Environ. Sci. Technol. 2003, 37, 5407-5413
Life Cycle Optimization of Automobile Replacement: Model and Application HYUNG CHUL KIM,† G R E G O R Y A . K E O L E I A N , * ,† DARBY E. GRANDE,‡ AND JAMES C. BEAN‡ Center for Sustainable Systems, School of Natural Resources and Environment, University of Michigan, 430 East University, Dana Building, Ann Arbor, Michigan 48109-1115, and Industrial & Operations Engineering Department, University of Michigan, 1205 Beal Avenue, Ann Arbor, Michigan 48109
Although recent progress in automotive technology has reduced exhaust emissions per mile for new cars, the continuing use of inefficient, higher-polluting old cars as well as increasing vehicle miles driven are undermining the benefits of this progress. As a way to address the “inefficient old vehicle” contribution to this problem, a novel life cycle optimization (LCO) model is introduced and applied to the automobile replacement policy question. The LCO model determines optimal vehicle lifetimes, accounting for technology improvements of new models while considering deteriorating efficiencies of existing models. Life cycle inventories for different vehicle models that represent materials production, manufacturing, use, maintenance, and end-of-life environmental burdens are required as inputs to the LCO model. As a demonstration, the LCO model was applied to mid-sized passenger car models between 1985 and 2020. An optimization was conducted to minimize cumulative carbon monoxide (CO), nonmethane hydrocarbon (NMHC), oxides of nitrogen (NOx), carbon dioxide (CO2), and energy use over the time horizon (1985-2020). For CO, NMHC, and NOx pollutants with 12 000 mi of annual mileage, automobile lifetimes ranging from 3 to 6 yr are optimal for the 1980s and early 1990s model years while the optimal lifetimes are expected to be 7-14 yr for model year 2000s and beyond. On the other hand, a lifetime of 18 yr minimizes cumulative energy and CO2 based on driving 12 000 miles annually. Optimal lifetimes are inversely correlated to annual vehicle mileage, especially for CO, NMHC, and NOx emissions. On the basis of the optimization results, policies improving durability of emission controls, retiring high-emitting vehicles, and improving fuel economies are discussed.
Introduction For the last few decades, automotive fuel consumption and exhaust emissions per mile in the United States have decreased significantly due to improvements in engine technology, emission controls, and efficient vehicle design. This progress has been primarily driven by regulations, * Corresponding author telephone: (734)764-3194; fax: (734)6475841; e-mail:
[email protected]. † Center for Sustainable Systems. ‡ Industrial & Operations Engineering Department. 10.1021/es0345221 CCC: $25.00 Published on Web 11/01/2003
2003 American Chemical Society
including Federal tailpipe emission standards, and Corporate Average Fuel Economy (CAFE) standards. Despite this recent progress, vehicle emissions of carbon monoxide (CO), oxides of nitrogen (NOx), hydrocarbons (HC), and particulate matter (PM) are still significant in most urban areas. In addition to increasing vehicle miles traveled (VMT), this discrepancy can be explained by continuing use of old vehicles and the increasing age of vehicles in use. The expected median lifetime of an average car has increased from 12.5 yr for model year 1980 to 16.9 yr for model year 1990 (1). Evaluating the optimal life of an automobile poses a challenging resource and environmental management problem. Extending the service life of an existing automobile avoids the additional resource investments and environmental impacts associated with the production of a new vehicle. For example, the production of a mid-sized 1995 automobile was found to consume 125 000 MJ of primary energy (equivalent to 20 barrels of oil), 1500 kg of iron ore, and 230 kg of bauxite and to generate 3000 kg of solid waste and 210 kg of air pollutants (2, 3). Product life extension is an important, well-recognized green engineering strategy, which can be achieved by enhancing vehicle durability and serviceability (4, 5). On the other hand, replacement of older, inefficient product with newer, more efficient product is another important mechanism for reducing environmental impacts. While product life extension limits manufacturingrelated impacts, adopting newer technology can enhance environmental performance during the product-use stage. The use phase of the 1995 mid-sized vehicle was responsible for 86% of the life cycle energy, 87% of the CO2, and between 90% and 95% of the regulated auto emissions (CO, NMHC, and NOx) (2, 3). Vehicle scrappage programs are designed to accelerate retirement of older vehicles. To encourage the adoption of newer, cleaner technology, these programs recruit and scrap old, high-polluting vehicles with some compensation to the vehicle owners. Scrappage policies also aim at encouraging new technologies and supporting the car manufacturing industry (6, 7). European countries, such as Greece and France, as well as some local governments in the United States and Canada have used incentives to implement such policies during the 1990s. In the United States, the Union Oil Company of California (UNOCAL), in cooperation with the California Air Resources Board (CARB) and the California Department of Motor Vehicles, first introduced a scrappage program in June 1990 in the Los Angeles Basin area, offering $700 for retiring eligible vehicles. As a result, over 8000 cars, accounting for more than 2% of pre-1970s model year vehicles in this area, were retired (6, 8). From a holistic environmental perspective, however, most scrappage programs fail to consider the complete life cycles of vehicles. Scrappage programs focus on the emissions of vehicle operations but fail to consider other sources of emissions: dismantling old vehicles and producing new vehicles. A recent study by the European Conference of Ministers of Transport (ECMT) concluded that the negative impacts of scrappage programs arise from the increased emissions and resource consumption from the additional vehicle scrapping and production processes (6). Wee et al. also argue that scrapping old cars would result in a net increase of life cycle energy use and CO2 emissions unless fuel economies improve considerably for the cars manufactured in the future (9). Thus, the benefits or emission credits created through scrappage programs might be exaggerated since the negative impacts of vehicle production and disposal were ignored. VOL. 37, NO. 23, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5407
Vehicle retirement is primarily decided by economic considerations, including repair cost, market price, and scrap price of a used vehicle. While the average service life of vehicles is increasing, the environmental consequences of such trends are not yet well understood. For example, is a 17-yr life for a 1990 model car optimal from an environmental perspective? The present study determines optimal product lifetimes using life cycle assessment (LCA), a comprehensive environmental measurement tool, and dynamic programming, an engineering optimization tool. To determine the optimum life of a product from an LCA perspective, a novel life cycle optimization (LCO) model is developed. The inputs to this mathematical model consist of a collection of life cycle inventories (LCIs) describing materials production, manufacturing, use, maintenance, and end-of-life environmental burdens as functions of product model years and ages. For demonstration of the LCO model, optimizations of the vehicle replacement decision for mid-sized passenger car models between 1985 and 2020 were conducted. Among the environmental categories, carbon monoxide (CO), non-methane hydrocarbon (NMHC), oxides of nitrogen (NOx), carbon dioxide (CO2), and energy consumption were selected as objectives for these optimizations. CO, NMHC, and NOx are regulated auto emissions that can have a significant impact on local and regional air quality. On the other hand, CO2 and energy consumption are global issues influenced by CAFE standards in the United States. On the basis of the model runs, correlations between input parameters and optimal vehicle lifetimes are discussed. In addition, policy implications of the optimization results for scrappage programs, vehicle designs, auto emission, and fuel economy regulations are discussed. The Supporting Information of this paper provides detailed analysis on the LCI modeling used for the optimizations.
Life Cycle Optimization (LCO) Model The LCO model is based on a dynamic programming method, and the input data to this model are determined by life cycle assessment (LCA). Life Cycle Assessment (LCA). LCA provides the environmental profile of a system and evaluates the distribution of the burdens and impacts across life cycle stages. LCA uses a systematic and comprehensive approach to assess environmental burdens associated with products, and it has been used as an analytic tool for pollution prevention, life cycle design, and optimization modeling (10-12). A vehicle life cycle consists of the following generic stages: materials production, manufacturing, use, maintenance, and end-oflife. The environmental burden of each stage shows different profiles for various categories. For instance, according to the U.S. Automotive Materials Partnership (USAMP) study, which measured the life cycle inventory (LCI) of a generic 1995 mid-sized car, the use phase contributes 85% of the total life cycle energy consumption (see Figure 1) based on 120 000 mi of service life. On the other hand, the use phase contributes only 19% of the total solid waste produced while the materials production phase contributes 58% (2, 3). Dynamic Programming. Dynamic programming is a collection of mathematical tools used to analyze sequential decision processes. Dynamic programming seeks the particular sequence of decisions that best satisfies a decisionmaker’s criteria. In a dynamic programming model, the time horizon of the problem is the period of time over which the decisions are made. This horizon is divided into intervals, or epochs, such that one decision is made for each epoch. At each epoch, a decision is made that changes the state of the system. The state is the particular set of characteristics of the system that is analyzed over the time horizon. Using vehicle replacement as an example, a state can be defined by a vector 5408
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 23, 2003
FIGURE 1. Life cycle energy consumption for a 1995 generic vehicle based on 120 000 mi of driving (2).
FIGURE 2. Schematic example of the life cycle optimization (LCO) model based on four policies. B1-B4 represent the final environmental burdens for the four policies. (i,j) that represents model year i and vehicle age j. In the business sector, dynamic programming is typically used to help a decision-maker choose the sequence of decisions that causes the least possible cumulative cost over time (or the greatest possible reward). This best sequence, the optimal path identifies the best decision to be made at each decision epoch over an entire time horizon. In the present study, dynamic programming is used to assess the optimal path based on environmental criteria. The decision-makers in this study include both individual car owners and policy makers. Model Construction. Figure 2 is a schematic example of the LCO model applied to vehicle replacement. The y-axis depicts the cumulative environmental burden of a criterion (e.g., CO, NMHC, NOx, CO2, or energy consumption), while the x-axis represents time. For the purpose of the present study, the initial vehicle is assumed to be produced at time 0, and a new model vehicle with a different environmental profile is introduced at time Ta and Tb. Decisions to keep or replace vehicles are made at the points marked by black dots. Materials production and manufacturing environmental burdens are shown as a step function at the time a vehicle is produced. The slope of each line segment represents an energy efficiency or emission factor of a vehicle depending on the criterion to be minimized. The slopes tend to increase with time, indicating deteriorations of emission controls or energy efficiencies. Assume that, at time 0, a decision maker tries to minimize the environmental burden of a criterion within the time horizon N based on information the decision maker has regarding the environmental performance of future vehicles. The decision maker seeks a solution of the form “Buy a new
vehicle at the start of year 0 and keep it for R years and retire it; then buy a new vehicle at the start of year R and keep it for β years and retire it, etc.” As an example, consider four policies depending on the decisions at Ta and Tb. It is assumed that retiring a vehicle and buying a new vehicle occurs simultaneously. (1) If the vehicle owner keeps the initial vehicle throughout the time horizon N, the cumulative environmental burden (B) will result in B1. The slope change between Tb and N represents vehicle deterioration expected for older cars. (2) If the vehicle owner replaces the initial vehicle with a new vehicle at time Ta and keeps the new vehicle until N, the cumulative environmental burden (B) will result in B2. (3) If the vehicle owner replaces the initial vehicle with a new vehicle at time Ta and replaces this second vehicle again at Tb, the cumulative environmental burden (B) will result in B3. (4) If the vehicle owner replaces the initial vehicle at time Tb with a new vehicle and keeps the new vehicle until N, the cumulative environmental burden (B) will result in B4, which is the minimum possible outcome. With this hypothetical example, policy (4) is the optimal policy, and the optimal vehicle lifetimes are Tb and N-Tb. However, in a real-world problem with a longer time horizon, the number of possible policy choices is often enormous. If a decision maker seeks an optimal replacement policy during a time horizon N with a new vehicle at the beginning of year 0, and the vehicle replacement decisions are made at the beginning of every year from year 1, the number of possible outcomes is 2N. In addition, the environmental profiles of N different model years need to be considered based on vehicle age. The LCO model provides an efficient algorithm to find an optimal policy, and the dynamic LCIs determine the environmental profiles of each vehicle model year and age. (See section “Model Application” later in this paper.) A mathematical model to find optimal vehicle retirement policy, as presented in Figure 2, is constructed using the following notation: n, first year of the study; N, last year of the study; M, maximum physical life of a vehicle; BM(i), environmental burden (hereafter referred to as burden) of the materials production of model year i vehicle; BA(i), burden of the manufacturing of model year i vehicle; BU(i,j), burden of the vehicle use during year j of model year i vehicle’s service; BR(i,j), burden of the maintenance during year j of model year i vehicle’s service; BE(i,j), burden of the end-oflife stage of model year i vehicle retired at the end of year j; and u(i,j), burden of purchasing (producing) a new vehicle at the start of year i and keeping it for j years. For any model year i, u(i,0) ) 0 and represents the case in which a new vehicle is not purchased in year i. Therefore
{
u(i,j) ) j
BM(i) + BA(i) + BE(i,i + j - 1) +
∑ (B (i,k) + B (i,k)) U
R
if j > 0
k)1
if j ) 0
0
where xi is the decision variable representing the number of years owning vehicle of model year i. For each criterion, this model seeks to minimize the burden from the life cycle of model years n to N by deciding how long to keep each vehicle before purchasing a new vehicle. The dynamic programming optimality equations are constructed as follows. Let f (i) be the minimum possible burden accumulated from the start of year i through the end of year N given that a purchase is made at the start of year i. Then
{
f (i) )
min
xi ∈ {1, 2, ..., M}
0
{u(i,xi) + f (i + xi)} ∀ i ) n, ..., N ∀i>N
A computer program using the C language was coded to implement this model using the dynamic LCI data discussed in the following section. The features and assumptions of the LCO model can be summarized as follows. (i) The present study does not discount environmental burdens because discounting environmental burdens is both complicated and controversial in methods and assumptions. (ii) The present model describes single replacement/ retirement scenarios in which one vehicle is replaced by another vehicle. Applying this model to more complicated scenarios, such as replacing one vehicle with multiple vehicles, is not attempted here. (iii) The present model assumes that a vehicle is driven constant mileage per year, regardless of model year or age. For example, this scenario would apply to a household vehicle used consistently for commuting or to a Federal fleet car. A Federal sedan is typically driven 12 000 mi/yr for 7 yr until sold to others (1). The U.S. EPA has estimated the average vehicle miles traveled (VMT) of the U.S. car fleet as a function of vehicle age for the MOBILE6 emission model. It ranges, for example, from 14 900 mi at age 1 to 5700 mi at age 20 (13, 14). However, changes in the vehicle mileage per year, if any, might be abrupt transitions associated with changes of owners and driving purposes. Moreover, the present model requires equivalent annual VMT between old and new model vehicles upon vehicle replacement. More sophisticated, stochastic models would be required to include variable mileage and are left to future research.
Model Application As indicated in the mathematical model description, the input data to the LCO model consist of a collection of single-year environmental profiles for five life cycle phases: materials production, manufacturing, use, maintenance, and end-oflife. These environmental profiles are modeled as functions of model years and/or ages of vehicles, which are called “dynamic LCIs” in the present study. Three basic factors that influence the environmental performances of vehicles have been defined: regulatory/social factors, technology improvements, and deterioration of a vehicle (see Figure S1 in the Supporting Information). For most model years, these dynamic LCI factors are related to each other. For example, technology improvements may enhance the durability of emission control systems such as the catalytic converter. Accordingly, this may change vehicle deterioration behavior. In addition, seven LCI parameters reflect the evolutions of these dynamic LCI factors: recycled content, materials use, VMT, energy intensity, fuel economy, emission factors, and component reliability. These parameters are determined for each model year and/or age within the range of the optimizations. The dynamic LCIs for generic U.S. mid-sized passenger cars with internal combustion engines between model years 1985 and 2020 were analyzed in detail separately and are available in the Supporting Information. The dynamic LCIs build upon the LCI analysis of a generic vehicle conducted by the U.S. Automotive Materials Partnership (USAMP). Since the goal of the USAMP study was to develop a comprehensive LCI of North American generic vehicles, the geographic system boundary of the study covered primarily the United States and Canada (2, 3). The dynamic LCI parameters of the present study also focus on the U.S. auto and materials industry. In additon, major assumptions associated with the dynamic LCI modeling, such as auto parts recovery and materials recycling, are consistent with the USAMP study. Large uncertainties are associated with the dynamic LCI parameters for the future model years, but this is inherent with forecasts of this nature. To obtain the best likely parameters available, a variety of literatures and resources from industries, universities, and government institutes were VOL. 37, NO. 23, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5409
analyzed. Implications regarding vehicle replacement policy based on other parameters will be discussed later. The dynamic LCI parameters for generic mid-sized passenger cars used for the present study are highlighted here. More detailed procedures to model dynamic LCIs from these parameters are presented in the Supporting Information of this paper. Material Use. The total weight of a mid-sized car remained nearly unchanged between 1985 and 1995 with a variance of approximately 4% or less (15). Using the analyses of the University of Michigan Transportation Research Institute (UMTRI) and the Energy Laboratory at MIT, the present study assumes that the total weight will decrease 17% between 1995 and 2020 and that the weight of aluminum in a generic mid-sized vehicle is expected to increase 71% between 1985 and 2020 (16, 17). Energy Intensity. Analysis of the actual historic trends shows that the energy intensity for primary aluminum and steel production decreased 26% and 12%, respectively, between 1985 and 1998. Forecasts by the U.S. Department of Energy (DOE), Energy Information Administration (EIA) suggest that energy intensity will further decrease 14% and 24% for aluminum and steel, repectively, between 1999 and 2020 (18, 19). Emission Factors. Historical trends of automotive emissions were determined for each model year and vehicle age based on actual data, including the EPA’s long-term Federal Test Procedure (FTP) surveys (20). Regulatory schedules, other analysis in the literature, and actual trends through the mid-1990s were used for predicting future vehicle emissions. Fuel Economy. For historical trends, the EPA’s salesweighted fuel economies were used (21). The reference case scenario of the U.S. DOE, EIA’s Annual Energy Outlook 2001 was selected to forecast the fuel economies of average new cars between 2000 and 2020 (19). According to these sources, fuel economies will increase from 27.0 to 32.5 mi/gal between 1985 and 2020 for an average new car. As an example, Figure 3 shows the dynamic LCI profiles of NOx and CO2 for the use phasesthe predominant life cycle phase for a vehiclesbased on driving12 000 mi annually. Profiles for other life cycle phases and criteria are presented in Figures S2-S6 of the Supporting Information. While the grams per mile levels of the NOx increase with the age of vehicles, CO2 emissions remain nearly unchanged with vehicle age for each model year. In particular, the regulated automotive emissions (CO, NMHC, and NOx) from the 1980s model years increase sharply with vehicle age, primarily due to the low durability of catalysts and computer controls. The improved technology and stricter regulations have remarkably enhanced the durability as well as the initial performance of emission control systems for the 1990s model years (20). Moreover, the emission profiles for model years beyond 2000 show that Tier 2 and further regulatory programs will additionally reduce emissions from both new and old vehicles. The environmental burdens for material production, manufacturing, maintenance, and end-of-life phase are also expected to decrease 20-40% between model years 1985 and 2020. The LCO model was applied to the scenario of generic U.S. mid-sized passenger cars to evaluate the optimal lifetimes and to recommend future policies based on the dynamic LCIs determined. The model years for the optimizations are set between 1985 and 2020, and the maximum physical life of a mid-sized passenger car is assumed to be 20 yr. The optimizations were conducted to minimize CO, NMHC, NOx, CO2, and energy consumption criteria. Other objectives, such as particulate matter (PM), methane, and air toxics are not considered, primarily due to data limitations. For instance, since the current Federal emission standards 5410
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 23, 2003
FIGURE 3. Dynamic LCIs of vehicle-use phase, BU(i,j), for select model years (MY) based on driving 12 000 mi annually. do not regulate PM for cars, estimations regarding the PM profiles of past and future model year cars are highly uncertain.
Results and Discussion Optimal Lifetimes for Mid-Sized Cars. Table 1 gives the optimization results of generic mid-sized model scenarios. The optimal lifetimes and cumulative environmental burdens for the future model year cars are reasonable estimates based on best available forecasts for the dynamic LCI parameters. The numbers in the third column represent the optimum lifetimes of cars in the order from left to right. The optimal set of lifetimes for the energy/CO2 objectives in Table 1 can read, for example, “Keep the model year 1985 car for 18 years and retire it at the end of 2002, then buy a model year 2003 car and keep it for another 18 yr until 2020 in order to minimize energy/CO2 emissions when driving a mid-sized passenger car 12 000 mi/yr”. (In the text, “[ ]” represents the set of optimal lifetimes.) The identical results for the energy and CO2 objectives can be attributed to fossil fuel combustion, which accounts for the majority of both energy consumption and CO2 emissions for a car. It is also notable that the regulated automotive emissionssCO, NMHC, and NOxsare optimized with shorter lifetimes than the optimal lifetimes for the energy/CO2 objectives. As can be seen in Table 1, the optimal car life becomes shorter with increasing annual VMT, especially for CO, NMHC, and NOx emissions. Most optimal lifetimes for 24 000 mi of annual VMT scenario are less than half of the optimal
TABLE 1. Optimal Vehicle Lifetimes and Cumulative Environmental Burdens for a 36-yr Time Horizon between 1985 and 2020 cumulative environmental burdens objective minimized
VMT (103 mi/yr)
energy/CO2
6 12 24
CO
optimal vehicle lifetimes (yr)
energy (103 GJ)
CO2 (105 kg)
CO (106 g)
NMHC (105 g)
NOx (105 g)
18, 18 18, 18 8, 10, 8, 10
1.80 3.34 6.45
1.16 2.18 4.25
1.88 4.95 7.62
3.26 6.18 9.67
2.86 6.52 14.2
6 12 24
5, 5, 12, 14 3, 3, 4, 6, 6, 7, 7 2, 2, 2, 2, 2, 3, 3, 4, 4, 4, 4, 4
2.02 3.84 7.34
1.29 2.46 4.75
1.44 2.76 5.29
3.16 4.29 10.2
2.49 4.54 12.3
NMHC
6 12 24
7, 15, 14 6, 6, 10, 14 4, 4, 6, 6, 8, 8
1.91 3.53 6.71
1.23 2.29 4.40
1.53 2.96 6.02
2.87 4.07 8.74
2.51 4.47 21.3
NOx
6 12 24
10, 12, 14 5, 5, 6, 6, 14 4, 4, 4, 4, 4, 4, 4, 4, 4
1.90 3.65 6.97
1.22 2.36 4.54
1.49 2.86 5.64
2.93 4.14 9.50
2.40 4.32 11.0
sophisticated multi-objective tradeoff analysis based on the current result was conducted in separate research (22).
FIGURE 4. Cumulative environmental burdens accrued during the 36 yr of time horizon when adopting energy/CO2-optimum and COoptimum policies with 12 000 mi of annual VMT. The cumulative optimal environmental burdens are normalized to 1 for each criterion. lifetimes for the scenario of 6000 mi of annual VMT. This can be attributed to the growing dominance of use phase emissions and energy consumption as well as a high deterioration rate caused by the increasing annual VMT. In other words, as the VMT increases, driving a new, loweremitting, and efficient car becomes more important, while the additional emissions from retiring an old car and producing the new car become relatively insignificant. On the other hand, since automobile life cycle emissions are dominated by the use phase, the cumulative environmental burdens increase roughly proportionally to annual VMT. Figure 4 compares the environmental burdens that will accumulate during the 36-yr term of optimization years (1985-2020) when adopting energy/CO2-optimum and COoptimum replacement policies based on 12 000 mi/yr VMT. Each axis represents a cumulative environmental burden, and the optimum (minimum) environmental burden for each criterion is normalized to 1. As can be seen, the energy/ CO2-optimum policy [18, 18] will create far more CO, NMHC, and NOx emissionssregulated auto emission criteriasthan the CO-optimum policy. The increment ranges between 51% (for NOx) and 79% (for CO). On the other hand, the COoptimum policy [3, 3, 4, 6, 6, 7, 7] keeps the other four criteria of environmental burdens comparable to their optimums increasing CO2 by 13%, energy by 15%, and NOx and NMHC by 5% from each optimum. As shown in Table 1, the NMHCoptimum policy [6, 6, 10, 14] and the NOx-optimum policy [5, 5, 6, 6, 14] do not significantly inflate other environmental burdens either. These results imply that policies that shorten vehicle lifetimes such as scrappage programs may significantly reduce CO, NOx, and NMHC emissions with moderate increases in CO2 emissions and energy consumption. A more
Determinants of Optimal Lifetime. Distribution of Environmental Burdens across Life Cycle Stages. In classical cost optimization with dynamic programming, the tradeoff between fixed and marginal cost plays an important role. For the LCO model, the fixed burdens are created during the materials production, manufacturing, and end-of-life stages, while the marginal burdens are created during the use and maintenance stages. The USAMP study shows that, for model year 1995 with lifetime 120 000 mi, fixed environmental burdens account for 11.8% of total life cycle CO2; 3.3% of CO; 8.1% of NMHC; and 8.4% of NOx (2, 3). Optimal lifetimes tend to be longer as the ratio of fixed to marginal environmental burdens becomes greater. Technology Improvement with Model Year. Technology improvements make vehicles cleaner and more efficient. Since the 1960s, major technological innovations for emission controls, such as catalytic converters, exhaust gas recirculation, and computer-based sensors and engine controls have been developed to meet Federal certification standards (23). Currently, clean vehicle initiative programs are guiding new technologies, such as transitional low-emission vehicles (TLEVs), low-emission vehicles (LEVs), ultralow-emission vehicles (ULEVs), and zero-emission vehicles (ZEVs). Moreover, manufacturers have been exploring new propulsion systems and fuels, such as hybrid electric vehicles and hydrogen fuel cell vehicles. Scenarios with such leapfrog technological innovations, as compared with the moderate forecasts of the present study based on conventional internal combustion cars, may dramatically reduce optimal vehicle lifetimes, especially for the energy/CO2 categories (24). On the other hand, in a slowly improving technology scenario, frequent vehicle retirements may increase overall environmental burdens within a time horizon due to the new environmental burdens associated with producing new vehicles. Efficiency improvements in materials processing, vehicle assembly, vehicle maintenance, and recycling across optimization years may also influence optimal vehicle lifetimes. Vehicle Deterioration with Age. For the present study, vehicle deterioration is described as increasing emissions per mile. While the regulated automotive emissions (CO, NMHC, and NOx) increase with vehicle age, CO2 emissions and fuel economy are not likely to deteriorate. Deterioration effects may have contributed to the short optimal lifetimes for the regulated automotive emissions compared with the lifetimes for the energy/CO2 criteria. In addition, as shown in Table 1, the late 1990s and future model cars generally have longer optimal lifetimes than earlier models when VOL. 37, NO. 23, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5411
determined by regulated automotive emissions. This can be explained by improving durability of emission controls with newer model years. Vehicle emission deterioration is categorized in two different ways: the degradation of normalemitters and the increasing number of high-emitters associated with the malfunctioning of emission controls. In fact, studies indicate that both normal- and high-emitter deterioration have decreased remarkably for the 1990s model years and that the durability of emission controls will continue to improve with future models (12, 20). Regulatory/Social Factors. These factors are closely linked to technological improvements and vehicle deterioration behaviors in a variety of ways. Regulations including CAFE and Federal tailpipe emission standards have been influencing fuel economy and emission control technologies for new vehicles. Federal certification tests also regulate emission deterioration of old cars by measuring exhaust emissions at 100 000 or 120 000 mi. In addition, many states in the United States are implementing Inspection and Maintenance (I/M) programs to improve used cars’ emissions. In fact, the emission factors of cars driven in I/M program areas are known to be lower than cars driven in non-I/M areas (25). Thus, in terms of regulated auto emissions, the optimal lifetimes of cars in I/M areas are expected to be longer than those in non-I/M areas. Also, the VMT of individual vehicles primarily depends on the vehicle type and purpose (e.g., household vehicles and government fleet vehicles) (1). In the United States, the average annual miles per household vehicle in 1995 was 12 200 mi, based on self-reported data, and 11 800 mi, according to odometer-based data (26). On the other hand, cars owned by the business sector were driven an average of 22 780 mi, while government cars were driven an average of 12 895 mi in 2000 (1). Thus, as can be predicted by the optimization results, optimal lifetimes for business cars may be shorter than those for household and government cars, when determined by environmental burdens. Policy Implications. In the ideal vehicle design and retirement schemes, optimal lifetimes determined for different environmental criteria would be equal to each other, and the actual vehicle retirement schedule would exactly match with this single optimal lifetime. However, the optimization results for the model years 1985-2020 show that the optimal lifetimes vary considerably with criteria and differ from the median real-world lifetimess12.5 and 16.9 yr for model years 1980 and 1990, respectively (1). In particular, optimal lifetimes determined by some regulated auto emission categories are unrealistically short (e.g., 2-4 yr optimal lifetimes, in Table 1, when determined by CO for a generic car driven 24 000 mi annually). Overall, optimization results suggest that there exists substantial potential for improvement in current retirement practices, emission regulations, and vehicle designs. The present study demonstrates that accelerating or delaying vehicle retirement can reduce different environmental burdens. Car scrappage programs, which try to retire high-emitting old vehicles using incentives, are a well-known example. The main targets of scrappage programs have been vehicles more than 10 yr old in the European countries and more than 15 yr old in the United States. The reimbursement in a scrappage program is provided either for a scrapped car or upon replacement of the scrapped car with a new or cleaner vehicle. In either case, scrappage programs are known to boost car sales as well (6, 27). Increased car sales can cause negative impacts on the environment by requiring additional vehicle manufacturing (6, 9). Therefore, as highlighted in the present study, it may be critical to optimize age, mileage, and emission factors of vehicles based on the target emissions for a successful scrappage program. The present study uses the average emission profile of model years to determine optimal scrappage timing of a generic car. However, if the 5412
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 23, 2003
emissions of a specific car are measured properly, the present model can provide more precise optimal scrappage timing for the car. Several studies have revealed that scrappage programs pose negative impacts in terms of CO2 emissions and energy consumption associated with additional vehicle production (6, 9). In fact, Table 1 and Figure 4 show that shortening car lifetimes increases CO2 emissions and energy consumption. The negative impacts associated with early automobile retirements seem insignificant at first as compared with the negative impacts of keeping old vehicles. However, further analysis regarding the tradeoffs between the regulated auto emissions and CO2/energy will be necessary. The optimal lifetimes for the regulated auto emission criteria for late 1980s and early 1990s model years are 3-6 yr (equivalent to 36 000-72 000 mi), based on driving 12 000 mi annually. These optimums are shorter than those for energy/CO2 objectives and shorter than most of real-world scrappage age targets. The short optimal lifetimes for the regulated auto emissions are associated primarily with the reductions achieved in new-vehicle emissions and the deterioration or failure of emission control systems. On the other hand, deterioration seems insignificant for the energy/ CO2 objectives throughout the optimizations. The optimal lifetime for these objectives (18 yr) is longer than those of real-world cars, based on driving 12 000 mi annually. Thus, policies should focus on the improvement of energy efficiencies in new cars, where major gains can be made. However, the optimal lifetimes for energy/CO2 should perhaps be longer in a realistic forecast. The EIA forecast for new car fuel economy (32.5 mpg by 2020), which has been used for the optimization, is based on the assumptions that gasoline prices will be higher than current prices and that advanced vehicle technologies will be developed (19). If these assumptions are optimistic, then a plausible alternative is continued constant fuel economy, consistent with CAFE standards and average car fuel economy being nearly unchanged for the last 15 yr. In this scenario, where the performance does not improve, the optimal lifetimes for energy/CO2 would be even longer. As perhaps shown by the very short optimal lifetimes determined by the regulated automotive emissions (CO, NMHC, and NOx), very strong regulations and technological developments, however, can result in shorter optimal lifetimes than real-world lifetimes. In the context of this analysis, this leads to unnecessary investment in new technologies as vehicles in the real world may last longer than is optimal if appropriate scrappage programs are not implemented. In applying these policies, environmental impacts from other life cycle phases need to be considered as well. Policies improving vehicle durability or fuel economy may result in increased emissions during vehicle production and scrapping. This is due to the complex links among the life cycle stages, dynamic LCI factors, and those dynamic LCI parameters defined in the present study. If, for example, fuel economies are improved through the use of energy-intensive materials beyond what is already analyzed in the present study, then the materials production becomes a more important factor in assessing the optimal lifetime for energy/ CO2. Rather than the optimal lifetime for energy/CO2 decreasing due to fuel economy improvements, optimal lifetime actually may increase due to the energy-intensive materials.
Acknowledgments Support for this work was provided by NSF under the 1999 Technology for Sustainable Environment (TSE) Program Grant BES-9985625. The authors gratefully acknowledge the contributions to the present study from Marc Ross and
Jonathan Bulkley. The authors also thank Kevin Cullen and Ronald Williams at General Motors for providing valuable information for the present study.
Supporting Information Available Detailed dynamic LCIs for U.S. mid-sized cars between model year 1985 and model year 2020; past trends and future forecasts for major environmental parameters, such as fuel economies and emission factors, in addition to the dynamic LCI methodology; and a summary of model results. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) DOE. Transportation Energy Data Book, 22nd ed.; U.S. DOE, Center for Transportation Analysis, Oak Ridge National Laboratory: Oak Ridge, TN, 2002. (2) Second Peer Review Final Report: Life Cycle Inventory Analysis of a Generic Vehicle; U.S. Automotive Materials Partnership Life Cycle Assessment Special Topics Group in Conjunction with Ecobalance and National Pollution Prevention Center; 1999. (3) Sullivan, J. L.; Williams, R. L.; Yester, S.; Chubbs, S. T.; Hentches, S. G.; Pomper, S. D. SAE Trans. 1999. (4) Life Cycle Design Guidance Manual: Environmental Requirements and The Product System; U.S. Government Printing Office: Washington, DC, 1993; EPA600-R-92-226. (5) Stahel, W. R. In The Greening of Industrial Ecosystems; Allenby, B. R., Richards, D. J., Eds.; National Academy Press: Washington, DC, 1994; pp 178-190. (6) Cleaner Cars: Fleet Renewal and Scrappage Schemes; European Conference of Ministers of Transport: 1999. (7) Stoffer, H. Automot. News 2002, January 7, 39. (8) Retiring Old Cars: Programs To Save Gasoline and Reduce Emissions; Office of Technology Assessment, U.S. Congress: Washington, DC, 1992. (9) Wee, B. V.; Moll, H. C.; Dirks, J. Transport. Res. D-Tr. E. 2000, 5, 137-143. (10) Keoleian, G. A. In Pollution Prevention Handbook; Freeman, H. M., Ed.; McGraw-Hill: New York, 1995; pp 253-292. (11) Life Cycle Design of a Fuel Tank System; U.S. Environmental Protection Agency, Office of Research and Development, National Risk Management Research Laboratory: Cincinnati, OH, 1997; EPA 600/R-97/118. (12) Kim, H. C.; Keoleian, G. A. Vehicle Replacement Policy using Life Cycle Optimization Modeling; Center for Sustainable Systems, University of Michigan: Ann Arbor, MI, 2001.
(13) Update of Fleet Characterization Data for Use in MOBILE6s Final Report; U.S. Goverment Printing Office: Wasington, DC, 1998; EPA420-P-98-016. (14) Fleet Characterization Data for MOBILE6: Development and Use of Age Distributions, Average Annual Mileage Accumulation Rates and Projected Vehicle Counts for Use in MOBILE6; U.S. Government Printing Office: Washington, DC, 1999; EPA420-P-99-011. (15) Binder, A. K., Ed. Ward’s Automotive Yearbook 2000, 62nd ed.; Ward’s Reports: Detroit, 2000. (16) Delphi X: Forecast and Analysis of the North American Automotive Industry for 2004 and 2009, Vol. 1 Technology; Office for the Study of Automotive Transportation, University of Michigan Transportation Research Institute: Ann Arbor, MI, 2000; UMTRI2000-3-1. (17) On the Road in 2020: A life-cycle analysis of new automobile technologies; Energy Laboratory, Massachusetts Institute of Technology: Cambridge, MA, 2000; Energy Laboratory Report EL 00-003. (18) Energy Information Administration. Changes in Energy Intensity in the Manufacturing Sector 1985-1994; 1998; http://www.eia.doe.gov/emeu/mecs/mecs94/ei. (19) Energy Information Administration. Annual Energy Outlook 2001 with Projections to 2020; 2001; http://www.eia.doe.gov/oiaf/ aeo. (20) Austin, S.; Ross, M. Presented at the SAE 2001 World Congress, Detroit, MI; SAE International: Warrendale, PA, 2001; No. 200101-0229. (21) Light-Duty Automotive Technology and Fuel Economy Trends 1975 Through 2000; Office of Transportation and Air Quality (OTAQ), U.S. EPA: Washington, DC, 2000; EPA420-R00-008. (22) Grande, D. E.; Bean, J. C.; Keoleian, G. A.; Kim, H. C.; Zhou, H. To appear in Proceedings of the International Conference of Industrial Engineering and Engineering Management, Shanghai, People’s Republic of China, December 2003. (23) Ross, M.; Goodwin, R.; Watkins, R.; Wang, M. Q.; Wenzel, T. Real-World Emissions from Model Year 1993, 2000 and 2010 Passenger Cars; American Council for an Energy-Efficient Economy: Washington, DC, 1995. (24) Kim, H. C.; Keoleian, G. A.; Spatari, S.; Bulkley, J. W. Presented at the Total Life Cycle Conference, Detroit, MI; SAE International: Warrendale, PA, 2000. (25) Wenzel, T. Environ. Sci. Policy 2001, 4, 377-389. (26) 1995 NPTS Databook; U.S. Department of Transportation, Federal Highway Administration: Washington, DC, 2001; ORNL/ TM-2001/248. (27) Bohn, J. Automot. News 1992, February 24, 32.
Received for review May 25, 2003. Revised manuscript received July 29, 2003. Accepted August 20, 2003. ES0345221
VOL. 37, NO. 23, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5413
