Decarbonizing real estate portfolios considering optimal retrofit investment and policy conditions to 2050

Summary Retrofitting existing buildings is crucial for achieving Net Zero emissions. Institutional real estate owners play a key role because of their significant ownership, especially of large buildings. We utilize an interdisciplinary approach to evaluate cost-optimal decarbonization conditions for three Swiss real estate portfolios owned by a global institutional investor. We leverage a bottom-up optimization framework for building asset retrofitting, scaled to the portfolio-level, to study the effect of policy scenarios and implementations. Results indicate that achieving Net Zero necessitates significant investments, largely through thermal energy efficiency measures and low-CO2 energy systems, as early as possible to avoid locked-in emissions. Owners will be challenged to smooth long-term capital investments, pointing to a potential liquidity crisis. Consequently, hard-to-decarbonize assets are unable to reach regulatory benchmarks largely because of lingering embodied emissions. To lower transition risk, we recommend that policymakers move toward average CO2 benchmarks at the real estate portfolio-level, emulating automotive fleets.


INTRODUCTION
Decarbonizing the buildings and construction sector, responsible for 36% of energy consumption and 39% of energy-related CO 2 emissions globally in 2018, 1 is crucial to achieve the 1.5 C climate goal. Various science-based targets for buildings exist which are in line with global IPCC pathways. 2,3 One such target by the World Green Building Council (WGBC) 4 encompasses buildings' whole-life operational and embodied carbon footprint and aims for Net Zero by 2050. Continued urban growth, largely in developing economies, is expected to double global floor area by 2050, 5 presenting difficulties to achieve the target.
Two challenges exist concerning decarbonization: (1) Assuring that new buildings are efficient, resilient, energetically renewable, while being constructed with low-CO 2 footprint materials, and (2) addressing the aging existing building stocks of developed economies, such as Europe's, where 90% of buildings are still expected to stand in 2050. 6,7 With new building regulations in Europe approaching Net Zero operational CO 2 by 2030, 8 targets are primarily threatened by low retrofitting rates (<1% annually). 9 The rate of deep retrofits must increase up to 3%, 10,11 encompassing a combination of energy efficiency (EE), renewable energy (RE), and complementary technologies (e.g. heat pumps and batteries). For buildings, these options are both commercially-available and commonplace. 12 Institutional real estate owners play a crucial role to achieve a Net Zero building stock because of their (1) significant ownership, (2) large investment shares, (3) centralized decision-making, and (4) available capital. [13][14][15] Owners are being pressured to incorporate Environmental, Social, and Governance (ESG) criteria, and CO 2 performance specifically, into investment strategies. 16,17 Voluntary reporting mechanisms 14 currently focus only on operational CO 2 emissions-Scope 1 (direct emissions from combustion) and Scope 2 (grid imports of electrical and thermal energy)-but embodied emissions of materials and technologies (Scope 3) is ignored. 18 Considering CO 2 is compelling real estate owners to develop bespoke action plans for specific portfolios. 4 Owing to the uniqueness of each building asset, they must systematically re-evaluate retrofitting potentials, technological options, and estimate policy-relevant transition risk. 19 Policies in the energy, climate, and real estate market domains influence real estate owners' investment decisions in retrofits. 20 The complex interactions of these policies, and their possible developments, challenge policymakers to set consistent and coherent policy mixes 21 toward Net Zero. Taken together, both actors' decisions are interdependent for meeting long-term CO 2 targets: (1) Owners' alignment of unique asset strategies across portfolios, and (2) policymakers' settings of policy conditions across various instruments. In this study, we take an interdisciplinary approach to deliver both policy and managerial perspectives toward the research question: Under which conditions can real estate portfolios be cost-optimally decarbonized? We do so by optimizing long-term retrofitting investment strategies to evaluate the real estate portfolio Net Zero transition.
Portfolio action plans navigate the large decision-space regarding what retrofitting project to do, when to prioritize investments, on which building (where), and how that impacts economic and environmental performance in future conditions. Extant methodologies used to develop asset to portfolio retrofitting plans, 22,23 prominently the Carbon Risk Real Estate Monitor (CRREM), 24,25 are largely limited in temporal, technological, and spatial dimensions. Generally, they utilize benchmarks and top-down retrofitting assumptions, but do not: (1) accurately account for each assets' context at the portfolio-level, 26 (2) develop investment strategies to 2050 subject to trade-offs of cost and Scope 1-3 CO 2 , 27-29 (3) consider the interactions of the large set of technological options on buildings' energy demand and supply, 30 and (4) evaluate future developments such as binding policies, economic contexts, and technological improvements. 31,32 Without considering these aspects, the available option-space is limited, making it difficult to find an optimal solution to the complex real estate investment and decarbonization decision-making problem.
We utilize a two-step approach: first, we use MANGOret (Multi-stAge eNerGy Optimization -retrofitting), a bottom-up building optimization model framework relevant for comprehensive retrofitting planning. 33 The model conducts a multi-objective cost (Net Present Value) and lifecycle CO 2 optimization for all aspects of asset retrofitting strategies: materials to technologies and systems. The asset strategies are fed into a portfolio optimization model to generate cost-optimal decarbonization plans to 2050 (see STAR Methods for details with graphical depiction in Figure 1).
We incorporate a dynamic dataset including (1) climatic Relative Concentration Pathways (RCPs), (2) techno-economic and context parameter developments, and (3) policy conditions. To better establish  iScience Article the energy modeling-policy interface, 34 we evaluate policy conditions by constructing three comprehensive building sector policy mix 21 scenarios along with further analyzing the influence of an automotive fleetinspired policy implementation approach for building portfolios. We differ from existing international-level analyses, 35,36 national-level analyses [37][38][39][40][41][42] and scenario analyses, [43][44][45][46] which largely consider cost optimization toward generalized energy or emissions benchmarks, by taking an owner's investment perspective under various policy conditions.
Our case study, presented in Table 1, encompasses 2020 data for three real estate portfolios of global institutional investors, with all assets domiciled in Switzerland. The three portfolios are diversified across real estate markets with building types, uses, sizes, and ages, being representative of other European institutional real estate investor portfolios. Overall, we consider an aggregated portfolio of 235 assets comprising 600 buildings with an average age of 45 years, with the uses: multi-family residential homes (55%), office (25%), retail (15%), hotel (3%), and other (2%). The portfolios have a value of nearly EUR 6.25b and 1.55 million m 2 floor area.
Our results indicate that future-looking policy scenarios present significant cost and CO 2 emission tradeoffs for real estate portfolio decarbonization. Moving toward fleet-level CO 2 benchmarks away from the current Building Energy Code ''one-size-fits-all'' approach, similar to the average requirements for automotive fleets, could help to cost-effectively reach low-CO 2 for real estate portfolios and reduce stranded asset risks.
We find that deep retrofits are urgently necessary to decarbonize, requiring increased investments from owners into envelope thermal EE investments and RE-based heating systems. The increased capital expenditure required to avoid a ''carbon bubble'', defined as real estate assets not strategized to meet decarbonization goals consequently locking-in emissions, 47 points to a potential liquidity crisis in the industry. The carbon impacts largely lie with embodied emissions of retrofits, for which there are currently few technological alternatives, thus making achieving Net Zero heavily reliant on offsetting as a last resort.

Policy conditions description
We develop two sets of policy conditions: (1) scenarios and (2) implementation approaches. The three policy scenarios consider the long-term evolutions of over ten different policy instruments relevant for European buildings from the energy, climate, and real estate domains with national energy scenarios. As different policy regimes can influence investment strategies, we used the prominent rationale used in scenario development methods 48,49 to partner the current goal of Net Zero 2050 (NZ-50) with two extreme scenarios-Business-as-usual (BAU) and Net Zero 2040 (NZ-40). The policy scenarios and instruments are presented in Table 2.
The implementation approaches pertain to how policies toward decarbonization are enforced in the building fleet. Next to the presently used regulation on a building-level in all Building Energy Code (BEC) regimes globally and in Europe, 8 we take inspiration from automotive fleet-level CO 2 regulations to differentiate two policy implementation approaches for performance-based policies-thermal energy efficiency, CO 2 performance, renewable heating, and on-site electricity production requirements. For example, automobile manufacturers are regulated at both (1) standard for individual vehicle types (e.g. cars, vans, and trucks) and (2) fleet-wide average requirements such as Europe's 2020 goal of 95 gCO 2 /km. [50][51][52]  We use this as an example for the real estate context: this would mean enforcing performance metrics for (1) each building in an asset versus (2) a weighted average value for the entire portfolio. In this article, the building-level policy approach is implemented at the level of a real estate asset which is often just one building but could also comprise several similar buildings next to each other. The fleet-level policy approach is implemented at the level of the real estate portfolio, which is an aggregation of assets. This approach could be expanded to a larger scope to consider a fleet of buildings in a city, region, and country. Taken together, we first optimize retrofitting strategies for each asset on a multi-objective basis from minimum-cost to minimum-CO 2 considering the three policy scenarios. Next, we optimize portfolio plans based on the optimal asset strategies considering the two policy implementation approaches.

Policy influence on optimal decarbonization strategies
The influence of policy conditions on the cost to CO 2 optimal portfolio strategies is shown in Figure 2. The option ranges, termed Pareto fronts, have strategic points which represent 30-year performance. Owing to the attractive cost-effective emissions reductions from Min-cost toward the left in the Pareto, in the following we term the fourth strategy (middle) as the Baseline. As reducing emissions becomes more expensive, for ease of recognition we colloquially term the fifth point as the Min-regret strategy, the sixth point as the Feasible low-CO 2 strategy, followed by the Min-CO 2 strategy. We find that climatic RCP scenarios have little impact on portfolio costs and emissions, generally differentiating costs +/À 4% within each policy scenario Pareto. Here we present results only for RCP 4.5, with RCP 8.5 insights presented in Figure S1.
Comparing the extreme strategies (Min-cost BAU versus Min-CO 2 NZ-40, both fleet-level), total costs without considering incentives increase 45% (with incentives, 25%) with subsequent lifecycle emissions iScience Article reductions of 47%. In other words, the modeled scenario futures present a large option-space with cost and CO 2 implications for owners' investment strategies. To commit to an optimal strategy, an owner has to ''believe'' that this future would be possible, especially a low-CO 2 one that entails high capital expenditure (CAPEX) on efficient and renewable retrofits. This affirms importance of reliability and clarity of future policy developments to aid investment decision-making.
Toward more stringent policy scenarios, the Pareto fronts get tighter both from a cost and CO 2 perspective. However, the option-space is far more reduced by the policy implementation approaches than the policy scenarios. Within each scenario, the building-level policy implementation Paretos (gray) are always sub-optimal in relationship to the fleet-level (colors): For similar levels of emissions in the Feasible low-CO 2 NZ-40 strategy, the fleet-level approach reduces costs by EUR 48m (8%). Although, because of the higher flexibility, the fleetlevel also presents higher emitting yet lower-cost solutions. As with other energy assets, 53,54 a fleet-level approach presents more attractive solutions for achieving low-CO 2 cost-effectively. Such an approach shows promise for innovating beyond performance requirements regulated solely at the building-level.

Beyond benchmarks toward fleet-level approaches
As real estate portfolios comprise diverse assets, Figure 3 presents a granular view of key asset cost, CO 2 , and energy consumption metrics for the NZ-40 scenario Baseline (building-level) and Feasible low-CO 2 (fleet-level). We compare these strategies as they have the same CO 2 emissions, but the Baseline (building-level) strategy has 9% higher cost. Results for Baseline (fleet-level) and Feasible low-CO 2 strategies (building-level) are presented in Figures S2 and S3. iScience Article Unlike automobiles, each building asset is unique because of its distinctive starting conditions and existing systems, presenting different energy and CO 2 reduction potentials. The fleet-level policy approach grants a certain decision-making flexibility at the asset-level, as shown in the top-right quadrants of Figure 3A where there are fewer ''expensive and CO 2 -heavy'' assets. This way, the fleet-level approach optimally prioritizes which assets are strategically ''hard-to-decarbonize'' 55 and ''low-hanging-fruit''. The relative movement by the low-hanging-fruits to decrease CO 2 outweighs the cost increases from the hard-to-decarbonize buildings, overall leading to a lower-cost portfolio strategy for the same emissions. Figure 3B visualizes the definitions of assets' movement between the strategies: The hard-to-decarbonize assets which increase in CO 2 but reduce CAPEX (values smaller than À2) and low-hanging-fruit which decrease in CO 2 but increase CAPEX (peak between À1.7 and 0).
In other words, regulating at defined benchmarks for CO 2 and energy consumption with a ''one-size-fitsall'' for all building assets pushes owners toward higher investment in low-CO 2 building technologies, even with similar total CO 2 . Policymakers could play a role in cost-effective decarbonization by allowing owners more investment flexibility to leverage portfolio decisions considering each asset's optimal option space. Otherwise, owners face a higher risk of stranded assets.
In the NZ-40 scenario, with the stringent combination of thermal EE (<24 kWh/m 2 ) and renewable heating (>80%), some assets could be labeled as stranded on a CAPEX-basis (>14 EUR/m 2 /a) or on a CO 2 -basis (>6 kgCO 2 /m 2 /a). Stranding refers to assets which are outliers in a portfolio strategy under a certain policy scenario, considered as two standard deviations from the mean. Owners might consider selling stranded assets to a non-GRESB reporting owner-brown-spinning 56 -or otherwise choose to demolish and rebuild, leading to significantly higher emissions than retrofitting. 18 Such situations present the danger of overall increasing building sector emissions and could be prevented through the fleet-level approach. This could be especially beneficial for valuable assets for which there are few cost-effective options to reduce energy and CO 2 because of the contextual situation: building use, construction quality, or historical protection. iScience Article increases in OPEX (1%), energy (2%), with both operational (14%) and embodied (4%) emissions. However, it must be noted that total emissions remain similar between fleet-and building-level policy implementations.
The asset distributions of embodied emissions and energy consumption are much wider than operational emissions for two reasons: (1) The limited decision space between envelope retrofitting technologies which dictate thermal energy demands, and (2) the techno-economic attractiveness to reduce operational CO 2 by adapting energy systems' design and operation.
Another perspective of strategic flexibility for real estate portfolio decarbonization is represented in the Figure 4 merit order. Moving toward more stringent low-CO 2 strategies shows divergence between chosen asset strategies, with the many assets in the left to middle of the curves gradually increasing in costs to achieve lower emissions. The pronounced difference lies on the right side of the merit order, with a fewer number of assets contributing a disproportionate amount of cost.
Five large assets (160,000 m 2 floor area, 10% of portfolio) contribute to around 13% for both cumulative CO 2 emissions and costs in the Baseline BAU and Feasible low-CO 2 NZ-40 strategies. However, the absolute values for the strategic differences for these five assets differ significantly by 16 ktCO 2 and EUR 13m: Baseline BAU (40 ktCO 2 , EUR 46m) versus Feasible low-CO 2 NZ-40 (24 ktCO 2 , EUR 59m). Although these five assets present large contributions to total portfolio cost and CO 2 budgets, nevertheless on a per floor area basis, they are considered CAPEX low-hanging-fruits with very high operational emissions and energy demands. Generalized, the portfolio approach tends to increase investments in large assets to reduce their CO 2 footprints while letting smaller buildings increase their emissions. This gives important insights for carbon transition risk with certain assets contributing differently toward portfolio decarbonization.  Our results show that it is imperative that immediate action is taken to reduce energy-use and CO 2 -intensity of existing buildings to achieve low-CO 2 cost-optimally. Figure 5 shows the costs incurred in optimal strategies across the various policy scenarios and under the fleet-level policy implementation approach. Higher costs are largely because of early investments in envelope retrofits (for owners, the largest CAPEX category with over 34% of total costs in Min-regret strategies) and low-CO 2 energy systems (17% of total costs). Optimal low-CO 2 strategies take advantage of the opportunity to achieve more energy efficient buildings right away owing to possible thermal energy reductions of up to 50-80% from deep retrofits ( Figure S6). In addition, the strategies benefit from adequately-sized RE-based systems and thus avoid a ''carbon bubble'' by re-strategizing away from inefficient and fossil-reliant buildings.
In comparison, OPEX retains a relatively constant share of total costs (38%) across strategies which are usually paid by tenants. Incentives for retrofitting and conversion technologies in the NZ-40 and NZ-50 scenarios buffer some of owners' net CAPEX investments. Although net CAPEX is still higher between the BAU and NZ-40 for Baseline (15%) and Min-regret (9%), it is lower by 16% in the Min-CO 2 .
Although the presented low-CO 2 optimal strategies could reduce the size of the carbon bubble, they could still position the real estate sector at the risk of a liquidity crisis. Owners will be challenged to smooth portfolio CAPEX investments over the years to avoid large ''spikes'' with many buildings necessitating retrofits  Figure 5B, the optimal solution demonstrates a nearly double 2021 CAPEX spike.
To reach Net Zero, the large impact of embodied emissions will need to be managed as they account for the majority in low-CO 2 strategies (50-60%). Insulation materials and window components, with energy technologies such as solar PV, are the main culprits and are difficult to avoid because of technologically immature alternatives 57 and lacking circular economy systems. 58 The 2021 CAPEX spike also yields an embodied ''carbon spike'' 59 in low-CO 2 strategies. This is largely because of attractive solar PV installations in the Min-regret BAU strategy with deep retrofits in the Min-regret NZ-40 strategy.
Although our results demonstrate the techno-economic feasibility of decarbonization for all scopes on a life cycle basis, nevertheless significant emissions remain. Reliable offsetting measures will have to be taken to achieve Net Zero. Illustratively, offsetting the remaining 209 ktCO 2 in the Min-regret NZ-40 strategy with Direct Air Capture would increase owners' costs by EUR 63m (16%) assuming an average future cost of 300 EUR/ton CO 2 abated. 60 Although embodied emissions have the highest shares, the long-term decarbonization of a portfolio relies on operational emission reductions. In the following, we break down the contributing elements for both operational and embodied emissions.

Operational CO 2 decarbonization pathways
In Figure 6, we present the operational CO 2 emission reduction pathways, highlighting the Baseline and Feasible low-CO 2 strategies for the BAU and NZ-40 scenarios. For all strategies, operational emissions reduce throughout the 2021-2050 time horizon as buildings conduct retrofit interventions on the demandand supply-sides. Comparatively to the Baseline BAU pathway, the Feasible low-CO 2 NZ-40 decarbonization pathway demonstrates a cumulative reduction of 51%. The majority of these emission reductions happen early in the horizon (42% difference in 2021) because of deep retrofits.
Operational CO 2 emissions pathways are highly influenced by the electricity grid decarbonization within each policy scenario (Data S1), key for reducing building sector Scope 2 emissions. 45 Owing to the overwhelming conversion of over 88% of assets in Baseline BAU toward RE-driven heat pumps and biomass boilers in the first five years, Scope 2 emissions greatly outweigh Scope 1. These imminent changes in . Even in the Baseline BAU strategy, the impact of thermal electrification can be seen with the ''emissions burden'' being moved outside of the building. Here, over 82% of operational emissions are from electricity and 5% from fossil fuels, whereas the majority of emissions for the buildings' today come from fossil combustion (Scope 1).
Similar to operational emissions, heating demands cumulatively reduce between 43 and 59% over the horizon ( Figure S4). As few envelope retrofits are conducted right away in the Baseline BAU strategy, the heating demand pathway results can be more directly benchmarked to the GRESB reported value of 65 kWh/ m 2 . Here, the Baseline BAU strategy begins in 2021 at 51 kWh/m 2 and the Feasible low-CO 2 NZ-40 strategy at 37 kWh/m 2 . Throughout the horizon, electricity demands increase slightly because of (1) occupancybased norms governing lighting and plug-loads, and (2)  Low-CO 2 building technology packages Figure 7 presents the owners' near-term CAPEX investments to 2025 for specific technologies and components, with their associated embodied emissions footprints for two points in the BAU policy scenario. In the Feasible low-CO 2 BAU strategy, the vast majority of investments are toward solar PV (33%), woodaluminum windows (22%), and heating systems (18%), accounting toward 27%, 30%, and 17% of embodied emissions respectively. Li-ion battery storage accounts for 5% of investments and contributes to 7% of embodied emissions. Comparatively, the Baseline BAU strategy has lower CAPEX and embodied emissions in the first five years because of later investments in window retrofits, solar PV, and heating systems which are pushed toward end-of-life. The 30-year embodied emissions differ by 16% between the strategies ( Figure 5).
By modeling the long-term evolutions of (1) energy carrier prices, (2) CO 2 taxes, (3) technological learning from RE technologies with (4) prospective policy measures, our results demonstrate a more ''level playing field'' for high-CAPEX, RE-based systems becoming more attractive than incumbent fossil fuel technologies (low-CAPEX, high-OPEX). This aligns with recent studies 40,61,62 indicating a paradigm shift away from fossil fuels toward RE-based and sector-coupled energy systems because of the superior technoeconomics.
Specifically, optimal solutions rely heavily on ground-& air-sourced heat pumps (G & ASHPs) with biomass boilers for winter peaks coupled with solar PV and batteries. These low-CO 2 heating systems account for a minor share of CAPEX and embodied emissions (<10%) in the Feasible low-CO 2 BAU strategy. Although many assets have existing oil and gas boilers (assumed to be paid off), even in the Baseline BAU strategy these systems are seldom used representing stranded energy systems.
The speed of the paradigm shift is strongly impacted by stringent EE regulations and attractive incentives in the NZ-40 scenario. For example, attractive policies for solar PV such as feed-in-tariffs and incentives result in an investment spike in 2021 even in the Baseline BAU strategy. Here, over 42% of CAPEX investments in the first five years are toward solar PV. Furthermore, in the NZ-40 scenario, facade stone wool insulations play a much larger role in CAPEX. iScience Article

DISCUSSION
In this study, we optimize portfolio retrofitting strategies toward cost-effective decarbonization to 2050. We explore strategically valuable policies and managerial insights for the transition toward whole-life Net Zero CO 2 portfolios by considering optimal investments in building assets under various policy conditions, moving away from single policy instrument analyses. 63,64 Considering policy mix scenarios shows that for the extreme strategies Min-cost BAU and Min-CO 2 NZ-40 (fleet-level), whole-life CO 2 emissions to 2050 can be reduced by 47%, with lifecycle operational CO 2 reducing by 69%. The range of results in between demonstrates the importance for policymakers from city to national-levels to provide policy certainty to alleviate transition risk.
Owing to their distinctive starting conditions and existing systems, building assets have varied energy and CO 2 reduction potentials, and therefore costs. Scaling bottom-up asset strategies to portfolio plans, assets lie in a distribution from hard-to-decarbonize to low-hanging-fruit. In this light, we explore a policy innovation to regulate portfolio CO 2 -performance as a building fleet, similar to automotive industry regulations. Moving away from a ''one-size-fits-all'' approach for performance-based policies on EE and CO 2 benchmarks could reduce transition costs 8% with equal emissions. Although a fleet perspective means that dirty assets still remain, it could prove important to give owners flexibility to manage the high-cost transition in time to 2050. Such flexibility could also increase acceptance of low-emission building standards and help break political deadlocks. Figure 7. Technology packages' contribution to CAPEX and embodied emissions Utilizing scenario inputs, the MANGOret optimization model develops optimal strategies (presented as aggregations in the Pareto fronts) for unique building assets. These strategies are fed-in to the portfolio-level model which optimally chooses the best strategies toward its own Pareto front. Depending on the policy alignment approach, the model is constrained differently with regards to performance-based requirements at the building-or fleet-levels.

OPEN ACCESS
iScience 26, 106619, May 19, 2023 11 iScience Article Decarbonized real estate portfolios require a significant number of deep retrofits earlier-than-planned, comprising complete re-insulation of building envelopes coupled with RE-based heating systems. Significant CAPEX into retrofits in the near-term could create capital liquidity and workforce adequacy issues, such as the burgeoning ''green'' buildings and construction workforce supported in the upcoming European Renovation Wave. 6 Owners used to smoothing CAPEX over the years could try to sell (brown-spin) or redevelop assets which are flagged as stranded. This could potentially further grow the carbon bubble in the building sector, necessitating both owner and policymaker urgency.
Operational Scope 1 & 2 CO 2 emissions can be optimally reduced by over half between extreme strategies. Even in minimum-cost strategies, we show a paradigm shift toward RE-based heating systems comprising heat pumps and biomass boilers coupled with solar PV. Many assets with existing fossil fuel boilers seldom use them because of the techno-economic inferiority, flagging stranded energy systems. Although Scope 1 emissions are low, Scope 2 emissions with regards to electricity grid decarbonization play an important role. On the other hand, embodied Scope 3 CO 2 emissions hold the largest emissions share in low-CO 2 strategies (50-60%) and are largely impossible to avoid. There is a need to develop commercially-available low-embodied emission options for the key culprits: insulation materials, window frames, with lowering values for solar PV and batteries. To reach Net Zero targets, significant offsetting will be required.

Limitations of the study
Our results are subject to several limitations. First, all optimizations conducted in this study are deterministic and assume perfect foresight for all future-looking data. Future work could focus on uncertainty and sensitivity analyses to ''stress test'' asset and portfolio strategies for various future uncertainties such as climate change risks 65 and the recent geopolitical developments for European fossil fuels. 66 Second, we rely on building archetypes and government databases for building asset-level datapoints which are subject to inaccuracies. Third, there is no assessment of the contributions of each individual policy instrument to the overall impact of the policy mix on optimal strategies. Similarly, the fleet-level policy implementation approach needs to be further explored as there is currently no precedent to the knowledge of the authors. Fourth, our work lacks an assessment of the financing mechanisms and distributional aspects of retrofitting real estate assets relating to the landlord-tenant split-incentive. Lastly, our study does not fully consider the potential co-benefits of building sector decarbonization, such as infrastructure resilience, air quality, water quality, and human health, which could improve socio-economic attractiveness.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: iScience Article of the energy balance'' in the model formulation allows to consider the interdependencies of technology scheduling, for example, how the timing and design of an envelope retrofit would impact the heating system design and operation.
The demand-side comprises thermal envelope energy efficiency with over 10 retrofitting packages for Facade, Roof, and Windows, consisting of various technology options which vary on a cost vs. embodied emissions basis. Namely, these are various options for insulation materials: oil-based extruded and expanded polystyrene (XPS and EPS, respectively) along with mineral stonewool, and windows: plastic and wood-Aluminum frames. These retrofitting technologies also have a depth element -the minimum Building Energy Code or target green building label (e.g. LEED, BREEAM, DGNB, Minergie, etc.).
The supply-side comprises energy supply, conversion, and storage, modeled in the popularized Decentralized Multi-Energy Systems (D-MES) framework considering multiple energy carriers: heating, electricity, natural gas, biomass, oil, and District Heating (DH). 76,77 The candidate technologies include: electrically-driven Air-Source Heat Pumps (ASHP), Ground-Source Heat Pumps (GSHP), fuel oil, natural gas, and biomass boilers, gas-fired Combined Heat and Power (CHP) engines, and DH. In terms of renewable energy technologies, only solar Photovoltaic (PV) panels are considered due to urban constraints. Additionally, Hot Water Thermal Storage Tanks (HWTS) and lithium-ion batteries are considered to store thermal and electrical energy, respectively. Buildings can import all energy carriers but can only export electricity as they are grid connected. Existing technologies (e.g. boilers) and DH connections for each building are also included, based on data availability. Non-energy components critical for real estate owners' retrofitting budgets are also included, namely: kitchens, bathrooms, and piping. To provide better accuracy for intervention timing, the model considers the component condition degradation utilizing the Schroeder method. 78 This study considers three real estate portfolios with all building situated in Switzerland. While we leverage datasets existing for Switzerland, a similar model approach can be used in other European countries with similarly available open-source data relevant for building retrofitting.
The MANGOret framework 33 requires a small set of building-specific data from real estate owners: address, construction year, renovation year, and last year's energy demands. Other techno-economic (e.g. technology CAPEX learning curves and energy carrier price evolutions) and environmental parameters (e.g. Electricity grid decarbonization and CO 2 tax evolution per policy scenario) with long-term projections are all given in the original model formulation and visually presented in the Data S1. MANGOret is a deterministic optimization framework which advances beyond the traditionally single-stage building energy optimization models using approaches such as life-cycle costing. 79 Other methodologies have recently been explored to make building energy decision-making problems more tractable such as artificial neural networks based on machine learning techniques. 80,81 We leverage an archetypal energy demand database to reference demands of various retrofitting packages for many unique buildings. The database consists of over 2,100 Swiss archetypes varied by building type, geographic zone, and age categories, which were simulated for from 2020-2060 in 10-year time-steps for 3 climate change scenarios based on two RCPs: 4.5 C, and 8.5 C. We utilize the CESAR (Combined Energy Simulation And Retrofitting) tool 82 built on the standard building energy software EnergyPlus. 83 The archetypes were developed based on OpenStreetMap 67 and government building statistics databases. 68,69 Solar irradiance data for each building location are taken from the Renewables.ninja API which connects to the MERRA-2 database. [70][71][72] We cluster the energy demand and solar irradiance time-series parameters using the k-medoids peak + typical day clustering approach, 84 in this study for two peak days for electricity and heating demand, and five typical 'normal' days, resulting in seven total typical days.
We optimize each asset location A l on a multi-objective basis, minimizing cost and CO 2 , for each year y for the entire time horizon 2021-2050. The multi-objective optimization outputs the Pareto front, consisting of seven points. Seven Pareto points are chosen as typical from other optimization studies to present sufficient strategic options (five) between the two objective extremes (minimum-cost and minimum-emissions) for the portfolio-level optimization.
While operating emissions can be easily determined and accounted for based on the energy consumption of the assets in the portfolio, accounting for embodied emissions is not as clearly defined. In our model, the embodied emissions of a technology are assigned to the year in which the technology is installed. We choose this accounting approach because we argue that on a physical basis, the CO 2 emissions embodied in materials and technologies were already released into the atmosphere by the time of installation in the building. Therefore, spreading these embodied emissions evenly over the component lifetime, as done in Switzerland, 85 or using time-dependent weighting factors (e.g. emissions released now have a greater impact than those at the end-of-life) as done in the French RE2020, 86 is a non-physical accounting measure which does not address the urgency of decarbonization.
The asset value is calculated with the industry-accepted Discounted Cash Flow (DCF) methodology: rental revenues less the costs. Rental revenues are calculated based on value-added investment formulated based on legally-mandated rental calculations.

Portfolio-level optimization
We replicate the decision-making approach of the real estate multi-year planning process within the MANGOret framework. We do so by nesting the asset-level optimization within the portfolio-level optimization in a two-step approach. First, as previously described, we optimize each individual asset on a multi-objective cost and CO 2 basis. 33 Here, we formulate a smaller portfolio-level model. Based on the seven optimal Pareto points pp for each asset A l , the portfolio optimization conducts the same multi-objective optimization, minimizing cost and CO 2 , by choosing the optimal asset-level strategies for the portfolio-level planning. It does so by utilizing a binary variable, Y sol l;pp , which references the chosen asset strategies total cost, CO 2 , and value.