On the estimation of interior walls in the district-scale Life Cycle Assessment of buildings

Abstract

The large-scale Life Cycle Assessment (LCA) of buildings supports decision-making processes in reaching the climate goals of the building sector. Yet, the considerable amount of data required for detailed assessments hinders the method. This is particularly the case for buildings’ interior, which oftentimes surpasses the normative cutoff criteria’s negligibility threshold, and thus has to be considered in building LCA. In light of interior walls’ contribution to buildings’ overall environmental impacts, and a lack of cohesive determination methods thereof, this article proposes a parametric approach to complement an existing district-scale LCA workflow for residential buildings. The approach is applied to an exemplary residential district, using a Latin Hypercube Sampling (LHS) method, to establish a range of interior wall areas and resulting effects on LCA calculations. Thereafter, the outcome is compared with the results of a preexisting approach and validated project data. The novel method is capable of significantly reducing the tool’s overestimation of interior wall areas from 400% to a value between 7% and 53%. However, it is evident that the area estimation in terms of Life Cycle Inventory (LCI) needs to be complemented with a more accurate material setup, which could be differentiated by building archetypes.

1. Introduction

Climate neutrality goals have been consistently sharpened to accelerate the full decarbonization of all industries, including the building sector (Bakardjieva Engelbrekt 2022). It is herein a common practice to increase buildings’ energy efficiency and to reduce their carbon footprints through refurbishment measures. Such measures either reduce the thermal transmittance (“U-value”) of the building envelope, increase the proportion of renewable energy used for the building operation, or both (Fahlstedt et al. 2022). The prioritization and incentivization of building refurbishments potentially facilitate the fulfillment of CO₂ emission reduction goals by allocating resources to the refurbishment of the most inefficient buildings. This can be achieved by conducting a Life Cycle Assessment (LCA) of the building stock, including the inventory of building elements and energy carriers (Life Cycle Inventory, LCI), and its environmental implications (Life Cycle Impact Assessment, LCIA). The same approach can be applied in early design stages to forecast the environmental impact of new buildings. Potential stakeholders and users of building stock LCA include urban planners for the conceptualization of retrofit plans and prioritizations, policymakers for the allocation of refurbishment subsidiaries, and real estate investors for portfolios’ compliance with Environmental, Social, and Governance (ESG) criteria (Park and Jang 2021) stipulated by e.g. the European Union’s Taxonomy (Linciano et al. 2022), or the US American Securities and Exchange Commission (SEC) (Cifrino et al. 2023). Methods have been developed to link single building models described in various data formats with preexisting environmental indicator data, such as OneClickLCA (One Click LCA Ltd 2021), BIM2LCA (Horn et al. 2020), and eLCA (Brockmann 2019). However, building LCA requires a considerable amount of data which is scarcely available and is aggravated on an urban scale (Lotteau et al. 2015). This data scarcity has motivated the ongoing development of Geographic Information System (GIS)-based district-scale LCA approaches such as ubem.io (Ang et al. 2021), Teco (Schildt et al. 2022), and urbi+ (Harter et al. 2021). The framework ubem.io supports the input of .geojson (Butler et al. 2016), and shapefile extension files, including .shp and .shx (ESRI 1998). Teco and urbi + employ building models in the City Geographical Markup Language (CityGML) (Gr¨öger et al. 2012), which is the most commonly available input data taxonomy for Building Energy Performance Simulations (BEPS) (Malhotra et al. 2021). While the development has been gaining considerable momentum, it has also revealed a few shortcomings in terms of data input availability as well as input and output accuracy. One of these deficiencies is the misrepresentation of interior walls in terms of LCI and LCIA due to either a lack of input data from the building model, inaccurate assumptions, or omission of necessary data. Interior walls are a considerable part of buildings’ LCA since their contribution can make up significantly more than 1% or even 5% of the material and energy flows in LCI and LCIA (Valencia-Barba et al. 2023; Schildt et al. 2022), thereby surpassing the negligibility threshold, or so-called cutoff criteria, offered by current international LCA standards that otherwise allow to omit specific material and energy flows that are deemed as insignificant (International Organization for Standardization 2006). This means that relevant environmental implications including the Global Warming Potential (GWP) in LCA calculations and potential recycling or re-use of materials of buildings are not considered. Moreover, there is a significant impact of Mechanical, Electrical and Plumbing (MEP) components in building LCA (Rodriguez et al. 2020). The layout of interior walls influences e.g. the piping and related material use, which is a potential aspect of building LCA refinement. To solve this issue, the aim of this research is (i) to outline the current state of research in the determination of interior walls’ LCI and LCIA. The contribution proceeds with (ii) the description of the ongoing development of Teco, particularly focusing on its underlying approach for the determination of interior wall areas. The development of a novel method for the aforementioned determination into Teco is explained. This is followed by (iii) an in-depth analysis of an exemplary residential district use case, where previously published research is complemented with the new method. The quality of the new results is discussed in detail (iv). This contribution concludes with (v) the summary of the novel method’s qualities and limitations, as well as current and future steps for further development and generalization. These steps are intended to answer the research question if and to which extent a parametric approach for the determination of interior wall areas in district-scale building LCA leads to an increase in the LCA’s output quality.¹

1.1. CityGML in district-scale LCA

The CityGML data format of the Open Geospatial Consortium (OGC) (OGC 2023) can be used to store and exchange virtual 3D city models (Gr¨oger et al. 2012). It is subdivided into a range of modules defining the semantics and topographic information of buildings (Gr¨oger et al. 2012). The Building Module is relevant for the scope of this contribution, as it is employed to represent a single building and its exterior and interior structure. CityGML models building elements with a surface geometry representation. (Malhotra et al. 2021; Malhotra 2023) In this context, the encoding standard’s version 2.0 uses a Level of Detail (LoD) concept (Gr¨oger et al. 2012), with LoD1 and 2 being the most available models (Malhotra et al. 2021).

Figure 1 gives an overview of the amount and accuracy of information represented in the varying CityGML LoDs.

Fig. 1. Overview of information in different CityGML LoDs, according to Malhotra et al. (2019).

There is no representation of interior architecture or constructive elements within the building from Levels 0 to 3 in CityGML 2.0. The CityGML standard has been extended with a range of Application Domain Extensions (ADEs). The Energy ADE is predominantly used in UBEM and BEPS (Agugiaro et al. 2018; OGC and Sig3D 2019). The version 2.0 of this extension provides modules for Construction and Materials and Building Occupancy. These modules add thermo-physical properties and user schedules to building representations, thus enabling Urban Building Energy Modeling (UBEM) for the purpose of BEPS (Reinhart and Cerezo Davila 2016; Nouvel et al. 2015). Propositions have been made to complement the Energy ADE with energy systems and environmental information to facilitate building LCA on a district scale (Mailhac et al. 2017; Schildt et al. 2021). It should be noted that Energy ADE-related data are rarely publicly available (Malhotra et al. 2021). This low granularity of building models in UBEM encourages the Reduced Order Modeling (ROM), where the number of input parameters is reduced while preserving a sufficient output quality. This purported bottom-up approach (Ali et al. 2021) considers every individual building in a dataset to find correlations between input and output values (Swan and Ugursal 2009; Ali et al. 2021). The method is commonly facilitated by the employment of building archetypes, i.e. statistical representations of buildings that are classified by a limited range of building properties, including age class, geometry, usage, and building type (Lauster 2018). Such an approach inherently leads to a high degree of sensitivity in LCA simulations, where the broad range of environmental impact factors for each life cycle module and phase exponentiates the result uncertainty (De Jaeger et al. 2020; Schneider-Marin et al. 2020). Common archetype data sources such as UrbanReNet, the German Federal Ministry of Transport, Building and Urban Development (“Bundesministerium für Verkehr, Bau und Stadtentwicklung”, BMVBS), and the European Typology Approach for Building Stock Energy Assessment (TABULA) generally do not include information about interior walls (Federal Ministry of Transport, Building and Urban Development 2010; Institut Wohnen und Umwelt 2012; Hegger and Dettmar 2014). This is due to the private nature of building interiors that are difficult to determine empirically. Also, the lack of non-load bearing interior walls’ contribution to a building’s structural integrity allows for a broader range of material layer setups. Teco and urbi + principally demonstrate functional workflows for using CityGML building models in district-scale building LCA. However, neither tool is currently able to accurately represent interior walls in terms of mass (LCI) and environmental implications (LCIA). It is due to either interior walls being out of the LCA scope or a flawed estimation method for their geometrical, thermo-physical, and environmental properties (Harter et al. 2021; Schildt et al. 2022).

1.2. Interior wall LCI and LCIA

Building LCA software, such as OneClickLCA (One Click LCA Ltd 2021) or eLCA (Brockmann 2019), generally includes the determination of interior walls’ LCI and LCIA. Building certification systems, e.g. by the German Sustainable Building Council (“Deutsche Gesellschaft für Nachhaltiges Bauen”, DGNB) (German Sustainable Building Council (DGNB) 2020a, 2020b), Leadership in Energy and Environmental Design (LEED) (U.S. Green Building Council (USGBC) 2021a, 2021b), Building Research Establishment Environment Assessment Method (BREEAM) (Building Research Establishment (BRE) 2019), or Minergie (MINERGIE 2023), stipulate the consideration of interior walls for the LCA scope of both new and existing buildings, and recommend the usage of common building LCA software. However, it should be noted that building certificates are naturally acquired by building owners who are in possession of relevant information, including architectural plans.

There is a range of building LCA case studies that highlight the influence of interior wall configurations on LCA simulation results, in particular the material layer setup and service lives (Benetto et al. 2018; Buyle et al. 2019; Valencia-Barba and G´omez-Sober´on 2019; Schildt et al. 2022; Valencia-Barba et al. 2023). Such studies often rely on project-related data for validation, or consider the early design phase of new construction. While the research findings provide insights into the environmental impact of interior walls, it does not aid in the LCI and LCIA estimation of existing buildings’ interior, if architectural plans and construction information are unavailable. For such purposes, few approaches exist. The Swiss building standard association Minergie (MINERGIE 2023) differentiates floor plans in three categories of interior wall lengths for the classification of new and existing buildings’ standards. The total length of interior walls is a function of the total Net Leased Area (NLA), and a rough assessment of the number of interior walls. The resulting interior wall area may be reduced by certain percentage terms depending on the scope of refurbishment measures. The guideline, however, states that this classification may prove difficult in an accurate estimation of interior walls’ embodied energy (Zertifizierungsstelle MINERGIE-ECO 2016). In the “Tool for the Energy Analysis and Simulation for Efficient Retrofit” (TEASER) (Remmen et al. 2017; Lauster 2018), the interior wall areas are estimated based on typical lengths and widths of different building usage zones stated in the Swiss Standard SIA 2024 (SIA 2006). As an extension to the TEASER framework, the software Teco employs the same estimation approach. The tool uses the material setup according to Table 1 for interior walls. While it appears to accurately reflect the thermo-physical properties needed for the UBEM and BEPS of buildings’ heat loads, it has been shown to vastly overestimate the interior wall area and environmental impacts in a residential district use case (Schildt et al. 2022). the Thus, presented research proposes to develop and integrate a parametric approach in Teco to narrow down the range of buildings’ interior wall areas and environmental impacts, which will be elaborated on in the following sections.

Table 1. Exemplary material layers for interior walls used in Teco and TEASER+.

Material	Thickness [m]
Gypsum	0.015 m
Solid brick	0.15 m
Gypsum	0.015 m

2. Materials and methods

This section describes the development of the parameterized approach for the determination of interior walls’ LCI and LCIA in the context of Teco. The new method will be demonstrated on an exemplary residential district. Section 2.1 sets out the development of Teco based on previous work on the TEASER enrichment tool. In Section 2.2, the novel parametric approach is explained in detail. This section concludes with the outline of an exemplary residential district in Section 2.3. It includes the LCA scope, i.e. considered building elements and environmental indicators.

2.1. From TEASER to Teco

TEASER is an open enrichment framework for building stock energy modeling (Remmen et al. 2018). The tool enables the generation of Modelica models for single or multiple buildings by using libraries such as AixLib (Müller et al. 2016), Buildings (Wetter et al. 2014), BuildingSystems (Nytsch-Geusen 2016), or IDEAS (Jorissen et al. 2018). It uses the characteristics year of construction and usage from the input data model to enrich geometrical information with building archetypes. Such archetypes are generally derived from empirical data (cf. Section 1.1; Ballarini et al. 2014). TEASER generates reduced order Resistance-Capacitance (RC) models with a single homogeneous thermal zone for individual buildings (VDI 2015; van Treeck 2010). TEASER uses the information from the buildings’ geometry and respective archetypes to generate simulation models that are compatible with the Modelica environment Dymola (Dassault Syst`emes 2020), to compute the annual heat load in an hourly resolution. The modeling process of TEASER involves the instantiation of the building elements ’ceiling’, ’door’, ’floor’, ’foundation’, ’interior wall’, ’exterior wall’, ’rooftop’, and ’window’, with material layers where applicable.

Previous versions of TEASER provided an interface for CityGML LoD1 & 2 building models as input data. However, the current version does not support the CityGML input. The development of TEASER + has been initiated as a consequence (Malhotra et al. 2019). The tool extends TEASER with an interface for the most commonly used CityGML LoD0-3 models (Malhotra et al. 2021) as input, and exports CityGML Energy ADE v.1.0 (Agugiaro et al. 2018) and Modelica simulation models as output.

TEASER + has been further extended as Teco for the determination of LCI and LCIA on an urban scale (Schildt et al. 2022). This includes a range of environmental indicators, including the GWP. In Teco, LCA data are added to the enrichment architecture of TEASER+. The tool uses the O^¨ KOBAUDAT database (BMI 2021), which focuses on the German building stock. The life cycle phases are adapted in accordance with the environmental product declaration standard DIN EN 15804 (DIN 2020), thus distinguishing between (A) production and construction, (B) operation and maintenance, (C) disposal, and (D) recycling potential. For modules (A), the maintenance phases in (B), (C), and (D), Teco complements the building elements’ and material layers’ volumes determined in the model generation process with LCA data from O^¨ KOBAUDAT, using the material’s reference flow and environmental indicator (International Organization for Standardization 2006). Table 2 shows an exemplary calculation for a hypothetical external wall insulation.

Table 2. Exemplary calculation of environmental impacts in Teco.

Variable	Value
aextwall,insulation	100 m^2
textwall,insulation	0.1 m
GWPinsulation	2 kgCO2 /m^3
GWPextwall,insulation	Πi = 20 kgCO2

For the operational phases in (B), Teco multiplies the operational heat energy demand yielded from the dynamic heat load simulation with LCA datasets for energy carriers, and an overall efficiency factor of the heating system. Furthermore, a building’s electricity demand Q_el is computed with Equation (1)(1) $$\begin{array}{cc}{Q}_{\mathit{\text{el}}}& =(q_{\mathit{\text{el}},b}+q_{\mathit{\text{el}},l})·A·t\\ & =(63\frac{\mathit{\text{Wh}}}{m^{2}d}+10\frac{\mathit{\text{Wh}}}{m^{2}d})·\mathit{\text{NLA}}·50a\end{array}$$(1) . (1) $$\begin{array}{cc}{Q}_{\mathit{\text{el}}}& =(q_{\mathit{\text{el}},b}+q_{\mathit{\text{el}},l})·A·t\\ & =(63\frac{\mathit{\text{Wh}}}{m^{2}d}+10\frac{\mathit{\text{Wh}}}{m^{2}d})·\mathit{\text{NLA}}·50a\end{array}$$(1)

Equation (1)(1) $$\begin{array}{cc}{Q}_{\mathit{\text{el}}}& =(q_{\mathit{\text{el}},b}+q_{\mathit{\text{el}},l})·A·t\\ & =(63\frac{\mathit{\text{Wh}}}{m^{2}d}+10\frac{\mathit{\text{Wh}}}{m^{2}d})·\mathit{\text{NLA}}·50a\end{array}$$(1) considers guideline values from DIN V 18599 (DIN 2018) for the electricity demand of appliances q_el,b and lighting q_el,l and multiplies them with the NLA and number of years in the life cycle scope. The resulting electricity demand is further multiplied with another LCA energy carrier dataset from O^¨KOBAUDAT to determine the total amount of nonrenewable primary energy with respect to electricity (PENRT_elec). Figure 2 displays the overall enrichment architecture in Teco.

Fig. 2. Overview of the enrichment architecture in Teco (Schildt et al. 2022).

TEASER + and Teco employ the identical interior wall estimation method as TEASER. This means that interior walls, ceilings, and floors are modeled as classes to be aggregated into a lumped thermal storage mass in the RC model. (Remmen et al. 2017). Moreover, only one thermal zone is modeled, leading to the consideration only of typical living rooms. The actual NLAs, storey heights, and interior walls’ areas are derived from guideline values of the Swiss standard SIA 2024 (SIA 2006; Lauster 2018). This standard defines typical room parameters for different types of usage, including geometry, ergonomics, and occupancy. For instance, a living room, a bedroom, or a kitchen are defined as having typical dimensions of 4 m x 4 m x 2.5 m (L x W x H). As mentioned earlier, such an approach has led to a significant overestimation of the interior wall area of an exemplary residential district (cf. Section 1.2). This will be assessed in detail in a juxtaposition of results in Section 3.

2.2. Parametric approach for interior walls in LCA

The state of research and practice shows that there is currently no cohesive and well-calibrated method for determining interior wall areas in UBEM for LCA purposes. Thus, the authors present an approach where the aforementioned area (A_iw) is a function of a building’s NLA A_building, its number of storeys n_storey and height thereof h_storey, ratios of horizontal to vertical wall lengths r_wall,x|y, shares of areas for different room types s_roomi, and ranges thereof r_sroom (see Equation (2)(2) $$A_{\mathit{\text{iw}}}=f(A_{\mathit{\text{building}}},n_{\mathit{\text{storey}}},h_{\mathit{\text{storey}}},r_{\mathit{\text{wall}},x|y},s_{{\mathit{\text{room}}}_i},r_{s_{{\mathit{\text{room}}}_i}})$$(2) ). (2) $$A_{\mathit{\text{iw}}}=f(A_{\mathit{\text{building}}},n_{\mathit{\text{storey}}},h_{\mathit{\text{storey}}},r_{\mathit{\text{wall}},x|y},s_{{\mathit{\text{room}}}_i},r_{s_{{\mathit{\text{room}}}_i}})$$(2)

For the scope of this contribution, the approach is limited to Single Family Houses (SFH) to enable the juxtaposition of simulation results with previous research. It is assumed that floor plans of SFH exhibit typical patterns of room types, adjacencies, and numbers of storeys. Thus, generic floor plans for an SFH of up to four storeys are abstracted as undirected graphs (Rahman 2017). Figures 3 and 4 show the graph abstraction for each floor.

Fig. 3. Abstraction of generic SFH first and second floor plans as undirected graphs.

Fig. 4. Abstraction of generic SFH basement and attic floor plans as undirected graphs.

The first and second floor plans are assumed as the minimum configuration of any SFH. Buildings with three storeys above ground are assumed to have an attic. The room types and respective variables are introduced here to follow the esthetics of archetypical SFH as introduced by TABULA (Institut Wohnen und Umwelt 2012), to enable the setup of a parametric solution space of the overall interior wall area. This should allow for the estimation of interior wall areas and, ultimately, their environmental implications analogously to the other building elements of the low-order modeling approach of TEASER + and Teco, to find solutions that are geared toward the low-order, district-scale view of buildings.

Every vertex or node in Figures 3 and 4 represents a room with the property ‘area’ (a_roomi) as in Equation (3)(3) $$\begin{array}{c}{a}_{{\mathit{\text{room}}}_i\mathrm{ }}∊[A_{\mathit{\text{building}}\mathrm{ }}·s_{{\mathit{\text{room}}}_i\mathrm{ }}·r_{s_{{\mathit{\text{room}}}_{i,\mathrm{\text{min}}}}},A_{\mathit{\text{building}}\mathrm{ }}·s_{{\mathit{\text{room}}}_i\mathrm{ }}·r_{s_{{\mathit{\text{room}}}_{i,\mathrm{\text{max}}}}}]\\ =\left[a_{{\mathit{\text{room}}}_{i,\mathrm{\text{min}}}},a_{{\mathit{\text{room}}}_{i,\mathrm{\text{max}}}}\right]\end{array}$$(3) . Note that the hallway area on every floor is defined as the difference between the respective floor’s NLA, and the areas of all rooms (Equation (4)(4) $$\begin{array}{c}\begin{array}{c}{a}_{{\mathit{\text{hallway}}}_{{\mathit{\text{floor}}}_j}\mathrm{ }}∊[{\mathit{\text{NLA}}}_{{\mathit{\text{floor}}}_{j\mathrm{ }}\mathrm{ }}‐\sum _{i=1}^n{a}_{{\mathit{\text{room}}}_{i,{\mathit{\text{floor}}}_j,\mathrm{\text{max}}}},\\ {\mathit{\text{NLA}}}_{{\mathit{\text{floor}}}_{j\mathrm{ }}\mathrm{ }}‐\sum _{i=1}^n{a}_{{\mathit{\text{room}}}_{i,{\mathit{\text{floor}}}_j,\mathrm{\text{min}}}}]\end{array}\\ =[a_{{\mathit{\text{hallway}}}_{{\mathit{\text{floor}}}_{j,\mathrm{\text{min}}}}\mathrm{ }},a_{{\mathit{\text{hallway}}}_{{\mathit{\text{floor}}}_{j,\mathrm{\text{max}}}}}]\end{array}$$(4) ). Moreover, each node has at least one connection, i.e. edge, to at least one other room, representing the length of an either horizontal or vertical wall as an expression of the rooms’ adjacency, l_{iw,roomi|roomj,x∨y} (see Equation (5)(5) $$\begin{array}{c}{l}_{\mathit{\text{iw}},{\mathit{\text{room}}}_i|{\mathit{\text{room}}}_j,x\vee \mathrm{y}}∊\left[\sqrt{\frac{a_{{\mathit{\text{room}}}_{i,\mathrm{\text{min}}\mathrm{ }}\mathrm{ }}}{r_{\mathit{\text{wall}},x|y,\mathrm{\text{min}}}}}‐l_{\mathit{\text{door}}},\sqrt{\frac{a_{{\mathit{\text{room}}}_{i,\mathrm{\text{max}}\mathrm{ }}\mathrm{ }}}{r_{\mathit{\text{wall}},x|y,\mathrm{\text{max}}}}}‐l_{\mathit{\text{door}}}\right]\\ =[l_{\mathit{\text{iw}},{\mathit{\text{room}}}_i|{\mathit{\text{room}}}_j,x\vee \mathrm{y},\mathrm{\text{min}}},l_{\mathit{\text{iw}},{\mathit{\text{room}}}_i|{\mathit{\text{room}}}_j,x\vee \mathrm{y},\mathrm{\text{max}}}]\end{array}$$(5) ). Every wall is assumed to have one door on average. (3) $$\begin{array}{c}{a}_{{\mathit{\text{room}}}_i\mathrm{ }}∊[A_{\mathit{\text{building}}\mathrm{ }}·s_{{\mathit{\text{room}}}_i\mathrm{ }}·r_{s_{{\mathit{\text{room}}}_{i,\mathrm{\text{min}}}}},A_{\mathit{\text{building}}\mathrm{ }}·s_{{\mathit{\text{room}}}_i\mathrm{ }}·r_{s_{{\mathit{\text{room}}}_{i,\mathrm{\text{max}}}}}]\\ =\left[a_{{\mathit{\text{room}}}_{i,\mathrm{\text{min}}}},a_{{\mathit{\text{room}}}_{i,\mathrm{\text{max}}}}\right]\end{array}$$(3) (4) $$\begin{array}{c}\begin{array}{c}{a}_{{\mathit{\text{hallway}}}_{{\mathit{\text{floor}}}_j}\mathrm{ }}∊[{\mathit{\text{NLA}}}_{{\mathit{\text{floor}}}_{j\mathrm{ }}\mathrm{ }}‐\sum _{i=1}^n{a}_{{\mathit{\text{room}}}_{i,{\mathit{\text{floor}}}_j,\mathrm{\text{max}}}},\\ {\mathit{\text{NLA}}}_{{\mathit{\text{floor}}}_{j\mathrm{ }}\mathrm{ }}‐\sum _{i=1}^n{a}_{{\mathit{\text{room}}}_{i,{\mathit{\text{floor}}}_j,\mathrm{\text{min}}}}]\end{array}\\ =[a_{{\mathit{\text{hallway}}}_{{\mathit{\text{floor}}}_{j,\mathrm{\text{min}}}}\mathrm{ }},a_{{\mathit{\text{hallway}}}_{{\mathit{\text{floor}}}_{j,\mathrm{\text{max}}}}}]\end{array}$$(4) (5) $$\begin{array}{c}{l}_{\mathit{\text{iw}},{\mathit{\text{room}}}_i|{\mathit{\text{room}}}_j,x\vee \mathrm{y}}∊\left[\sqrt{\frac{a_{{\mathit{\text{room}}}_{i,\mathrm{\text{min}}\mathrm{ }}\mathrm{ }}}{r_{\mathit{\text{wall}},x|y,\mathrm{\text{min}}}}}‐l_{\mathit{\text{door}}},\sqrt{\frac{a_{{\mathit{\text{room}}}_{i,\mathrm{\text{max}}\mathrm{ }}\mathrm{ }}}{r_{\mathit{\text{wall}},x|y,\mathrm{\text{max}}}}}‐l_{\mathit{\text{door}}}\right]\\ =[l_{\mathit{\text{iw}},{\mathit{\text{room}}}_i|{\mathit{\text{room}}}_j,x\vee \mathrm{y},\mathrm{\text{min}}},l_{\mathit{\text{iw}},{\mathit{\text{room}}}_i|{\mathit{\text{room}}}_j,x\vee \mathrm{y},\mathrm{\text{max}}}]\end{array}$$(5)

Evidently, each interior wall’s length is derived from the respective room’s area and wall ratio. To ensure model consistency, the sum of interior wall lengths per direction on each floor is assumed to be the respective exterior wall length (see Equation (6)(6) $$\sum _{i=1}^n{l}_{\mathit{\text{iw}},{\mathit{\text{room}}}_i|{\mathit{\text{room}}}_j,x\vee \mathrm{y}}=l_{\mathit{\text{ew}},x\vee \mathrm{y}}$$(6) ). (6) $$\sum _{i=1}^n{l}_{\mathit{\text{iw}},{\mathit{\text{room}}}_i|{\mathit{\text{room}}}_j,x\vee \mathrm{y}}=l_{\mathit{\text{ew}},x\vee \mathrm{y}}$$(6)

Furthermore, the model considers a wall length minimum as a boundary condition for plausibility (see Equation (7)(7) $$l_{\mathit{\text{iw}},i,x\vee \mathrm{y}\mathrm{ }}\ge l_{\mathit{\text{iw}},i,\mathrm{\text{min}}}\ge 0\forall i\in I,\mathit{\text{with}}\mathrm{ }I:=\mathit{\text{number of rooms}}$$(7) ). The threshold value is at least zero. (7) $$l_{\mathit{\text{iw}},i,x\vee \mathrm{y}\mathrm{ }}\ge l_{\mathit{\text{iw}},i,\mathrm{\text{min}}}\ge 0\forall i\in I,\mathit{\text{with}}\mathrm{ }I:=\mathit{\text{number of rooms}}$$(7)

The sum of all interior wall lengths is multiplied by h_storey to determine the total interior wall area. (8) $$A_{\mathit{\text{iw}}}=\left(\sum _{i=1}^n{l}_{\mathit{\text{iw}},i,x}+\sum _{i=1}^n{l}_{\mathit{\text{iw}},i,y}\right)·h_{\mathit{\text{storey}}}$$(8)

The updated workflow of Teco involving the novel interior wall estimation method is illustrated in Figure 5, highlighting the sources of information (Institut Wohnen und Umwelt 2012; OGC and Sig3D 2019).

Fig. 5. Updated workflow of Teco including interior walls’ parametric solution space.

In particular, interior walls’ areas and ultimate contributions to buildings’ LCA are determined by a parametric solution space as defined by Equations (2)(2) $$A_{\mathit{\text{iw}}}=f(A_{\mathit{\text{building}}},n_{\mathit{\text{storey}}},h_{\mathit{\text{storey}}},r_{\mathit{\text{wall}},x|y},s_{{\mathit{\text{room}}}_i},r_{s_{{\mathit{\text{room}}}_i}})$$(2) to (8)(8) $$A_{\mathit{\text{iw}}}=\left(\sum _{i=1}^n{l}_{\mathit{\text{iw}},i,x}+\sum _{i=1}^n{l}_{\mathit{\text{iw}},i,y}\right)·h_{\mathit{\text{storey}}}$$(8) . The input stems from basic, archetypical building characteristics such as the existence of a basement and/or attic, as well as user input for room areas and wall length ratios. Table 3 illustrates the value ranges for the aforementioned wall length and floor area ranges (Equations (2)(2) $$A_{\mathit{\text{iw}}}=f(A_{\mathit{\text{building}}},n_{\mathit{\text{storey}}},h_{\mathit{\text{storey}}},r_{\mathit{\text{wall}},x|y},s_{{\mathit{\text{room}}}_i},r_{s_{{\mathit{\text{room}}}_i}})$$(2) and (7)(7) $$l_{\mathit{\text{iw}},i,x\vee \mathrm{y}\mathrm{ }}\ge l_{\mathit{\text{iw}},i,\mathrm{\text{min}}}\ge 0\forall i\in I,\mathit{\text{with}}\mathrm{ }I:=\mathit{\text{number of rooms}}$$(7) ). These ranges have been defined to include a user’s reasonable assumptions on floor plans while retaining their parametric adaptation. This results in a process that outputs a solution space where the upper limit can be used for a conservative estimate in terms of the overall building LCA. Also, it can be refined if more detailed information is available. For instance, it is assumed that living rooms make up a substantial share of the first-floor plan (s_lr), however their exact share (rs_room), and length ratios of adjacent walls (r_wall,x|y), can only be estimated as a range. If more detailed information is available, the user is able to narrow down the respective ranges, or to set absolute values. Yet, it should be considered that some values stem from the building models of the CityGML files used in the overall workflow (see Figure 5).

Table 3. Functional parameters for wall lengths and floor areas.

Variable	Value
rwall,x|y	∈ [0.35, 1.65]
sbasr1	0.35
sbasr2	0.35
slr	0.30
sr	0.20
sk	0.15
sbar,1st	0.10
sbr1	0.20
sbr2	0.20
sbr3	0.20
sbar,2nd	0.15
sattr1	0.35
sattr2	0.35
rsroomi	∈ [0.90, 1.10]
liw,min	0

Latin Hypercube Sampling (LHS) is used to create samples for the evaluation of the input variables’ ranges and the determination of respective outputs (McKay et al. 1979). It should be noted that there is a broad range of sampling methods (Macdonald 2009), with LHS being subject to further advancements (Shields and Zhang 2016). For the scope of this work, we have chosen LHS since it has shown to be comparatively efficient in generating large samples, and to create uniformly distributed samples by partitioning each input variable range into equally probable intervals (Loh 1996; Manteufel 2000). Other methods and respective algorithms might prove to be more time-efficient. Yet, the runtime is seen as negligible by comparison to the overall simulation length of Teco and TEASER+, the use of which is the scope of this contribution. For a nuanced evaluation of the presented method’s performance, the algorithm’s runtime will still be investigated in Section 3.

Furthermore, the well-established operability of LHS serves a convenient implementation. The authors use the “lhs” module from the “pyDOE” (pyDOE 2023) package in Python to extend Teco with the parametric study for interior walls. For this purpose, n_samples is defined as the number of samples to be created for each input variable. Samples that violate the boundary condition in Equation (7)(7) $$l_{\mathit{\text{iw}},i,x\vee \mathrm{y}\mathrm{ }}\ge l_{\mathit{\text{iw}},i,\mathrm{\text{min}}}\ge 0\forall i\in I,\mathit{\text{with}}\mathrm{ }I:=\mathit{\text{number of rooms}}$$(7) are automatically excluded from the evaluation. The sampling yields a vector of total interior wall areas, $\overrightarrow{A}$_iw, representing the output set from the sampled input variables. $\overrightarrow{A}$_iw is then used in Teco’s workflow to determine a range of environmental impacts for all considered materials (see Section 2.1). Note that LHS is conducted for one building of each type found in the district, respectively. The results are then upscaled by the number of present houses of that type.

2.3. Exemplary residential district application

The considered residential district consists of 104 SFH in western Germany, which are currently under construction within the scope of an ongoing research project. Each SFH belongs to one of four types of houses that are distinguished by NLA and number of storeys. Table 4 gives an overview of the relevant information of the district, which has been provided by industrial partners within the research project. Figure 6 shows the CityGML building models that have been created using CityBIT (Malhotra 2021). For the determination of the buildings’ elements’ LCI, measurements have been obtained from confidential architectural plans. For this contribution, the scope of the presented LCA, the case study focuses on the influence of the developed parametric approach on the determination of interior wall areas, and their consecutive influence on the overall environmental impacts (GWP). The results obtained from Teco’s previously published and newly developed workflow will be compared with the detailed validation results yielded from the aforementioned project information. Again, the considered building elements are exterior and interior walls, roofs, foundations, interior floors, and windows. The interior wall estimation method developed by Minergie (see Section 1.2) is deemed to require information that is generally not available in CityGML datasets and is thus discarded for the present application.

Fig. 6. Visual representation of the use case district using FZK viewer IAI/KIT (2021).

Table 4. Basic parameters of the district use case.

House type	I	II	III	IV
Year of construction	2022
No. of buildings	20	27	16	41
NLA [m^2]	167	180	141	172
Building height [m]	6.15			9.05
Storeys	2			3
Storey height [m]	2.50
Roof type	Flat roof
PENRTelec [kWh/m^2a]	8.76
GWPPENRT,elec [kgCO2eq/m^2a]	1.57
External wall	10 cm wood layer 18 cm insulation
Internal wall	10 cm wood layer 8 cm insulation
Roof	13 cm wood layer 20 cm insulation
Foundation	3 cm footfall sound insulation 8 cm insulation 30 cm concrete 10 cm perimeter insulation underfloor heating
Floor	3 cm footfall sound insulation 13 cm wood layer 6 cm insulation underfloor heating
Windows	Triple glazing (25% of façade) wooden framing
Stairs	Wooden stairs
Doors	Wooden front door Aluminum garage door
Utilities	Heat transmission station 10 kWh electrical heat pump 500 L steel buffer storage 2 × 70 kW central electrical heat pumps

3. Results

Figure 7 illustrates the previous results obtained from Teco’s heuristic LCA approach, and the manual validation using project-related information.

Fig. 7. Area (LCI) and GWP (LCIA) per building element in district from Teco and manual validation (Schildt et al. 2022).

All values, including those related to module D, are displayed as absolutes for visual clarity. The estimation of interior walls’ areas, GWP, and resulting proportion thereof in the overall LCIA, vary significantly between Teco and the manual validation. Teco’s result of the district’s total interior wall area exceeds the actual value by about 400%. The tool overestimates the building element’s GWP for modules A to C, and underestimates module D, by up to 1000%. This, in turn, influences the relative contribution of interior walls to the district’s GWP. While according to Teco, the building element accounts for 22% of all GWP emissions in modules A to C throughout the district life cycle, the validation shows that this value is no more than 2%. Similarly, module D either accounts for 7% or 12% of all recycling-related advantages.

Figure 8 displays the distribution of total interior wall area yielded from the novel method. The median result “med” is about 42% higher than the true value “µ” obtained from project information for all chosen numbers of samples “n”. The algorithm’s runtime “O” increases significantly by any increase of created samples n. Figure 9 shows the aforementioned distribution by each building type of the district. Type III exhibits the lowest deviation between med and µ of about 7%, with the true value lying within the estimation margin. The remaining types’ median values deviate by 40% to 53% from the true value. Again, the runtime O increases significantly by the number of samples n, with building type IV exhibiting the longest runtimes overall. The interior walls’ GWP distribution yielded from Teco’s material archetype (see Figure 10) shows a vast deviation of the results from the true value. Modules A to C are overestimated by about 100%, whereas the median for D is about 4900% higher than µ. The analogous calculation using the actual material setup retrieved from architectural plans naturally leads to a deviation that is proportionate to the presented approach’s interior wall area estimation (see Figure 11).

Fig. 8. Total interior wall area distribution by number of LHS sample.

Fig. 9. Interior wall area distribution by building type and number of LHS samples.

Fig. 10. Distribution of GWP inferred from interior wall area by modules and number of LHS samples with Teco’s material setup.

Fig. 11. Distribution of GWP inferred from interior wall area by modules and number of LHS samples with the actual material setup.

4. Discussion

The presented method has led to a significant reduction of the interior wall area in the district LCI from 400% to a range between 7% and 53%, when considering the median of each computation. The true value µ is within the samples’ boundaries only in the case of building type III (see Figure 8). Any increase in the number of samples does not lead to a consistent pattern change in the estimation margin. Nevertheless, the runtime of the developed algorithm increases at least polynomially. This is particularly the case for building type IV, which can be traced back to an increase of variables and dimensionality due to an additional floor (3 instead of 2). With each building type (or archetype) representing one LHS process, the number of samples needs to be considered for reasons of practicality, and balanced against any gain in accuracy. For the presented exemplary district results, no convergence of values can be identified, i.e. an increased number of samples does not seem to lead to a reduction of minima and maxima, nor number of outliers. Thus, for the scope of the presented research and use case, n_samples ≤ 200 appears to be a sensible choice.

The resulting distribution of GWP inferred from the interior wall area shows a tremendous deviation from the true value between 100% (modules A to C) and 4900% (module D), and between 7% and 53% when using the true material layer setup. This emphasizes the difference between the material layers in Teco’s archetypes and the actual construction, exhibiting vastly different environmental impacts, respectively. It is noteworthy that the new approach assumes that building IV has a developed attic as a third floor, whereas the architectural plans prove otherwise. While the scope of the research project has presented the authors with a comparatively rare opportunity for the validation of an abstract building model with highly detailed plans, this is generally not the case. A heuristic district-scale LCA tool, using a ROM approach for UBEM and BEPS, should rather function without this regular comparison, and give a rough assessment. Thus, assumptions on buildings’ interior should be generic and archetype-oriented. In case of Teco’s TABULA archetypes, no SFH has a third floor that is not an attic.

In terms of LCA output quality, a conservative estimate of environmental impacts is generally favorable to avoid any distortion of e.g. CO₂ emission statements, i.e., GWP. However, a generally vast overestimation prevents the comparability between buildings and districts, thus counteracting the intention of prioritizing buildings for refurbishments and incentives. It is elusive to define a threshold percentage term for this matter. Nevertheless, the achieved reduction of interior wall area overestimation from 400% to a range between 7% and 53% supports the tradeoff between conservatively estimating districts’ environmental impacts, and retaining an appropriate degree of comparability between different buildings and districts. In light of the well-defined boundary conditions (see Equations (3)(3) $$\begin{array}{c}{a}_{{\mathit{\text{room}}}_i\mathrm{ }}∊[A_{\mathit{\text{building}}\mathrm{ }}·s_{{\mathit{\text{room}}}_i\mathrm{ }}·r_{s_{{\mathit{\text{room}}}_{i,\mathrm{\text{min}}}}},A_{\mathit{\text{building}}\mathrm{ }}·s_{{\mathit{\text{room}}}_i\mathrm{ }}·r_{s_{{\mathit{\text{room}}}_{i,\mathrm{\text{max}}}}}]\\ =\left[a_{{\mathit{\text{room}}}_{i,\mathrm{\text{min}}}},a_{{\mathit{\text{room}}}_{i,\mathrm{\text{max}}}}\right]\end{array}$$(3) to (8)) and parameter value ranges (see Table 3), it is improbable that any other sampling method where samples are uniformly distributed would yield significantly different results. However, the algorithm’s runtime might be different. With the comparatively longer simulation runtime of the overall Teco workflow, the presented runtimes are seen as negligible for the scope of this work.

5. Conclusion

The presented parametric method is a novel approach to use Latin Hypercube Sampling (LHS) as a sampling algorithm in finding a viable solution space for interior walls’ areas and environmental implications. It significantly outperforms the current estimation method of interior walls in the workflow of TEASER + and Teco in terms of accuracy (see Section 4), enabling a more precise determination of LCI and LCIA of buildings on a district scale. This tackles the issue of fulfilling the normative requirements of LCA by allowing stakeholders to actively consider interior walls in their calculations (see Section 1), while still delivering a conservative estimate. The input parameter ranges can be adapted in case more specific information is available.

However, it should be emphasized that the overall feasibility of the estimation method strongly depends on the use case and available information. For instance, a new construction development involves detailed floor plans and/or Building Information Modeling (BIM) models, as was used for the validation of the presented materials (see Section 2.3). This allows for a meticulous determination of interior wall areas and materials, rendering the presented method unnecessary. Conversely, any application where the ultimate goal is retrofitting, such as policymaking, planning of urban retrofit and dismantling, or real estate investments, oftentimes has no such detailed information available. Instead, estimation methods such as the one presented are beneficial to identify potentials and priorities for sustainability-related purposes including recycling, re-use, and the decarbonization of the building stock. However, it should be noted that building use type changes stemming from retrofitting projects, such as the conversion of buildings to student housing projects, involves significant changes to interior wall setups not necessarily reflected in CityGML building taxonomy updates.

Subsequently, the refinement of the input parameters that do not stem from CityGML files (see Figure 3) is only sensible in case statistical information is available. For example, archetypical building data with details on floor plan layouts, or architectural descriptions of different building age classes, could be used to refine the value ranges with respect to local characteristics. This may include the validation of parametric solutions with empirical on-site data.

The initial research question, to which extent a parametric approach for the determination of interior wall areas in district-scale building LCA leads to an increase in the LCA’s output quality (see Section 1), can be answered as follows. A few uncertainties hinder the specification of input parameter ranges and thus expand the solution space, leading to an ultimate overestimation of the interior walls’ environmental impacts. Yet, the presented method offers the possibility to refine these input parameters with further research. Moreover, it has proven to significantly reduce the error margin of Teco’s workflow as demonstrated with the exemplary district.

Future work on this method should include the generalization of its application field, i.e., buildings’ use, building archetypes, and location. Current work focuses on the inclusion of different types of residential as well as office buildings in Germany. The applicability of the overall method in other parts of the world, including TEASER+, Teco, and the presented approach for interior walls, can be achieved by implementing equivalents of the employed TABULA archetypes (see Section 2.2). In case of the United States of America, such information could be derived from the Prototype Building Models provided by the American National Standards Institute (ANSI)/American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE)/Illuminating Engineering Society (IES) Standards 90.1 and 90.2 (ANSI and ASHRAE 2018; ANSI, ASHRAE and IES 2022; U.S. Department of Energy 2023). The relationship between interior walls and MEP components (see Section 1) needs to be investigated to infer more accurate estimations of piping and other parts.

Finally, the presented approach should be applied on more exemplary districts to assess the output quality and feasibility on a larger scale. This may include the showcased purpose of LCA as well as any other retrofit-oriented planning processes.

Nomenclature
ADE	=	Application Domain Extension
ANSI	=	American National Standards Institute
ASHRAE	=	American Society of Heating, Refrigerating and Air-Conditioning Engineers
BEPS	=	Building Energy Performance Simulation
BIM	=	Building Information Modeling
BMVBS	=	Bundesministerium für Verkehr, Bau und Stadtentwicklung
BREEAM	=	Building Research Establishment Environment Assessment Method
CityGML	=	City Geography Markup Language
DGNB	=	Deutsche Gesellschaft für Nachhaltiges Bauen
ESG	=	Environmental, Social, and Governance
GIS	=	Geographic Information System
GWP	=	Global Warming Potential
IES	=	Illuminating Engineering Society
LCA	=	Life Cycle Assessment
LCI	=	Life Cycle Inventory
LCIA	=	Life Cycle Impact Assessment
LEED	=	Leadership in Energy and Environmental Design
LHS	=	Latin Hypercube Sampling
LoD	=	Level of Detail
MEP	=	Mechanical, Electrical, and Plumbing
NLA	=	Net Leased Area
OGC	=	Open Geospatial Consortium
PENRTelec	=	Total amount of nonrenewable primary energy with respect to electricity
ROM	=	Reduced Order Model(ing)
SEC	=	Securities and Exchange Commission
SFH	=	Single Family House
TABULA	=	Typology Approach for Building Stock Energy Assessment
TEASER	=	Tool for the Energy Analysis and Simulation for Efficient Retrofit
UBEM	=	Urban Building Energy Model(ing)

Data and software availability statement

The presented tools Teco and TEASER + are available on GitHub: https://github.com/RWTH-E3D/Teco, accessed on February 27, 2024; https://github.com/RWTH-E3D/TEASERPLUS, accessed on February 27, 2024. The next version of Teco will include the implementation of the presented estimation method for interior walls of different residential building types and office buildings. The validation data stem from project-related plans that are subject to a Non-Disclosure Agreement that allows for the anonymized input in the scope of publicly funded research.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was financially supported by BMWK (German Federal Ministry of Economic Affairs and Climate Action), promotional reference 03EWR010B.

Notes

1 This manuscript is an extended version of a previously published contribution to the Conference Proceedings “2022 Building Performance Analysis Conference and SimBuild co-organized by ASHRAE and IBPSA-USA”.