$Sylvia\ Marinova^{a}$; $Sebastiaan\ Deetman^{a}$;$Ester\ van\ der\ Voet^{a}$; $Vassilis\ Daioglou^{b,c}$
$^a$ CML – Institute of Environmental Sciences, Leiden University, Einsteinweg 2, 2333 CC, Leiden, the Netherlands
$^b$ PBL Netherlands Environmental Assessment Agency, PO Box 30314, 2500 GH, The Hague, the Netherlands
$^c$ Copernicus Institute of Sustainable Development, Utrecht University, Princetonlaan 8a, 3584 CB, Utrecht, the Netherlands

Received 30 January 2019, Revised 25 October 2019, Accepted 2 November 2019, Available online 7 November 2019.

Handling editor: Yutao Wang

Table of Contents

Abstract

Huge material stocks are embedded in the residential built environment. These materials have the potential to be a source of secondary materials, an important consideration for the transition towards a circular economy. Consistent information about such stocks, especially at the global level, is missing. This article attempts to fill part of that gap by compiling a material intensities database for different types of buildings and applying that data in the context of a scenario analysis, linked to the SSP scenarios as implemented in the global climate model IMAGE. The database is created on a global scale, dividing the world into 26 regions in compliance with IMAGE. The potential use of the database was tested and served as input for modelling the housing and material stock of residential buildings for the period 1970–2050, according to specifications made for the SSP2 scenario. Six construction materials in four different dwelling types in urban and rural areas are included. The material flows related to those stocks are estimated and analysed in a companion paper (also exploring commercial buildings) by Deetman et al. (2019). The results suggest a significant increase in the material stock in housing towards 2050, particularly in urban areas. The results reflect specific patterns in the material contents across the different building types. China presently dominates developments in the global level building stock. The SSP2 projections show a stock saturation towards 2050 for China. In other regions, such as India and South East Asia, stock growth is presently just taking off and can be expected to become dominant for global developments after 2050. The database is created to be used as input for resource and climate policymaking as well as assessment of environmental impact related to residential buildings and assessment of possibilities for urban mining. In the future, we hope to extend it as new data on materials in the built environment become available.

1. Introduction

The demand for primary materials has increased significantly during the last decades, driven by industrialisation and economic development. The demand for raw materials is forecasted to continue growing with the increase in global population and affluence (OECD, 2013), resulting in a growing in-use stock of materials. An important share of these materials is related to residential buildings. The residential building sector accounts for 30–50% of the material consumption, forming a massive material stock which increased during the past years and is expected to expand further (Steger and Bleischwitz, 2011). Demographic changes and increased Gross Domestic Product (GDP) are expected to lead to a growth in the demand for floor area and construction materials respectively (OECD, 2013), both per capita and in an absolute sense. The built environment is associated with considerable environmental impacts related to the construction and operation of buildings, ranging from the extraction and transformation of resources to the increased energy demand of the in-use buildings (Augiseau and Barles, 2017). At the same time, it represents a huge urban mine of valuable raw materials for secondary resource providers. As yet, there is little insight into these stocks. However, knowledge of these stocks and their dynamics is essential information for a transition towards the circular economy (Müller, 2006; Krausmann et al., 2017).

Individual estimations of stocks and flows of building materials on a national and regional level have been performed and described in various studies. However, so far, there is little harmonisation: each of these studies has its focus, uses its data and makes its own methodological choices. Material Flow Analysis (MFA) is the methodology widely used to quantify the materials flows and stocks in the built environment. The two main approaches of material stock assessment can be described as bottom-up and top-down (Auping et al., 2014; Urge-Vorsatz et al., 2012). The top-down approach calculates stocks at the aggregate level, as the result of net-additions-to-stock of a material over a period of time. The bottom-up approach** divides the stock into categories of products or applications and estimates the stock by characterising each of its components with a material intensity ratio (e.g. kg/m2).

Over the past years, efforts have been made to explore the dynamics of the stock (Müller, 2006; Olaya et al., 2017; Hashimoto et al., 2007). For instance, Müller (2006) applied stock dynamics modelling to forecast the resource demand simultaneously with the related waste generation. Hu et al. (2010) (Hu et al., 2010) used Müller’s dynamic stock model as a basis for the development of an MFA model which represents the changes in the residential buildings floor area in use in China between 1900 and 2100. A number of studies developed this approach further and explored the material composition of the stock while taking into consideration the generation of construction and demolition waste (Hashimoto et al., 2007; Hu et al., 2010; Reyna and Chester, 2015; Aksoezen et al., 2016), as well as the spatial distribution of the stock (Heeren and Hellweg, 2018; Tanikawa and Hashimoto, 2009; Kleemann et al., 2016; Koutamanis et al., 2018).

Recognising the importance of the environmental implications of material demand, researchers assessed the relationship between the material stock and negative impacts related to the built environment such as energy consumption and greenhouse gas emissions. The most recent studies employ a Life Cycle Assessment (LCA) approach and take into account material and energy flows in addition to emissions related to the life cycle of the building itself (Stephan et al., 2012; Yang et al., 2018; Nemry et al., 2010).

In recent years, a small but growing number of studies has been conducted with the purpose to record, store and analyse information on the material content of the built environment. For example, Gontia et al. (2018) (Gontia et al., 2018) developed a material intensities database of residential buildings in Sweden. The study explores 46 buildings and separates them according to their building type, construction type and construction period. In addition, Kleemann at al. (2016) (Kleemann et al., 2016) developed a material content database in order to investigate the building stock in Vienna, Austria. Another study compiling a material database along with the investigation the total material stock and flows resulted from demolition waste is Miatto et al. (2019) (Miatto et al., 2019), recording detailed information of material intensities of buildings in one city (Padua, Italy). Besides, Heeren at al. (2019) (Heeren and Fishman, 2019) compiled a material intensities database on a global scale by extracting information from 33 studies and recording approximately 300 data points from those studies.

Studies like these recognise the importance of the material stock but address it in individual case studies at various scale levels. On the global scale, the available literature associated with material stocks in the built environment is limited, and lacking in detail. To address this gap, this article aims to summarise the existing knowledge on the residential building stock composition, to integrate it into a global level material content database, and to test the usability of the database by applying it in a global material stocks model for residential buildings. To facilitate using these data for scenario assessments, we do this in relation to the IMAGE (Integrated Model to Assess the Global Environment) Integrated Assessment modelling suite as used by PBL (Netherlands Environmental Assessment Agency) for the assessment of global level climate change scenarios (Stehfest et al., 2014a; Doelman et al., 2018; van Vuuren et al., 2017; O’Neill et al., 2017).

This article has the following objectives:

• Review of existing studies using different approaches to identify the material content in residential buildings.
• Compile a database of materials used in the construction of residential buildings at the global level in accordance with the IMAGE regions.
• Test the applicability of the database in a scenario context by modelling the past, present and future material stock in residential buildings based on IMAGE data and the materials database, using a bottom-up approach.

This paper focuses on stocks of materials in residential buildings. In a companion paper (Deetman et al., 2019), we add two pieces of research:
• A stock assessment of materials in various types of commercial buildings
• An assessment of inflows and outflows related to both residential and commercial buildings: the stock of building materials, and waste streams related to demolition.

2. Methodology and data

2.1. The building stock model

In order to assess the practical applicability of the database, we apply a stock model which aims to determine the in-use stock of construction materials used in the built environment and makes estimations of their future stock. In this paper, we focus on the in-use stock of residential buildings. The starting point for the stock estimations is the total Useful Floor Area (UFA) specified for 26 world regions, as projected by the IMAGE model and described by Daioglou et al. (2012) (Daioglou et al., 2012). Section 2.2 describes this in more detail. The UFA is translated into material stock for the period between 1970 and 2050 by using material intensities per square meter UFA. Similar to Müller’s model (Müller, 2006) the main drivers in the system are population and lifestyle in terms of UFA per capita.

The building stock model distinguishes between urban (including suburbs) and rural areas, as well as different types of residential buildings: detached houses, row houses, apartment buildings and high-rise buildings (Van Beers and Graedel, 2003; Stephan, 2013; Carre and Crossin, 2015). The additional variables that feed the model are the distribution of the population over the different dwelling types, the total UFA per building type for the 26 regions, and the material quantity per building type expressed in terms of kg/m2 UFA.

As mentioned above, the urban/rural distinction is made in the IMAGE-TIMER projections, while the distribution over the different dwelling types is calculated based on national statistics (Residential Energy Consumption Survey (RECS) - Energy Information Administration, 2019; Australian Government Australian Bureau of Statistics, 2019; Eurostat, 2019). The material intensity is based on the existing literature, reviewed and documented in a material intensities database. The different calculations steps and data sources are discussed in the next section and are illustrated in Fig. 1.

Fig. 1. Outline of the material stock analysis model. The rectangles represent the variables or drivers, the squares illustrate the stock and the hexagon is a calculation step. The data sources are written in bold.

2.2. The IMAGE model and the SSP scenarios

The IMAGE integrated assessment model (IAM) assesses the interactions between human development and natural systems. The model identifies the impact of energy consumption, land, water and other natural resources use on the natural environment and explores policy options concerning sustainable development, climate change and land use. IMAGE generate scenarios of socio-economic developments based on a set of drivers (Stehfest et al., 2014b). The model operates on a global scale, dividing the world into 26 regions. IMAGE has various sub-models on agriculture, land use, energy systems etc. One of these sub-models is the IMAGE Energy Regional model or TIMER. TIMER simulates the composition and dynamics of the energy system and projects its potential future trajectories and greenhouse gas emissions. The main variables used as input for the model are population and sectoral activity (e.g. GDP, Private Consumption) as they are identified as the most important drivers of energy demand (Van Vuuren et al., 2007). The model developed in this paper makes a similar assumption and regards population and lifestyle (in this case expressed as Useful Floor Area (UFA) per capita) as a driving force for material demand.

IMAGE, together with similar IAMs, is used to project the narrative of the Shared Socio-Economic Pathways (SSP) scenarios (Riahi et al., 2017). The SSP scenarios describe five different trajectories of socio-economic development of the world and are used as a basis for assessing climate change and sustainable development at the global level (van Vuuren et al., 2017). Each trajectory has a baseline variant which includes future developments without considering extra climate mitigation policies. Each also has variants with different levels of ambition for reaching climate policy targets. All of the SSP scenarios can be linked to an estimation of the building and material stock. For the purpose of the current study, we consider only the SSP2 baseline scenario which assumes moderate population growth, economic and technological development and contains no specific efforts towards sustainable development. The SSP2 scenario is regarded as the “middle of the road” SSP scenarios, as it projects present trends and developments into the future (KC and Lutz, 2017).

We use the IMAGE framework, and especially the TIMER model, to generate driving force data for stocks of materials in residential buildings. The TIMER model has detailed representation of the development of long term and global residential energy demand across urban and rural households, calibrated to historical data. The residential energy demand is linked to changes in demographic and economic development, as well as lifestyle parameters. These, in turn, affect “intermediate indicators” also generated by TIMER, including household sizes and residential floorspace (Daioglou et al., 2012).

Using the IMAGE framework has the advantage of enabling a link to globally recognised scenarios. Reciprocally, adding our model of the built environment to TIMER will enable to assess development scenarios on their consequences for resource requirements as well as environmental impacts in one modelling endeavour.

2.3. Data, variables and calculations

Fig. 1 illustrates the calculation procedures as well as data sources. A full representation can be found in the supplementary information of the companion paper by Deetman et al. The model involves four processes, depicted by hexagon and squares in the figure. The starting point for the material stock analysis model is IMAGE/TIMER. TIMER provides population numbers for all 26 regions over the period 1970–2050, divided into urban and rural population. TIMER also provides UFA per capita for the 26 regions and the period 1970–2050. By multiplying those, we obtain total UFA per region over time, divided into rural and urban population.

This information needs to be detailed further by allocating the thus obtained UFA data over the four different housing types (detached houses, row houses, apartment buildings and high-rise buildings). TIMER does not deliver that information. We derived multipliers for this allocation process based on population and housing statistics on the one hand, and our literature database on material intensities of buildings on the other hand. Statistics tell us the number of the population living in the four different types of dwellings ($P_{u/r}$), which we recalculate into the share of the population ($Fp_{u/r}$). The literature database provides, for the buildings investigated, square meters of UFA per house of each type, which we recalculate into $m^2 UFA/cap$ and which we assume is an indicator for lifestyle (L) which is building type-specific (d,r,a,h). The UFA/cap varies across dwelling types and is different for rural and urban areas. These we use as weighting factors to obtain multipliers or allocation factors for distributing the total UFA over the different dwelling types (R).

The stock or the UFA ($S_{ufa}$) is then obtained by multiplying the total UFA which we obtained from IMAGE-TIMER by these allocation factors R. This is done per region and per year over the 1970–2050 period, and for urban and rural areas. Finally, the housing stock in terms of square meters is multiplied with the material intensity data in terms of kg/m2 of the different materials, to determine the materials stocks (Sm), also per region and per year, and for six different materials.

2.3.1. Population

The population data – historical numbers and projections until 2050 for each for the 26 regions – are extracted from TIMER, based on the SSP2 scenario, as mentioned earlier. The description of the regional classification can be found in Appendix A (Figure A.1). The historical data are based on United Nations’ data and the projections on the assumptions made by the International Institute for Applied Systems Analysis (IIASA) (KC and Lutz, 2017). The future population projections are based on various assumptions related to economic, educational, policy and technical development of the globe and the individual regions, and are shaped by demographic rates and migration flows (van Vuuren et al., 2017; Dellink et al., 2017).

2.3.2. Distribution of population over the different dwelling types

Based on the difference in the construction materials, construction practices and even climate conditions, the residential buildings can significantly vary. To increase the reliability of the database and facilitate its usage, we chose to break down the stock of residential buildings into types. Four different building types were identified for the purpose of the study: detached houses, semi-detached/row houses, apartment buildings and high-rise buildings. The distinguishing of the dwelling types is based on a study conducted by Kumar et al. (2015) (Kumar et al., 2015), which recognises the most common types of buildings in Canada and is supported by data from national housing statistics. We made the distinction between apartments and high-rise buildings based on the number of floors: apartment buildings are defined as comprised of separated units within a building with a maximum of four storeys. This assumption is made based on the difference in the construction of buildings above four storeys and the need for reinforcement (steel) which leads to changes in the material composition (Engineering students’ guide to multi-storey buildings, 2019). We exclude the informal dwelling types typical for many of the developing countries due to lack of data on the material quantities in these buildings. We acknowledge their importance and we hope more building types will be incorporated in the database in the future.

Furthermore, we acknowledge that not only the materials but also the average house size, in terms of square meters per capita, is different for each building type (i.e. detached houses tend to be more spacious than apartments) and even between urban and rural areas. We attempt to account for this by applying a weighted disaggregation of the stock based on regionally specified average per capita floor space, as found in reviewed literature discussed in Section 1. For further information on the disaggregation, please refer to the Supplementary Information (SI).

Finally, another important note is that we consider that the population is not equally distributed throughout the four different housing types. In order to calculate the percentage of people living in different types of dwellings, data on the distribution of the population by dwelling type is collected from statistical sources. This type of information is not available at the global level. We used statistical information from Europe, North America, Australia and Japan, respectively from Eurostat, EIA’s Office of Energy Consumption and Efficiency Statistics (Residential Energy Consumption Survey (RECS) - Energy Information Administration, 2019), the Australian Bureau of Statistics (Australian Government Australian Bureau of Statistics, 2019) and Statistics of Japan (tat Japan and 2015 Populat, 2015). The data can be found in Table 1.

2.3.3. Lifestyle

The floor area per capita is one of the United Nation’s indicators to trace the progress towards meeting the goals of the Global Strategy for Shelter (UN-HABITAT, 2013). We, therefore, use the floor area (UFA) per capita as an indicator of lifestyle. An increase in this indicator implies an improvement in the living conditions of the population.

The per capita floor area (L) provided by the IMAGE-TIMER model increases towards 2050 for all regions. The minimum values found in the set are for urban India: 7 m2 per capita for the present (average of 2000–2015), increasing to 16 m2 per capita towards the end of the modelling period (2035–2050). On the other end of the scale we found rural United States, currently showing an average of 57 m2 per capita, which increases to 63 m2/cap towards mid-century.

Both the population share and the average per capita floor space for different building types (four building types in urban and rural areas) are static assumptions. This is most probably not realistic; however, we have no grounds to make different assumptions. In the scenario calculations, changes in this distribution originate only from urbanisation: an increased share of urban areas.

2.3.4. Materials intensity database

No official statistical datasets are available for material quantities in buildings. There is, however, a modest body of studies focused on the material contents in residential buildings (Gontia et al., 2018; Miatto et al., 2019; Heeren and Fishman, 2019). We used these studies to create our database, by translating this information into material intensity indicators (Im): the material content per square meter UFA. We included six materials in our database: concrete, steel, aluminium, copper, wood and glass.

In order to create the database, a list of publications was reviewed and a total of 56 studies was selected from this list to be included. The studies selected for the database were chosen based on their purpose, spatial and time scale, materials studied and the level of detail. The literature research process is described below.

First, we reviewed publications focused on compiling material intensities data for buildings. For the purpose of the current paper, we used two studies containing material intensities databases (Gontia et al., 2018; Heeren and Fishman, 2019). One of the studies has a national scope as it includes a database of material contents in residential buildings in Sweden (Gontia et al., 2018). The other presents a database at the global level (Heeren and Fishman, 2019).

Second, we explored MFA studies including flows and stocks of residential buildings and construction materials. The studies investigating stocks were preferred given that the current paper focuses on stocks and does not look into material flows. Eleven studies from this category were included in the database (Hashimoto et al., 2007; Reyna and Chester, 2015; Tanikawa and Hashimoto, 2009; Gontia et al., 2018; Miatto et al., 2019; Van Beers and Graedel, 2003; Stephan and Athanassiadis, 2018; Condeixa et al., 2017; Johnstone, 2001; Mesta et al., 2018; Huang et al., 2013). Some of them assess materials in the existing in-use building stock at a local scale (Reyna and Chester, 2015; Miatto et al., 2019; Condeixa et al., 2017; Johnstone, 2001). Other papers have a more general character as they focus on the material stock in entire countries (Hashimoto et al., 2007; Van Beers and Graedel, 2003). The assessment of the spatial distribution of the material stock using Global Information System (GIS) is the purpose of four of these studies (Tanikawa and Hashimoto, 2009; Kleemann et al., 2016; Van Beers and Graedel, 2003; Mesta et al., 2018). Estimating the material in- and outflows by using data on the material demand together with construction and demolition activities is the aim of several other papers (Stephan and Athanassiadis, 2018; Huang et al., 2013).

Third, studies dealing with the LCA of residential buildings were reviewed. We identified various studies conducted with different purposes in the broad context of energy efficiency, efficient use of materials and the possibilities for their recovery (Reyna and Chester, 2015; Yang et al., 2018; Stephan, 2013; Carre and Crossin, 2015; Kumar et al., 2015; Johnstone, 2001; Evangelista et al., 2018; Oyarzo and Peuportier, 2014; Ortiz-Rodríguez et al., 2010; Henry et al., 2014; Ezema and Olotuah, 2015; Nemry and Uihlein, 2008; Asif et al., 2005; Cuéllar-Franca and Azapagic, 2012; Buyle et al., 2015; Pajchrowski et al., 2014; Atmaca and Atmaca, 2015; Stephan and Stephan, 2014; Asif et al., 2017; El Hanandeh, 2015; Pinky Devi and Palaniappan, 2014; Bansal et al., 2014; Sharma and Marwaha, 2015; Ramesh et al., 2012; Shukla et al., 2009; Lee et al., 2017; Jeong et al., 2012; Lee et al., 2015; Chen et al., 2001; Li et al., 2016; Su and Zhang, 2016; Jia Wen et al., 2015; Abd Rashid et al., 2017; Utama and Gheewala, 2009; Utama and Gheewala, 2008; Suzuki et al., 1995; Rauf and Crawford, 2015; Carre, 2011; Aye et al., 2011; Fay et al., 2000; Bhochhibhoya et al., 2017; Zhang et al., 2014; Reza et al., 2014). Whenever they included data on the material composition of buildings, we used them in our database. We found that the most abundant sources are the LCA studies estimating the environmental impact of residential buildings (Kumar et al., 2015; Evangelista et al., 2018; Oyarzo and Peuportier, 2014; Ortiz-Rodríguez et al., 2010; Cuéllar-Franca and Azapagic, 2012; Jeong et al., 2012) and investigating the performance of the residential buildings in terms of energy as well the possibilities for energy optimisation using alternative materials (Kumar et al., 2015; Pajchrowski et al., 2014; Lee et al., 2017; Zhang et al., 2014; Blanchard and Reppe, 1998; Mosteiro-Romero et al., 2014).

Next to those studies, we included various publications specifying material intensities of particular houses. These publications were often useful since they are very specific and contain field data. The downside of these papers is that they lack representativeness. If necessary, we recalculated the data from these studies into kilograms of specific materials per square meter of UFA. The calculation steps can be found in the database in the Supplementary data. A full list with all of the papers can be found in the database as well (Appendix B).

We used those studies to compile a database containing material intensities (kg/m2) per building type and region. For 9 out of the 26 regions, the review did not yield any relevant studies. When no information was available for a specific region, material or housing type, we applied a global mean value based on the other regions with available data, as shown in Table 2 in the Results section.

Table 2. Mean values of the material content by housing type expressed in kg per m2 (Im). In the brackets, the number of data points is presented. They describe the material content in the 56 studies reviewed in our database. Some studies describe material content for multiple houses or case studies, thus leading to more than 56 data points in some cases.

Steel Concrete Wood Copper Aluminium Glass
Detached Houses 35.63 (87) 846.33 (104) 53.07 (121) 1.73 (13) 3.56 (19) 2.68 (43)
Row Houses 32.89 (8) 1208.13 (11) 34.97 (10) 0.01 (4) 0.23 (1) 1.07 (1)
Apartment Buildings 97.36 (53) 955.92 (84) 37.17 (82) 0.31 (22) 1.94 (17) 6.35 (33)
High-rise Buildings 116.98 (30) 910.21 (56) 54.48 (36) 0.01 (1) 2.20 (6) 4.42 (25)
Notes: For region-specific details, please see the SI.