Global Scenarios of resource and emissions savings from systemic material efficiency in buildings and cars


 Material production now accounts for 23% of global greenhouse gas (GHG) emissions. Resource efficiency and circular economy policies promise emission reductions through reducing material use, but their potential contribution to climate change mitigation has not yet been quantified. Here we present a high-resolution approach for tracking material flows and energy use of products throughout their life cycles, focusing on passenger vehicles and residential buildings. We estimate future changes in material flows and operational energy use due to increased yields, light-weight designs, material substitution, increased service efficiency, extended service life, and increased reuse and recycling. Together, these material efficiency strategies can reduce cumulative global GHG emissions until 2060 by 16-39 Gt CO2e (passenger vehicles) and 28-72 Gt CO2e (residential buildings), depending on climate policy assumptions. The use of wood and more intensive use are promising strategies in residential buildings. Ride sharing and car sharing are best for residential buildings.

For residential buildings: Directly from IEA statistics: https://www.iea.org/data-andstatistics?country=WORLD&fuel=Energy%20consumption&indicator=Share%20of%20total% 20final%20consumption%20(TFC)%20by%20sector queried on September 15, 2020. "Explore energy data by category, indicator, country or region": Energy topic: Energy consumption. Indicator: Share of total final consumption (TFC) by sector. Region: world. Sector: Residential: Time: 2018, Value: 21% Method: Gross domestic product (GDP) was deliberately not chosen as scenario driver nor model input to enable a high-resolution service-level framework that can be used to depict future low carbon lifestyles instead of aggregate demand modelling (e.g. modeling useful energy demand as aggregated function of GDP), for which there is little evidence for decoupling (Haberl et al. 2020).
Only in rare cases where there is a long enough time series, GDP-based extrapolations for the building stock were used for determining future service level in the SSP2 scenario. This approach was used for the USA and Japan.

Scenario and model framework
Services are linked to material cycles via the stock-flow-service nexus (Haberl et al. 2017) . The scheme starts with the energy service cascade to relate values to services to functions to products (and their operation) (Kalt et al. 2019), stock-driven modelling to translate product in-use stock demand into production of new and recycling of old products (Müller 2006), new and old products to material flows via dynamic material flow analysis (MFA) (Brunner and Rechberger 2016), and the material flows to the energy demand and related GHG emissions via environmental extensions as done in previous work Modaresi et al. 2014).

Fig. SI1-1 (next page):
Calculation scheme for the use phase (here shown as 'product stocks'). Stock levels are determined from historic stocks and scenarios following different storylines. The stockdriven model then determines the age-cohort decomposition of the in-use stock as well as product inflows and outflows and the associated material content. With the total stock broken down into different age-cohorts by the stock-driven model, the function and energy flows of the use phase can then be determined (cf. below) by applying the following parameters in turn: intensity of operation and intensity of use (for service flows) and energy intensity and energy carrier split (for energy use of the use phase). The indices are as follows (cf. RECC config table and RECC index table): t: time, c: agecohort, r: region, g: good/commodity/product, S: scenario (SSP, RCP, and/or RE), V: service category, n: energy carrier, t0: starting time of prospective assessment (2015).The red section of this figure is our interpretation and implementation of the energy service cascade (Kalt et al. 2019).

ODYM-RECC model
The ODYM-RECC model (open dynamic material systems model for the resource efficiency and climate change mitigation project) is a modular depiction of major end-use sectors and the material cycles for the climate-relevant bulk materials (Pauliuk and Heeren 2020) (https://github.com/YaleCIE/RECC-ODYM). Its system definition  comprises the use phase of materials (in products) and the material cycle stages mining, primary production, manufacturing, waste management and scrap recovery, and remelting/recycling as well as an energy supply scenario.
ODYM-RECC generates a set of what-if scenarios (Börjeson et al. 2006) for the climate-relevant end-use sectors and bulk material cycles against different socioeconomic, technology deployment, and climate policy backgrounds. It does so by applying a mass-balanced framework for the material cycles (Brunner and Rechberger 2016). It allows us to study the impacts of a broad spectrum of sustainable development strategies on the material cycles and identify trade-offs and constraints. It does not assess the likelihood of realisation of any of the scenarios studied but checks if mass balance constraints (e.g. by long product lifetimes or limited scrap supply) render some scenarios unfeasible from a material cycle point of view. ODYM-RECC is a multi-layer model depicting products, materials, chemical elements, energy flows, and emissions, with mass balance across all processes down to the individual chemical element. ODYM-RECC has six modules that quantify the system in Fig. 0.1 by translating a given service scenario into product stocks, inflows and outflows (module 'use phase UP', using stock-driven modelling (Müller 2006), product outflows into scrap and recycled materials (module 'waste management and recycling WR', using parameter equations), product inflows into material demand and fabrication scrap (module 'manufacturing MF' using parameter equations), material demand into primary production and related impacts (module 'primary production PP', using environmental extension factors), and by determining the chemical element composition of al stocks and flows (module 'material-element composition ME', using mass balance). Finally, the energy consumption and environmental pressure and impact indicators are calculated (module 'energy and extensions EX').
For the RECC project, 35 data aspects (time, age-cohort, process, material, chemical element, waste/scrap, environmental extension, socioeconomic scenario…) were defined and each of the 104 model parameters has a specific data model that links it to the data aspects. For example, the parameter for the product lifetime extension potential has the three aspects 'product', 'region', and 'scenario. The parameter for the future stock levels needed has the four aspects 'scenario', 'product', 'region', and 'time'. The resolution of each data aspect is defined in the model configuration file, a summary is given in Table SI1-1. The model parameters are linked to the system variables (stocks and flows shown in Fig. 0.1) via the model equations, which are grouped into the five ODYM-RECC modules. The parameters are divided into three groups: socioeconomic parameters such as future population, service demand, or intensity of operation of stocks (e.g. vehicle-km per year), technology parameters like energy efficiency of stock operation of the future emissions intensity of energy supply, and resource efficiency parameters describing both the potential for resource efficiency at the different stages of the system (green boxes in Fig. 0.1), and the speed of implementation of these potentials under different socioeconomic and climate policy scenarios.
Each RE strategy can be implemented separately or as part of a cascade of strategies. The model allows for calculating the impact of one strategy at a time (sensitivity analysis) or a bundle of strategies in different orders of implementation, each for different socioeconomic and climate policy scenarios.

The ODYM-RECC Database
The ODYM-RECC v2.4 database contains 104 model parameters of two to six dimensions each. Parameters range from static values (direct emissions of combustion by MJ of energy carrier) to highly detailed highly uncertain datasets (e.g., the future energy carrier split of buildings by region, time, and operation mode (heating/cooling/hot water).
The ODYM-RECC database was compiled as a community effort involving a large number of experts. Its scope is unprecedented in the industrial ecology community. Data templates and project wide classifications were used to facilitate the compilation of the various types of information.
Depending on data availability, we applied several pathways of data compilation, which are listed and described in detail below.
 Extract mostly socioeconomic parameters from existing scenario models (scenario reference)  Compile own plausible scenario estimates for socioeconomic parameters in line with the different scenario narratives where established model framework results are not available (group consensus scenarios)  Extract process-, product, and material-specific data from the engineering and industrial ecology literature (bottom-up data)  Extract quantitative estimates of resource efficiency strategy potentials, mostly related to prototypes and case studies, from the literature (strategy potentials)  Simulate energy consumption and material composition of a number of building and vehicle archetypes with specialised software, which are then used as bottom-up product descriptions with and without implementation of RE strategies (archetype descriptions)

Scenario reference
For the socioeconomic parameters the Shared Socioeconomic Pathways (SSP) database and model results as well as available data from the World Energy Outlook and Energy Technology Perspectives models were used wherever possible, e.g., for future population, future GHG intensity of energy supply, or the drive technology mix for vehicles (Riahi et al. 2017;O'Neill et al. 2014; OECD/IEA 2010a; IEA 2015; OECD/IEA 2017). The data were extracted from available databases (like the SSP scenario database hosted at IIASA: https://www.iiasa.ac.at/web/home/research/researchPrograms/Energy/SSP_Scenario_Data base.html) or shared by colleagues, then parsed and reviewed by the RECC team, then aggregated, disaggregated, and interpolated to fit the ODYM-RECC project-wide classification.
For each parameter file the data gathering process is documented both in the respective template files in the RECC database (if only Excel was used), in custom scripts (for more comprehensive datasets) and in the data log files archived under https://github.com/YaleCIE/RECC-data.

Group consensus scenarios:
For some parameters like the future stock levels or the split of residential buildings into different types no detailed SSP-consistent scenario calculation was available that we could refer to. Hence we assumed a set of plausible target values for a number of socioeconomic parameters in line with the storylines of the individual socioeconomic scenarios. This process is commonly used when translating broad storylines into high product and regional resolution and sector-specific parameters, cf. Riahi et al. (2017) and Grübler et al. (2018). The target values for 2020, 2030, 2040, 2050, and 2060 chosen and the rationale for their choice are documented in scenario target tables, one for each parameter. From there, the target values are read, interpolated, smoothed with a moving average, and exported in ODYM format to be directly used in the ODYM-RECC model. The documentation for the individual parameters is archived in https: The open model and data framework allow for third parties to modify the scenario assumptions and to run calculations with custom parameters and storylines.
Bottom-up data: For the energy intensity, emissions intensity, and material composition of products and processes detailed but representative product or process descriptions were compiled from the literature and available databases. These data include the material composition and specific energy consumption of vehicles and buildings, e.g., (Hawkins et al. 2013; Reyna and Chester 2014; Marcellus-Zamora et al. 2016), the loss and recovery rates for the manufacturing and waste management industries e.g., Liu et al. 2012), and the specific energy consumption and process emissions for the manufacturing, waste management, and primary material production industries (Wernet et al. 2016; OECD/IEA 2010b; IEA 2015; OECD/IEA 2017). While the data can be regarded as representative of current average global technology, their main limitation is that they are static and no information on their change under different socioeconomic and climate policy scenarios, in particular, is given. To become more realistic a scenario reference was made wherever possible (cf. above), e.g., for the changing GHG intensity of the supply of different energy carriers, for which a combination of MESSAGE IAM results and IEA Energy Technology Perspective results was used. Also, for the average GHG intensity of primary metal production a scenario analysis based on ecoinvent was calculated to take into account scenario-dependent changes of the GHG intensity of electricity generation.
Resource efficiency Strategy potentials: For some parameters, including the improvement potentials for fabrication scrap, end-of-life recovery efficiency of scrap, re-use of steel components in buildings, or product lifetime extension, previous estimates can be used . The other strategies were covered by the scenario formulation approach described above.

Archetype descriptions:
Here, 'archetype' refers to an idealized description of the physical properties (energy intensity of operation and material composition) of a product with a certain functionality, assuming typical user behaviour in a given region.
For passenger vehicles, drive technology, segment (car size), and material design choice together determine the archetypes' material composition, and the three properties above plus the assumed driving cycle determine its specific operational energy consumption (specific = per km driven).
For residential building, building type, energy standard, material intensity (conventional or lightweight design), material design choice, and stylized climate conditions (heating and cooling degree days by region) together determine the archetypes' material composition and specific operational energy consumption (specific = per m2).
For the final product categories residential buildings and vehicles, the product-specific simulation tools BuildME (https://github.com/nheeren/BuildME), GREET (https://greet.es.anl.gov/) and FASTSim (https://www.nrel.gov/transportation/fastsim.html) were used to model the archetype descriptions by deriving model estimates for both the material composition and energy intensity of operation for different building and vehicle configuration. For each of the nine building and six vehicle types four archetypes, representing maximal potential for change, were simulated: a standard product without special consideration of material efficiency, downsizing, or material substitution, a downsized product, a product with ambitious material substitution, and a downsized materialsubstituted product.
For a detailed description and definition of all model aspects, the classifications used for them, the system variables and parameters, the model equation and their division into modules and the data compilation, (dis)aggregation and formatting process, we refer to the ODYM-RECC model documentation.
The ODYM-RECC database is formatted in standardised spreadsheets and archived on Zenodo (dataset DOIs [to be inserted for final publication]).

Model resolution
The information presented here is a summary only. The full info about the resolution of the RECC project is documented in the Master classification file, which is part of the project's database: Chemical Elements, dimension: Element:

MaterialProductionProcess, dimension: Process:
 One (average) primary production process for each material.

ManufacturingProcess, dimension: Process:
 One average manufacturing process for each product/good Waste management process, dimension: Process:

RECC 2.4 database and scenario drivers
Here, we list the changes in the main drivers and the country/region-level plots.                      Main stock and service parameters, passenger vehicles, by country/region. Figures SI5-1 -SI5-20 below show the temporal development of the passenger vehicle fleet (total number of vehicles by drive technology, left y axis) and the annual change of the total fleet size (stock change, right y axis).

Additional Results and Discussion
Detailed description of Figure 1: Passenger vehicles, top row: The striking difference in development is that emissions in the Global North and its constituting regions decline steadily over time, in all scenarios. In the Global South, they tend to peak between 2030 and 2040 (2°C scenario) or continue to increase (NoNewClimPol scenario). This behavior is a direct consequence of contraction and convergence of service levels: Global North service levels decline and are provided more efficiently, allowing for absolute decoupling of GHG emissions. Global South service levels increase substantially in most regions (current car ownership rate in Sub-Saharan Africa is ca. 18 cars per 1000 people), and fuel shift, low carbon energy supply, and ME can counter this strong growth trend only in the 2°C scenario and only after 2030. For China, the trend reversal can happen earlier, between 2025 and 2030.
In the SSP1 (easy mitigation and adaptation) scenario, the relative impact of full ME is similar in both climate policy scenarios, because both non-ME baselines are material-intensive, with a 15 year product lifetime and substantial improvement potential for end-of-life scrap recovery especially for plastics but also for the other materials. The shift to smaller segments and light cars and the car-and ride-sharing strategies are important contributors as well.
Residential buildings, bottom row: Here, all regions show a huge GHG emissions reduction potential, as -despite strong growth in the regions of the Global South -the emissions reductions from increased energy efficiency in buildings outpace all assumed growth. Still, especially in the NoNewClimPol scenario, 2050 emissions without ME are nowhere near carbon neutrality, which is an aspiration for the easy-to-mitigate residential building sector in some strategy portfolios (zero emission buildings). Here, the ME strategies can prove particularly effective in achieving deep emissions cuts: In some regions, including the global total, Global South, China, India, and Sub-Saharan Africa, The 2°C-plus-ME GHG emissions are only a fraction of the 2°C-no-ME emissions, and for some regions (for SSP1: Global South, Other Asia, Middle East-Northern Africa, and India), forest carbon uptake associated with residential building timber use even leads to total negative emissions.
Regarding the forest regrowth modelling: There are of course some big assumptions here: that sustainable harvest exists and is possible at the scale required, that forests would not have added carbon to storage without timber removal, and there is no opportunity cost of the timber no longer being available for other carbon-saving measures.                   4.5. Impact of ME on primary and secondary material production over time 4.6. Impact of ME on use phase energy consumption over time

4.7.
Share of electricity and hydrogen in use phase energy consumption over time 4.8. Passenger-km are delivered over time Additional discussion and context Edelenbosch et al. (Edelenbosch et al. 2020) recently published highlighting the lack of attention towards demand-side solutions in IAMs, noting that there is great uncertainty in future energy/cap, and pointing out the potential for energy efficiency in transport and buildings in particular.
Not only through our assessment of ME efficiency strategies here, which as noted is absent from the IAM framework, we also contribute important knowledge with respect to future drivers of service and energy demand per capita, responding to their "challenge [for modelers] to better understand drivers of future energy efficiency and service demand, that contribute to the projected energy demand".

Other supporting material
The supporting information SI2 provides a detailed description of the ODYM-RECC v2.4 model as well as a summary of the input data. The data gathering is documented in the respective data templates of the 104 model parameters, which are available via [link to final dataset version on Zenodo will be inserted prior to publication!]. The ODYM-RECC v2.4 model is available under a permissive license via https://github.com/YaleCIE/RECC-ODYM. The model results here were calculated by running the ODYM-RECC scripts of commit no. 7bd4e46 with the data in the archive linked above.