Decarbonisation Index (DCI): an LCA-based key performance indicator for the automotive industry

Light-duty vehicles (LDV) and vans contributed 8% of global direct emissions in 2021 (IEA 2022). Automotive original equipment manufacturers (OEMs) have incorporated climate change in their strategies, aiming at contributing to slow down anthropogenic carbon emissions caused over the life cycle of their products and services. Ford Motor Company, Mercedes-Benz AG, Volkswagen AG, BMW Group, General Motors, Groupe Renault, and Toyota Motor Corporation (i.a.) have committed themselves to meeting verified science-based carbon reduction targets (SBTi 2022). When looking into the future, this commitment is especially crucial as the demand for, e.g. urban passenger transport in LDVs in Asia is estimated to increase by 73% between 2015 and 2035 (OECD 2022). The automotive industry thus needs to find sufficient measures to cut carbon emissions while their production output is increasing.

Life cycle assessment (LCA) studies have been applied in the automotive sector for nearly three decades to compare different technical solutions regarding their potential environmental burden (see, e.g. Gradin et al. 2020, Bushi et al. 2019; Lombardi et al. 2017; Muñoz et al. 2006; Saur et al. 1996). Furthermore, LCA studies of vehicles are a powerful tool for OEMs to identify carbon hotspots and consequently reduction measures on the product level. The studies published by OEMs are externally verified to be in line with the ISO 14040/14044 requirements and are usually published for the market launch of a new model. LCA studies describe, however, only a snap-shot of the emissions caused by one vehicle model in a certain point of time and a specific geographic region. Furthermore, mostly generic data sourced from LCA databases is used to model the product carbon footprint and the studies are rarely updated. Due to the immense effort of generating a high-quality LCA for a product as complex as a vehicle, LCAs are so far only publicly available for a certain share of vehicle models in an OEM’s portfolio. Though not a product but a fleet perspective on life-cycle carbon emissions is necessary to develop decarbonisation strategies on an organisational scale. The interplay of shares of different vehicle powertrains, vehicle sizes, etc. in different markets matters in order to develop the most CO₂e efficient reduction measures. These encounter adjustments of the vehicle portfolio or realising measures along the product life-cycle from supply chain through production to use as well as recycling.

From a carbon point of view, the scope 1–3 emissions according to the GHG Protocol (WBCSD and WRI 2004) take on an “LCA + ” perspective as additional indirect emission sources such as employee commuting and business travel are included (Fig. 1). In contrast to vehicle LCA studies, more specific, up-to-date and regionalized data is used to produce the scope 1–3 emissions inventory. Where possible, the absolute emissions reporting to the Carbon Disclosure Project (CDP) reporting relies on measured primary data for the full fleet of vehicles sold in the respective reporting year. Due to legal regulations, this is the case especially for scope 1 and 2 emissions and for tank-to-wheel emissions of the vehicles according to defined test cycles. For the remaining life cycle phases (supply chain, well-to-tank, recycling), the CDP reporting depends on input from LCA studies and life cycle inventory (LCI) databases (see, e.g. Agyei Boakye et al. 2023). Emissions of the other scope 3 categories are mainly calculated based on generic data sources and have—despite single exceptions (logistics)—merely indirect linkage to the physical vehicle life cycle (Neef 2020).

The yearly updated inventory can also be externally verified, e.g. on limited or reasonable assurance level (see, e.g. (BMW 2020; Toyota 2020; VW 2022). This corporate carbon accounting is the prerequisite for a targeted reduction in emissions: knowledge of sources and amount of past emissions on an organisational scale (Damert et al. 2017). However, long-term strategic decisions cannot solely be based on past carbon performance but need additional modelling of future carbon emissions to evaluate the effect of potential reduction measures (Schaltegger and Csutora 2012).

The United Nations German Global Compact Network recommends the use of a key performance indicator (KPI) to monitor the progress towards achieving absolute carbon reduction targets (Deutsches Global Compact Netzwerk (DGCN) 2017). Likewise, establishing a KPI to internally monitor the effectiveness of reduction measures and support their implementation is recommended by Busch (2010) and Zvezdov and Schaltegger (2015). KPIs are therefore related to both past-accounted and future-modelled carbon emissions. Such a decarbonisation KPI is not only complex in its calculation but also in the expectations of different stakeholders it is supposed to meet. For example, an OEMs’ strategy department expects stability of calculation principles. Though OEMs’ decarbonisation practitioners expect an “evolving KPI” that is adaptable to changing methodological requirements and data specificity, external stakeholders like non-governmental organisations expect both a transparent calculation methodology and the exact depiction of real-world emissions right from the start. Can a decarbonisation KPI meet all these demands? In this study, we discuss this issue among others and address the following research question: How can existing carbon accounting and LCA approaches be combined to a life-cycle-based decarbonisation KPI for the automotive industry?

The above described approaches of producing vehicle LCAs on product level and absolute emission inventories on fleet level currently applied by OEMs are both static and past-oriented. The here developed decarbonisation KPI needs to combine this static and past-oriented perspective for monitoring of past emissions with a modular and future-oriented perspective that allows for scenario-building. As such, external transparency of the OEM’s decarbonisation progress for stakeholders as well as internal monitoring, planning and steering of fleet compositions and reduction measures would be facilitated.

The KPI’s functional unit of “t CO₂e/vehicle” would allow for an operationalisation of decarbonisation targets in harmony with the internal key processes of OEMs which are product-oriented. In line with Burritt et al. (2011), corporate carbon management schemes should be set-up in a resource-efficient way. The decarbonisation KPI should therefore, whenever possible, draw on existing data sources and calculation approaches of the past emissions reporting. A dynamic modularity of the KPI calculation model for future emission modelling is necessary to compute scenarios with different fleet compositions and reduction measures to decide on the most efficient strategy to meet the decarbonisation target.

Dynamic because today’s absolute past-oriented emissions reporting is calculated for each scope and category separately. As such, the relation of the parameters influencing emissions caused during the life-cycle of vehicles on fleet level cannot be shown by changing an input parameter at one point in the model because no holistic model yet exists. Though this is necessary to model the KPI’s development depending on the input received from the fleet planning departments which, in turn, depend on the respective fleet emission legislations, for example, an increasing share of battery electric vehicles (BEV) will lead to increasing material supply chain emissions but a lower fleet emission average. These correlations must be depicted in a dynamic model. The calculation model’s modularity is important because it also matters (to stay with the example) in which market, by which brand and in which year the share of BEVs increases. Here, the crucial input parameter “CO₂e-intensity of the electricity mix” must be connected to the market, time, brand and life-cycle-phase-specific emissions modelling. When looking exemplarily at the VW Group and its diverse brands, an electrified SUV BEV produced and used in China in 2023 potentially causes higher total carbon emissions and thus a higher effect in the overall Group KPI result than a compact BEV in the EU in the same year: (a) because the CO₂e-intensity of the Chinese electricity grid mix is higher than the EU one; (b) because a SUV BEV would be built with a higher battery capacity, higher vehicle weight and higher electricity consumption per kilometre driving distance than the compact BEV; (c) because more vehicles are sold in China than in the EU; and (d) because sales numbers differ. The example illustrates that the modularity of the model must encompass even the component level depending on identified carbon hotspots in the supply chain (in this case the battery cell production). As such, supplier-specific data instead of the generic data used in vehicle LCAs can be used to calculate reduction potentials on fleet level. Below, we describe which other existing approaches can help to achieve this resource-efficient dynamic modularity needed for the decarbonisation KPI.

In the following case study, we want to show an exemplary application of the DCI by using publicly available data of the Volkswagen Group as well as scenarios derived from publicly available studies. The data available from the company is suitable for this case study as it covers a variety of brands with different vehicle portfolios and market activities. Like this, the KPI and its use can be shown on different aggregation levels. The DCI’s applicability for developing life-cycle-based decarbonisation strategies is shown in an exemplified emission projection for 2035. The modelling options, data specificity and sources pursued in this case study are marked with an “X” in Table 1 (for past emissions/2015) and Table 2 (for future emissions/2035) and are indicated for each data point. The modelling options listed are not exhaustive but are meant to give an idea of the numerous possible options to calculate a DCI depending on the available data and the pursued steering model. The options are grouped along the phases of the DCI plus basic input data that is used throughout all phases. The data specificity options indicated are both on vehicle level (component, gearbox, model, powertrain, segment) and on fleet level (market, brand, group). Additionally, it can be indicated whether the data used is reporting-year specific and, in Table 2 only, whether interpolated data is used to model future emissions. Data sources are distinguished in internal, public, supplier and other. Option A1 in Table 1 can be read as follows: the number of vehicles used for the 2015 DCI calculation is powertrain, segment, market and brand-specific. As the information is sourced from the VW Annual Report 2015, it is reporting year-specific and based on public sources. In turn, option B1 is only indicated to be powertrain and segment-specific because reference vehicle LCAs are used to group all vehicles into their powertrains (ICE/BEV/PHEV) and segment (regular/large). These vehicle LCAs are not reporting year-specific because they have been published in another year. For one example in Table 2, modelling option 4A can be read as follows: the emissions calculation of battery cell manufacturing with renewable energy sources is based on component (i.e. battery cells), segment (i.e. battery capacities of reference vehicle LCAs) and powertrain (i.e. BEV) specific data and includes interpolated data (projected emission factors from an LCA database). In Table 2, two general modelling paths are shown. Either static data from past reported DCIs or dynamic data based on projections and interpolation between data points can be used.

The modelling results presented in the above case study mainly rely on LCA-sourced data and assumptions. Environmental impact data from LCA databases represents sector average or most likely figures. As no measurement is completely certain, such data needs to be considered including its uncertainty ranges, i.e. the parameter uncertainty (Bamber et al. 2020; Bałdowska-Witos et al. 2020). Additionally, the model (in this case, the DCI methodology) resp. its results need to be tested for model uncertainty. Below, we conduct a global sensitivity analysis of selected input parameters regarding their effect on the modelled DCI 2015 figure (Fig. 3) by means of the Monte Carlo (MC) simulation and the software Crystal Ball (10,000 repetitions). Like this, we can identify the parameters which contribute most to the uncertainty of modelled results. Across industry sectors, scientific disciplines and LCA research, the MC simulation for conducting stochastic uncertainty or sensitivity analysis is most commonly used (Bałdowska-Witos et al. 2020).

Chosen input parameters to be tested for their combined effect on the DCI results were assumed to have triangular distributions. The reference New European Driving Cycle (NEDC) use phase emission averages are estimated to have an uncertainty range of a generically set − 10% and a specific + 21%. The + 21% are based on the average deviation of the NEDC and the Worldwide harmonised Light Duty Test Cycle (WLTC) emission values in the EU (Dornhoff et al. 2020). The assumed uncertainty of the modelled lifetime kilometrage of 200,000 km is assumed to be + / − 15% based on Weymar and Finkbeiner (2016). A generic + / − 10% uncertainty range for the vehicle curb weights and the share of scope 1–2 emissions on overall production emissions was assumed due to lack of published specific data. The used supply chain and recycling emission factors are derived from vehicle LCA studies. These studies are based on ca. 40,000 processes modelled within the LCA software GaBi which are connected to specific emission factors within an LCI database. Each factor has its own uncertainty. Unfortunately, the authors of the vehicle LCAs used in this case study did not indicate an overall uncertainty of their published LCIA results. Therefore, the supply chain and recycling emissions factors are also assumed to have a generic + / − 10% uncertainty range.

Figure 8 shows the results of the MC simulation taking into account the described uncertainty ranges for the chosen parameters. The modelled arithmetic mean value of the MC simulation is 38.85 t CO₂e/vehicle and is thus 2.6 t CO₂e/vehicle higher than the benchmark value (36.2 t CO₂e/vehicle, see Fig. 3). The calculated standard deviation is 2.88 t CO₂e/vehicle. The 95% confidence interval is indicated with 33.87 t CO₂e/vehicle at the lower end and 44.87 t CO₂e/vehicle at the upper end. The main drivers for the DCI’s model uncertainty are the parameters with the highest rank correlation coefficients: the NEDC tailpipe emissions value of the small ICE reference vehicle (0.51), the lifetime kilometrage (0.33), the NEDC tailpipe emissions value of the large ICE reference vehicle (0.06) and the vehicle curb weights (0.01).

As shown in the above analysis, the fleet emission averages are crucially defining OEM’s reported GHG emissions over the life cycle of their products if ICEs dominate the fleet. A change of legislation regarding driving cycles (like in the EU in 2017 (Dornhoff et al. 2020)) is therefore an external requirement that affects OEMs’ reported GHG emissions majorly. Re-calculating past reported emissions becomes necessary when the methodology changes between reporting years. Otherwise, the development of emissions cannot be interpreted correctly.

In order to address the robustness of parameters that are (not yet) legislated, OEMs could, e.g. include market-specific measured vehicle lifetime kilometrages instead of the generic 200,000 km. In this exemplary DCI calculation, only publicly available data was used as input to the model. When applying the DCI method internally, OEMs will have access to more accurate and specific data regarding the fleet emission averages and curb weights per, e.g. brand-powertrain-gearbox combination thus addressing these parameters with higher certainty.

Although the model’s sensitivity to the chosen parameters is comprehensible, the complexity of the tested model and the variety of input parameters with further non-included uncertainties should be kept in mind when interpreting the DCI modelling results. This holds especially true for projected future DCI figures when more input parameters (like, e.g. electricity mixes for charging electrified vehicles or battery technologies) are based on forecasting models with further assumptions.

The DCI combines the static perspective of past emissions monitoring and reporting with the dynamic perspective of OEMs long-term vehicle sales planning. Existing data sources and approaches from the past emissions reporting (to, e.g. the CDP platform) can be used to calculate the DCI resource efficiently. Likewise, the KPI’s modular calculation facilitates scenario-building for future emissions throughout the vehicles’ life cycle on a fleet level while taking reduction measures on product and component level into account. As such, the most efficient strategy to achieving a carbon reduction target can be identified and used as a decision basis by the OEM managers. In this study, even with only the basic DCI calculation and data available, the most impactful reduction measures (a) electrification of the fleet, (b) renewable energy sources for charging and (c) the battery cell manufacturing were identified.

The DCI thus meets the methodological requirements developed for its derivation (see Sect. 2). But does it also meet the expectations of the different stakeholders we brought up in the introduction? The DCI cannot be a “stable” KPI in a way that its calculation principles are set in stone. The MC analysis showed, i.a. that the DCI result is susceptible to change depending on the standardised driving cycle taken as a basis for the calculation. The (automotive) decarbonisation field is developing constantly with new standards and regulations on the way. Examples include a required carbon footprint for EV batteries in the EU starting 2024 calculated with a specific set of rules (European Parliament 2022; European Commission 2023b) and the currently updated guidance for the transport sector of the Science Based Targets initiative (SBTi) (SBTi 2023). The GHG Protocol confirms that the adoption of standards helps to reduce the uncertainty of inventory data. It also stresses the process of improving the inventory data while providing a data basis suitable for decision-makers inside and outside of the company (WRI; WBCSD 2011). Consequently, change is inherent when setting up and working with a decarbonisation KPI. Nonetheless, the decision-making basis for OEM strategists and external stakeholders must remain stable in a way that the same decision would be made disregarding the “calculation premises” that are applied to calculate the DCI. Here, “calculation premises” refer to the list of calculation methods and data used to calculate the DCI for a specific year. Examples are the driving cycles applied in the use phase, clipping rates for certain components or specific data obtained from suppliers versus generic secondary data. Accordingly, the overall DCI result in t CO₂e per vehicle can differ, but the conclusions remain the same.

This dynamic development regarding regulations and standards is mirrored and fuelled by the development process happening inside the company setting up their DCI. At first, a rough estimate of emissions (like in this study) to deduct the most pressing measures is calculated. Subsequently, the OEM’s decarbonisation practitioners will opt for exchanging the generic database with specific data to calculate the reduction potential of further measures. Like this, a higher degree of “transparency” is introduced to the DCI model. But what will the effects be? Will a specific aluminium supplier have a higher or lower carbon footprint than the European aluminium mix data that is currently sourced from the LCA databases? It is hence possible that more data transparency leads to higher DCI values at first. Though OEMs should not be afraid to pursue this higher level of calculation accuracy, only with such a (stepwise) introduction of supplier-specific data can crucial reduction measures in the component production be identified, measured, their effect be reported and a deep decarbonisation of the material supply chains be achieved.

Consequently, more departments outside the decarbonisation and LCA departments will become part of the “DCI team” by providing specific data and using the KPI as a steering instrument themselves. These are, e.g. the fleet planners, the procurement, the logistics and production departments. Vehicle-specific data for the use phase is already standardised based on driving cycles and the established reporting to the authorities (in many markets). Though, especially the material supply chain was shown to be the hotspot phase to realise reduction measures in an electrified fleet (Fig. 7). Gathering specific data from suppliers for every reporting year is a challenge for OEMs and suppliers alike regarding data quality and IT infrastructure. Thus, the need for new cross-industry standards rises. An example for an ongoing supply chain related data initiative is Catena-x (Catena-x 2023).

The described developments inside and outside the company lead to constantly changing sets of “calculation premises” for the DCI. When internal insights in processes and available data evolve and external requirements for reporting change (e.g. the legislative change from NEDC to WLTC in the EU), the DCI calculation premises must be adapted in order to reflect the current methods and data environment. Like this, however, analysing the actual decarbonisation progress happening on the OEM’s product is difficult to impossible because the calculation premises differ for each reporting year. Therefore, re-calculating already published DCI results of past reporting years is necessary to be able to assess the OEM’s decarbonisation effort. This is especially important when a decarbonisation target (e.g. a Science Based Target) is set by the OEM. Here, the base year, the current reporting year and the target year must be calculated with the same calculation premises. In order to keep these adjustments to a minimum, a threshold could be used to define the magnitude of change compared to the last set of calculation premises that make re-calculation necessary. Another way to prevent numerous adjustments, to increase transparency for external stakeholders and to facilitate the external verification of the DCI results on, e.g. reasonable assurance-level for publication within the OEM’s annual or sustainability report are again standardisation or harmonisation initiatives. Ongoing examples for such vehicle LCA related initiatives are TranSensus LCA (Fraunhofer LBF 2023) and UNECE GRPE IWG LCA (UNECE 2022). Though even when faced with these complex methodological developments, OEMs should not hesitate to implement the DCI starting with basic calculation premises and data sets. It will be a continuous improvement process regarding the accuracy of results, but the deduction of most pressing measures is possible right from the start.

The financial aspect of developing and implementing an OEM’s decarbonisation strategy is outside the scope of this study. Still, OEM managers will not base their decisions on fleet compositions and operationalisation of reduction measures only on the carbon reduction leverage indicated by the DCI but also on the costs. Therefore, future research could address how the DCI methodology can be coupled with an internal carbon pricing system. The cooperation with the financial departments is especially important when considering the dependence of the automotive industry’s decarbonisation success on developments in other industrial sectors. How should high and long-term investments be handled that are necessary to, e.g. support the retrofitting of steelworks or the provision of renewable electricity? Harpankar (2019) provides an overview of carbon pricing approaches which could be used as a basis to develop an internal carbon pricing system in accordance with the DCI methodology. Likewise, the DCI could be adapted to heavy-duty vehicles and two-wheelers to cover the whole product range of OEMs.

The here developed KPI “Decarbonisation Index” (DCI) is applicable by any OEM regardless of their company structure, powertrain portfolio or market coverage to monitor past emissions and to model future emissions by using the common denominator of t CO₂e/average vehicle. The DCI combines the product-level view of the vehicle LCA with the fleet-level view necessary to develop a life-cycle-based decarbonisation strategy for meeting a GHG reduction target. The DCI’s modular approach facilitates the use of both generic LCA data and supplier-specific data on component level depending on the OEM’s access to data and overall purpose of using the KPI for internal steering and/or external reporting. The DCI is an “evolving KPI” as its methodological basis is constantly changing due to OEMs’ own data requirements and insights or new external standardisation initiatives. Nonetheless, the DCI is also a stable and reliable KPI as past reported emission values can be adjusted by using the current methodological set. As such, both external stakeholders and OEM managers can analyse the decarbonisation progress and deduct the most impactful reduction measures although the absolute DCI values differ between methodological sets. An OEM’s DCI result also depends on the decarbonisation efforts in coupled sectors such as the energy sector. A close cooperation between OEMs and energy providers is therefore crucial. Lastly, the standardisation and digital exchange of material and component carbon footprints between OEMs and suppliers is necessary to achieve a deep decarbonisation of the supply chains and is likely to shape the DCI results in the future.