different building heights within different ring roads, and the heat map on the right shows the total RECE. c RECE proportion across different building eras within different ring roads, and the heat map on the right shows the total RECE.

Spatial analysis revealed two distinct emission regimes in Beijing (Fig. 2b, c). Residential buildings within the 3^rd Ring Road, comprising 15% of construction area, contributed to 19.5% of annual RECE, with pre-1980s residential buildings (56% of emissions) and low-rise buildings (36%) being disproportionately impactful. Beyond the 4^th Ring Road, a contrasting pattern dominated: post-2000s high-rises (>18 floors) covering 69% of residential area emitted 14.2 Mt CO₂ annually (64% of city total). Given space constraints, prioritizing taller buildings was more favorable for reducing carbon emissions and commuting time.

High spatial heterogeneity of community-RECI

Community Residential Electricity Carbon Intensity (community-RECI) represented remarkable spatial heterogeneity across Beijing’s 54,000 grid cells (100 × 100 m resolution), with the average of 43 kg CO₂/m² and the largest of 644 kg CO₂/m² (Supplementary Fig. 4). Statistical analysis confirmed highly significant differences in both RECE and RECI distributions among communities (p-value < 0.01, Supplementary Fig. 5). This heterogeneity primarily reflected the complex interplay between construction eras, community landscapes and green infrastructures (Fig. 3). The traditional low-rise aging community, like urban villages and hutongs, paradoxically showed elevated community-RECI despite modest per building RECE, due to their extreme building density and limited green space (e.g. the Baijiatuan Village at 16.7–69.8 kg CO₂/m², 3.7% above the city average). Middle-aged communities exemplified by Youyi Community demonstrated the peak RECI values at 82.3 kg CO₂/m² (ranging from 10.8 to 330.5 kg CO₂/m²), nearly doubling other communities, resulting from dense middle-rise buildings with minimal amenities. In contrast, newer high-rise residential communities showed how strategic buildings can mitigate RECE. While the per building RECE can reach 532 t CO₂ (2.8 times of the city average) across Tiantongyuan Community, the largest residential community in Beijing with over 700,000 residents, its community RECI was only 40% above the average, thanks to its extensive green and recreational spaces. The Taiyanggongyuan Community further demonstrated this principle, achieving a RECI of 38.9 kg CO₂/m² (only 47.3% of Youyi Community’s value) through vertical growth coupled with ecological design, proving that building height and green infrastructure can synergistically reduce community RECI despite increased absolute carbon emissions.

Fig. 3: Residential electricity carbon intensity across different residential communities.

Distinguished by construction era, building height and internal conditions, four typical communities were studied: The Low-aging Community with pre-1980s buildings (Baijiatuan Village), the Middle-aging Community with 1980s and 1990s buildings (Youyi Community), the High-Rise Community with Green Space and post-2000s buildings (Tiantongyuan Community) and the High-Rise Community with post-2000s buildings (Taiyanggongyuan Community). The location of the four communities in Beijing was presented in supplementary Fig. 9. a The street and satellite images of the four studied residential communities. The community name and its average floor were written above. The street and satellite images are freely available to the public (https://map.baidu.com). b The 3D bar chart of construction eras. c The 3D bar chart of RECE. d The 3D bar chart of RECI.

Climate change increases future RECE in Beijing

Our multi-model ensemble analysis, incorporating 22 general circulation models (GCMs) under different SSP scenarios (2021-2100), projected distinct future RECE trajectories for Beijing (Fig. 5a). Monte Carlo simulations confirmed model robustness (p-value < 0.001), though identified temperature sensitivity and building age as key uncertainty sources(Fig. 4). Projected annual RECF exhibited scenario-dependent patterns: an “invert-U” trajectories under SSP 126 with post-2075 declines, versus monotonic increases in other higher-emission scenarios (Fig. 5b). By 2100, SSP-126 could limit RECE growth to 3.4% ~ 6.1% above current levels, while SSP-585 could drive pre-1980s buildings to 78.8 kg CO₂/m² annually, underscoring aging buildings’ climate vulnerability and the 4.6–41.9% mitigation potential achievable through low-carbon pathways.

Fig. 4: Uncertainty analysis of simulated RECF projections in 2100.

We compare the RECF results in 2100 of all the 22 climate models across construction eras and future climate scenarios. The height of the bar chart presented the average value of the 22 models and the error bar on the top presents the differences between models. The specific values of the 22 models are displayed in a dot plot next to error bar.

Fig. 5: Projected annual RECE and RECF under different SSPs.

a The historical and future annual total RECE in Beijing from 2000 to 2100. b The future annual RECF for different construction eras from 2021 to 2100.

All scenarios showed moderate RECE increases until 2060, followed by dramatic divergence. SSP126 projections peaked at 23.64 Mt around the 2070 s before declining (Supplementary Fig. 6), yielding cumulative savings of 52 ~ 212 Mt carbon emissions between 2021 and 2100, equivalent to 2 ~ 10 years of current RECE in Beijing. Despite associated economic costs, the SSP126 pathway could generate $233 ~ 935 million of direct electricity savings, along with $41 ~ 137 million of carbon mitigation at current carbon price of $7.84/t CO₂. Conversely, the annual RECE was projected to increase by 21% ~ 52% in Beijing under SSP585, with post-2000s buildings accounting for 58.7% ~ 64.3% of this increase. Additionally, SSP585’s projected 74 additional extreme hot days (≥30 °C) could contribute 12.5 Mt carbon emissions, equivalent to 56% of Beijing’s current annual RECE.

Projected emission increases varied geographically (5.4–45.7%, Fig. 6), revealing two primary concern areas: (1) aging communities within the 3^rd Ring Road facing 3.4–20.4% RECE growth due to elevated RECF, and (2) new high-rise communities beyond the 5^th Ring Road projected to contribute 15 Mt carbon emissions annually by 2100 under SSP585. Extremely high-rise (>18 floors) buildings outside the 5^th Ring Road emerged as particularly critical future emission sources, though their per-area intensity remained lower than aging core districts.

Fig. 6: Future RECEs within each Ring Road in 2100 under four SSPs.

a The RECE proportion across different building heights within different ring roads in Beijing. b The RECE value across different building heights within different ring roads in Beijing. c The RECE proportion across different construction eras within different ring roads in Beijing. d The RECE value across different construction eras within different ring roads in Beijing; Each quarter circle shows the result of one scenario. The location of different ring roads in Beijing is presented in Supplementary Fig. 7d.

Residential building renewal can offset the climate change impact on RECE in Beijing

All three renewal strategies (the 2000CS, ULEB, and NZEB) can effectively reduce residential electricity consumption compared to aging buildings, while differentially mitigating climate change impacts on RECE. Under the SSP126 and SSP245 scenarios, all strategies showed positive effects from 2050 onward (Fig. 7a), yet only the NZAB renewal strategy can completely neutralize climate change impacts by 2100 under SSP585. Full implementation could reduce Beijing’s annual RECE by 4.2–9.8 Mt CO₂ (13.7–35.9% of current levels) by 2100, with cumulative SSP126-period reductions reaching 230.2–1867.8 Mt CO₂ (equivalent to 10-85 years of current emissions). These results demonstrated that while all renewals offer benefits, only NZEB can provide sufficient resilience against worst-case climate scenarios.

Fig. 7: Annual RECE under different building renewal strategies and its associated RECE mitigation performance compared with Climate Change scenario (CC) under different SSPs.

a The future annual RECE of the CC scenario and three renewal strategies under different SSPs. The upward red arrow indicates more carbon emissions than current level, and the downward blue arrow indicates that residential building renewal can completely counteract the negative impact of climate change on RECE. b Shows the “competition” between CC scenario and different renewal strategies under different SSPs. The grey dotted line shows the RECE increment of each year compared with present level. The other color dotted lines show the RECE reduction compared with CC scenarios. If the color dotted line is above the grey dotted line, it means that the RECE reduction due to residential building renewal is higher than the RECE increment due to climate change, and thus the residential building renewal can counteract the effect of climate change. The “winner” was marked in red, and it means that NZEB can counteract the climate change effect under all SSPs; ULEB can counteract the climate change effect in SSP126, SSP245, and SSP370; 2000CS can only counteract the climate change effect in SSP126 and SSP245.

The 2000CS strategy can temporarily offset RECE increases during 2060-2080 under SSP370, but proved inadequate post-2080 as climate impacts intensified. The ULEB strategy offers superior but still incomplete carbon mitigation, failing to fully compensate climate change impacts after 2086 under SSP585. In contrast, the NZEB strategy could achieve substantial mitigations (11.85 Mt CO₂ by 2100 under SSP585; 384.4 Mt CO₂ cumulative), through its long-term efficacy required reassessment beyond 2100. These findings underscore that while residential building renewal can significantly mitigate RECE growth (particularly NZEB), carbon-neutral policies remain essential to achieve China’s dual carbon goals. This is especially evident under high-emission scenarios, where even advanced strategies yield diminishing returns over time.

Discussion

The advanced low-carbon technologies in ULEB and NZEB renewal strategies could incur substantial upfront costs, requiring $97–116 billion more than the conventional 2000CS strategy for Beijing’s entire building stock, posing a formidable barrier for China’s low-carbon transition⁴². Globally, innovative financing mechanisms could demonstrate scalable solutions. The EU’s Renovation Wave Strategy allocates €672.5billion (37% for climate action) to address its €275 billion/year renovation gap, while the U.S. Property Assessed Clean Energy financing (PACE) program’s tax-based repayment system boosts retrofit adoption by 20–30%⁴³. However, without targeted subsidies, these programs risk exacerbating energy inequality, as low-income households show 40% lower participation rates^44,45. The large-scale renewal of China’s residential building stock must be underpinned by coherent and evolving policy and governance frameworks. Recent national strategies have demonstrated significant momentum: the 14th Five-Year Plan for Building Energy Efficiency and Green Building Development has explicitly promoted the ultra-low and near-zero energy buildings, while with more recent Action Plan for Accelerating Energy Conservation and Carbon Reduction in the Building Sector in 2024 has set a definitive 2027 target for their nationwide scaling. To translate this top-level ambition into local action, future efforts must align national carbon-neutrality goals with municipal urban planning imperatives. Thus, a multi-level governance approach should integrate targeted financial incentives, dedicated renovation subsidies, and community-centric retrofit programs to ensure an effective and equitable low-carbon transition across all regions.

Despite massive initial costs, the ULEB and NEZB strategies can yield substantial returns^46,47: projected SSP585 mitigation of 256/384 Mt CO₂ by 2100 generates $2.03/$3.02 billion at Beijing’s current $7.84/t CO₂ price, albeit just 0.2–0.3% of total costs⁴⁸. This ratio transforms dramatically under projected carbon price escalation ($66.5/t CO₂)^49,50,51, with benefits reaching $125–187 billion to fully offset renewal expenses. Comparative analysis reveals stark carbon pricing disparities: from $5–9/t CO₂ across Chinese regional markets (2023 average: $8.00) to Europe’s €109/t CO₂ peak⁵². When coupled with anticipated 18–25% cost reductions from technological learning⁵³, the economic case for renewal strengthens considerably, though spatial carbon price heterogeneity necessitates localized feasibility assessments.

Beyond direct carbon and economic gains, residential building renewal can further deliver substantial co-benefits across public health and social equity. Improved insulation and ventilation systems can improve thermal comfort and indoor air quality, while reducing hospitalizations for respiratory and cardiovascular conditions by 9–20%⁵⁴, with pediatric asthma admissions decreasing 21% in retrofitted households⁵⁵ - benefits particularly crucial for vulnerable demographics (children, elderly, and patients with chronic conditions). Targeted renewal interventions in disadvantaged communities can demonstrate even broader impacts, alleviating energy poverty by 40% through reduced utility costs, while simultaneously improving indoor environmental quality and household economic resilience^33,56. These findings position residential building renewal as a transformative intervention that concurrently addresses climate mitigation and social inequality. The compounded benefits - spanning public health improvements, energy poverty reduction, and economic savings - can prove especially impactful for marginalized populations worldwide, making urban renewal programs powerful tools for achieving both environmental sustainability and the UN Sustainable Development Goals.

Rapid urbanization in China has left 160,000 aging residential communities, which means that many aging residential communities require urgent renewal, with high energy intensity, outdated infrastructure, poor facilities, and acute vulnerability to climate extremes^57,58. These communities face compounding risks as urban heat island intensification exacerbates thermal discomfort^59,60. A threat exemplified by the 2010 Russian heatwave, which resulted in 55,000 deaths in similar aging communities⁶¹. Beijing exemplifies this challenge, as its aging communities are home to millions of residents and exhibit 2.3 times higher RECI than newer buildings⁴¹. Without intervention, these districts will become major carbon emission hotspots as cooling demands escalate under climate change^19,62,63. Beijing’s ongoing renewal of 301 aging communities, covering 7 million m² of construction area⁶⁴, demonstrated scalable solutions.

Our estimate showed the NZEB renewal strategy could mitigate 11.85 Mt CO₂ by 2100 under the SSP585 scenario, equivalent to a reduction of 10.10 kg CO₂ per m². Nationally, renewing China’s 3.5 billion underperforming residential units through NZEB renewal approaches may annually yield 35.36 Mt CO₂ reductions (53.4% of current aging-building RECE)^65,66. This aligned with global evidence showing 420 m² renewals can save 4508 kWh/year while cutting 4243 kg emissions⁶⁷, confirming renewals as among the most cost-effective residential decarbonization measures^68,69, especially in tropical regions^70,71. Although calibrated for Beijing, our analytical framework can inform assessments in other megacities confronting similar climate-energy pressures. Cities in temperate zones like Tokyo and Seoul can share comparable U-shaped temperature-electricity response functions^72,73. Extending this approach globally, future research could integrate building datasets (like the Global Building Atlas⁷⁴) with regional climate and socioeconomic data to quantify spatial variations in RECE and their responses to future climate change. Such efforts would enable systematic cross-city comparisons across climate zones, building typologies, and governance systems, substantially enhancing our understanding of global climate-energy-carbon feedbacks.

Emerging solutions like building-integrated photovoltaics (BIPV) face adoption barriers (costs, aesthetics)⁷⁵, while urban green spaces demonstrate dual benefits: mitigating heat islands through evapotranspiration and sequestering 11,972.08 (trees) to 5758.07 (shrubs) Mg CO₂ per ha over a 50-year period⁷⁶. Our findings advocated combining architectural renewals with green infrastructure - a synergistic approach that addresses both operational emissions and urban microclimates, creating a robust pathway toward China’s carbon neutrality goals.

Our RFCF findings aligned with previous studies while revealing methodological-driven variations (Supplementary Table. 5). Our results fall within the range reported by both Cai et al. and Beijing Civil Building Energy Consumption Report^77,78. However, the primary difference with Cai et al. stemmed from their reliance on detailed statistical data on residential electricity consumption⁷⁷, whereas Beijing Civil Building Energy Consumption Report is based on annual electricity consumption data from 1666 residential buildings⁷⁸. Notably, Xu et al.’s slightly higher results reflect their inclusion of electric vehicle charging loads - a factor excluded from our residential-sector focus⁷⁹. These comparative analyses underscored how methodological choices shape emission estimates.

Through triangulation with official statistics (Beijing Statistical Yearbook) and gridded consumption data from Chen et al.⁸⁰, our modelled 2015–2019 RECE results showed strong concordance (±10–15% variance, Supplementary Table. 6). This robust agreement across independent datasets, including macro-level annual reports and micro-scale spatial data, confirms the model’s empirical validity for Beijing’s context. The consistent <15% deviation threshold suggested reliable predictive capacity for policy applications.

The applicability of our Beijing carbon mitigation insights to other megacities is still contingent upon some critical urban characteristics. (1) Building typologies, including age, height, construction materials, and insulation standards: pre-1960 European buildings (representing >35% of inventory) demonstrate fundamentally different thermal performance than China’s modern high-rises due to inferior insulation standards⁸¹. (2) Urban morphology: Chinese studies confirmed that compact polycentric forms could reduce per capita emissions by 12-20% compared to sprawling configurations^{82,83,84,85,86}. (3) Local climate-energy interactions: These multidimensional variations necessitated careful calibration of our modeling framework to regional building typologies (age distributions, construction materials), spatial organization patterns, and microclimate conditions to ensure policy-relevant outcomes. We consequently advocate for city-specific integrations of building energy inventories, urban form analytics, and grid decarbonization trajectories in future studies.

Despite these strengths, several important sources of uncertainty remain in our analysis. While we quantified model uncertainty using 22 climate models and Monte Carlo simulations, our analysis demonstrates that the uncertainty of RECF is strongly influenced by building age and future climate scenarios. Under more extreme warming scenario (SSP585), pre-1980s buildings exhibit the highest mean RECF and the greatest uncertainty (79.1 ± 6.47 kg CO₂/m²), with a standard deviation nearly three times higher than low-emission scenarios (Fig. 4). However, other unmodeled factors would introduce additional variability: occupant behavior, policy-technology interactions, building occupancy dynamics, and demographic-microclimate couplings. Ürge-Vorsatz et al. found that differences in occupant behavior alone can account for 50% of the observed variability in residential energy use⁸⁷. Li et al. demonstrated that policy uncertainty and climate variability could lead to fluctuations in energy consumption of over 20% in the short term and larger deviations in the long term⁸⁸. This underscored the need for next-generation models incorporating stochastic human dimension parameterizations, policy scenario ensembles, and integrated demographic-energy modules to better constrain decarbonization pathways.

Our findings pointed out that temperature proved to be the dominant climatic determinant, and RECE was more sensitive during hotter months, and previous studies also confirmed the significant impact of temperature on residential electricity consumption^{72,73,89,90,91,92,93}. However, we recognize this approach simplifies complex meteorological interactions, such as humidity, solar radiation, cloud cover, and precipitation^72,94. For example, Ihara et al. revealed that air humidity could impact residential electricity consumption by 0.6–0.9 (W/floor-m²)/(g/kg)⁷², while Kang and Reiner observed increased electricity consumption during midday periods on rainy days⁹⁴. Though humidity effects proved marginal in Beijing’s temperate climate, we advocate for expanded meteorological parameterization in future studies to enhance global applicability, especially for coastal megacities experiencing combined heat-humidity extremes.

Our daily residential electricity consumption data collected from Haidian district, which aligns closely with the citywide average in per-household electricity consumption. Though Haidian District’s electricity data mirrors Beijing’s household average and encompasses diverse building stocks, three urban typologies require deeper investigation: (1) historic cores (Dongcheng/Xicheng) with unique thermal mass properties⁶⁰; (2) sprawling suburbs (Tongzhou) facing distinct retrofit challenges; and (3) traditional neighborhoods like Dashilar where microclimate-building interactions alter energy demand patterns by 15–30%⁶⁰. These variations underscore the necessity for multi-district validation using high-resolution urban energy data and microclimate-coupled energy models, when they are available.

Our analysis is constrained by its 2015–2019 temporal scope, deliberately chosen to avoid COVID-19 pandemic distortions while capturing recent consumption patterns. However, projecting electricity demand to 2100 introduces substantial uncertainties beyond our climate and socioeconomic scenarios. Although our scenario incorporates alternative future climate and socioeconomic pathways, other disruptive factors require consideration, including smart home technological advancement, aging population dynamics altering thermal comfort needs, appliance efficiency revolutions, and lifestyle transformations. Celik et al. showed that the smart home energy management can reduce neighborhoods peak residential electricity loads by 12% to 25% and lower total household electricity consumption by 10% to 17%⁹⁵. Sabrina et al. examined that the adoption of smart home systems in Saudi Arabia can lead to even 30% in household electricity use during peak periods⁹⁶. To address these, we advocate for dynamic Urban Building Energy Modeling (UBEM) - an approach validated by Lin et al. and Li et al. for its capacity to integrate technological innovation, behavioral adaptation, and policy impacts within urban energy systems^97,98,99.

Another important consideration concerns the interaction between building-level mitigation and power grid decarbonization. Under China’s national development scenario, the carbon intensity of the regional grid is projected to decline to 0.144–0.239 kg CO₂/kWh by 2050¹⁰⁰. Incorporating this decarbonization trajectory would further reduce residential building electricity carbon emissions, underscoring the need for future work to integrate diverse socioeconomic and grid decarbonization scenarios.

The study’s scope excluded operational emissions from residential natural gas consumption and embodied carbon from construction materials and processes. Electricity dominates the carbon emission for appliances, cooling, and lighting^101,102,103, and studies have confirmed its greater emission share compared to the morning-peak gas use¹⁰⁴; however, a full-cycle assessment is still demanded by integrating daily gas data and life-cycle assessment. Furthermore, as reinforced concrete engineering can embody up to 90% of construction-phase GHG emissions^105,106, future studies should incorporate embodied carbon. Moreover, this study also did not account for emissions embodied in electricity trade, particularly relevant for import-dependent regions like Beijing where 18–22% of power is grid-transferred¹⁰⁷. Given that 8–10% of global power-sector emissions are virtually traded¹⁰⁸, with interregional transfers displacing 15–20% of reported emissions in the U.S¹⁰⁹, future climate responsibility frameworks must develop network-based emission tracking systems and transnational compensation mechanisms to ensure equitable burden-sharing.

Methods

Residential building dataset in Beijing

We collected data on 430,014 building footprints with height information and 7567 residential building community footprints in Beijing for the year 2020 using the Amap (https://ditu.amap.com/). Besides, we also gathered the Point of Interest (POI) data from the website (www.udparty.com), which provided detailed information on building ages. This allowed us to compile a comprehensive dataset that includes information on community location, building age, construction area, floor count, and residential land area.

Based on architectural characteristics¹¹⁰, we categorized the residential buildings into four construction eras: the pre-1980s (built before the 1980s), the 1980s (built in the 1980s), the 1990s (built in the 1990s) and the post-2000s (built after 2000). In Beijing, distinct construction eras are strongly characterized by shifts in dominant building materials and thermal insulation standards. The pre-1980s residential buildings were constructed under China’s welfare-oriented housing system¹¹⁰ and typically featured solid clay brick exterior walls and fly ash ceramic roofs. These buildings from this era, commonly found in hutongs and urban villages, were largely uniform in design, featuring small dwelling units and now-outdated materials. The 1980s and 1990s residential buildings were primarily constructed with brick-concrete materials, such as clay bricks and lime mortar⁶⁵. Compared to earlier structures, they offered larger floor areas and usually included two or three bedrooms. From the 2000s onward, reinforced concrete became dominant, enabling the high-rise residential, high-density residential communities stipulated by the national Ministry of Housing and Urban–Rural Development. Consequently, the construction era can serve as a robust proxy for gauging spatial variation in building envelope performance and material composition across Beijing. Additionally, residential buildings were classified by height into four categories⁶²: the low-rise (≤3 floors), the middle-rise (3 ~ 7 floors), the high-rise (7 ~ 18 floors), and the extreme high-rise (>18 floors).

Beijing exhibited significant spatial heterogeneity in residential buildings (Supplementary Fig. 7). Aging residential buildings of the pre-1980s and the 1980s were primarily concentrated within the 4^th Ring Road, while newer post-2000s residential buildings were mainly located outside the 4^th Ring Road, with 61.7% situated beyond the 5^th Ring Road. The post-2000s residential buildings were generally higher, with a predominance of medium- and high-rise structures. Extreme high-rise residential buildings, which accounted for 52% of Beijing’s total residential construction area, were mostly located beyond the 5^th Ring Road, but these buildings only occupied 23.3% of the total residential land area. Notably, 3/4 of these extreme high-rise residential buildings were constructed after 2000. This heterogeneity in Beijing’s residential building stocks can enable a comparative analysis of RECE across both construction eras and building heights.

Daily residential electricity consumption dataset

From the power and residential department, we collected the database of daily household residential electricity consumption from 2015 to 2019. The choice of this period was motivated by both data availability and the intention to avoid the atypical electricity consumption patterns that emerged with the onset of the COVID-19 pandemic in 2020. These patterns may reflect behavioral anomalies that are not representative of long-term trends. This dataset included information from 2026 households across 31 residential buildings, with a total construction area of 176,962 m² in the Haidian District. These 31 residential buildings can represent all above four construction eras: three pre-1980s buildings (108 households and 7074 m² construction area), 14 buildings from the 1980s (770 households and 44,150 m² construction area), 11 buildings from the 1990s (806 households and 68,900 m² construction area), and three post-2000s buildings (342 households and 56,838 m² construction area). The pre-1980s residential buildings were uniformly built under the welfare-oriented housing system, resulting in minimal variation in housing attributes. Despite only three pre-1980s buildings being included in our dataset, the daily data from 108 households can adequately capture the relationship between electricity consumption and temperature, as confirmed by subsequent analysis. The post-2000s residential buildings tended to be taller with larger construction areas; even though only three buildings from this era were represented in our dataset, they accommodated 342 households. Thus, our collected daily electricity consumption data was sufficient to support subsequent functional analysis.

To ensure our results are representative of Beijing, we carefully assessed the characteristics of Haidian District in relation to the broader city. Haidian’s average annual per-household electricity use (3833 kWh/yr) is nearly identical to the citywide mean (4039 kWh/yr), with only a −5.1% deviation. This is the smallest deviation observed among all districts. Geographically, Haidian spans from the 2nd to the 6th Ring Roads, encompassing central, midtown, and peripheral neighborhoods, and is home to nearly 3.13 million residents. Its built stock is highly diverse, including a wide range of construction eras (from pre-1980s to post-2000s) and building height (low-, mid-, and high-rise), as well as a substantial stock of 6.47 millionm² of aging residential buildings currently undergoing renewal. Besides, the district’s geography, which includes both plains and hills, further reflects the overall urban and topographic complexity of Beijing. Given these factors, our daily residential electricity consumption dataset from Haidian district serves as a representative sample, enabling reliable analysis of electricity consumption patterns and policy implications at the city scale.

Residential electricity carbon footprint, emission, and intensity

To account for variations in residential electricity consumption across households, we employed the Residential Electricity Carbon Footprint (RECF) to capture differences across construction eras. The RECF was defined as the residential electricity consumption per unit of construction area (Fig. 8). Additionally, we adopted a 100 × 100 m (10,000 m²) grid resolution for calculating the Residential Electricity Carbon Intensity (RECI) for each community, in accordance with official spatial management standards. The “Grid based urban management system-Basic management grid division” issued by the Beijing Municipal Commission of Urban Management and national standard “Information system for digitized supervision and management of city-Part 1: Basic management grid” both specify that the grid size for central city districts should also be about 10,000 m². We defined the RECI as the ratio of the total residential electricity carbon emissions from all buildings within a grid cell to this grid cell area (Fig. 8). These standards support our use of a 100 × 100 m grid, which is well matched to typical community sizes and is widely used in municipal planning and management in Beijing.

Fig. 8: The concept of primary indicators used in this study.

The schematic illustrates the relationship between building characteristics, electricity consumption, and carbon emission indicators in urban residential communities. The residential communities’ type was marked in different colors: Green: High-rise community with well-developed green spaces and recreational infrastructure; Orange: Low-rise, high-density community with limited amenities; Blue: Mid-rise community with moderate population density and little green space.

For carbon emissions, we adopted the carbon emission coefficients from the “China Regional Grid Carbon Dioxide Baseline Emission Factor (OM)”, published by the Ministry of Ecology and Environment of China in 2019¹¹¹. These coefficients were derived using the methodology outlined in the “Tool to calculate the emission factor for an electricity system” by the Clean Development Mechanism of the United Nations¹¹². For 2019, we utilized the weighted average of emission coefficients from the preceding three years, adjusted according to the annual electricity output of the grid. Given that our dataset spans from 2015 to 2019, we applied the 2019 carbon emission coefficient for the North China region of 0.9419 tCO₂/MWh in our study. This coefficient considers inter-regional electricity trade (Supplementary Table 4), making it suitable for areas like Beijing that rely heavily on electricity imports. It was derived from net power generation, fuel types, and fuel consumption across all power plants within the system, which can be calculated using Eq. (1).

$$E{F}_{grid,y}=\frac{{\sum }_{i}(F{C}_{i,y}\times NC{V}_{i,y}\times E{F}_{C{O}_{2},i,y})}{E{G}_{y}}$$

(1)

$E{F}_{grid,y}$ is the carbon emission coefficient in year y (tCO₂/MWh); $E{G}_{y}$ is the total net power generation in year y; $F{C}_{i,y}$ is the total consumption of fuel i in year y; $NC{V}_{i,y}$ is the Low heating value of fuel i in year y; $E{F}_{C{O}_{2},i,y}$ is the carbon emission factor of fuel i in year y. This is from the IPCC Guidelines for National Greenhouse Gas Inventories in 2006; i is the type of fossil fuels consumed in the power system. Thus, the RECE can be calculated using Eq. (2).

$$REC{E}_{jt}=R{E}_{jt}\cdot \alpha$$

(2)

where RECE_jt is carbon emissions due to daily residential electricity consumption of building j on day t; RE_jt is the daily residential electricity consumption of building j on day t; α is the carbon emission coefficient.

To model RECE across different construction eras, we established the response functions based on daily outdoor temperature and corresponding RECF. The construction area of each building was calculated by multiplying the land area by the number of floors. The land area corresponded to the footprint area of the building. Thus, the RECF can be calculated using Eq. (3).

$$REC{F}_{jt}=\frac{REC{E}_{jt}}{{l}_{j}\cdot {h}_{j}}$$

(3)

where RECF_jt is the residential electricity carbon footprint of building j on day t; l_j is the land area of building j; h_j is height floor of building j.

To capture differences in residential communities, we used RECI to represent variations across different communities (Fig. 8). We generated a comprehensive set of 54,329 grid cells in Beijing. Consequently, the community RECI can be calculated using Eq. (4).

$$REC{I}_{kt}=\frac{REC{E}_{kt}}{gc}$$

(4)

where RECI_kt is the residential electricity carbon intensity of grid cell k on day t; RECE_kt is the carbon emissions of grid cell k on day t; gc is area of grid cell.

Temperature data and future scenarios

Our temperature dataset included current and future temperature projections. We collected current daily outdoor air temperature from 2015 to 2019 from the National Centers for Environmental Information (https://www.ncei.noaa.gov/). To match the locations of the residential buildings where we had collected daily electricity consumption data, we selected the nearest meteorological station. The current daily temperature data were then used to establish the temperature-RECF functions.

For future projection, we relied on data from the sixth phase of the Coupled Model Intercomparison Project (CMIP6), as outlined in the 6th IPCC assessment report (IPCC AR6). The CMIP6 provides climate projections under various Shared Socioeconomic Pathways (SSPs), and we used temperature projections from 22 general circulation models (GCMs) (further information in Supplementary Table 7). However, due to the relatively coarse spatiotemporal resolution of the CMIP6 climate models, directly applying them in calculations presents challenges. Additionally, the varied spatial resolutions of the 22 GCMs’ raw data introduce further uncertainty. To address this issue, we applied the NWAI-WG statistical downscaling method, developed by Liu and Zuo¹¹³, to uniformly downscale the raw data. Previous studies have successfully demonstrated the feasibility and accuracy of this method^{114,115,116,117}. Parts of future temperature data were presented in Supplementary Table 8.

Four future climate scenarios were selected, which combined the four SSPs and Representative Concentration Pathway (RCP): SSP126, SSP245, SSP370, and SSP585. Each SSP represented different development pathways of economic development, welfare, and ecological conservation. The four RCPs corresponded to different greenhouse gas concentration scenarios. We employed the RCP2.6, RCP4.5, RCP6.0, and RCP8.5 scenarios to explore a range of possible future climate scenarios. RCP2.6 represents a “very stringent” mitigation scenario, aiming to keep global warming well below 2°C above pre-industrial levels. RCP 4.5 RCP4.5 is an intermediate scenario, while RCP6.0 predicts a peak in emissions around 2080 followed by a decline. RCP8.5, on the other hand, represents a “business-as-usual” scenario, with continued emissions growth throughout the century. These combined scenarios are commonly used in climate studies and can help guide future development^118,119.

Modelling temperature-residential electricity carbon footprint functions

Amid evolving economic levels, technological progress, lifestyle changes, and socioeconomic factors play pivotal roles in determining residential electricity consumption. For instance, Jia et al. reported that higher incomes were associated with increased residential electricity consumption for services like laundry, while larger households drove higher lighting demands¹²⁰. By contrast, basic electricity uses such as entertainment, refrigeration, and cooking accounted for more than half of total annual residential electricity consumption and were less susceptible to socioeconomic variations¹²⁰. On the contrary, climatic factors and building characteristics emerged as significant determinants of residential electricity consumption^{72,73,89,90,91,92,93}. In developing our residential electricity carbon footprint (RECF) models, we carefully considered the potential influence of multiple climatic variables, including air temperature and humidity. To guide variable selection with empirical evidence, we conducted a correlation analysis between daily RECF and both humidity and temperature, using our daily electricity consumption and meteorological datasets from 2015 to 2019. As shown in Supplementary Fig. 8a–d, RECF across all building construction eras in Beijing exhibits only a weak correlation with humidity (Pearson’s r < 0.3; low R² values for both linear and polynomial fits). In contrast, Supplementary Fig. 8e–h reveals a pronounced “U-shaped” association with temperature, with Pearson’s r values around 0.4 and much higher R², particularly for polynomial fits. This robust relationship holds across all building eras. These empirical findings support our decision to focus on temperature as the principal climatic driver in our main modeling framework. We attribute the relatively minor effect of humidity to Beijing’s dry climate, with annual average humidity (~50%) significantly lower than in many tropical/subtropical cities, and a downward trend over recent decades (declining by 0.9% per decade overall, and 1.66% per decade in urban areas from 1976–2015)¹²¹. This context reduces the expected impact of humidity on electricity demand compared to wetter regions. Therefore, our analysis focused on temperature and construction eras, with the assumption that the relationship between temperature and RECF follows a consistent U-shaped curve across different construction eras.

The U-shaped relationship between temperature and RECF across all building construction eras revealed that the electricity usage decreased with rising temperature up to a certain point, then increased rapidly at higher temperatures. This pattern reflects the real-world shift from heating to cooling demand as temperatures move beyond the comfort range. Previous studies indicated that most Beijing residents perceive thermal comfort between 18 °C and 22 °C^117,122,123, and that air conditioner usage typically begins around 20 °C. Our empirical data also show a pronounced increase in electricity usage above this point, coinciding with the onset of the cooling season in Beijing (typically mid-May to mid-September). Thus, 20 °C serves as a logical and evidence-based segmentation threshold.

To capture the empirical relationship between temperature and residential electricity carbon footprint (RECF), we tested several modeling approaches, including linear, polynomial, and logistic regression. Model fitting results (Supplementary Fig. 8i–l and Fig. S9m–p) showed that for temperatures below 20 °C, linear regression achieved high R² values and Pearson’s r, indicating a robust linear relationship. For temperatures above 20 °C, logistic regression provided the best fit, with R² values ranging from 0.73 to 0.82, compared to polynomial and linear regression. Notably, the superiority of logistic regression was most evident for post-2000s buildings (R² 0.82 for logistic and 0.78 for polynomial). While polynomial and logistic regression models performed similarly for pre-1980s, 1980s, and 1990s buildings, logistic regression is preferable from a behavioral perspective, as it captures the saturation of electricity use at high temperatures when most households operate cooling systems at full capacity. In contrast, polynomial models may overestimate consumption increases at temperature extremes.

Based on this evidence, we adopted a segmented modeling approach with a 20 °C threshold, using linear regression for temperatures below 20 °C and logistic regression for temperatures above 20 °C for each construction era (i), to more accurately reflect the heating and cooling dynamics in Beijing’s residential sector. The temperature-RECF function can be calculated using Eq. (5).

$$REC{F}_{t}^{i}=\left\{\begin{array}{l}a{T}_{t}^{i}+b,T < 20^\circ {\rm{C}}\\ c+\frac{d}{1+{\left(\frac{{T}_{t}^{i}}{\delta }\right)}^{p}},T\ge 20^\circ {\rm{C}}\end{array}\right.$$

(5)

where RECF_tⁱ is the RECF under the construction era i on day t; T_t is the daily average temperature on day t; and the construction eras refer to the pre-1980s, the 1980s, the 1990s, and the post-2000s. The parameters of a, b, c, d, and δ were the estimated constants for these models. The temperature-RECF for all construction eras can be seen in the Supplementary Table. 3.

With these temperature-RECF functions and future temperature data, we can estimate future daily RECF for buildings from different construction eras, as well as the future RECE and RECI at both building and community scales. For different construction eras, their daily RECE_tⁱ can be calculated by multiplying RECF_tⁱ and their construction area S_i. Therefore, the total RECE_t in Beijing on a given day is the sum of RECE_tⁱ from all four construction eras, which can be calculated using Eq. (6).

$$REC{E}_{t}=\mathop{\sum }\limits_{i}REC{F}_{t}^{i}\cdot {S}_{i}$$

(6)

where RECE_t is the total RECE in Beijing on day t.

Residential building renewal

In China, the typical residential land tenure is 70 years, which means that residential buildings generally require renewal around this age. Residential building renewal refers to replacing existing aging buildings with new and well-designed residential buildings. Zero Energy Buildings (ZEBs) have been widely accepted^8,124 as a means to reduce energy consumption, improve energy efficiency, and transition to renewable energies. The three renewal strategies evaluated in this study represent a hierarchical framework aligned with China’s progressive carbon neutrality goals. These range from the 2000 Construction Standard (2000CS), a baseline program focusing on safety and essential insulation, to Ultra-Low Energy Buildings (ULEB), which achieve 50–65% higher efficiency than 2000CS and are promoted under Beijing’s 14th Five-Year Plan, and finally to Near-Zero Energy Buildings (NZEB), which meet the national technical standard (GB/T 51350-2019) in future. Together, these three pathways constitute the core strategies assessed in our analysis. Detailed specifications and design standards for ULEBs and NZEBs are available in various sources^{125,126,127,128}.

Regarding residential energy consumption, NZEBs can reduce energy consumption by 60% to 75%, compared to buildings meeting the 2015 energy standard, while ULEBs can achieve a 50% reduction (Supplementary Table. 1). The NZEB strategy also incorporates the photovoltaic (PV) retrofitting, along with systems like Ground Source Heat Pumps, Air source Heat Pumps, Solar Thermal Systems and Bio-fuel. The 2000CS strategy represented the traditional building renewal approach, primarily focusing on enhancing building quality and safety, with less emphasis on energy efficiency. The main differences among these renewal strategies lie in the choice of materials (e.g., concrete and steel), external wall material, thermal insulation, and energy use. For each renewal strategy, the Standard for Building Carbon Emission Calculation (GB/T 51366-2019) provides their corresponding carbon emission coefficients (Supplementary Table 2). In practice, Zhang et al. quantified their energy intensity for the year of 2016. For example, ULEB building consumes only 7.6 kg ce/m²/year, and NZEB building consumes 6.1 kg ce/m²/year⁸. Based on that, we first convert annual energy intensity (kg ce/m²) into annual carbon emissions (kg CO₂/m²) using the official emission factor from the National Development and Reform Commission national guidelines (https://www.ndrc.gov.cn/xxgk/zcfb/tz/201511/W020190905506438889540.pdf). We then calculate a correction coefficient by dividing the RECF value of ULEB and NZEB by the simulated RECF for 2016. This ratio was applied to our model-projected RECF values for future scenarios. This approach accounts for both empirical performance differences and temporal dynamics in baseline emissions. Thus, the future RECF of renewal strategies q can be calculated using Eq. (7).

$$REC{E}_{q,future}=\mathop{\sum }\limits_{i=1}\left(REC{F}_{o,future}\times \frac{REC{F}_{q,2016}}{REC{F}_{o,2016}}\times {S}_{q,i}\right)$$

(7)

where RECF_q,future is the future RECF under renewal strategies q; RECF_o,future is the model-projected RECF; RECF_q,2016 is RECF of renewal strategies q required based on the energy intensity from Zhang et al. RECF_o,2016 is the 2016 RECF from our model. The total future RECE under strategies q can be calculated as the sum of RECE across each construction era i. S_q,i is the construction area of construction era i after renewal strategies q.

The renewal expenditure E_q can be calculated by multiplying the construction area and the renewal cost (C). Yang et al. estimated the construction cost for NZEBs to range from 296 to 955 $/m², depending on climate zones⁸. For residential buildings, the cost of NZEB renewal was about 99 $/m² higher than the traditional 2000CS approach. Additionally, Zhang et al. estimated the incremental construction cost for ULEBs at around 11.3$/m^2,129. Therefore, the renewal expenditure E_qcan be calculated using Eq. (8).

$${E}_{q}=\mathop{\sum }\limits_{i}{S}_{q,i}\cdot {C}_{q}$$

(8)

Uncertainly analysis

The primary source of uncertainties in our study stemmed from future temperature projections, as our RECE estimates for the period 2021–2100 depend on these projections. Additionally, variations in residential building types contribute to the overall uncertainties, primarily due to limitations in available raw data. To address this uncertainty, we employed the Monte Carlo simulation method, running 2000 iterations to account for a range of extreme temperature values within a 95% confidence interval. To further ensure the robustness of our model and findings, we conducted difference tests to validate the dependability and stability of our findings under varying conditions.