Efficiency improvement and technology choice for energy and emission reductions of the residential sector

The residential sector currently accounts for one fifth of global energy use and corresponding greenhouse gas emissions, largely driven by increasing demand for space heating and cooling. Climate change mitigation action requires these to reduce, but the exact decarbonization strategies and their heterogeneity is unclear. We use a regional recursive dynamic energy system model with an explicit representation of residential energy use and building stocks to explore the contribution of this sector in long-term decarbonization pathways. The projections show that in a 2 ̊C scenario, global heating demand is expected to decrease from current levels by 18% and 64% by 2050 and 2100, respectively. However, due to increasing affluence in warmer regions, cooling demand is expected to increase by 112% and 201% respectively. Yet, direct residential emissions are almost eliminated by 2100. This is achieved by combining increased envelope efficiency and advanced heating technologies in a synergistic manner, where the adoption of high efficiency heating and cooling reduces the need for increased insulation, and vice versa. By combining these measures with rooftop PV, the net energy demand of many household types approaches zero. The exact residential sector strategies vary across different regions, depending on local climate, socio-economic, and building stock characteristics.


Introduction
The residential sector accounts for approximately 20% of 2017 global final energy use and is directly responsible for about 6% of energy related greenhouse gas (GHG) emissions. If upstream emissions from electricity are included, then residential buildings are responsible for almost 18% of global GHG emissions (IEA 2019). Globally, residential energy demand has been increasing steadily since 2000 at around 1% per year, driven by a growing population and increased demand of energy services, particularly for space and water heating, space cooling, and appliance use. These trends are expected to continue, given trends in population and household consumption and the expected impact on residential energy use and emissions (Lucon, Ürge-Vorsatz et al. 2014, Levesque, Pietzcker et al. 2018). Yet, to limit global warming within the targets of the Paris Agreements net anthropogenic GHG emissions, across all sectors, must fall to net-zero within the next few decades (Rogelj, Shindell et al. 2018).
The residential sector can contribute to the mitigation of global GHG emissions through a reduction of final energy demand and through a switch towards low or zero emission energy carriers. Existing studies have investigated the question of how changes in technology choice and consumption practices can reduce emissions (van Sluisveld, Martínez et al. 2016, Knobloch, Pollitt et al. 2018, Levesque, Pietzcker et al. 2019. These studies typically show that such lifestyle and technology changes could reduce total energy demand by up to 60% by the end of the century. Studies focusing on improving building envelope efficiency highlight the important role this can play, also limiting residential space heating and cooling demand by up to 60% (Edelenbosch, Rovelli et al. under review). Edelenbosch et al. (under review) highlights the importance of the make-up and turnover of building stocks, which are an important determinant of the rate of improvement of aggregate building efficiency as potential lock-in into low efficiency infrastructure. Finally, the residential sector can also contribute to decarbonization through rooftop solar photovoltaics (PV). It has been estimated that rooftop (PV) could provide 30% of global, or 70% of urban household electricity demand. This would fundamentally shift the role of buildings from energy consumers to so-called 'prosumers' (Poponi, Bryant et al. 2016), allowing for a broader energy-system decarbonization.
While the above studies highlight the potential of individual measures, it is unclear how a full strategy to reduce emissions looks like. For instance, in the above studies highlighting the importance of behavioural change and technology adoption, the improvements in building efficiency through increased insulation is not explicitly dealt with, with exogenous assumptions on how this may develop (Knobloch, Pollitt et al. 2018). In fact, the decision to invest in efficient technology depends on decisions on building envelope efficiency (in new buildings or renovation of existing buildings), and vice-versa. This raises the possibility for synergies and trade-offs between different measures, such as the reduced motivation to invest in extremely high building envelope efficiency if heating demand is produced from ultra-high efficiency technologies such as heat pumps. Furthermore, the dependence of energy use and associated emission reduction strategies on local climate and socio-economic characteristics is not explicitly investigated.
In the present study we conduct a techno-economic analysis of the global residential sector and its long-term pathways consistent with the Paris Agreement for different world regions. We use a recursive dynamic energy system model with an explicit representation of the residential building stocks and energy demand, the TIMER-REMG model. We use this model to investigate the interplay between -and adoption of -three key mitigation measures: building envelope efficiency improvement, heating/cooling technology choice, and investment in rooftop PV. We project the adoption of different mitigation measures and how they influence the energy demand and emissions of the residential sector in the 21 st century, as well as their contribution to meeting the Paris Agreement.
In Section 2 we give a brief description of the updated TIMER-REMG model, how the mitigation measures and building stock accounting are modelled, and the scenarios run in this study. In Section 3 we present the results of our simulations focusing on the expected developments on building stocks and the projected energy use and emissions of the residential sector in baseline and climate change mitigation scenarios. We also investigate the role different energy use and emissions mitigation measures play in scenarios consistent with strict climate goals. Section 4 discusses the main uncertainties of the results, as well as their implications in broader socio-economic and energysystem contexts. Finally, Section 5 summaries the main conclusions of our analysis.

Model
This analysis uses an updated version of the TIMER energy system model, which forms the energy system component of the IMAGE Integrated Assessment Model (IAM) (Stehfest, van Vuuren et al. 2014). TIMER is a recursive dynamic energy system model representing the global energy system, projecting developments in energy supply, conversion, and demand ( van Vuuren, Stehfest et al. 2017, van Vuuren, Stehfest et al. 2018. The model is driven by exogenous projections of population, GDP, value added, and household expenditures. For the residential sector, a stylized bottom-up submodel (REMG) projects the final energy demand for five end-use functions: cooking, lighting, space heating, space cooling, water heating, and appliances. The model has an explicit representation of 26 world regions with different climatic and socio-economic conditions, as well as five income quantiles for urban and rural households. Thus, TIMER-REMG captures the different energy use characteristics and investment motivations of 260 representative households. The motivation, formulation, and background data of the TIMER-REMG model are available in Daioglou, van Ruijven et al. (2012).
The updated model used in this analysis includes four key additions: accounting of building stocks throughout the projection period including construction and decommissioning of capital, calculation of investments in thermal insulation of these stocks at construction or renovation, inclusion of heat pumps as a space heating technology, and the possibility for households to invest in rooftop solar PV. The implementation of each of these is described in the following sections, with further details available in the Supplementary Material.

Residential Stocks and Insulation
The REMG model projects regional residential floorspace based on economic development and population density, calibrated to historic data of floorspace development (Daioglou, van Ruijven et al. 2012). Resultant changes in floorspace demand, together with regional building lifetime drive the stock accounting. The TIMER-REMG model starts its calculations in 1971 with an annual timestep, thus providing a 50-year "spin-up" period for building stocks.
The model determines the thermal efficiency of the building envelope (i.e. the U-values which denote the building envelope thermal conductivity, measured in W/m 2 /K) by applying different levels of insulation. These can be applied either for new buildings or at a later stage via renovation. The model includes six insulation levels representing different use of thermally resistant materials on walls, windows, floors, and roofs. Each of these levels have their own capital costs and their thermal conductivity has been scaled for regional climate characteristics (Petersdorff, Boermans et al. 2005). The decision to invest in insulation depends on (i) the annualized capital cost of insulation levels, and (ii) the possibility for lower fuel costs due to reduced heating and cooling demand. Concerning renovations, we assume that only stocks which have been in place for at least 15 years can be renovated, and the technical lifetime of renovation investments are limited by the remaining lifetime of the building stock. Thus, the discount rate of investments in renovations increases as buildings approach the end of their lifetime.
Following, the motivation to invest in insulation depends on regional and income class dependent discount rates, household floorspace, and heating and cooling energy demand. The market shares of the insulation levels (both for new constructions and renovations) are allocated based on their relative competitiveness using a multinomial logit function. Regional climate characteristics affect both the techno-economic parameters of insulation levels, as well as heating/cooling costs which are a prime motivator for investing in insulation. As discount rates and heating and cooling demand vary across income groups, investments in insulation are skewed towards richer households. Finally, in climate policy scenarios where a price is attached to the emitted CO2 (see Section 2.2), the fuel costs for heating and cooling increases. The extent of this increase depends on the CO2 intensity of heating technology use and the makeup of electricity generation. Accordingly, these additional costs increase the competitiveness of higher insulation levels due to their lower heating/cooling demands.

Heating and cooling technologies
Projections of useful energy demand for space heating and cooling is based on the methodology developed and presented in Daioglou, van Ruijven et al. (2012) and (van Ruijven, van Vuuren et al. 2011). Space heating demand is modeled as a function of floorspace (m 2 /capita), population (capita), and heating degree days (HDD), based on Isaac and van Vuuren (2009). The useful heating demand intensity (kJ/m 2 /HDD) is calibrated to historic U-values and energy use for space heating. Its future development is linked to the improvements in building envelope efficiency (i.e. the U-values of different building stocks described above). Residential final heating demand is supplied through eight possible energy carriers (coal, fossil liquids, natural gas, hydrogen, traditional biomass, modern bioenergy, district heat, and electricity) each with a representative technology with associated costs (fuel and capital) and conversion efficiency. Specifically, for electricity, two technologies are included: Resistance heaters, and heat pumps. Thus, in total nine heating technologies compete for market shares of heating energy demand based on their relative costs using a multinomial logit function. The model is designed to explicitly represent the energy-ladder as households opt for cleaner (but more expensive) heating fuels as they become richer, with the poorest households heating with traditional biomass, assumed to have a zero cost.
Similarly, cooling demand is also a function of population, cooling degree days (CDDs), and cooling intensity which is linked to improvements in building envelope efficiency. Unlike heating, cooling demand can only be met via electricity with three possible cooling appliances: fans, air-coolers, and air-conditioners. Households which lack electrification do not satiate any of their cooling demand. The ability of households to invest in cooling appliances depends on their household expenditures (calibrated to historic cooling appliance data), with the poorest households only fulfilling part of their cooling demand using fans. With increasing household expenditures, it becomes possible to invest in air-coolers and air-conditioners, fulfilling previously unmet cooling demand. The ownership as well as the energy use of air conditioners are related household expenditures, local CDDs and an assumed efficiency improvement.

Rooftop photovoltaic
The model includes a dynamic representation of residential rooftop PV including their potential, investments, and dispatch (Gernaat, de Boer et al. 2020). The regional potential of rooftop PV surface area is determined by relating projections of floorspace area to roof area, corrected for factors affecting suitability such as shading, architectural features and orientation. This is then combined with the regional solar energy flux, to get the technical potential of rooftop PV energy supply. Using data on module costs, operation and maintenance costs, and marginal load factors we can determine the supply curves of levelized costs of rooftop-PV electricity. By including endogenous learning-by-doing and changes in floorspace, we get projections of the economic potential through regionally explicit PV cost-supply curves. Consequently, geographic variation of the marginal costs is represented across 26 regions with varying solar irradiation (technical potential), while social variation is represented across 10 income groups with differing floorspace area (technical potential) and discount rates (economic potential).
These supply curves allow for a household investment decision to either buy electricity from the grid or invest in a PV system. A multinomial logit equation is used to determine the share of residential electricity demand, which is met by rooftop PV, by comparing the marginal PV levelized cost with the cost of grid-based electricity. Thus, investing in rooftop PV benefits from larger households, locations with higher irradiation, higher electricity demand, relatively higher costs of grid-based electricity, and lower discount rates enjoyed by wealthier households.

Scenarios
The scenarios used in this study are introduced in Table 1. We use the IMAGE implementation of the SSP2 baseline as our reference scenario (Fricko, Havlik et al. 2017. We also present the results for the residential sector according to an RCP2.6 climate target, representing a 2˚C climate target. Throughout the manuscript, these scenarios are referred to Baseline and 2˚C respectively. The 2˚C scenario follows an emission price projection (determined in IMAGE) across all GHG emission sources (fossil fuels, industry, and AFOLU) applied globally which results in a cost-effective pathway meeting an emission budget consistent with a 2˚C global mean temperature increase. It is important to note that all scenarios assume the same climate, thus climate feedbacks on energy demand and supply do not differ across scenarios. We have done this to isolate economic effects of climate policy on investment decisions and to allow for comparability between the Baseline and 2˚C scenarios.
We also project two sensitivity cases to isolate and decompose the role of insulation and technology choices (heating, cooling, rooftop PV). These scenarios act as references to determine the effect of specific mitigation actions. In these reference cases we force the heating and cooling technology choice of the Baseline projection but allow the model to determine the decision of insulation that is influenced by the carbon tax trajectory of the 2 o C case. In the first sensitivity case insulation was limited only to new buildings (InsulNew) and in the second it is allowed on the entire building stock (InsulAll). The results of these scenarios can be used to isolate the energy and emission mitigation effect of insulation, renovations, and technology improvement. We specifically investigate the following:  In the Supplementary Data further information is provided including demand of "useful energy" (i.e. joules of heat demanded), final energy per energy carrier, and emissions including indirect energy system emissions (or indirect sequestration in case bioelectricity production combined with carbon capture and storage). Final energy carriers available to the residential sector are coal, fossil liquids, natural gas, hydrogen, traditional biomass, modern bioenergy, district heat, and electricity. Unless otherwise stated, all final energy results are "gross" residential energy demand, thus possible generation from rooftop PV is not deducted from heating or cooling electricity demand, and electricity generated from rooftop PV is presented separately. We do this to provide a clearer picture on energy demand structure. The difference between heating/cooling electricity demand and rooftop PV generation is the net electricity demand of the residential sector.
Emissions only account for carbon dioxide from energy use and are determined by attaching emission factors based on the carbon content of primary energy carriers. Specifically, for modern bioenergy the emission factor includes land-use change emissions and requirement of nonrenewable energy in the conversion of biomass into secondary energy carriers based on ). Indirect energy system emissions depend on the makeup of electricity, district heat, and hydrogen production, all of which are influenced by carbon prices and assumptions on technology availability including carbon capture and storage (Krey, Guo et al. 2019, Daioglou, Rose et al. 2020). Africa regions. These regions also show the greatest relative improvement in their U-values in the Baseline, relative to the current situation. This is largely because they start from very high inefficient building envelopes and they witness the greatest marginal growth in floorspace leading to significant opportunities to invest in efficiency in new buildings. The projections also show the importance of local climate conditions. Colder regions (OECD, Reforming Economies) are projected to have significantly lower U-values than warmer regions, as heating demand forms a significant financial burden. Reforming Economies are projected to have the lowest U-values, even though they have a lower GDP than the OECD region, because of their local climate which is significantly colder than the OECD. When focusing on specific regions represented in the IMAGE model Canada, Russia and Ukraine reach U-values below 0.5 W/m 2 /K, which is typically regarded as a very high efficiency building envelope. Table 2 shows annual investments in building envelope efficiency (in new buildings and renovations) for 2030, 2050, and 2100, as well as the average renovation rate for the 2020-2050 and 2020-2100 periods. The figures in brackets are the investments in insulation for renovations only. As shown, already in the Baseline, there are significant investments in insulation via both renovations and new constructions. The OECD sees the highest renovation rate, as this region has an older standing stock, as well as a colder climate. However, investments increase significantly in other regions over time as they get richer. This is also reflected in the higher renovation rates for the 2020-2100 rather than 2020-2050 period as the proportion of older building stocks increases in the latter part of the projection period (see Figure S1 in the Supplementary Material). To put these numbers into context on current investments in envelope efficiency, in 2016 total investment in building envelopes (residential + services) was 69.3 B$, most of it in the form of renovations 1 , growing 12% compared to 2005 (IEA 2017). These compare with our global renovation investments in 2030 of 108.6 B$ in the Baseline and 160.6 B$ in the 2˚C scenario, indicating increasing renovation investments, particularly in the OECD and Asia. In the 2˚C scenario, cumulative investments in insulation in renovations and new constructions increases across all regions. This is also reflected in the increased renovation rates. Global investments in renovation are projected to be 160.7 B$ in 2030 (i.e. a 48% increase compared to the Baseline), increasing to 202.6 B$ in 2050 and 317.9 B$ in 2100. However, as also reflected in the Uvalue projections of Figure 1, the greatest difference between the Baseline and 2˚C scenarios are in the OECD and Reforming Economies regions. This is because of their colder climates and higher household expenditures. For these regions most of the spending on efficiency is projected to happen in the shorter term, to renovate existing structures, with annual investments falling over time. On the other hand, for most other regions, annual investments are expected to increase over time as (i) their building stocks increase, and (ii) households become wealthier and investment in renovation becomes more financially attractive. Overall, the Middle East & Africa show the lowest renovation rates as they have milder climates and, on aggregate, lower household expenditures.

Energy and Emissions for Heating and Cooling
Final energy demand for heating and cooling for the Baseline and 2˚C scenarios are shown in Figure  2. Also shown are the electricity available for heating and cooling from residential rooftop PV (presented as a negative demand), as well as the emissions arising from residential and cooling (excluding indirect emissions from electricity or district heat production). Final energy demand per region and energy carrier are available in the Supplementary Data. As expected, total heating and cooling demand depends a lot on local climate characteristics. In the Baseline, while heating demand is expected to slightly increase in the Middle East and Africa, and Latin America -due to increasing incomes and overall floorspace -globally it is expected to plateau. This is due to increases in building envelope efficiency in colder regions (see Figure 1), as well as a shift towards more efficient heating fuels over time (particularly natural gas and/or electric heaters, see Supplementary Data). Contrastingly, cooling demand is projected to increase across all regions, with the increase being particularly pronounced in Asia, the Middle East and Africa, and Latin America. This is due to the expectation that the many households in these regions will reach household expenditures which will allow them to transition from the use of fans to larger air-conditioning units to meet their cooling demands. Combined with growing floorspace, this leads to a large increase in cooling final energy demand.
In the 2˚C scenario, final energy demand for both heating and cooling drops significantly compared to the Baseline. The use of emission prices in this scenario promotes increased building efficiency as shown in Figure 1, which reduces the overall heating and cooling demand. Emission prices also drive fuel switching towards cleaner and more efficient heating fuels, particularly electric resistance heaters and heat pumps, further driving down final energy demand. As expected, the use of fossilbased heating fuels is completely phased out in the 2˚C scenario. Even though heat pumps are extremely efficient, due to their higher costs they are projected to be used primarily in regions with significant heating demand (OECD, Reforming Economics, and to a lesser extent Asia and Latin America). In warmer regions, the 2˚C scenario shows a significant drop in cooling demand, again due to a combination of investments in appliance and building envelope efficiency. Globally, compared to the Baseline, in the 2˚C scenario total heating demand is projected to fall by 19% and 56% in 2050 and 2100 respectively. For the cooling demand the drop is 7% and 47% respectively. Compared to final energy demand today, in the 2˚C scenario global heating demand is expected to decrease by 18% and 64% by 2050 and 2100 respectively. On the other hand, cooling demand is expected to increase by 112% and 201% respectively, but lower than the projected increase in the absence of climate policy.
The (direct) emissions follow final energy demand for heating, as all emissions from cooling are indirect (i.e. emissions take place at electricity generation). Since in the 2˚C scenario long-term heating is almost completely electrified (with some use of district heat or hydrogen), direct emissions drop to near zero. In the absence of an emission price, global heating emissions are projected to increase to levels about 17% higher than today by 2050, subsequently falling to approximately 10% of current levels. The use of electricity, district heating or hydrogen as heating fuels puts pressure on indirect emissions, with these being more than double the direct emission by 2100 in the Baseline. However, in the 2˚C scenario, due to extensive use of renewables and bioenergy with carbon capture and storage, indirect emissions fall to very low levels, even going negative after 2050 ( van Vuuren, Stehfest et al. 2017, van Vuuren, Stehfest et al. 2018).

Figure 2. Final energy for residential heating and cooling (coloured bars, left axis), and direct residential emissions (points, right axis). Rooftop PV electricity generation shown as negative demand (left axis). Baseline and 2˚C scenarios.
Figure 2 also shows the generated electricity from residential rooftop PV, as a negative demand. This is gross generation and thus it can also be used for other residential energy services (appliances, cooking, lighting), or for export to the grid. Investment in rooftop PV is promoted in the 2˚C scenario as aggregate increases in the price of electricity make this investment worthwhile, leading to more rapid buildup in rooftop PV capacity compared to the Baseline. However, investments in efficiency lead to an overall lower electricity demand in the 2˚C scenario, and so the total rooftop PV generation in 2100 is slightly lower than the Baseline. Investments in increased building and heating/cooling efficiency, as well as rooftop PV makes households into so-called "prosumers". By combining investments in appliance and building envelope efficiency, as well as rooftop PV, households may become "zero-energy", or even "positive energy" by becoming net exporters of final energy. Figure S5 of the Supplementary Material shows how urban and rural households of different income quintiles approach this status in the 2˚C. As shown, richest households get closest to meeting their energy needs (including heating, cooling, cooking, appliances, and lighting) as they can invest in high efficiency (appliance and building envelope) and have larger floorspace allowing for increased rooftop PV generation. By 2100 it is projected that the richest urban households in Latin America become "positive energy households" as they also benefit from a climate which limits their final energy demand.

Technology factors contributing the energy demand and emission reductions
As shown in Section 3.2, decisions concerning investments in heating/cooling technologies as well as building envelope efficiency dictate the projected energy demand and emissions. In the 2˚C scenario the extra price on emissions leads to further investments on all fronts. Figure 3 outlines the contribution of (i) insulation in new buildings, (ii) improved insulation in existing buildings via renovation, and (iii) investment in more efficient heating/cooling technologies, to reducing final energy demand and emissions between the Baseline and 2˚C scenario. The figure also shows the effect of renovation in the Baseline, giving a better understanding of the total impact of renovations.

On a global scale, it is shown that the application of more insulation in new buildings accounts for most of the reduction in secondary energy demand. On a regional scale is can be seen that regions
where older buildings make up most of the building stock (see Figure S1 in Supplementary Material) or most of the final energy demand is for heating, renovations also contribute to reduced energy demand. For the OECD region, while renovation rates are amongst the highest (see Table 1), the improvement on final energy is limited. This is because OECD households on aggregate start with higher quality buildings than in order regions, leading to lower marginal gains from renovation. This conclusion has also been observed with detailed national studies (van den Wijngaart and van Polen 2020). The Reforming Economies and Asia benefit the most from renovations (in both the Baseline and 2˚C scenarios) as they have a large existing stock with low efficiency levels in the Baseline (Asia) or experience increasing household expenditures combined with very cold climates (Reforming economies). Conversely, in the Middle East and Africa as well as Latin America renovations have a more limited role since most of their energy demand is for cooling, which benefits less from high levels of insulation. Furthermore, as the Middle East and Africa, and Asia, are projected to have a large increase in residential floorspace in the future (see Figure 1), they benefit more from applying improved insulation in new buildings. The effect of increased efficiency in heating and cooling technologies reduces final energy demand mostly in colder regions as the efficiency gain from electrifying heating (through resistance heaters and heat pumps) can lead to large final energy demand reduction.
Concerning emission mitigation, while increased insulation does contribute, most of the projected mitigation is expected to come from changes in cooling and heating technologies. The electrification of heating in the OECD, Reforming Economies, and Asia, leads to steep declines in direct emissions. Similarly, the use of more efficient air-conditioning units in warmer regions plays a very important role.
These results show an emergent result from our recursive system-dynamic model formulation. The model highlights that neither insulation nor technology change dominates energy and emission mitigation efforts, that is, neither option is applied to its maximum. Rather, an equilibrium is reached where improved levels of both heating/cooling efficiency and building envelope efficiency are applied. The modeled households benefit from a self-limiting feedback between these two, where improved heating/cooling efficiency reduces the need (or the economic incentive) for increased insulation, and vice versa. In this system, the applied emission price (which is calculated on total energy and land use greenhouse gas emissions) acts as a pressure on how far this equilibrium is pushed.

Discussion
This paper presents an updated version of the TIMER-REMG model, part of the IMAGE Integrated Assessment Model (IAM). The improvements presented here are an important step forward for how IAMs represent the residential sector, which do not typically track changes in building stocks, their energy efficiency, and the interaction between different energy and emission mitigation options. Thus, existing residential mitigation strategies projected by IAMs are at very aggregate levels, limiting their explanatory power. The explicit representation of different income levels, insulation levels, renovation and construction, selection of heating and cooling technologies, and rooftop PV, allow for an improved understanding of the dynamics, synergies, and tradeoffs faced by this sector in emission mitigation strategies.
The additional detail of the model also imposes significant data requirements and associated uncertainty. A sensitivity analysis conducted on several parameters (building lifetimes, insulation cost, learning rates and technological improvement) identified that investments in building envelope efficiency are most sensitive to assumptions on the shape of the representative household (see Table S1 in Supplementary Material) (Mikropoulos 2018). Limited information exists about the variation of multiple parameters across countries and time, with the model based on limited datasets with a strong geographical focus (Petersdorff, Boermans et al. 2005, ENTRANZE 2008. Regional data availability also limits the level at which the model can be calibrated. In the above scenarios, TIMER-REMG is ultimately calibrated to reflect historic (1971-2018) data on final regional residential final energy demand (IEA 2017). Additionally, floorspace, household size, appliance and air-conditioning ownership, and heating fuel demand are calibrated for each region and income level, subject to data availability (Daioglou, van Ruijven et al. 2012). Residential building stock and building envelope efficiency are calibrated to limited country level data, particularly for the European context (ENTRANZE 2008). Parameterization, calibration, and data availability issues can benefit in the future from ongoing "big-data" activities concerning the built environment (i.e. https://insights.sustainability.google/).
Our reported results for current building stock age profile and thermal efficiency are in line with other similar studies (Edelenbosch, Rovelli et al. under review). Concerning the potential to reduce residential heating and cooling energy demand, our results are broadly in line with other global modelling studies which show that this energy demand can be reduced from current levels through the use of advanced technologies, but not completely eliminated (Knobloch, Pollitt et al. 2018, Levesque, Pietzcker et al. 2018, Gambhir, Rogelj et al. 2019. These studies highlight that complete decarbonization of the residential sector is possible through increased electrification -combined with decarbonization of electricity supply to avoid indirect emissions. The conclusion that decarbonization of the residential sectors depends on a synergistic use of both improvements in thermal insulation and technological improvements in heating and cooling, corroborates the results from more detailed national modelling exercises (Filipidou andJimenez Navarro 2019, van den Wijngaart andvan Polen 2020). These studies also concur with our observation that that the extent of required insulation improvements depends a lot on local climate characteristics, with warmer regions not benefiting significantly from marginal improvements. Concerning our result that renovations lead to a modest improvement on total energy demand, particularly in the OECD, this result concurs with a recent detailed study for the Dutch building sector, which concluded that increasing the insulation levels of the entire Dutch building stock would lead to a reduction in heating demand between 7-27%. These low levels of savings result from the fact that a large portion of the Dutch building stock already have high insulation levels, leading to low marginal gains (van den Wijngaart and van Polen 2020). The potential importance of rooftop PV is in line with conclusions from the IEA which has stated that there is technical potential to meet up to 70% of urban electricity demand, with strong regional variation (Poponi, Bryant et al. 2016). This is in line with our results where rooftop PV contributes significantly to meeting household's electricity demand, even leading to certain households becoming "energy positive" (see Figure S5). However, PV generation is very sensitive to price developments (both PV and grid-electricity), and as shown in Gernaat et al. (2020), a 50% decrease in PV costs could lead to a 3-fold increase in generation.
The TIMER-REMG model does not include certain important dynamics which would affect the results. Of particular importance is the so-called tenant-landlord effect in rented households, where landlords and tenants have limited motivation to invest in efficiency measures, due to their limited personal benefits and lack of long-term returns, respectively. Furthermore, the model assumes that the investment in efficient appliances and building envelopes automatically leads to reduced energy demand and/or emissions. However appropriate usage is of central importance to ensure that these benefits are reaped. Finally, while the model includes financial motivations which affect technology adoption dissagregated across income quantiles (heating/cooling demand, fuel costs, capital costs, discount rates), it does not include social dynamics of technology adoption. Previous studies that included an explicit representation of consumer archetypes (early adopters, laggards) and social influence effects showed these to influence technology diffusion ).
The updated TIMER-REMG can act as a basis for improved understanding of mitigation strategies and costs, as well as a starting point for the analysis of the energy and material demand of the residential sector. By explicitly modelling the construction and decommissioning of building stocks the model can be used to project material demand futures . By making an explicit connection to material use, the model can be used to explore the potential negative emissions arising from the substitution of steel and cement with wooden or bio-based structures which in principle lock-up atmospheric carbon if produced appropriately (Churkina, Organschi et al. 2020, Favero, Daigneault et al. 2020, Pomponi, Hart et al. 2020). Additionally, model can be expanded to disaggregate between different residential building strategies (detached, hi-rise, terraced). This could then be supplemented with Multi Regional Input Output (MRIO) databases and Life Cycle Assessment (LCA) inventories to better understand environmental impacts (beyond energy and GHG emissions) of different urban settlement types and material use strategies. Given that by 2050 it is projected that residential floorspace will increase by approximately 50%, decisions made today may have long-lasting effects on energy, emissions, material use, and broader environmental impacts.
Finally, it is important to note that the results presented are based solely on a "no policy" baseline, and a scenario which meets a 2˚C target by applying a globally uniform price on GHG emissions from all sources. Neither of these scenarios are realistic and thus our projections should not be interpreted as forecasts. Rather they aim to highlight the key dynamics of the sector and the potential emissions mitigation of different strategies. It is important to investigate the effect of specific policies (subsidies, building codes), how they may benefit or burden poorer households, possible economic side-effects such as the free-rider and rebound effects. Furthermore, it is important to investigate the potential synergies of the technological changes highlighted here with demand side management options and lifestyle changes (Pedersen, Hedegaard et al. 2017, Ellsworth-Krebs 2020, Ershad, Pietzcker et al. 2020.

Conclusions
The results presented in this paper aim to show the potential role of technology and energy efficiency decisions in reducing energy demand and emissions from residential heating and cooling. The updated model used includes key dynamics including building-stock turnover, regional climate characteristics, and economic decision making. Several conclusions can be drawn from the results.
Investments in insulation play an important role in reducing energy demand from residential heating and cooling. These investments are expected even in scenarios with no additional climate policy; however, they are significantly increased in a scenario meeting a 2˚C target. Renovation is most important in regions where existing building stocks are of relatively low-quality given local climate characteristics. Globally, investments in improved building envelope efficiency are shown to have to increase to more than $450 billion per year, almost half of which is in renovations, if efficiency levels are to be in line with the 2˚C target. Most of these investments must be made in regions with higher heating demand (OECD, Reforming Economies, Asia). Regions which are projected to increase their residential floorspace in the coming decades (particularly Asia, the Middle East and Africa, and Latin America) are shown to in efficient new buildings in the 2˚C scenario, indicating that they risk getting locked-in into poor infrastructure if these investments are not made early on.
For greenhouse gas emission mitigation, fuel switching accounts for more than half of the emission reduction. Investing in efficient building envelopes can reduce the energy demand for heating and cooling. Compared to today, in the 2˚C scenario global heating demand is expected to decrease by 18% and 64% by 2050 and 2100 respectively. On the other hand, cooling demand is expected to increase by 112% and 201% respectively, but lower than the projected increase in the absence of climate policy. However, meeting the extremely strict emission targets require a near complete decarbonization of the buildings sector. In the 2˚C scenario, global residential heating reduces its direct emissions by 90%, with the OECD and Reforming economies achieving almost 100% reduction. In this light, moving to electrified heating in combination with the complete decarbonization of the power system is required. Heat pumps offer a particularly attractive solution, despite their high upfront investment costs due to their extremely high efficiency. As such they both reduce total final energy demand and direct emissions.
Households meet emission requirements for a 2˚C scenario by combining efficiency improvements and fuel switching. Investments in further insulation approximately double aggregate building envelope efficiency compared to today. The greatest improvements are projected to be in poorer regions, while colder regions have the highest absolute efficiency (see Figure 1). However, even at very high emission prices, extremely high building envelope efficiency levels (leading to U-values below 0.5 W/m 2 /K) are not extensively adopted. These investments take place in-tandem with fuel switching and increased efficiency in heating and cooling technologies. The results show that an equilibrium is reached where improved levels of both heating/cooling efficiency and building envelope efficiency are applied, but neither to their maximum level. Households benefit from a selflimiting feedback between these two, where improved heating/cooling efficiency reduces the need for increased insulation, and vice versa. In this system, the applied emission price (which is calculated on total energy and land use greenhouse gas emissions) acts as a pressure on how far this equilibrium is pushed.
Regions with greater space heating demand benefit more from insulation and electrification of heating. Accordingly, they have the greatest potential to limit their energy demand and emissions throughout the 21 st century. Warmer regions are expected to have an increasing final energy demand as currently un-fulfilled cooling demand is met, driven by expected increases in household expenditures. Thus, while their energy demand is expected to increase, they can still benefit from improvements in cooling efficiency and, to a lesser extent, building envelope efficiency. Distribution of regional household expenditures across rich and poor households is also shown to be very important. Poorer households may face difficulties in investing in technologies and insulation with very high upfront costs, thus keeping them in a low-efficiency, high-demand, high-emission situation. On the other hand, richer households can lower the energy demand and emissions, and even invest in rooftop PV (further benefiting from their larger floorspace) to reduce their net energy demand. This would further exacerbate economic inequalities as richer households can invest their way out of long-term energy and emission costs.
The energy use of the residential sector can become net-zero -or the sector can even become a net producer. The model projects that households produce 5 EJ/yr in 2050, increasing to 35 EJ/yr in 2100 in the mitigation scenario. By combining the electrification of heating (and other energy services), energy efficiency, and the adoption of rooftop PV, the model projections show that households approach, and in some cases surpass, their own energy demand. It is shown that the richest households benefit from this since they are most able to invest in efficiency measures and rooftop PV. Furthermore, they benefit from their large floorspace, increasing their rooftop-PV potential. Warmer regions reach net-zero status first as they benefit from greater solar irradiation and lower heating demand. The increased emergence of "prosumer" households in the projected 2˚C scenario highlights the importance of power grids and electricity storage technologies which would be able to manage the increased mismatch and variability of supply and demand.