Life cycle environmental benchmarks for Flemish dwellings

To reduce the environmental effects caused by building construction and operation, life cycle assessment (LCA) is increasingly applied. In recent years, national building regulations have implemented LCA requirements to support building life cycle impact reduction. A key element in these regulations are environmental benchmarks which allow designers to compare their building designs with reference values. This study aims to develop bottom-up life cycle environmental benchmarks that represent the range of environmental impact results achieved with conventional construction in Flanders, Belgium. For this purpose, the study investigates the potential of using a database of building energy performance calculations. Specifically, this study considers 39 residential buildings identified as representative of the Flemish energy performance of buildings database of 2015–2016, applying modifications to establish scenarios that are still relevant in 2025. The buildings are assessed with the Belgian LCA tool TOTEM to calculate an aggregated environmental score based on the European product environmental footprint (PEF) weighting approach and including 12 main impact categories. In addition to the aggregated score, the climate change (CC) indicator is analysed individually. In view of the benchmarks, variations were applied to the 39 original buildings in terms of heating system and materialisation. The variation in heating system included changing gas boilers to electric heat pumps to comply with upcoming (2025) Flemish building regulations. The variations in building materials included three sets of conventional Flemish building element compositions that were applied to generate a wider spread of impact results as a basis for benchmarks. Benchmark values were derived through a statistical analysis of the 117 modelled variants: a best-practice value (10th percentile), reference value (median) and limit value (90th percentile). For the environmental score, the benchmark values are 86, 107 and 141 millipoints per square meter of gross heated floor area (GHFA) (mPt m−2GHFA), respectively; and for CC, the benchmark values are 844, 1015 and 1284 kg CO2-eq m−2 GHFA. Finally, the study discusses the representativeness, implications and limitations of the final benchmarks and benchmark approach.


Introduction 1.Building environmental impact assessment and benchmarking
The world is currently dealing with multiple threats to our environment, such as climate change (CC), biodiversity loss and increasing pollution of air, water and soil.The building sector plays a significant role in this, as building construction and operation require considerable amounts of energy and resources and generate large amounts of waste.Building construction and operation are for example responsible for over one third of global CO 2 emissions globally, indicating there is an urgent need to implement reduction measures in this sector [1].
Life cycle assessment (LCA) enables the evaluation of various impacts caused by a building over its entire life cycle.Currently, LCA is typically used voluntarily to compare the impact of multiple variations of a design and/or identify its environmental 'hotspots' and reduce the impact accordingly.However, in recent years several countries have been exploring the implementation of LCA in building regulations to enforce building impact reduction [2][3][4][5][6][7][8][9][10].France for example has implemented mandatory limit values for embodied and operational carbon of residential buildings in the RE2020 regulation, after a five-year test phase of the voluntary label E+/C− [2,3].In Denmark, all new buildings are required to perform an LCA, and a threshold applies for the embodied carbon of buildings over 1000 m 2 [4].In the Netherlands, building LCA has been mandatory since 2013, and five years later a limit value was implemented for an aggregated environmental (embodied) cost which includes 11 different impact indicators [5,6].Furthermore, the Nordic countries are collaborating on harmonisation of their LCA methods and are collectively investigating how to implement LCA requirements in building regulation [7].Norway and Sweden require new buildings to report embodied carbon, and Sweden and Finland will implement mandatory threshold values in the coming years [8,9].Outside of Europe, Canada has currently the most advanced regulatory measures, having implemented mandatory reporting of and reduction targets for embodied carbon of structural materials of major construction projects since 2023 [10].
An essential component in these regulations is the assessment and benchmarking of building life cycle impacts.Such benchmarks support the development of targets and mandatory threshold values that buildings should comply with in the long and short term, respectively [11].Additionally, benchmarks help building stakeholders understand the environmental performance of a building in a broader context [12,13].Hence, accurate definition of building environmental benchmarks is crucial to support building impact reduction strategies.
Similar developments in building LCA have been emerging in Belgium.Like other European countries, Belgium developed a national building LCA method called 'Environmental Profile of Buildings (MMG)' which is in line with the European LCA standards EN 15804+A2 and EN 15978 [14,15].The MMG method involves environmental data and life cycle scenarios for construction products that depict the Belgian construction sector [16].Since 2018, the MMG method is available as a web-based tool called TOTEM (Tool to Optimize the Total Environmental impact of Materials) [17], which is supported by the three Belgian regions.TOTEM is currently used on a voluntary basis to compare the environmental impact of building variants, but it is lacking benchmarks that represent the range of impact of current Belgian construction practice.

Approaches to building environmental benchmarks
Based on the benchmark aim and the data source, two main approaches are distinguished to define environmental benchmarks for buildings.The first is a top-down approach which defines long-term benchmarks for buildings with the aim of fulfilling global environmental goals or policy targets.The second one is a bottom-up approach in which benchmarks are derived from a statistical analysis of the building stock, i.e. by evaluating either representative real buildings or generic, fictive archetypes [18].All national policies mentioned earlier employ bottom-up benchmarks.
Most studies in the literature also apply the bottom-up approach, albeit with different types of building datasets.Several studies apply a data-driven approach and derive benchmarks from real case studies or LCA results, often collected from other studies or from databases of existing rating tools [19][20][21][22][23][24][25][26][27][28][29][30][31][32].Consequently, these datasets generally include the results of LCAs with varying scopes and methods and different background data [29][30][31][32][33][34].Moreover, as these datasets contain buildings that were designed to be low-impact or to achieve a certification, they may not be representative of common construction practices [28,32,34].Alternatively, several studies apply an archetype approach to derive benchmarks from a limited set of buildings representative of a particular building stock [35][36][37][38][39][40][41].Some of them include multiple variations of the buildings, e.g. in terms of construction type, energy performance level and climate zone [37][38][39][40][41].By applying variations to the same set, the influence of material choices and possible mitigation strategies can be explored.Finally, these approaches can be combined or supplemented with other strategies, such as building stock modelling.For example, Verellen and Allacker [42] developed a building stock model by combining a building-by-building approach, i.e. modelling each building of a stock, with an archetype approach for the purpose of benchmarking.

Energy performance of buildings (EPB) database
Under the European energy performance of buildings directive (EPBD), Member States are obligated to implement minimum energy performance requirements for new buildings and extensive renovations [43].Consequently, Member States generally require energy performance calculations for new buildings, implying they have extensive databases of EPB calculations including data on geometry and energy performance of new buildings.Furthermore, the data collected from these assessments have been used to investigate the cost-effectiveness of energy performance levels in many Member States [43].EPB databases provide valuable input data for LCA as well, but are currently not used to assess building life cycle environmental impacts, let alone for developing environmental benchmarks.Possible hurdles are the presumably incomplete building geometry data (i.e.only building thermal envelope is included), the relatively large size of the databases making it challenging to manage, or the fact that building life cycle impacts have only recently become of interest in European policy.However, the revised EPBD [44] will impose mandatory reporting of life cycle carbon emissions, making it relevant to explore the link between EPB and LCA calculations and with benchmarking accordingly.

Aim and scope
This study aims to reveal how building data collected in the context of energy performance regulations can support the development of life cycle environmental benchmarks for buildings.The EPB database of new residential buildings in Flanders (Belgium) is used as a case study.Like many EU countries, the Flemish region has conducted regular cost-effectiveness studies on the EPBs [45][46][47][48].At the start of this research, the most recent cost-effectiveness study for residential buildings dated from 2017 [49].The study selected a set of single-family houses (SFHs) and multi-family houses (MFHs) from the Flemish EPB database of 2015-2016 that were considered representative in terms of energy and geometry.Whereas the cost-effectiveness study explored the financial cost of energy performance levels, this paper focuses on benchmarking the life cycle environmental impact of the cases by performing LCA.
In a preliminary study [50], five sample cases were selected from the representative SFH to test the workflow between the EPB data and the LCA modelling and compare the results to an earlier benchmarking study [51] of Flemish dwellings.In this paper, the full set of representative buildings is used to derive benchmarks for the environmental impact of new residential buildings using the Belgian TOTEM tool.Firstly, some adaptations are made to the original cases (i.e. from 2015-2016) to align them with Flemish EPB 2025 requirements [52] and then evaluate how these adaptations affect the benchmarks.Secondly, material variants are applied to the adapted cases to extend the range of embodied impact results as a basis for the benchmarks.The cases already represent a wide range of possible operational energy results for dwellings that follow the projected 2025 requirements.This paper therefore focuses on expanding the embodied impact results in view of developing benchmark values that represent low environmental impact, average impact, and high impact dwelling materialisations.
Benchmarks are calculated both for the embodied and operational impact separately and for the life cycle impact in total.As opposed to most of the benchmarks in the abovementioned national policies, this paper focuses not just on carbon but considers all 12 main impact categories recommended by the European EN 15804+A2 standard [14].

Methods
This section describes the steps taken to develop benchmarks for Flemish dwellings, including the selection of representative dwellings, the assessment method, and the final benchmark definition.Further, it presents the dwellings themselves and elaborates on the approach developed to generate results as a basis for the benchmarks.

Representative Flemish dwellings from the EPB database 2.1.1. Selection procedure for representative dwellings
The buildings used in this study include 30 SFHs and nine MFHs selected from the EPB database of residential buildings submitted in 2015-2016.The 39 buildings were selected for a cost-effectiveness study of energy performance levels in Flanders in the context of the European EPBD [49].The study identified 30 SFHs and 28 apartment units (part of nine different MFHs) that could be considered representative of the overall pool of 4134 SFHs and 1244 MFHs in the database, respectively, in terms of geometry and energy performance.
In the cost-effectiveness study [49], the representative dwellings were selected in two steps.In the first step, the SFHs and MFHs were subdivided into clusters and sub-clusters based on geometric and energy parameters.Two clustering approaches were applied.In the first approach ('geometric clustering'), the databases were divided into five clusters based on GHFA and compactness (i.e.volume/heat loss area ratio), which were each further subdivided into five sub-clusters based on window area (in m 2 ) and the overheating indicator (in Kelvin-hour/year or Kh yr −1 , EPB method).Figure 1 shows the distribution of the selected SFHs across the five big clusters and across five sub-clusters within two of the big clusters.The second clustering approach ('all-in clustering') was performed to enable 'peculiar' cases to form clusters (e.g.high compactness but poor energy performance) by considering the following parameters: -GHFA (m 2 ) -compactness (m −1 ) -window area (m 2 ) -overheating indicator (Kh yr −1 ) -thermal and energy performance level ('K-value' and 'E-value') -air tightness: leakage rate in m 3 hm −2 at a 50 Pa pressure difference ('v 50 -value') -factor representing the quality of execution of the ventilation system ('m-factor') -reduction factors for demand-based ventilation ('f reduc ') and for preheating ('r preheat ') -efficiency or seasonal performance factor of the central heating system -energy for heating, domestic hot water (DHW), cooling, and energy gains from photovoltaic (PV) panels.
The all-in clustering approach resulted in 30 SFH clusters and 31 MFH clusters.In a second step, cases were selected in the cost-effectiveness study that could be considered as representative of all defined clusters.They applied a least-squares method to identify cases that were as close as possible to a cluster centre of both a geometric cluster and an all-in cluster, resulting in 30 SFH and 28 apartment units (from nine different MFH).Whereas the entire database consisted of an SFH-apartment ratio of 77%-23%, a 50/50 ratio was adopted as this had been, and still is, the trend over the past years [49,53].In a third step, the representativeness of the selected cases for the database was validated by performing two-tailed z-tests and cumulative frequency analyses for the parameters considered in both clustering approaches.Hence, the selection contains a variety of operational energy results obtained under the 2015 EPB requirements.Contrarily, the materialisation of the buildings is never considered in the selection procedure, implying that the buildings are not necessarily representative in terms of materialisation and therefore embodied impact.Nevertheless, the cases cover the wide range of GHFA, compactness and window area of the entire database, which are parameters that have been suggested to have significant influence on not only the operational but also embodied impact [54][55][56][57].
In the cost-effectiveness study, a few of the initially selected cases were omitted because of unusual or incomplete data, and some of the larger cases were replaced by smaller ones based on stakeholder feedback.This paper considers the 39 cases originally identified as representative, therefore deviating from the selection presented in the final report of the cost-effectiveness study [49].

Characteristics of the representative dwellings
Figure 2 shows simplified representations of the selected cases and the spread of dwelling types, number of storeys, construction types and heating systems among them.A table containing more detailed geometry information (e.g.GHFA, volume, building element ratios per m 2 GHFA, etc) is provided in the supplementary data.Among the SFHs and MFHs there is a significant variety of geometries and characteristics (e.g.roof types, basements, number of storey floors, construction type).The 30 SFHs include the three typical Flemish dwelling types, more specifically, 12 detached houses (40%), 10 semi-detached houses (33%) and eight terraced houses (27%).Note that the shares of detached, semi-detached and terraced houses in the selection is different from the ratios in the entire database (39%, 47%, 14%, respectively).The MFHs range from two-storey buildings with two apartment units to a nine-storey building with 24 apartment units.One of the MFHs (IDA6) consists of eight studio units.Further, all 9 MFHs and 25 of the SFHs are solid construction (i.e.bricks and concrete), and the remaining five SFHs have timber frame construction.

LCA method
As the benchmark values are intended to be applicable in the Flemish regulatory context, the Belgian LCA-tool TOTEM was used.The reference study period (RSP) in TOTEM is 60 years.The functional unit is 1 m 2 of GHFA (m 2 GHFA) of building, which is in line with the EPB software, hence the EPB energy use data did not have to be converted.The system boundaries include the following stages (according to EN 15978 [15]): A1-3 product stage, A4-5 construction process stage, B2 maintenance, B4 replacement, B6 operational energy use, and C1-4 end of life stage.The building models included the following elements (according to BB/SfB classification [58]): (13.)+ floor on grade, (16.2)+ basement wall, (21.)+ external wall, (22.)+ internal wall, (23.)+ storey floor, (27.)+ roof, (31.) external window and door, (32.) internal door, (53.) water supply, (54.) gas supply, (55.) space cooling, (56.) space heating, and (57.) ventilation.The preliminary study [50] did not include the embodied impact of technical installations (i.e. the latter five categories).For the MFHs, the building models included the MFH as a whole, i.e. including all apartment units as well as shared spaces.
The environmental data of generic building components in TOTEM are based on building material and process records from ecoinvent v3.6 [59].The life cycle impact assessment (LCIA) covers all 12 main impact categories recommended by the European EN 15804+A2 standard [14].The LCIA results include both the results of each impact indicator individually as well as an aggregated environmental score.The aggregated score is calculated based on the European product environmental footprint (PEF) weighting method, which includes a hybrid evidence-and judgement-based weighting scheme, and is expressed in millipoints per functional unit (mPt/FU) [60].This study considers both the aggregated environmental score and the individual indicators and pays particular attention to the CC indicator.Note that this study as well as TOTEM apply the PEF method to comply with European recommendations, but that PEF is one of many possible weighting schemes for LCA results.Furthermore, PEF does not consider the hotspots in building life cycle impacts, i.e. the environmental indicators most relevant to buildings, but is a general weighting method for product LCA.

Conversion of EPB input data to TOTEM LCA model 2.3.1. Spreadsheet template for consistent workflow
The data of the cases were provided in the form of anonymised EPB files, including all building data required to perform an energy performance calculation with the EPB software.To conduct an LCA of the buildings, the data had to be converted to input data for the Belgian LCA tool TOTEM.The input data required for TOTEM are the building GHFA and the amount and composition of all building elements covered in the current version of the tool, presented in section 2.2.For operational energy, the yearly final energy use for heating, cooling (if applicable), DHW and auxiliaries calculated by the EPB software can be used as input in TOTEM.
Because the EPB files are limited to data needed for an energy performance calculation, two critical data gaps had to be tackled: (1) the amount and composition of internal building elements and (2) the materialisation of finishings and foils that are omitted in EPB because of their negligible influence on the thermal performance.To ensure consistency in the modelling of the cases in TOTEM and to fill the identified data gaps, a detailed workflow was defined in the form of a spreadsheet template.More information about the workflow and filling of data gaps is presented in the preliminary study [50].

Additional scenario to align with 2025 EPB requirements
The Flemish EPB requirements are updated almost yearly, hence several of the selected buildings might not be in line with the current (2023) requirements.Currently, the EPB requirements in force up until 2025 have been defined.A critical change is that from 2025 new buildings cannot connect to the natural gas network anymore and will therefore have to provide heating by either electric heat pumps, district heating, biomass-fuelled boilers and/or direct electric heating [52].
As shown in figure 2, 23 of the SFH and all but one unit of the MFH include a condensing gas boiler for heating and DHW.As this study aims to calculate benchmarks for new buildings, which will soon no longer allow gas boilers, the cases were modelled in an additional scenario that includes a heat pump.The Flemish Government states that the 2025 requirements will be implemented in a way that at least a heat pump should be installed in new residential buildings, which it will support by providing subsidies [61].An electric air/water heat pump was chosen as this is currently the only installation available in TOTEM that is in line with the 2025 requirements.Moreover, this type was also applied in the case that only has a heat pump (ID14) and three of the four cases with a hybrid system.Specifically, the air/water heat pump of ID14 was applied to all SFHs, and the air/water heat pump of the upper unit in IDA8 was applied to all MFH units.The SFH heat pump has a COP of 3.6 and a water temperature of 45 • C, which is also the maximum allowed temperature from 2025 [52].The MFH heat pump has a COP of 3.5 and a water temperature of 35 • C. Accordingly, high temperature radiators were replaced by low temperature radiators.
Other changes to the requirements since 2015 include lower maximum U-values for some building elements and updated requirements for solar energy systems.As TOTEM currently does not include the latter, the additional scenario does not consider energy gains nor embodied impacts resulting from solar energy systems.Regarding the U-values, this issue was overcome by applying the material variants, presented in section 2.4.Besides, many of the cases already fulfil the 2025 requirements.Moreover, a preliminary analysis [62] of five sample cases showed that increasing the insulation thickness of building elements to achieve the 2025 requirements has negligible influence on the life cycle impact of the buildings.Since this scenario is only used for comparison with the results of the material variants and not for the benchmarks itself, the building elements were not modified in the updated scenario.

Material variants: definition and application
As mentioned in section 2.1.1,no claims can be made about the representativeness of the materialisation and thus the representativeness and/or the spread of embodied impact of the cases.To develop a set of building materials that is both representative of current Belgian construction practice and generate a wide range of impact results, three material variants were applied to all cases.The three variants are defined as a combination of building elements with a low environmental impact, average impact and high impact in terms of aggregated environmental score.The building elements were derived from a statistical analysis of the impact results of the predefined building elements included in TOTEM.Most elements in this library are typical for Belgian residential construction.
For each building element type, all elements in the TOTEM library that are applicable for new residential construction with a U-value closely aligning with the EPB requirements were considered (i.e.no more than 0.04 W m −2 K −1 lower or 0.02 W m −2 K −1 higher).For walls, storey floors and flat roofs, a distinction was made between solid and timber elements.Hence for every building element type, either one set or two (i.e. one solid and one timber) sets of building elements that were considered applicable were derived.For each set, the elements closest to the 10th percentile, median and 90th percentile were identified.The selected elements were then combined into the following groups: three groups of solid elements, i.e. a group of the 10th percentile, median and 90th percentile elements; and, likewise, three groups of timber elements.The 10th and 90th percentiles were chosen over the minimum and maximum because it excluded extreme results, it excluded elements that are uncommon (although possible) in residential construction, it resulted in element groups that were more physically plausible, and it is consistent with benchmark definition in general [11].
Within each group, small adjustments were made to some of the building elements to create element groups that are physically logical.For example, the structure of the storey floor and flat roof was matched, and the internal wall and ceiling finishes were harmonised across all elements.
Table 1 clarifies the different variants that have been considered in this study.As explained in section 2.3.2, a first variant applied to the cases is the 2025 scenario (Or25) where the original building materialisation is maintained, but gas boilers are replaced by an air/water heat pump.This variant is compared with the original buildings (Or15) to evaluate the effect of the 2025 EPB requirements on the life cycle impact of buildings.
The material variants were applied to the heat pump scenario (Or25) of each building.This ensures that the benchmarks derived from these variants are still relevant when the Flemish 2025 EPB requirements come into force.For each case, its original construction type was maintained.It is assumed that the building geometry would have been different if the buildings were designed for another construction type, which implies that a mixed combination (e.g.originally solid case with timber materialisation applied) would be less relevant.Therefore, the solid material combinations were only applied to the cases that are originally solid constructions, and the timber material combinations were only applied to the timber constructions.Hence, for each building, three material variants were considered (Var10P, VarMed, Var90P).Knowing that building geometry can significantly influence the building impact [63], the relationship between building element ratios and the variant results was further investigated.

Benchmark types and definition
The benchmarks were derived from a statistical analysis of the building variants.Specifically, the analysis includes the three material variants (i.e.Var10P, VarMed, Var90P) for all 39 cases, hence 117 buildings in total.Based on the 117 buildings, the following benchmark values are defined: -best-practice value: 10th percentile: ambitious level of performance achieved when applying good practices in terms of both embodied and operational impact; -reference value: median: representative of conventional Belgian construction practice (2025) when no thresholds nor reduction targets are imposed; and -limit value: 90th percentile: lowest acceptable performance, obtained by inattentive design in terms of both embodied and operational impact.
The abovementioned definitions are based on benchmark definitions in the literature [18,64] and the ISO 21678 standard on building sustainability benchmarking [11], and made explicit in the context of this paper.

Results
Firstly, the LCA results of the buildings as originally modelled and with the variation in heating and DHW system are compared.Secondly, the results of the application of material variants are presented, including both the impact of the applied building elements themselves and the results of their application to the buildings.Lastly, the benchmark values derived from the material variants are presented and compared to the buildings in their original materialisation.

Original buildings and updated scenario 3.1.1. Aligning with EPB 2025 requirements: replacing gas boilers with heat pumps
Figure 3 shows the life cycle environmental score of the buildings as originally implemented in EPB (Or15), i.e. the majority including condensing gas boilers for heating and DHW, and of the variant in which the boilers are replaced by air/water heat pumps (Or25).From top to bottom, the results are sorted from lowest to highest reduction in environmental score between Or 15 and Or 25.Note that the uppermost six buildings already included a heat pump and therefore show the same results for both scenarios.The results for CC are provided in the supplementary data.
For the cases that do not include a heat pump in Or15, the lowest reduction observed is 1% for MFH IDA6 (studios), and the highest reduction is 17% for semi-detached house ID23.Excluding the cases that already employed a heat pump, the average reduction in environmental score is 10%.The apartments show on average the lowest reduction of 7%, while the detached, semi-detached and terraced houses show an average reduction of 11%-12%.Figure 3 further reveals that the magnitude of the reduction is not directly correlated to the magnitude of the environmental score of Or15, nor to the GHFA or compactness of the buildings.
The total reduction in environmental score is for all cases the result of an increase in embodied and decrease in operational environmental score.The increase in embodied environmental score is on average 16% and ranges from 6% for detached house ID17 to 38% for IDA6, the MFH with studios.This increase is caused by the higher embodied impact of the heat pump, which is around 16 times higher than the impact of the gas boiler in TOTEM.The average reduction in operational environmental score is 34% and ranges from 27% for semi-detached house ID01 to 40% for terraced house ID24.Further, the MFHs show on average the largest increase in embodied impact (20%) and the largest reduction in operational impact (37%).This is the result of the lower water temperature in the heat pumps applied in the MFH.A further analysis showed lower reductions in operational energy when the SFH heat pump was applied to one of the MFH.

Embodied, operational and life cycle impact per dwelling type
Figure 4 presents the spread of the embodied, operational and life cycle environmental score per dwelling type for both the original Or15 and updated Or25 buildings.In both scenarios, the MFHs have the highest embodied impact, but the lowest operational impact.The MFH's life cycle impact is similar to the life cycle impact of the detached houses.For the SFHs, the difference between the three dwelling types becomes more apparent in the Or25 scenario for the life cycle and operational impact.
The increase in embodied and reduction in operational impact cause a shift in their relative shares to the total impact in the Or25 scenario.Whereas in the original Or15 scenario the average ratio is 49/51 for the SFHs and 54/46 for the MFHs, in the Or25 scenario this shifts to 59/41 and 69/31 for the SFHs and MFHs, respectively.The embodied impact represents less than 50% for about half of the Or15 cases, and is more than 50% for all Or25 cases, with up to 71% for five MFH.The supplementary data shows that this shift is even more noticeable for CC impact individually.As shown by Röck et al [65], this is a typical trend observed with advancing building energy efficiency standards, as these lead to higher thermal insulation levels and integration of renewable energy systems and hence give rise to building embodied impacts.

Contribution of impact indicators to the aggregated environmental score
Figure 5 presents the contribution of the individual impact indicators to the aggregated environmental score of all buildings in the Or15 and Or25 scenarios, sorted by ascending CC impact of the Or25 variants.For all Or25 buildings, the most important indicator is Depletion of abiotic resources (ADP), ranging from 27% to 35% of the environmental score.CC is for all cases the second highest indicator (23%-27%), followed by particulate matter emissions (8%-21%) and eco-toxicity (8%-11%).Other indicators contribute less than 5% each.
The results show a significant shift in the relative contributions of indicators between Or15 and Or25.For the cases with gas boilers, the most important indicator is CC (34%-40%) followed by ADP (27%-32%).However, for none of the buildings has ADP increased in absolute terms.On the contrary, ADP has reduced around 3%-14%.Switching to heat pumps causes an increase in all indicators except CC, ADP and photochemical ozone formation.
Figure 5 further reveals that the environmental score of the Or25 buildings largely follows the same increasing trend of the CC, with some notable irregularities.Remarkably, the most prominent irregularities are caused by the timber buildings.For the two buildings with the lowest CC, the timber case ID13 shows an 11% higher environmental score than solid case ID19.Similar differences are observed for the other timber cases and their neighbouring cases in the graph.Timber construction generally has a lower CC impact than solid construction, but a higher impact in the categories particulate matter and land use [66], which is also observed here.Note however that this not only caused by the construction type, but also the geometry and therefore the element quantities per building.

Building material variants 3.2.1. Building element variants and their life cycle impact
Tables 2 and 3 present the three building material variants for solid and timber construction, respectively.All combinations represent conventional Belgian construction practices but show a wider range of impacts than the elements of the original cases.Figure 6 reveals the positioning of the life cycle environmental score of the selected building elements in the TOTEM element library for the three material variant combinations.For elements part of the building envelope, an operational impact is calculated based on the transmission losses, which is related to the elements' U-values.
Compared to the median element, the difference with the 10th percentile element ranges from −2% (external window) to −72% (party wall, solid), and the difference with the 90th percentile element ranges from +2% (flat roof, timber) to +204% (pitched roof).Across all element types, the 90th percentile can be from 1.3 times (flat roof, timber) to 4.7 times (pitched roof) larger than the 10th percentile element.The relative differences are smaller for the timber elements than for the solid elements, except for the storey floors.Furthermore, in most cases the selected timber elements perform better than the selected solid elements, except for the 10th percentile and median of the loadbearing internal walls and the median of the storey floors.A table presenting the life cycle and embodied environmental score, and the relative difference between the three selected elements of all element types is provided in the supplementary data.

Life cycle environmental score of the building material variants
Figure 7 shows the life cycle environmental score of all buildings with original materialisation (Or25 scenario) and with the three material variant combinations applied.From to bottom, the buildings are grouped per dwelling type and then sorted from smallest to largest life cycle environmental score of the original building.
Most of the original buildings are closest to, and generally slightly larger in environmental impact than the median variant.None of them perform better than their 10th percentile variant nor worse than their 90th percentile variant.One MFH, one detached house, three semi-detached houses, and eight out of the nine terraced houses have an environmental score below the median variant.For 10 out of those 12 buildings, the difference with the median is less than 3%.For the buildings that perform worse than their median variant, differences ranging from less than 1% to 20% are observed.The building that is closest to its 10th percentile   variant is semi-detached timber house ID14, with an original impact of 4% more than the percentile variant.On the opposite end, the closest to its 90th percentile variant is the 24-unit MFH IDA9, being 2% lower.
As intended, the range of results has expanded by application of the material variants.For the original buildings, the highest environmental score (IDA6, 140 mPt m −2 GHFA) is 1.7 times higher than the lowest  4 summarises the average relative differences in life cycle environmental impact for all buildings, per dwelling type and per construction type.The percentages represent average increase or reduction in environmental score.
On the 10th percentile variants are 18% lower than the original buildings and 31% lower than the 90th percentile variants.The latter are on average 18% higher than the original buildings.Compared to the original buildings, the apartments show on average the biggest reduction (24%) obtained by the 10th percentile variants and the lowest increase (+13%) obtained by the 90th percentile.Contrarily, the semi-detached and terraced houses show the lowest average reduction (14%-15%) and biggest increase (22%) for the 10th and 90th percentile variant, respectively.Comparing solid and timber construction types, both show very similar relative differences between the variants.

Influence of building geometry on embodied impact
To understand the impact of the building geometry on the embodied impact, the geometric characteristics are further investigated.The building element ratios per building are included the supplementary data.Figure 8 shows the environmental score of the variants as a function of the compactness and in relation to the GHFA.
The only element that seems to have a significant impact on the building embodied impact is the windows.The window-to-floor ratio is 12%-21% for the (five) cases with the smallest difference between Var10P and Var90P, while it is 17%-39% for the (five) cases with the biggest difference.No such direct correlations can be traced for the other elements.That is, apart from windows, big differences in embodied impact between variants cannot be explained by a big ratio of one particular element.Only for the detached buildings, i.e. which do not have party walls, the differences appear bigger for dwellings that have a relatively high pitched roof to GHFA ratio.This is caused by the big difference between the 10th percentile and 90th percentile element for both party walls and pitched roofs, as shown in section 2.4.For semi-detached and terraced houses, the influence of the roof is likely less noticeable due to the inclusion of party walls.
Both the cases with the lowest Var10P and/or the cases with the highest Var90P show a mix of dwelling types, sizes (GHFA), geometries (building element ratios), compactness, and construction types.The same holds for the cases that show the biggest difference between these two variants.For the SFHs, the buildings with lower-impact variants generally have more storeys than the buildings with high-impact variants.Specifically, the buildings with high-impact variants are one and two-storey dwellings, while the buildings with low-impact variants include two or three storeys.Furthermore, the SFH that achieve the highest impact have relatively low GHFA and compactness.This observation, i.e. large houses having low life cycle impacts per m 2 , has also been found in other studies [54,63,67].For large buildings, the total impact is divided by a bigger floor area, resulting in a lower impact per reference unit.The use of the floor area as reference unit is hence discussed in literature as it favours larger houses although they have a bigger impact in absolute terms.This paper additionally highlights how a functional unit of floor area poses an issue in the context of benchmarks, since the benchmarks should ultimately drive impact reduction in the building sector, which is only obtained if building impact reduces in absolute terms (and not solely per m 2 floor area).

Benchmark values
Table 5 presents the benchmark values for the environmental score and the CC indicator, for the life cycle impact as well as the embodied and operational impact separately.Note that the sum of the embodied and Table 5. Benchmarks derived from the statistical analysis of the whole dataset of 117 variants: best-practice (10th percentile), reference (median) and limit (90th percentile) values for the life cycle, embodied and operational environmental score and climate change.The statistical values were directly derived from the set of 117 scenarios, i.e. no weighting was applied in view of the representativeness of the dataset of the actual building stock.operational limit value is higher than the life cycle limit value, and for the best-practice and reference value the opposite is true.The differences between the benchmarks are bigger for the environmental score than for CC.For the life cycle environmental score, the best-practice value is 19% lower than the reference value and 39% lower than the limit value.For life cycle CC, the differences are 17% and 34%, respectively.The same holds for the embodied impact.The embodied environmental score's best-practice value is 26% lower than the reference value, and 51% lower than the limit value, while for CC these differences are 21% and 43%, respectively.

Best-practice
The benchmark values were also calculated separately for the SFHs and MFHs, and for the different SFH types.These results are provided in the supplementary data.For the life cycle environmental score, all SFH benchmarks are the same as the values presented in table 5, and for MFHs they deviate 1-3 mPt m −2 GHFA.The distinction per SFH type gives differences of up to 14 mPt m −2 GHFA compared to the values in the table, with higher values for the detached and semi-detached houses and lower values for the terraced houses.The latter obtain both lower embodied and lower operational impact because of the lower quantity of enclosing wall allocated to the terraced houses (i.e.only half of the party walls) on one hand and the lower heat loss area on the other hand.
Figure 9 shows the environmental score of the original cases Or25 grouped per dwelling type.The best-practice (green), reference (yellow) and limit (red) value obtained from the statistical analysis of the building from the three variants are indicated by vertical lines.For the original buildings, the colours of the bars show whether they are below the best-practice value (green), between best-practice and reference (yellow), between reference and limit (orange) or higher than the limit value (red).
The only building that already achieves the best-practice value is terraced house ID19, which is 2 mPt m −2 GHFA below.On the other hand, the environmental score of the MFH with studios IDA6 is equal to the limit value.Of the remaining 37 buildings, 15 are in between the best-practice and reference value, and 21 are in between the reference and limit value, and one terraced house (ID12) equals the reference value.Both for the MFHs and detached houses, the majority are higher than the reference value.For the MFH, two are below the reference value, i.e.IDA2 with two apartments and IDA7 with eight apartments.For the detached houses, four of the twelve are below the reference value.For the semi-detached houses, five of the ten are higher than the reference value, and for the terraced houses only one out of eight are higher than the reference value.

Discussion
Firstly, the benchmark values as well as their objective are critically compared to the literature.Then, the implications of the benchmark method and assumptions and the application of the benchmark values are discussed.Finally, this section reflects on the limitations of the benchmark approach and data and presents recommendations for future research.

Comparison with benchmarks in the literature
In figure 10 the obtained benchmark values are compared to values in the literature reviewed by Trigaux et al [18].The review analysed benchmark values from regulations, labelling systems, rating tools and research.A statistical analysis of the reviewed benchmarks for residential buildings resulted in median values of 5.7 and 13.3 kg CO 2 -eq m −2 .yearfor embodied and life cycle CC, respectively.
For comparison, the benchmark values in table 5 are divided by the RSP of 60 years.For life cycle CC, the limit, reference and best-practice results are 21.4,16.9 and 14.1 kg CO 2 -eq (m 2 yr) −1 , respectively.For embodied CC, this gives 14.1, 10.2 and 8.0 kg CO 2 -eq (m 2 yr) −1 .Hence the life cycle and embodied CC benchmarks obtained in this study are 1.3 and 1.8 times higher, respectively, than the median values derived in the review.On top of the variations in scope and system boundaries, this difference is primarily caused by the fact that the review also includes top-down benchmarks derived from global or national goals, providing more ambitious target values [18].Furthermore, it is important to note that among all benchmarking systems, whether derived bottom-up or top-down, benchmarks are derived for a specific purpose.Bottom-up approaches, such as the ones referred to in section 1.2, are often applied to derive benchmarks that are representative of an existing or new building stock.This study, however, did not aim to develop benchmarks representative of the Flemish stock, but rather benchmarks that display the range of impact achievable with current building practice.The approach applied in this study is therefore less suitable for calculating values representative of actual Hence, on top of the differences in benchmark method presented by Trigaux et al [18], the purpose of benchmarks is at least as important when comparing with and interpreting benchmarks.Note, however, that the benchmark purpose will be reflected in the benchmark method.

Relevance of technical installations
The comparison of the original Or15 with gas boilers and updated Or25 scenarios with heat pumps in section 3.1 reveals the influence of these heating systems on the LCA and consequently, on environmental benchmarks.
The comparison validates that installing heat pumps instead of gas boilers in new buildings results in significant reduction in life cycle impacts, both in aggregated score and CC individually.Eliminating gas as a source for heating and DHW can therefore be an important part of the decarbonisation of the building sector.Nevertheless, it is important to note that the reduction in life cycle impacts is obtained by a large reduction in operational impact that compensates for the relatively smaller increase in embodied impact.This highlights the importance of considering both the operational and embodied impact of alternative technical installations in buildings.This is especially relevant when evaluating the impact of systems for heating and DHW that were developed in view of energy reduction, as their embodied impact might partly compensate the reduction in the operational impact.
In view of developing benchmarks for new buildings, the comparison demonstrates that the assessment should consider the changing building regulations in the future.In this case, the shift from gas boilers to heat pumps for heating and DHW resulted in reductions up to 20% and 50% in life cycle environmental score and CC, respectively.So even though gas boilers are still conventional practice today, the benchmarks derived from such a scenario would not be representative of new construction likely to occur in future years.
Related to that is the decarbonisation of the electricity mix.A study by Norouzi et al [68] found that a 'steady progression' in decarbonisation of the UK electricity mix could result in a 54% reduction of a building's life cycle CC.Although electricity mix is inherently country-specific, the study demonstrates that developments in decarbonisation of the electricity mix significantly affect building impact and benchmarks.In addition, it is important to note that decarbonisation of the electricity mix will further amplify the shift in impact indicators already observed by swapping the heating system.

Implications and application of the benchmarks 4.3.1. Level of ambition
Analysis of the potential correlation between building geometry (i.e.building element ratios) and LCA results showed that various geometries can achieve a big reduction in life cycle impact.Buildings with the biggest difference between their high-impact variant Var90P and low-impact variant Var10P include different dwelling types of different sizes and geometries (e.g. with or without basement, with flat and/or pitched roof, etc), and both solid and timber construction.Also, the lowest environmental scores were achieved by varying building geometries, implying that if a threshold value is defined, there is still sufficient design freedom to achieve this benchmark.However, buildings that have the highest impact in the Var90P variant generally have a relatively low compactness, low GHFA and fewer floors.Therefore, these buildings would be the first to be required to reduce their life cycle impact if a threshold value were in force.As argued by Stephan and Crawford [54], a functional unit that considers house size or number of inhabitants should be considered in policy to avoid large buildings being favoured and hence to effectively reduce the absolute impacts caused by buildings.Another approach would be to impose reduction targets instead of threshold values so that large buildings would eventually undergo bigger absolute impact reductions.
The fact that the majority of the original Or25 buildings are above the reference value does not necessarily imply that the reference value is too ambitious.These buildings have never been obligated to perform an LCA so has never been awareness of their life cycle impacts.Therefore, it rather suggests that the buildings could surely have reduced their environmental impacts if the designer would have been conscious about the impact of their design decisions on the building life cycle impact.

One benchmark for residential buildings
Similar to the current Flemish EPB regulation and Flemish energy performance classes, one benchmark could be defined for all residential buildings, i.e. based on both SFHs and MFHs (note however that the EPB regulations apply to apartment units instead of apartment buildings).This study showed relatively small differences in life cycle impact between SFHs and MFHs.However, the difference will presumably be bigger if the scope of the LCA was expanded.Although the LCA in TOTEM did include walls, floors, and openings of shared spaces, it did not include stairs, lifts, or foundations, which would probably have a bigger influence on the embodied impact of the MFHs.In addition, TOTEM does not include electricity from lighting and appliances (e.g.lifts or household appliances), which would likely cause a bigger increase in operational impact for MFHs than SFHs.On the other hand, MFHs might benefit from collective systems for heating and DHW, which were not considered in this study.Hence, in addition to the evolution of building practice, developments in LCA methods and tools should be considered in future benchmark definition.

Embodied and operational or life cycle benchmarks
The literature shows a preference for a life cycle approach over separate assessments for operational and embodied impacts for buildings [54,65,[69][70][71][72][73][74][75].In terms of benchmarking, life cycle benchmarks offer greater flexibility by allowing designers to balance embodied/operational impact as preferred [70].Yet, as presented in section 1.1, most regulatory measures currently include separate benchmarks, typically by implementing embodied benchmarks alongside existing operational benchmarks.
The effect of implementing life cycle or separate benchmarks depends on the benchmark values adopted as thresholds.Because the percentiles and the median were calculated separately for the embodied, operational and life cycle impact, the obtained values stem from different buildings.As a result, the sum of embodied and operational benchmarks does not equal the life cycle benchmark.The life cycle benchmark for the limit value was smaller than the sum of the separate benchmarks, implying that in this case a life cycle benchmark would be more ambitious.Contrarily, the life cycle benchmarks for the reference and best-practice value would be less ambitious than if separate benchmarks were considered.Hence, an approach which includes a mandatory limit for the life cycle impact and indicative (i.e.reference and best-practice) values for embodied and operational impact separately could be beneficial.

Limitations and future research
In addition to limitations and recommendations suggested in the previous sections related to benchmark and LCA scope and approach, the representativeness of the results for new buildings may be questioned.This study did not analyse the extent to which the cases (dating from 2015) are still relevant for new (2025) buildings in terms of design and thus embodied and operational impacts.Applying this study's benchmark approach to the buildings from the most recent (2023) Flemish cost-effectiveness study [76] would enable an evaluation of the relevance of the outdated cases.In addition, this study was limited to typical Flemish housing types (detached, semi-detached, terraced houses and apartments) and therefore disregarded innovative forms of housing (e.g.MFH with shared living spaces, assisted living facilities, etc).Innovative housing could be a way to limit building impact per housing unit and should therefore be evaluated in future research.
Regarding the level of ambition, two more things should be noted.Firstly, the TOTEM element library only includes a limited set of innovative (e.g.bio-based) building materials.Innovative materials and construction practices could provide lower environmental impacts [66,[77][78][79][80] and would enable definition of more ambitious benchmarks.Secondly, the bottom-up benchmarks should be complemented by top-down benchmarks to reveal their level of ambition with respect to global or national goals.Moreover, definition of top-down benchmarks for the different impact categories will reveal the environmental hotspots (i.e.overshoots in certain categories) that are now potentially overlooked due to their small contribution to the aggregated score.Hence, comparison of bottom-up and top-down benchmarks may reveal that the applied weighting method does not accurately reflect which indicators should be prioritised, providing a potential basis for a (distance-to-target) weighting method adapted to building LCA.Besides, top-down benchmarks are typically defined on a per person or per building basis and hence consider dwelling size (either directly or indirectly, i.e. defining a maximum footprint per person), providing insights into maximum desired sizes to stay within the targets.
Related to the representativeness and level of ambition, is the dynamics of benchmarks.As building practices will evolve, so will building environmental benchmarks.In this regard, Lützkendorf and Balouktsi [81] argue the need to define the benchmark temporal validity, i.e. the time frame within which benchmarks remain valid for their intended application.The temporal validity can for example be limited to a point in time in which critical changes in building practices as well as in socio-economic and/or political conditions are expected [81].In this paper, prospective (2025) building regulations and current building practices and conditions were applied to develop benchmarks valid from 2025, but no temporal limit of validity was identified.Future work may therefore identify critical developments and moments to reevaluate building benchmarks, considering policy ambitions to further push environmentally sustainable building design.
Lastly, a limitation related to the applied LCA approach is the use of ecoinvent process-based LCI data.Process-based data is considered to underestimate the actual impact of materials due to incomplete system boundaries [82].Especially in relation to top-down benchmarks, high accuracy of environmental impact calculation is desired to assess how far we currently are from defined goals and ambitions.For this purpose, hybrid LCI that supplements process-based LCI with input-output-based LCI could be applied [83].

Conclusion
This paper developed a novel approach to derive life cycle environmental benchmarks for residential buildings.The approach was applied to dwellings used for the cost-effectiveness study of the Flemish region in Belgium in the context of the European EPBD.The approach consisted of applying low-impact, average-impact and high-impact material variants representative of conventional construction to the case study buildings to obtain a wide range of environmental impact results.A statistical analysis of the results was performed to derive benchmark values: a best-practice (10th percentile), reference (median) and limit value (90th percentile).
Using the Belgian building LCA tool TOTEM, benchmarks were calculated for an aggregated environmental score that includes 12 impact indicators, and for the CC indicator individually.For the environmental score, the best-practice, reference, and limit values are 86, 107 and 141 mPt m −2 GHFA, respectively; and for CC, the benchmark values are 844, 1015 and 1284 kg CO 2 -eq m −2 GHFA, respectively.A comparison of the benchmarks with the LCA results of the original buildings (adapted to the EPB requirements of 2025) demonstrated that the benchmarks are feasible and require most dwellings to improve to achieve the reference value.Nonetheless, a critical analysis of the implications of the developed benchmarks was presented.
Comparison of the original buildings including a gas boiler for heating and DHW with a prospective scenario for 2025 including an electric heat pump demonstrated the relevance of anticipating future building regulations and developments when defining benchmarks.Furthermore, the 2025 scenario revealed a shift from operational to embodied impact as well as shifts in impact indicators.The latter is especially relevant to keep in mind if top-down benchmarks would be defined for individual indicators, as this reveals for which indicators buildings are (at risk of) transgressing the identified targets.
In addition to the development of top-down targets, this paper discussed several topics for future research, such as analysis of innovative materials and installations, new ways of living, and prospects for building practice in general.
Finally, this paper emphasised that the benchmarking method as well as the purpose of benchmarks is essential for interpreting them.The benchmark approach in this study aims at representing the range of impacts of new dwellings with conventional construction practice, rather than providing an accurate representation of a stock.The approach is therefore considered relevant for developing indicative benchmarks for LCA tools and building environmental regulations.

Figure 1 .
Figure 1.Distribution of the selected single-family houses (a) across the five clusters based on gross heated floor area (GHFA) and compactness and (b) across the sub-clusters within two of the initial clusters (i.e.blue and green cluster).The colours of the (sub-)clusters are random and are not related in the right and left figure.Adapted with permission from De Pauw et al [49].

Figure 2 .
Figure 2. (a) Simplified representation and IDs and (b) spread of dwelling types, number of storeys, construction types and heating and domestic hot water (DHW) systems of the cases.The outline and fill colours of the representations in (a) correspond with the colours of the construction types and heating systems, respectively, in (b).* Hybrid 1: heat pump backed up by gas boiler for hydronic heating and DHW; * Hybrid 2: heat pump for hydronic heating and DHW backed up by electric resistance heating.

Figure 3 .
Figure 3. Gross heated floor area (GHFA) and compactness (lines) and life cycle environmental score (bars) of the buildings as originally implemented in EPB (Or15) and of the variant with heat pump (Or25).The buildings are sorted by ascending reduction in life cycle environmental score between Or15 and Or25.The x-axis for GHFA is scaled to more clearly show the range of the SFH, therefore omitting the GHFA of IDA7 (914 m 2 ) and IDA9 (2332 m 2 ).

Figure 4 .
Figure 4. Spread of embodied, operational and life cycle environmental score per dwelling type, for the buildings as originally implemented in EPB (Or15) and of the variant with heat pump (Or25).

Figure 5 .
Figure 5. Per case, contribution of the individual impact indicators to the aggregated environmental score of the original building Or15 (upper bar) and the building variant with heat pump Or25 (lower bar).The cases are sorted by ascending climate change impact of the Or25 variant.

Figure 6 .
Figure 6.Positioning in the TOTEM library of the building elements applied in the building material variants.For the walls, storey floors and flat roofs, the upper and lower coloured dots are the selected solid and timber elements, respectively.

Figure 7 .
Figure 7. Life cycle environmental score of all buildings in original materialisation (Or25) and in the three material variants (Var10P, VarMed, Var90P).The cases are grouped per dwelling type and sorted by ascending life cycle environmental score of Or25.

Figure 8 .
Figure 8. Environmental score of all buildings in original materialisation (Or25) and in the three material variants (Var10P, VarMed, Var90P) as a function of compactness (x-axis) and in relation to the gross heated floor area (bubble size).

Figure 9 .
Figure 9. Life cycle environmental score of all buildings in original materialisation (Or25), grouped per dwelling type and sorted from smallest to largest life cycle environmental score of Or25.

Figure 10 .
Figure 10.Comparison of the benchmark values for climate change (CC) with benchmarks from the literature reviewed by Trigaux et al [18].

Table 1 .
Building variants derived from variations in heating system and material choices.

Table 2 .
Building element combinations that make up the three building material variants for solid construction.

Table 3 .
Building element combinations that make up the three building material variants for timber construction.

Table 4 .
Average relative differences in life cycle environmental score between the material variants for the different dwelling types and construction types.GHFA).Considering all variants of all buildings, the highest environmental score (ID21 90th Percentile variant, 183 mPt m −2 GHFA) is 2.4 times higher than the lowest score (ID19 10th Percentile variant, 76 mPt m −2 GHFA).Table