The Horizon 2020 project ProSUM (2015-2017) resulted in the creation of the Urban Mine Platform (UMP). This RMIS – vehicle raw materials application summarises work undertaken to update and expand the original ProSUM data on the fleet and composition of vehicles.
The study uses dynamic stock and flow modelling to quantify the number of vehicles in the EU27+3 and a selection of their associated metals over time. The stock flow modelling which accounts for the number of vehicles is combined with mapping of vehicle metal compositions over time.
The stock-flow model employed for modelling the number of vehicles employs blended data from EuroStat, ICCT, and several other data sources. The stock represents the vehicles in use in the fleet. The flows are inflows or outflows to and from the stock. The inflows to the stock are new vehicles and imported used vehicles entering the fleet. The outflows are vehicles deregistered from the fleet, which are followed by flows of used vehicles (a) exported from the EU27+3, (b) reported as recycled or (c) of unknown whereabouts.
Data from Eurostat includes “drivetrain type”. The dataset here covers vehicles below 3.5 tons, being passenger cars - M1, vans - N1 and unknown. The dataset represents these three types of vehicles that are further categorized by drivetrain type, cylinder capacity, and weight category (see Table 1).
Stocks for vehicles in each country in the EU27+3 area are calculated for each vehicle key and for every year of data up until 2020, which is the last year of complete available data. This alongside age distribution data allows a complete snapshot of the entire stock of vehicles for each year modelled.
Flow data is calculated from the same data sources, with the exception that the age of all new vehicles is assumed to be zero. Trade flows are calculated from COMEXT data, and, where the age distribution of the traded vehicles is not in the data, it is inferred from the change in stock data. The net-registered trade flows are in the order of 1% of new vehicles put on market every year and vastly smaller than the existing stocks.
The operation of the model happens in two stages. First, the model calculates the stock of vehicles at time t+1 by taking the stock at time t and adding the flows of new vehicles and imports and subtracting the flows of exports. This calculation is done for every vehicle key and country. The difference between this value and the observed stock gives the number of ’deregistered’ vehicles. This process is repeated for all years with historical data. De-registrations compared to the stock are used to fit a ’death function’ for each country, modelled as a Weibull survival curve.
Second, when the model switches into future ’projection’, where there is no data on stocks and flows, the deregistered flow is calculated from these lifespan distribution functions and is added into the calculation to predict the stock at time t+1.
The final step is to calculate the distribution of ELVs. The data for recycled vehicles (i.e. ELVs reported to Eurostat) only includes the total number of vehicles.Since the deregistered vehicles are calculated by the model at a vehicle key and cohort level, it is assumed that the number of recycled vehicles had the same vehicle key and cohort composition as the deregistered vehicles. Notably, this number is significantly smaller than the calculated number of deregistered vehicles, but the gap is decreasing as ELV statistics improve.
The model calculations are run for each vehicle key and country for the years 2000-2022. The stocks and flows modelled are: new vehicles placed on the market, internal imports of second hand vehicles, internal exports of second hand vehicles, external (outside EU) imports of second hand vehicles, external exports of second hand vehicles, fleet in use, deregistered vehicles recorded as recycled and deregistered vehicles not recorded as recycled.
Vehicles are categorized using vehicle keys defined by four characteristics (see Table 1). This provides 675 different possible vehicle keys. Additionally, the vehicles are categorised by their cohort, which is the individual year they are placed on the market, between 1980 and 2023. The majority of these vehicle keys represent unlikely or in some cases impossible combinations of characteristics (e.g "cylinder capacity" is not relevant for BEV). Moreover, many parameters do not change over time or across categories, and some of them vary in a determined way.
Each parameter considered in the model needs to be determined for each possible vehicle key of each cohort. In consequence, quantifying those parameters relies on data collection for specific cases, for trends in time and across categories, as well as robust working hypotheses in order to interpolate and extrapolate the parameters where no specific data is available. This is performed by quantifying relevant parameters (e.g. the total mass of electrics and electronics) for an average or example vehicle, and then extrapolate this to other vehicle categories using information about key differences that depend on selected characteristics of the vehicles. For example, it is often relevant to scale the mass of materials or components per vehicle by the vehicle weight category. In general, when there is no known reason to believe that a systematic difference between vehicle categories for a particular parameter exists, data across those categories are assumed to be identical.
For each parameter, the level of differentiation between different vehicle keys and cohorts depends on its relevance and data availability. The most detailed data (e.g. mass of aluminium alloys per vehicle) are differentiated by three characteristics: vehicle drivetrain type, vehicle weight or cylinder capacity, and vehicle cohort. The least detailed data (e.g. mass fractions of elements in aluminium alloys) are not differentiated between different vehicle types at all.
The data follows a hierarchical structure: each vehicle is composed by components and materials, that are themselves composed by (chemical) elements. Some components and materials are called after elements, such as "wrought aluminium" or "high strength steel", but are in fact alloys that contain the elements "Al" or "Fe". Moreover, not all the components and materials of a car are taken into account, which means that if the mass of a vehicle is higher than the sum of the mass of its components and materials in the model. The rest is composed by components that are not considered in the study, such as tires, wind shields, batteries or car seats.
For each component and material, we quantify two parameters: the mass of the component or the material per vehicle, and the mass fraction of each element in each material and component. With these two parameters, we can easily calculate the mass of each element contained in material or component per vehicle and the total mass of each element per vehicle.
Table 2 presents the full lists of materials, components and elements as well as which elements are quantified in each material/component. Power electronics here includes the inverter, DC-DC converter, on-board charger and related controllers.
Most of the data is based on the work realized in the frame of the Horizon 2020 project ProSUM (2015-2017), which resulted in the creation of the Urban Mine Platform (www.urbanmineplatform.eu). Additional information are collected from market reports, scientific literature and official statistics.
A detailed description of the methodological approach used for each material, component and element is available in section 2.2 of the accompanying JRC report: Amund N. Løvik, Charles Marmy, Maria Ljunggren, Duncan Kushnir, Jaco Huisman, Silvia Bobba, Thibaut Maury, Theodor Ciuta, Elisa Garbossa, Fabrice Mathieux, Patrick Wäger, Material composition trends in vehicles: critical raw materials and other relevant metals. Preparing a dataset on secondary raw materials for the Raw Materials Information System, EUR 30916 EN, Publications Office of the European Union, Luxembourg, 2021, ISBN 978-92-76-45213-3, doi:10.2760/351825, JRC126564.
The data prepared in this study have variable degrees of uncertainty due to differences in data availability. Due to the diversity of the primary data (e.g. collected using very different methods) and the small amount of data available, statistical analysis to estimate uncertainties is not feasible.
In general, the relative uncertainty is higher for components, materials or elements with lower mass per vehicle. This is both due to the quantity of data available to inform our estimates as well as the inherent uncertainties of primary data collection. For example, chemical analysis is more challenging for metals that occur in low mass fractions (e.g. Au and Ag) compared to metals that occur at higher mass fractions (e.g. Cu). We consider data on steel and aluminium content the most reliable data, since it is ultimately based on a comprehensive analysis of many different representative vehicle models. The least reliable data are the mass fractions of rare metals, such as precious metals and rare earth elements, due to the small amount of primary data, and potentially high variability between different vehicle models. For example, observations of gold content range from around 0.2g per vehicle (Widmer et al. 2015) up to 7g per vehicle (Cullbrand and Magnusson, 2011). The lower end estimates almost certainly include some underestimation due to not capturing all relevant components, while the higher end estimates might be high-end vehicles with much higher content than the average car. This range gives an indication of the variability between different vehicle models. However, the uncertainty for the dataset presented here, which represents average values for the entire vehicle fleet, is considered lower.
In Table 3, we present a qualitative indication of the uncertainty for each element and some additional comments about interpretation of the final results.
Please cite the content of this website as: Maria Ljunggren, Amund N. Løvik, Duncan Kushnir, Charles Marmy, Jaco Huisman, Silvia Bobba, Thibaut Maury, Theodor Ciuta, Elisa Garbossa, Fabrice Mathieux, Patrick Wäger, 2022, RMIS – Raw materials in vehicles . Further details about the vehicle content data are available in Løvik et al. (2021) as described above.
This work is carried out in the context of the JRC 2021-2022 institutional work-programme, through its project “SUstainable and secure sourcing of Raw materials for future competitive and low carbon European strategic VALue chains (SureVAL)”. The work is supported by the service contract number 722515.