Combining Agent-Based Modeling and Life Cycle Assessment for the

Article pubs.acs.org/est

Combining Agent-Based Modeling and Life Cycle Assessment for the Evaluation of Mobility Policies Querini Florent* and Benetto Enrico Public Research Centre Henri Tudor, 6A Avenue des Hauts-Fourneaux, L-4362 Esch sur Alzette, Luxembourg S Supporting Information *

ABSTRACT: This article presents agent-based modeling (ABM) as a novel approach for consequential life cycle assessment (C-LCA) of large scale policies, more specifically mobility-related policies. The approach is validated at the Luxembourgish level (as a first case study). The agent-based model simulates the car market (sales, use, and dismantling) of the population of users in the period 2013−2020, following the implementation of different mobility policies and available electric vehicles. The resulting changes in the car fleet composition as well as the hourly uses of the vehicles are then used to derive consistent LCA results, representing the consequences of the policies. Policies will have significant environmental consequences: when using ReCiPe2008, we observe a decrease of global warming, fossil depletion, acidification, ozone depletion, and photochemical ozone formation and an increase of metal depletion, ionizing radiations, marine eutrophication, and particulate matter formation. The study clearly shows that the extrapolation of LCA results for the circulating fleet at national scale following the introduction of the policies from the LCAs of single vehicles by simple up-scaling (using hypothetical deployment scenarios) would be flawed. The inventory has to be directly conducted at full scale and to this aim, ABM is indeed a promising approach, as it allows identifying and quantifying emerging effects while modeling the Life Cycle Inventory of vehicles at microscale through the concept of agents.

■

INTRODUCTION Rationale. Internal Combustion Engine Vehicles (ICEVs) are major contributors to air pollution,1 anthropogenic global warming,2 and oil depletion, while Electric Vehicles (EVs) are often promoted as a way to reduce these burdens. Yet, they have not managed to represent a significant part of the sales: in the European Union, 13 970 EVs (34 in Luxembourg) were sold in 2012, representing 0.1% of total passenger cars sold this year.3,4 Nonetheless, the Grand Duchy of Luxembourg has set an objective of 40 000 EVs circulating in the country in 2020.5 To achieve this goal, a 5,000€ State incentive (“prime CAR-e”) was initially proposed in 2012 but will end up after December 2014. The progressive deployment of an ambitious charging infrastructure from now to 2020 has also been decided. The assessment of these policy actions, among others aiming at targeting these objectives, has however to account for the whole range of environmental effects, beyond greenhouse gas emissions and from a lifecycle perspective. State of the Art. Life Cycle Assessment (LCA) is a universally recognized methodology aiming at quantifying the environmental impacts of products and technologies throughout their lifecycle. LCA studies of electric mobility are numerous (see chapter S6 in Supporting Information, SI) but their results are often discordant, mainly because they depend on the specific hypotheses and local spatial-temporal contexts adopted (see ref 6 for a thorough review of literature). Also, © 2015 American Chemical Society

LCAs are done at the technology level (e.g., LCA of one specific EV model) and then scaled up to assess mobility scenarios, without properly considering and simulating the actual mechanisms and effects that policies might generate on the mobility system and especially on the users. While providing interesting insights on the potential impacts of EVs (namely on which phase of the lifecycle to focus on), Querini and Benetto7 have hypothesized that the LCA models described in the literature might not be fully appropriate to assess the impacts of mobility policies. A specific LCA approach, called consequential LCA (CLCA) has been proposed to assess the environmental consequences of decisions (changes) affecting systems and markets at the margin.8 The C-LCA rationale has then been generalized and applied to the simulation of the direct and indirect consequences of strategic/policy decisions on large scale systems, not necessarily assuming marginal conditions.9,10 Irrespective of the scale of the consequences (marginal or not), the research question tackled in this paper is about the simulation of the evolution of a complex system (such as mobility) in terms of direct (and eventually indirect) effects triggered by policies. C-LCA is therefore a possible Received: June 30, 2014 Accepted: January 14, 2015 Published: January 14, 2015 1744

DOI: 10.1021/es5060868 Environ. Sci. Technol. 2015, 49, 1744−1751

Article

Environmental Science & Technology

Figure 1. Modeling principles of LCA.

social interactions. Furthermore, being a micro approach like LCA, ABM allows modeling every single vehicle independently. ABM has already been successfully used to assess, from an LCA perspective, the planting of switchgrass by farmers responding to policies.16 Independently from the assessment of environmental concerns, ABM is a proven and widely used approach to model user behavior in mobility systems,17,18 for instance to forecast daily travel behavior, such as commuting activities (an overview of this literature in this field can be found in19). To that extent, ABM is potentially a very interesting approach for deriving consequential LCIs. Aim and Scope−The Four Questions. In a companion paper,19 an ABM was developed in order to assess the deployment of EVs in Luxembourg and its French neighboring region (Lorraine), following a number of specific policy actions. This ABM allowed observing the use of cars by the agents. Starting from these results, the aims of the present paper are (i) to confirm the feasibility of using ABM for C-LCA at two levels: to produce foreground data for the LCI and to generate the data related to the deployment of EVs; (ii) to demonstrate the fallacy of the upscale of results from attributional LCAs of individual vehicles, based on hypothetical scenarios, in order to answer policy-related questions at national scale; (iii) by using the combined ABM-LCA approach to assess the impact of the Luxembourgish policies aiming at deploying EVs and to provide recommendations; and (iv) to investigate how the variety and the amount of results generated by ABM could be

methodological solution for the LCA of (large scale) policies, allowing overcoming the limitations of the conventional approach based on the upscaling of the LCAs of single vehicles. The added value provided by C-LCA has already been demonstrated, e.g., by using economic equilibrium models to derive consequential inventories.11−13 However,11 highlighted the lack of consideration of human behavior for the computation of consequences, which is indeed primordial in the case of mobility systems. The results of C-LCA of mobility policies being the result of the accumulation of a multitude of actors driven by a variety of rationales, Agent-Based Modeling (ABM) appears to be perfectly tailored to derive consistent foreground data for the life cycle inventory.14 ABM is characterized by the decentralized execution of autonomous entities called agents, which, among others, are capable of acting in an environment, communicating with other agents and are motivated by a set of tendencies.15 The intrinsic ability of ABM to model the agent−agent and agent−environment relationships makes it a potentially superior alternative to traditional scenario and economic modeling. In particular, the adoption of alternative vehicles is not only a process involving rational economic decisions but also complex mechanisms, such as personal tastes, resistance to change, learning process from experience, word of mouth effects among people, etc. These elements are difficult to integrate into a conventional macroscopic framework and favor the use of ABM, which is an effective approach to model complex phenomena involving 1745


Article


photochemical oxidant formation, and terrestrial acidification. Other potential impacts linked with land use and toxicity are not presented here, since the related inventories are not reliable enough (21 and22 have warned against the interpretation of these impact categories). ICEVs Production and End-of-Life (EoL). In the ABM, vehicles types are separated between battery-EVs, plug-in hybrid-EVs, gasoline-ICEVs, and diesel-ICEVs. All these vehicles belong to one of the following segments: A (mini), B (small), C (medium), D (large), E (executive), J (sport utility), M (multipurpose) and U (light duty converted to passenger cars). For every segment, the range of possible weights is calculated using the weight range of the three bestselling vehicles in France and Germany as a proxy and then weights are attributed using a probabilistic approach (with a distribution centered on the average weight). Then, vehicles are separated in two parts: powertrain and glider (i.e., the rest of the vehicle without the powertrain: chassis, wheels, body, seats, etc.). The glider production is calculated according to the inventory by22 and using the ecoinvent 2.2 database.23 The inventory is slightly modified, to take into account the progressive inclusion of lightweight materials (aluminum) in the glider (+20−40 kg between 2012 and 2020,24). The average weights of vehicles decrease between 2013 and 2020 by 110 kg for the average European car (C-segment),25 i.e., by 8%. The inventory of powertrain production is based on ref 22. The environmental impacts of the assembly are then linked to the total weight of the car. EoL impacts are calculated following the hypotheses of ref 26: recycled materials are directly introduced into the production phase of the vehicles, while the rest of the materials are either incinerated or landfilled. The impact of dismantling and shredding is proportional to the weight of the vehicle. Details are given in the SI. Battery-EVs and Plug-in Hybrid-EVs Production and EoL. Glider production is proportional to weight and is considered identical to ICEVs. Electric motor and transmission inventories are collected from ref 22 while Li-Ion battery is the average between refs 27 and 28. Then, glider weights and battery and electric powertrain characteristics for the 10 models of vehicles in the simulation have been modeled according to existing vehicles in the European market (cf. SI). For plug-in hybrid-EVs, only a small number of vehicles are commercially available in 2014 in Europe. Therefore, these vehicles have been used as a basis for C- (compact cars) and D-segment cars (family cars) and all other models available after 2014 have been modeled using these vehicles and extrapolations considering conventional ICEVs and battery-EVs. While the electric powertrain is dismantled and mostly recycled (still, some parts such as the neodymium magnet are not recycled, according to ref 29), the EoL of the Li-Ion battery is complex to assess. Here, we considered a recycling rate of 50%, which is the minimum required by EU Directive 2006/ 66/CE.30 Vehicles Use Phase. The potential environmental impacts that are caused by the use phase are associated with fuel and electricity consumption production, tailpipe emissions, noncombustion emissions (tire, brake wire, and windshield wiper fluid) and maintenance. Maintenance frequencies (see SI) are taken from ref 26. For battery-EVs, lifetime of the battery before being recycled and replaced is set to 80 000 km. Lifetime of vehicles is equal to 150 000 km (±50 000 km), which is the usual lifetime considered by carmakers in LCAs.26,31,32 Fuel and

useful for the LCA community. In our approach, the basic LCA models, handling four types of cars (battery-EVs, plug-in hybrid- EVs gasoline-ICEVs and diesel-ICEVs) are fed by detailed data provided by the ABM (e.g., characteristics and number of travels, charging patterns, auxiliary use, etc.), the latter running from 2013 to the dismantling of the last vehicle studied and including all vehicles (except commercial fleets) sold between 2013 and 2020 (included). Gasoline-ICEVs and diesel-ICEVs correspond to all vehicles respectively using gasoline and diesel as their only fuel and therefore cover both conventional vehicles and hybrids that cannot be charged on the grid and therefore act as conventional cars (with a lower fuel consumption). Battery-EVs are EVs driven only using the electricity from the grid (with a large battery to store the electricity) while plug-in hybrid EVs correspond to vehicles that use both electricity from the grid and gasoline.

■

MATERIALS AND METHODS Modeling Principles. In the agent based model,19 agents represent individuals using their cars for various purposes, selling or disposing of the vehicles to buy new and used cars. As shown in Figure 1, this model provides the necessary inventory data to derive the consequential LCIs of the policies. Each time an agent buys a new car, all detailed characteristics of the latter are known: type (gasoline-ICEV, diesel-ICEV, battery-EV or plug-in hybrid-EV), segment, consumption and weight. Then, the details regarding all the travels done during the use phase are calculated on an hourly basis: distance, type of driving (urban, road or highway), and external temperature. For battery-EVs and plug-in hybrid-EVs, charging of the battery is also known on an hourly basis. During the use phase, the cumulative mileage of the car is progressively updated until it reaches the maximum mileage set. Then the agent considers the possibility to sell the vehicle to another agent during the simulation. Finally, the car is dismantled. The impact of the policy considered is therefore calculated as the sum of all individual car LCAs carried out using the foreground data provided by the model, avoiding biases from an arbitrary upscaling, based on hypothetical replacements and travels. The overall LCA results are then compared to the LCA results of the changes that would have occurred without any policy implementation, i.e., in the case of a business as usual simulation of the evolution of the mobility system. Our approach is therefore similar to the one of16 for assessing the penetration of vehicles in the fleets. However, we went one step further as the ABM developed is not only used to derive the penetration of EVs in the fleet but also to assess precisely the use phase of the car (avoiding the use of averages). In other words, detailed consumptions and emissions for every travel of the agents are calculated, enabling taking into account the large variability of car models and uses at national scale. Sections 2.2 to 2.4 describe how the different vehicles’ life cycle phases were modeled, while section 2.5 presents the policies and scenarios retained. Although the case study retained is focused on a specific country scale (Luxembourg and Lorraine), the combined ABM/C-LCA approach as well as most of the data used could be replicated and applied to any other region and scenario, at least at European scale. Environmental impacts are calculated using ReCiPe 2008 methodology,20 considering the following midpoint impacts: Global Warming (GWP), fossil energy depletion, freshwater eutrophication, ionizing radiations, marine eutrophication, metal depletion, ozone depletion, particulate matter formation, 1746


Article


Figure 2. Individual vehicle details. Top left: sales repartition, per type (BEV: battery-EV, CIICEV, PHEV: plug-in hybrid-EV and SIICEV) and segment (A, B, C, D, E, J, M, U). Top right: sales repartition, per type and car status in the household (main or second). Bottom left: sales repartition, depending on first own country. Bottom right: number of owners per car, depending on first owner country.

battery-EV and 4,000€ for a plug-in hybrid-EV) are maintained but, as introduced in section 1.1, the CAR-e inventive is canceled in Luxembourg in 2015. Finally, scenario “S4. Additional policies” is the same as S3, except that the CAR-e incentive is maintained during all the simulation. All other scenario details are the same as in the reference scenario of the companion paper, regarding commuting distances, ability to charge vehicles at home (60% of people), etc.19

electricity production impacts are calculated using the ecoinvent database.23 The COPERT IV software33 is used to derive hot emissions factors for urban, road and highway Artemis driving cycles34 as well as cold emissions depending on external temperature. These emission factors are then used in conjunction with travel characteristics in order to assess in real time during the execution of the model the tailpipe emissions corresponding to the use of the car by the agent. Car consumptions per Artemis cycle are calculated using Fastsim software.35 For ICEVs, 128 different vehicles have been modeled in order to represent the available vehicles in Luxembourg and Lorraine. Consumptions are calculated in the same manner for battery-EV (using Fastsim and the 10 vehicles retained), except that the consumption is expressed in kWh of electricity. Plug-in hybrid-EVs use electricity until their battery is discharged and then use fuel and have thus three different consumption formulas (one for electricity consumption and two for fuel consumption, respectively, for full and empty batteries). When the vehicle is charging, electricity consumption on the grid is calculated using state of charge of the battery and charging time. All data are provided in the SI. Scenarios and Life Cycle Impact Assessment (LCIA). The calculations are conducted for the 8-year period between 2013 and 2020. Four scenarios have been retained. Scenario “S1. noEV” corresponds to the situation where no EV is available to the individuals and the mobility system evolves under business-as-usual conditions. Scenario “S2. BAU” (Business As Usual) assumes that financial incentives are canceled in 2014 and limited deployment of the charging infrastructure (only 10% of the users are able to charge the vehicles at the work place). Scenario “S3. Current policies” is the scenario with the default hypotheses: 50% and 10% of workers (respectively in Luxembourg and Lorraine) can charge their vehicles at work. Incentives in Lorraine (6,300€ for a

■

RESULTS Individual Data Regarding EV Adoption. Figure 2 shows detailed results of EV deployment in the fleet (for the scenario “S3. Current policies”). The top graph shows the sales repartition, depending on type and segment. It illustrates the fact that battery-EVs + plug-in hybrid-EVs represent about 10% of the sales in Luxembourg and Lorraine. However, it also shows that, even with all the segments covered after the year 2015, battery-EVs that are sold are still mainly small and mini cars and, to a lesser extent, medium and large cars. On the contrary, most plug-in hybrid-EVs sold are large cars, followed by medium, small and executive cars. Similarly, the top right graph of Figure 2 shows that battery-EVs are mainly used as secondary cars, i.e., for commuting to work and short distances. As a consequence, lifetime of battery-EVs is equal to 15 years on average, compared with 10 years for the other types of cars (and therefore invalidating the commonly accepted 10-year lifetime used in attributional LCAs of EV). However, plug-in hybrid-EVs are massively bought as main cars and only a few as secondary cars. Bottom left figure shows that 40% of batteryEVs and 50% of plug-in hybrid-EVs are sold in Luxembourg, even though there no incentive after 2014. Obviously, EVs are mostly bought by working people, inactive individuals representing a negligible fraction of the sales. Battery-EVs also represent a higher share of sales for cross-border 1747


Article


Figure 3. Individual vehicle impacts. Top left: average GWP of vehicles depending on type and segment. Top right: average ionizing radiation of vehicles depending on type and first owner country. Bottom left: average acidification depending on type and segment. Bottom right: average acidification, depending on vehicle type and purchase year.

acidification, depending on vehicle type and purchase year. While in 2013, battery-EVs appears to provide significant benefits compared to diesel-ICEVs, this gain is highly reduced when comparing vehicles sold and 2020 (>Euro 6 aftertreatment). Scenario Results. Figure 4 presents the environmental impacts of scenarios relatively to Scenario “S1. noEV”. Standard

commuters from Lorraine, who also mainly use diesel-ICEVs. Finally, the bottom right figure shows the number of owners during the lifetime of vehicles. The number of owner for battery-EVs is close to 1and, if the first owner is living in Luxembourg, the total number of owners for gasoline-ICEVs and diesel-ICEVs is around 3, while it is only 2 if the vehicle is sold in Lorraine. Therefore, the LCA of a car only used in Luxembourg would not be representative of the fleet. Individual Data Regarding EV Impacts. The Top left graph of Figure 3 presents the average GWP of vehicles, depending on vehicle type and segment. For ICEVs, GWP is strongly correlated with segment, since the bigger the car, the higher the GWP during the use phase. The top right graph presents ionizing radiations results, which are higher for battery-EVs than for other vehicles. The difference between vehicles charged in Luxembourg and vehicles charged in Lorraine is very important, though vehicles can have different owners and therefore can be charged, for instance, in Luxembourg during the first part of their life and then in Lorraine during the second half. The bottom left figure shows that acidification is very dependent on vehicle segment and type. Indeed, for gasoline-ICEVs, acidification can vary between 80 and 140 kg SO2‑eq , because of the difference of fuel consumption during the use phase. Battery production plays a strong role for battery-EVs and plug-in hybrid-EVs. The Esegment battery-EV (executive car with a large 90 kWh battery) show an acidification potential superior to its diesel-ICEV equivalent, while for the A-segment car (mini car), battery-EV appears less impacting than diesel-ICEV. Although most of the results show the same trend as the literature, acidification is in contradiction with refs 26 and 36. The reason for this discrepancy is that emissions considered in our study were calculated with COPERT and not based on regulations or measurement according to the official New European Driving Cycle (NEDC). Finally, bottom right graph shows average

Figure 4. LCA results of the 3 EV deployment scenarios, relative to the scenario without EV.

deviations are calculated by running the same scenarios 8 times (as described in ref 19) and take into account the variability of vehicles characteristics and uses. Depending on the scenario, battery-EV sales represent between 1.7% and 2.0% of total vehicles sold between 2013 and 2020 and plug-in hybrid-EVs between 6.4% and 7.8%. Figure 4 shows that, with the S2 scenario (business as usual), it is likely that the EVs deployed will generate higher ionizing radiations, metal depletion and particular matter formation. However, their potential benefits 1748


Article


Environmental Impacts, Vehicle Level. As shown in section 3.2, battery-EVs and plug-in hybrid-EVs are not replacing the same vehicles. While battery-EVs are more likely to be small cars replacing gasoline-ICEVs mainly used for commuting, plug-in hybrid-EVs are larger cars replacing dieselICEVs. Battery-EVs have the lowest GWP but this greatly varies, depending on the size of the battery and the consumption of the ICEVs. Conclusions are similar for fossil resource depletion. Higher ionizing radiations are a side effect of using nuclear electricity. Freshwater eutrophication and metal depletion are higher for battery-EVs because of the battery production. If the battery could last for the entire lifetime of the car, then this conclusion would still hold. Acidification is a good example of the bias introduced when scaling up attributional LCAs of specific vehicles to assess national policies. Indeed, when comparing A-segment cars, acidification is lower for battery-EVs, while this conclusion is not true for E-segment cars (model based on the Tesla model) because of the use of very large battery. This confirms that when assessing national policies affecting all car segments at full scale based on LCAs carried out on a reduced number of vehicle models and using simple functional units, significant bias is inevitably introduced. The ABM-LCA approach allows avoiding this shortcoming by gathering consistent simulations on a large number of vehicles types and related usages at microscale that can be further used to derive macro-scale impacts. Finally, considering that after 2020 unknown after-treatment regulations will be enforced, it is possible that Battery-EVs will only decrease fossil depletion and GWP, while increasing all other impacts because of their batteries. ABM/C-LCA Model Limitations. For the time being, the model only includes cars that are used for private purpose (but can be owned by individuals or leased by companies). The noninclusion of company fleets represents a shortcoming of the model, since they represent a good opportunity for EV deployment. However, their inclusion is not likely to drastically change the results at large scale, as these vehicles represent 14% and 2% of the circulating fleet, respectively in Luxembourg and Lorraine. Limitations are more linked with the ABM part than to the LCA model (see ref 19 for details). However, the LCA model is affected by high uncertainties regarding the battery technologies and electricity consumption of EVs in 2020. Moreover, environmental impacts of ICEVs are also complex to assess as the next aftertreatment standards are unknown. All these elements could drastically change the LCA results. Despite these limitations, the ABM shows that EV deployment will start in the last years of the simulation. This means that it is possible to assess the impacts of EV for the next 8 years, but not afterward when they will reach mass scale deployment and thus have significant impacts at national scale. This limitation is not inherent to our model but to all LCAs of future EVs: therefore, further researches should focus on assessing the uncertainties linked with long-term mobility LCAs. Moreover, the sensitivity analysis shows that the results would change significantly if the study was carried out in another country. Indeed, not only would the electricity mix change, but also the deployment of EVs in a fleet with different uses and characteristics. Therefore, to apply the method to another country, it is mandatory to run the model with suited hypotheses and data. Indeed, the coupling between ABM and LCA allows fitting very well an LCI to a given country, which therefore leads to results that can even be more difficult to

are highly uncertain. The uncertainties of EVs deployment are indeed very high for the base scenario where no external stimulus aims this deployment. For scenario S3, while the above-mentioned drawbacks are still likely to happen, it is also likely that EVs will help decreasing photochemical ozone, fossil fuel depletion and marine eutrophication. Scenario S4 shows similar results, but also with probable lower acidification and GWP. Sensitivity Analysis. The use phase of EVs is very sensitive to the source of electricity and therefore, we studied the possibility of generalizing the results to other countries that rely on coal electricity by selecting Germany as our test case. Results show an increase of all environmental impacts, even though the EV deployment is lower than in Luxembourg. For instance, results show an increase of 7% of acidification (−2.5% for Lorraine and Luxembourg) or +5% of GWP (−4% in Lorraine). Ionizing radiations are the only impact whose increase is lower for Germany than for Lorraine, thanks to the lower share of nuclear power. We also assessed the effect of batteries that would last the entire lifetime of EVs but this would not change the global trends that are observed (the detailed results are presented in the SI), since this mainly has an effect on metal depletion and freshwater eutrophication, which are twice as low but are still higher in scenario S3 than in S1.

■

DISCUSSION Scenario Results and Policy Implications. EVs allow significant decrease in GWP, fossil resource depletion, photochemical ozone formation, marine eutrophication, ozone depletion, and acidification while increasing other impacts, such as metal depletion, freshwater eutrophication, particulate matter formation, and ionizing radiations (the two latter for Lorraine especially). Although when looking at the vehicle level, plug-in hybrid-EVs appear to be have higher impacts than battery-EVs for GWP, fossil depletion, photochemical ozone formation, ozone, metal depletion, ozone depletion, and acidification, they are responsible for the larger part of the impact reduction observed at Luxembourg and Lorraine scale. This is due to the fact that their sales are three times higher than the sales of battery-EVs, which means that they replace three times more conventional ICEVs. Moreover, they replace larger ICEVs, which, as discussed in the next section, have higher potential impacts. Finally, although the impacts of the policies seem low (a few percents), they have large implications. For instance, EVs could allow decreasing GWP by 52 000 t CO2‑eq (S3) or increase ionizing radiations by 85 000 t U 235 −eq . As policy recommendation for Luxembourg, we would first advise to have larger infrastructure deployment and provide all necessary measures to ensure EVs deployment and use. Indeed, the very uncertain nature of EV deployment leads to high uncertainties on the environmental consequences, although it seems certain that batteries have negative effects, e.g., increased metal depletion. Therefore, a second recommendation would be to extend the lifetime of batteries, by for instance promoting their reuse in other applications before dismantling and recycling. Finally, considering the results obtained for the German mix, we would recommend to Luxembourg’s stakeholders to keep the renewable electricity policy, while France should decrease the share of nuclear power to limit the potential ionizing radiations impact. 1749


Article


impact considered, but we recommend pursuing them, while encouraging all solutions increasing battery lifetime and promoting renewable electricity. (iv) We think that the large amounts of data generated by the coupling of ABM and LCA could clearly be beneficial for the LCA community.

generalize to other situations. Finally, one way to improve the results would be to use an hourly electricity mix for the calculating the impacts of the use phase of EVs. However, this was not done here since Luxembourg is importing most of its electricity and the renewable electricity used by EVs is actually fossil energy compensated through purchases of green certificates. When applying to the model to other country, hourly mix should be explored as the ABM developed allows knowing the charging time of EVs on an hourly basis. ABM As a Novel Approach for C-LCA of Mobility Systems. On the basis of the results obtained for the case study of Luxembourg and Lorraine, ABM seems to be a very appropriate modeling approach for consequential LCAs of large scale policies. First, it allows assessing the environmental impacts of large complex system, by summing all the impacts of all individual vehicles consistently simulated a microscale. Therefore, it provides a practical way to combine conventional LCAs of given vehicle models (technologies), with large scale consequential inventories at the scale of a country. Then, ABM is particularly interesting for studying systems that are not purely economic driven. Indeed, car adoption and use by individuals rely on a number of complex effects and interactions and are often not the result of only an economic equation. Therefore, mobility systems can often present emerging effects, such as mass adoption of EVs because of a fad for a given brand or model, and for which ABM appears powerful. In our system, the penetration rate of EVs is not arbitrarily defined but as the results of the decisions taken by the agents considering their environment and other agents’ decisions. Finally, we think that, provided that sufficient data are available, ABM enables taking into account many specifics aspects of different countries (mobility needs, purchase power, personal attitudes, etc.) and further researches will aim at exploring the feasibility of using the model for larger case studies (for instance larger European States). However, ABM results are strongly depended on the hypotheses and data that are used to build and initialize the model. In the companion paper19 we showed that the number of EVs are very sensitive to many parameters and in this paper, 3 scenarios were derived to explore the ABM possible results. Yet, despite these uncertainties, the trends associated with the policies are similar and therefore, ABM appears to be efficient in qualifying the impacts of policies (increase or decrease, low or high). However, the large uncertainties from the ABM as well as those from the LCA model (quantified in this paper) prevent, for the moment, one from using ABM as a quantitative predictive tool. This conclusion rejoins the opinion developed by Miller et al.16 Finally, ABM and LCA are approaches that both use and generate large amount of data. The combination of the two leads to even more data that simultaneously have a high granularity and a large scale, which could represent a goldmine for C-LCA practitioners (see SI). Quick Answers to the Four Aims and Questions As Take-Home Message. This section sums up the conclusions to the four questions that were raised in the goal and scope (1.3) section: (i) We showed that it is feasible to couple LCA and ABM, even for complex large scale systems such as mobility. Here, the ABM is not only used for scenario elaboration but also to obtain a life cycle inventory that encompasses all the variability of the car characteristics and uses. (ii) Attributional LCAs of single vehicles cannot be scaled up to assess large scale policies as the conclusions on a single model can be misleading. (iii) The impact of policies aiming at deploying EVs have mixed consequences, depending on the

■

ASSOCIATED CONTENT

S Supporting Information *

Input LCA data and additional results. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Phone: +352 42 59 91-3379; e-mail: fl[email protected]. Author Contributions

Florent Querini designed the models, ran the simulations, interpreted the results, and wrote most of the manuscript. Enrico Benetto assured supervision of the scientific work, revised the draft versions of the manuscript, and contributed to the “Introduction” part. All authors have given approval to the final version of the manuscript. Funding

The research associated with the article is entirely funded by the FNR (Luxembourg National Research Fund) under the HELCAR grant (4886210). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work has been conducted in the framework of HELCAR (hybrid and electric cars life cycle assessment), a postdoctoral project entirely funded by the FNR (National Research Fund Luxembourg) which is gratefully acknowledged.

■

ABBREVIATIONS ABM agent-based model Battery-EV battery-powered electric vehicle C-LCA consequential LCA CIICEV compression-ignition (diesel) ICEV EoL end of life EV electric vehicle GWP global warming potential HVAC heating, ventilation and air conditioning ICEV internal combustion engine vehicle LCA life cycle assessment LCI life cycle inventory LCIA life cycle impact assessment NEDC new European driving cycle plug-in hybrid-EV plug-in hybrid electric vehicle SIICEV spark-ignition (gasoline) ICEV

■

REFERENCES

(1) European Environmental Agency. Transport emissions of air pollutants (TERM 003) − Assessment; EEA; 2013; http://www.eea. europa.eu/data-and-maps/indicators/transport-emissions-of-airpollutants-8/ds_resolveuid/V0JG744JKO. (2) European Environmental Agency. Greenhouse gas emission trends (CSI 010) − Assessment; EEA, 2013; http://www.eea.europa.eu/dataand-maps/indicators/greenhouse-gas-emission-trends/ds_resolveuid/ YYWDOO7KYN. (3) European Environmental Agency. Monitoring CO2 emissions from new passenger cars in the EU: summary of data for 2012; EEA, 2013; 1750


Article

Environmental Science & Technology http://www.eea.europa.eu/publications/monitoring-co2-emissionsfrom-new-cars/at_download/file. (4) STATEC Institut national de la statistique et des études économiques du Grand-Duché du Luxembourg Website; http://www.statistiques. public.lu/fr/acteurs/statec/index.html (in French). (5) Gouvernement du Grand-Duché du Luxembourg. Vers une mobilité plus durable − nouvelles initiatives dans les domaines de la mobilité électrique et de la mobilité basée sur le gaz naturel ; Présentation du Ministère de l’économie et du Commerce extérieur et du Ministère du Développement durable et des infrastructures, 2012 (in French); http://www.eco.public.lu/salle_de_presse/conferences_presse/2012/ 03/01_electromobilite/Conf__rence_de_presse_1_3_2012_-_Vers_ une_mobilit___plus_durable.pdf. (6) Hawkins, T. R.; Gausen, O. M.; Strømman, A. H. Environmental impacts of hybrid and electric vehiclesa review. Int. J. Life Cycle Assess. 2012, 17 (8), 997−1014 DOI: 10.1007/s11367-012-0440-9. (7) Querini, F.; Benetto, E. Definition and influence of functional unit in LCA of electric and plug-in hybrid vehicles applied to implantation scenarios. Life Cycle Management Conference (LCM). 2013, Gothenburg: Sweden. (8) Ekvall, T.; Weidema, B. P. System boundaries and input data in consequential life cycle inventory analysis. Int. J. Life Cycle Assess. 2004, 9 (3), 161−171 DOI: 10.1007/BF02994190. (9) Vázquez-Rowe, I.; Marvuglia, A.; Rege, S.; Benetto, E. Applying consequential LCA to support energy policy: land use change effects of bioenergy production. Sci. Total Environ. 2014, 472, 78−89 DOI: 10.1016/j.scitotenv.2013.10.097. (10) Halog, A.; Manik, Y. Advancing integrated systems modelling framework for life cycle sustainability assessment. Sustainability 2011, 3 (2), 469−499 DOI: 10.3390/su3020469. (11) Marvuglia, A.; Benetto, E.; Rege, S.; Jury, C. Modelling approaches for consequential life-cycle assessment (C-LCA) of bioenergy: critical review and proposed framework for biogas production. Renew. Sust. Energy Rev. 2013, 25, 768−781 DOI: 10.1016/j.rser.2013.04.031. (12) Dandres, T.; Gaudreault, C.; Tirado-Seco, P.; Samson, R. Macroanalysis of the economic and environmental impacts of a 2005− 2025 European Union bioenergy policy using the GTAP model and life cycle assessment. Renew. Sust. Energy Rev. 2012, 16 (2), 1180− 1992 DOI: 10.1016/j.rser.2011.11.003.. (13) Rege, S.; Arenz, M.; Marvuglia, A.; Vás quez-Rowe, I.Quantification of agricultural land use changes in consequential life cycle assessment using a partial equilibrium model: opportunities and threats.J. Environ. Inform., 2014. (14) Davis, C.; Nikolic, I.; Gerard, P. J. Integration of Life Cycle Assessment into Agent-Based Modelling. J. Ind. Ecol. 2009, 13 (2), 306−325 DOI: 10.1111/j.1530-9290.2009.00122.x.. (15) Ferber, J. Multi-Agent Systems. An Introduction to Distributed Artificial Intelligence; Addison Wesley: London, 1999. (16) Miller, S. A.; Moysey, S.; Sharp, B.; Alfaro, J. A stochastic approach to model dynamic systems in life cycle assessment. J. Ind. Ecol, 2013, 17 (3), 352−362 DOI: 10.1111/j.1530-9290.2012.00531.x. (17) Arentze, T. A.; Hofman, F.; van Mourik, H.; Timmermans, H. J. ALBATROSS: multiagent, rule-based model of activity pattern decisions. Transport. Res. Rec 2000, 1706 (1), 136−144 DOI: 10.3141/1706-16. (18) Arentze, T. A.; Timmermans, H. J. A learning-based transportation oriented simulation system. Transport. Res. B-Meth. 2004, 38 (7), 613−633 DOI: 10.1016/j.trb.2002.10.001. (19) Querini, F.; Benetto, E. Agent-based modelling for assessing hybrid and electric cars deployment. The case of Luxembourg and Lorraine. Transport. Res. A-Polym. 2014, 70, 149−161 DOI: 10.1016/ j.tra.2014.10.017. (20) Goedkoop, M.; Heijungs, R.; Huijbregts, M.; De Schryver, A. ; Struijs, J. ; Van Zelm, R. ReCiPe 2008. A life cycle impact assessment method which comprises harmonised category indicators at the midpoint and the endpoint level; Ruimte en Milieu, 2009; http://www.presustainability.com/download/misc/ReCiPe_main_report_final_2702-2009_web.pdf.

(21) Querini, F.; Morel, S.; Boch, V.; Rousseaux, P. USEtox relevance as an impact indicator for automotive fuels. Application on diesel fuel, gasoline and hard coal electricity. Int. J. Life Cycle Assess. 2011, 16 (8), 829−840 DOI: 10.1007/s11367-011-0319-1. (22) Hawkins, T. R.; Singh, B.; Majeau-Bettez, G.; Strømman, A. H. Comparative environmental life cycle assessment of conventional and electric vehicles. J. Ind. Ecol. 2013, 17 (1), 53−64 DOI: 10.1111/ j.1530-9290.2012.00532.x. (23) Frischknecht, R.; Jungbluth, N.; Althaus, H. J.; Doka, G.; Dones, R.; Heck, T.; Hellweg, S.; Hischier, R.; Nemecek, T.; Rebitzer, G.; et al. The ecoinvent database: Overview and methodological framework (7pp). Int. Life Cycle Assess. 2005, 10 (1), 3−9 DOI: 10.1065/lca2004.10.181.1. (24) European Aluminium Association. Aluminium in cars. Unlocking the light-weight potential; EAA, 2013; http://www.alueurope.eu/wpcontent/uploads/2013/10/EAA-Aluminium-in-Cars-Unlocking-thelight-weighting-potential_September2013_03.pdf. (25) Huss, A.; Maas, H.; Hass, H. Well-to-wheels analysis of f uture automotive f uels and powertrains in the European context. Tank-to-wheels report; European Commission, Joint Research Centre, Institute for Energy and Transport, 2013; http://iet.jrc.ec.europa.eu/about-jec/ sites/iet.jrc.ec.europa.eu.about-jec/files/documents/report_2013/ttw_ report_v4_july_2013_final.pdf. (26) Barat, A.; Dang, V.; Querini, F.; Morel, S. Fluence and Fluence ZE Life Cycle Assessment October 2011; Renault, 2011; http://www.renault. com/fr/lists/archivesdocuments/fluence-acv-2011.pdf. (27) Notter, D. A.; Gauch, M.; Widmer, R.; Wager, P.; Stamp, A.; Zah, R.; Althaus, H. J. Contribution of Li-ion batteries to the environmental impact of electric vehicles. Environ. Sci. Technol. 2010, 44 (17), 6550−6556 DOI: 10.1021/es903729a. (28) Majeau-Bettez, G.; Hawkins, T. R.; Strømman, A. H. Life cycle environmental assessment of lithium-ion and nickel metal hydride batteries for plug-in hybrid and battery electric vehicles. Environ. Sci. Technol. 2011, 45 (10), 4548−4554 DOI: 10.1021/es103607c. (29) Binnemans, K.; Jones, P. T.; Blanpain, B.; Van Gerven, T.; Yang, Y.; Walton, A.; Buchert, M. Recycling of rare earths: a critical review. J. Cleaner Prod. 2013, 51, 1−22 DOI: 10.1016/j.jclepro.2012.12.037. (30) European Comission. Directive 2006/66/EC of the European Parliament and of the Council of 6 September 2006 on batteries and accumulators and waste batteries and accumulators and repealing Directive 91/157/EEC (Text with EEA relevance). http://eur-lex.europa.eu/ legal-content/EN/TXT/?uri=CELEX:32006L0066. (31) Mercedes-Benz. Environmental certif icate A-Class; Mercedes, 2008; www3.mercedes-benz.com/fleet-sales/en/assets/documents/ Environmental_Certificate_A_Class.pdf. (32) Volkswagen.The Golf Environmental commendation − background report; Volkswagen, 2008; www.volkswagenag.com/vwag/vwcorp/ info_center/en/publications/2008/04/environmental_ commendation0.-bin.acq/qual-BinaryStorageItem.Single.File/Golf_ UP_Hintergrundbericht_GB.pdf. (33) Gkatzoflias, D.; Kouridis, C.; Ntziachristos, L.; Samaras, Z. COPERT 4: Computer programme to calculate emissions f rom road transport; EEA, 2007; http://lat.eng.auth.gr/copert/files/tech01.pdf. (34) André, M. The ARTEMIS European driving cycles for measuring car pollutant emissions. Sci. Total Environ. 2004, 334, 73−84 DOI: 10.1016/j.scitotenv.2004.04.070. (35) National Renewable Energy Laboratory. Fastsim 2012 Beta 2; http://www.nrel.gov/vehiclesandfuels/vsa/fastsim.html. (36) Warburg, N.; Forell, A.; Guillon, L.; Teulon, H. ; Canaguier, B. Elaboration selon les principes des ACV des bilans énergétiques, des émissions de gaz à effet de serre et des autres impacts environmentaux induits par l’ensemble des f ilières de véhicules électriques et thermiques, VP de segment B (citadine polyvalente) et VUL à l’horizon 2012 et 2020. Report for the French ADEME, 2013; http://www.ademe.fr/sites/ default/files/assets/documents/90511_acv-comparative-ve-vt-rapport. pdf (in French).

1751


Combining Agent-Based Modeling and Life Cycle Assessment for the

Recommend Documents