Environ. Sci. Technol. 2009, 43, 6320–6325
Linking Material Flow Analysis and Resource Policy via Future Scenarios of In-Use Stock: An Example for Copper MICHAEL D. GERST* Thayer School of Engineering, Dartmouth College, 8000 Cummings Hall, Hanover, New Hampshire 03755
Received March 20, 2009. Revised manuscript received July 1, 2009. Accepted July 13, 2009.
A key aspect to achieving long-term resource sustainability is the development of methodologies that explore future material cycles and their environmental impact. Using a novel dynamic in-use stock model and scenario analysis, I analyzed the multilevel global copper cycle over the next 100 years. In 1990, the industrialized world had an in-use copper stock about twice as large as the developing world and a per capita inuse stock of about six times as large. By 2100, the developing world will have an in-use copper stock about three times as large as the industrialized world, but the industrialized world will maintain a per capita stock twice that of the developing world. Under a scenario of no material substitution or technological change in copper products, global in-use stock in 2100 will be about as large as currently known copper resources. However, current scrap recycling trends and exploration will alleviate absolute supply pressure but not environmental impacts from decreasing copper ore grades. Additionally, unexpected emergent properties of dematerialization are observed from the in-use stock model that arise solely from the properties of stock dynamics, an infrequently discussed cause of dematerialization in the literature.
Introduction With the recent formation of the United Nations International Panel for Sustainable Resource Management (IPSRM) in 2007, new organized and international attention has been brought upon the important issues of resource management and related environmental impacts. As top priorities, the new panel has chosen to focus on assessing the environmental risks of biofuel production and metal mining and recycling. A crucial piece, among many, to addressing impacts of metal mining and recycling is the study of past and possible future metal cycle patterns. A metal cycle or any material cycle can be defined as the mathematical representation of the natural and anthropogenic (i.e., human made) states of a material organized into a network of stocks (or reservoirs) and flows. The concept (also known as material flow analysis) borrows heavily from many disciplines such as biogeochemical budgets in the geological sciences and capital stock and flow accounts in economics (1, 2). As an analytical framework, material flow analysis aims to understand the economy and its impact on the environment in terms of the material that is perturbed, * Corresponding author phone: 603-646-2488; fax: 603-646-2277; E-mail:
[email protected]. 6320
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 16, 2009
transformed, accumulated, and discarded in the course of fulfilling societal desires. It has clear application to policy analysis because it provides a flexible and consistent framework by which measurements of material efficiency or environmental impact can be measured and decisions and policies can be tested. Probably one of the first and most important precursor applications of this type of analysis was the United States President’s Materials Policy Commission (better known as the Paley Commission) of 1952 (3). Convened because of the concern for future material supply problems in the United States, the Paley Commission published 25 year demand scenarios for a large variety of materials and used these scenarios to make recommendations for domestic economic and resource policy. Novel to their approach and infrequently used in long-term material flow analysis since (see refs 4 and 5 for more recent examples) was the linkage of future material demand with the dynamics of services (e.g., housing) provided by material stock. As discussed in Gordon et al. (4) and Mu ¨ ller (5) as well as by Schipper et al. (6) in the energy analysis field, linking material and energy flows to services and in-use material stock allows for a deeper understanding of material and energy cycles and greater relevance to decision makers. In practice, applying this approach in a future-oriented study is more challenging than simply fitting a statistical model to energy and material demand and macroeconomic data. First, relating demand for services to in-use material stock by means of a predictive model requires specifying future microlevel decisions (e.g., household choice of dwelling size and materials) in a way that is tractable to analyze largerscale resource questions. Second, the collection of the necessary data on service demand and in-use stock is often problematic. As detailed in a review of in-use metal stocks by Gerst and Graedel (7), measuring in-use stock can be achieved by two methods, top down and bottom up. The top-down method takes long-term data on metal demand and metal discard and calculates in-use stock by adding the annual difference of demand and discard over time. It is a method that relies on more easily accessible data, but it is difficult to link to demand for services. In contrast, bottomup studies directly measure the stock in use at a discrete point in time in terms of metal contained in technologies. It is data intensive. As a result, it is typically calculated for a single year, but it directly links in-use stocks and services. As a consequence of the difficulties in calculating in-use stock and specifying a tractable level of model aggregation, most future-oriented studies of metal supply and demand have focused on forecasting by statistical calibration to more accessible historical demand data (8-15). This paper significantly builds upon previous service demand and material and energy flow research by (1) compiling a historical demand for a services data set, (2) introducing a world (with regional detail) material cycle model for copper, (3) and presenting scenario analyses over the 21st century.
Methodology The material cycle model for copper is based on the principle of mass balance. It considers the stocks and flows of 14 different copper-containing technologies (Table 1) such that demand, D, is specified by Di,j,t ) Si,j,t - Si,j,t-1 + Oi,j,t 10.1021/es900845v CCC: $40.75
(1)
2009 American Chemical Society
Published on Web 07/22/2009
TABLE 1. Copper-Containing Technologies end-use category buildings (residential and nonresidential)
infrastructure
transportation industrial equipment
technologies wiring piping heating, ventilation, and air conditioning (HVAC) refrigerator dishwasher clothes washer clothes dryer power distribution lines power transformers power plants telephone mainlines cars trucks industrial equipment
where S is the level of in-use stock, and O is the rate of discarded stock. The subscripts are indices that represent 1 of 14 (i) copper-containing technologies, 1 of 11 (j) world regions, and time t. In contrast to models centered on demand statistics, the methodology of this study is centered on the level of in-use stock. The rate of discarded stock is estimated from the vintage structure of in-use stock and average technology lifetimes, leaving demand to be calculated by eq 1. In-use stock is estimated from Si,j,t ) Ωi,j,t(•)Ψi,j,t(•)
(2)
where Ωi,j,t(•) is a function which calculates the in-use stock of a copper-containing technology (e.g., length of power distribution cable), and Ψi,j,t(•) is a function which calculates the copper content of the technology (e.g., kiligram of copper per length of power distribution cable). Both Ωi,j,t and Ψi,j,t are calibrated using historical data on demand for services and the copper content of technologies. As an example, consider that estimating Si,j,t for residential wiring and piping involves, among other things, estimating the percentage of dwellings in a region that contain piping or wiring (also called penetration rate of a technology). Inspection of historical data on residential wiring and piping penetration rates (Figure 1) indicates that the many microlevel decisions that lead to residential wiring or piping installation can be represented by a proxy indicator of regional affluence, the purchasing power parity adjusted gross domestic product per capita measured in constant dollars (GDP ppp per capita). A comparison of the top and bottom panels of Figure 1 clearly shows that households acquire wiring and piping with increasing levels of affluence, and that a household will typically install electrical wiring before piping. The observed patterns indicate that an abstracted wiring and piping penetration model of the residential dwelling stock of a region is a reasonable approximation of microlevel behavior at the regional level. For penetration rates, a standard saturation curve of the form of PRi,j,t ) [1 - exp(-RGDPj,t)]β
(3)
is appropriate, where PR is the penetration rate and GDP is the GDP ppp per capita. Similar regionally aggregated models are specified for each of the copper-containing technologies and are statistically fit to historical data. In total, this study found that in addition to GDP ppp per capita, the relevant macro socioeconomic variables for explaining patterns in Ωi,j,t and Ψi,j,t are population, average household size, and level of urbanization. For the sake of brevity, details on the functional forms,
FIGURE 1. Historical data on household wiring (top panel) and piping (bottom panel) penetration rates, aggregated by region. Data is shown on 5 year intervals. Each region includes at the least the years 1995 and 2000. Regional definitions are from the Intergovernmental Panel on Climate Change (IPCC). Data sources are listed in the Supporting Information. statistical fitting, and data sources are documented in the Supporting Information.
In-Use Stock Model Validation The methodology results in historical in-use stock values for each technology and region. Checking the validity of the results presents a challenge because current techniques do not allow direct measurement of in-use material stock in the same way as one would measure the metal content of a rock. As detailed in Gerst and Graedel (7), many studies have been published on the in-use stock of copper, although the methods used and spatial scale and level vary widely. A rough check on the reasonableness of the results can be made by comparing the literature values of in-use stock normalized by population to the modeled regional in-use stock in 2000 (Table 2). A zero-order inspection of the modeled per capita in-use stock values by region is as expected. The more affluent developed countries have much larger per capita reservoirs of copper than the developing countries; the world average is somewhere in the middle but closer to the developing world values because of the global distribution of population. The current state of the literature on metal in-use stock values is spatially diffuse and subject to its own inherent uncertainties. Little data is available for the developing world and only in the form of estimates at the city level. However, preliminary comparisons can be made to illuminate current uncertainty and data gaps. Using a simple statistical test (two-tailed t test at five percent level), the reasonableness of VOL. 43, NO. 16, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6321
TABLE 2. In-Use Stock Model Results for 2000 and Comparison with Literature percentage of in-use stockc
values from literatured
regiona
Sb
buildings
infracture.
transportation
industrial equipment
mean
AFR CPA NAM PAO WEU World
9 44 206e 170 154e 57e
40 (30) 58 (50) 45 (45) 41 (55) 46 (46) 46
39 (29) 26 (14) 29 (34) 34 (20) 35 (40) 32
6 (15) 1 (11) 11 (12) 9 (7) 8 (8) 6
15 (25) 15 (28) 15 (10) 15 (18) 15 (12) 15
36 30 231 232 177 43
SD
88 39 36 11
range
N
157-391 171-298 140-220 35-50
1 1 6 10 4 2
a Regions are from IPCC: AFR ) Sub-Saharan Africa, CPA ) Centrally planned Asia (includes China), NAM ) North America, PAO ) Pacific Organization for Economic Co-operation and Development (OECD) (includes Australia and Japan), and WEU ) Western Europe. b Values are the estimated in-use stock in killigrams per capita. c End-use categories correspond to those listed in Table 1. Nonparenthesized values refer to modeled values; parenthesized values refer to a calculated regional average taken from reported values in the literature. d Descriptive statistics of per capita in-stock values from the literature. SD is standard deviation. N is number of sample values. Values are taken from Gerst and Graedel (7). e At the 5% level, the hypothesis that S is the true in-use stock value, given the mean and SD from the literature (and assumption of normality), cannot be rejected. Test is not done for AFR and CPA because only one literature value is available.
TABLE 3. SRES Scenario Themes scenarios A1 scenario themes globalization environmental consciousness variables population growtha decline in average household sizeb urbanizationc economic growtha
A2
B1
B2
higher lower
lower lower
higher lower higher higher
low high
high low
low high
medium medium
low high low very high medium-high high
medium medium
a Total population growth and economic growth are from the original IPCC SRES. b Decline in average household size is calculated from the dwelling number calculations discussed in Methodology. c Urbanization scenarios are from ref 16, which was designed to be consistent with the original IPCC SRES scenarios.
the modeled values (for the developed regions and World) can be checked against the data in the literature, yielding a positive result for North America, Western Europe, and World regions. An explanation of these results lies in the uncertainty and regional specificity of the data used to calibrate the model. Information on wiring and piping copper content as well as floor space of nonresidential buildings is scarce and limited regionally to the United States. This explains the relative accuracy by which the model replicates the share of copper in buildings in North American and Western Europe but not in the other regions listed in Table 2. Because buildings are generally the largest reservoir of in-use copper, they are a significant source of uncertainty and an area where further data collection is needed. An additional significant source of uncertainty lies in the simplified power distribution model used (see Supporting Information for more detail). Here, an average system was defined for Europe and North America that greatly simplifies the complexity of actual distribution systems. Nevertheless, it is clear from Table 2 that the proposed methodology does provide a reasonable model of in-use copper stock.
Future Scenarios Models such as those specified in this study are useful in gaining insight about possible future outcomes. They have utility on their own as a way to generate material cycle 6322
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 16, 2009
FIGURE 2. In-use copper stock scenarios by region, developing (DEV) and industrialized (IND). Top panel shows scenarios for absolute amount of in-use stock. Bottom panel shows scenarios for in-use stock normalized by regional population. scenarios without regard to economic feedbacks of material price as well as a source of input data to more comprehensive economic models such as in Gordon et al. (4). Developing an economic model is beyond the scope of the present study, so the focus will be on presenting the results of the scenario analysis from the copper cycle model. Scenarios are useful in future-oriented studies because they allow for an internally consistent framework to define deeply uncertain assumptions about the future. Future inuse stock can be modeled by the calibrated functions that comprise Ωi,j,t and Ψi,j,t, and the uncertain exogenous inputs of population, average household size, urbanization, GDP ppp per capita, and copper content. In the following discussion, my exploration of the uncertainty of the future copper cycle is limited to macro socioeconomic variables. This amounts to keeping technological parameters such as the introduction of new copper-containing technologies, change in copper content of technologies, and material
most effect if concentrated in the building and infrastructure categories because these comprise the largest share of inuse stock (Table 2). The difference between absolute amounts and per capita in-use stock is perhaps the most drastic when comparing A1 and A2 scenarios. In 2100, both scenarios yield roughly equal in-use stocks but drastically different per capita values, with the A1 scenario having roughly double the per capita stock of the A2 scenario in both regions. The cause of these results can be investigated by decomposing the influences of growth of required in-use stock by ln FIGURE 3. Comparison of global copper cycle in 2100 to copper resources under the bounding scenario. Ranges of stock, cumulative extraction, and cumulative discard are from scenarios. Information on ore deposits is from Gerst (18) and Kesler and Wilkinson (19). The bounding scenario is not meant to be realistic but an exploration of the bounds of possible technological outcomes. In this case, the scenario is quantified by assuming no material substitution away from copper (beyond current patterns), no reduction in copper use per unit of technology, and no copper recycling. substitution constant over time. While such scenarios may be unlikely to occur, their creation allows for isolating the socioeconomic effects on the copper cycle. To explore the future copper cycle uncertainty with respect to model input variables (Table 3), I adopted the widely used scenario taxonomy from the IPCC Special Report on Emissions Scenarios (SRES) (17). Here, future uncertainty is framed by two scenario themes, degree of globalization and degree of environmental consciousness, which can take on either high or low values. Assumptions about future trajectories of the input variables are then made so that four internally consistent scenarios are produced. For the sake of simplicity, in-use stock scenario results are aggregated into developing or industrialized world regions. It is important to note that the in-use stock values shown in Figure 2 are required in-use stock values that are independent of past amounts of in-use stock; they do not yet take into account technology lifetimes. In absolute weight (Figure 2, top panel), the industrialized world starts with 167 Tg of in-use copper stock, compared to 85 Tg in the developing world, a ratio of roughly two to one. Normalized by population (Figure 2, bottom panel), the magnitude of this difference becomes even more striking with the industrialized region having an in-use copper stock of 132 kg per capita, compared to 21 kg per capita in the developing world, a ratio of roughly six to one. Clearly the affluence of the industrialized region has a significant influence on its relatively larger in-use copper stock, given its relatively smaller population. By 2100, a different picture emerges. In-use copper stock for the industrialized region ranges from 398 to 637 Tg, while stock in the developing regions ranges from 1067 to 1514 Tg. Normalized by population, these ranges are 283 to 401 kg per capita and 116 to 270 kg per capita for the industrialized and developing regions, respectively. In absolute terms, the developing region has significantly surpassed the industrialized region. However, in per capita terms, the industrialized region retains a nontrivial effect of affluence, dropping the 1990 per capita gap of six to one ratio to about two to one. If technological parameters were not held constant or if price effects were accounted for in an economic model, then absolute and per capita in-use stock would certainly be different. However, if these effects were roughly similar across regions, then much of the relative results between the developed and developing world would be similar. In absolute terms, changes in technological parameters would have the
( )
St+1 ) St
{
[
∑ ∑ ln( k
i)1
n
j)1
S(Ak)j,t+1, Ak*j,t)i,t+1 Si,t
)]
Si,t k
∑ i)1
Si,t
}
(4)
In eq 4, the separate influence of the k input variables (Table 3), Ak,t, for each technology i is calculated by multiplying the share of in-use stock for technology i at time t by the natural logarithm of the quantity of in-use stock of technology i as a function of the input variable of interest at time t + 1, with all other variables at time t, divided by in-use stock of technology i at time t. Repeating the described calculation for all technologies and input variables sums to the total growth rate of stock shown on the left of the equal sign in eq 4. Recalling the scenario attributes from Table 3, the A1 scenario of fast demographic transition (i.e., low population growth and fast decrease in household size) and very high growth in economic output and the A2 scenario of slow demographic transition and medium-high growth in economic output essentially explore directly opposite outcomes; both yield average annual in-stock growth rates of 2.2%. Global in-use stock growth for the A2 scenario is much more dependent on population growth than that of the A1 scenario, with average annual growth rates due to population of 0.94% and 0.46%, respectively. In contrast, growth in the A1 scenario is driven more by growth in economic output and fast decreases in average household size than that of A2 scenario, with average annual growth rates due to economic output and changes in average household size of 1.11% (A1) and 0.60% (A2) and 0.25% (A1) and 0.11% (A2), respectively. By definition, per capita means normalized by population, so the difference between the per capita in-use stock in the A1 and A2 scenarios can be explained by growth being much more concentrated on the absolute amount of people in the A2 scenario. Summarily, decomposing the results of Figure 2 yields insight that there is not one single dominating attribute of growth of in-use copper stock. As a generalization across all scenarios, growth in economic output is the most influential variable, meaning that increases in affluence in the industrialized and developing worlds will have a larger effect on the build-up of copper (and perhaps other material) stocks than population size.
Discussion Translating in-stock scenarios into use for concrete resource policy analysis is a step beyond this paper. However, firstorder analyses can be made to guide the direction of future research. One such exercise is to explore the outcomes of the system by create bounding scenarios. Implications for Copper Resources. In relation to copper supply, one can make simple long-term comparisons on the basis of known and predicted resources (18, 19) as of the beginning of the 21st century and the state of the copper cycle in 2100 (Figure 3) by constructing a bounding copper scenario in which all of the following occur: no material VOL. 43, NO. 16, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6323
FIGURE 4. Comparison of copper intensity of use (IU). Solid plotted lines are from this paper and cover 1950-2100. Four scenarios are shown for each region and the world. Ayres world IU is the intensity of use curve specified by Ayres et al. (14). Boxes and vertical lines show, respectively, the ranges of peak IU and IU asymptote, assumed for scenarios from Kapur (15). substitution away from copper (beyond current patterns), no change in copper use per unit of technology, no new copper-containing technologies, and no copper recycling. Such a scenario is admittedly unrealistic. However, its purpose is not realism but to show where the bound of possible outcomes exists. Here, resources are defined as any concentration of copper in the crust of Earth above that of common rock whether or not it is economic to extract now or in the future. Cumulative extraction and discard parameters of the copper cycle are estimated on the basis of assumed average lifetimes of copper-containing technologies, inefficiencies of copper mining and processing, and the principle of mass balance in the economy. In 2100, in-use copper stock will range from about 1450 to 2100 Tg, which compares to known copper resources as of the beginning of the 21st century of about 1450 Tg. Hence, in the bounding scenario, the amount of in-use stock in 2100 will most likely exceed the stock of known copper resources. Furthermore, cumulative extraction of copper resources will range from 5820 to 7730 Tg, which yields a nontrivial chance that extraction would have to extend beyond 1 km below the surface of Earth, the current depth of shallow open pit mining operations (19). While open pit mining to 1 km has its own environmental impacts, mining below this depth will likely involve underground mining, which has its own set of technical and environmental complications. The range observed for cumulative discards highlights the importance of scrap recycling in the resource management of copper and other materials. In the bounding scenario, where recycling does not occur, the resulting stock of copper discards reaches a range of about 3910 to 5190 Tg, at least a factor of 2 greater than in-use stock in 2100. If actual global recycling is around 50% (20), then the cumulative copper extraction scenario would be in the range of undiscovered porphyry ore less than 1 km deep, and the stock of cumulative discards would be in the same range as in-use stock in 2100. Because much of copper scrap originates from the building and infrastructure categories, policies aimed at improving scrap collection should focus on these areas. The actual outcome for the copper cycle will exist somewhere between the bounding scenario and no future use of copper and will, 6324
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 16, 2009
among other factors, be highly dependent on the relative price of copper compared to other potential substitute materials. Implications for Theory of Dematerialization. Historical and future changes in parameters of a material cycle are frequently described through a concept called dematerialization, a reduction in the intensity of use of a material (demand per unit of economic output). Dematerialization is thought to occur via four generalized pathways: (1) transition in economies from agriculture to industrial to service-based, (2) saturation of bulk markets for materials, (3) material substitution, and (4) technological change in the amount of material required per unit of service (21). These four factors of intensity of use are difficult to measure over time; as a result, authors (12, 14, 15) typically resort to calculating demand by calibrating a single-equation model with the aggregate statistic of metal demand per unit of economic output as a function of economic output per capita. Overwhelmingly, authors assume that the factors of intensity of use lead to a model whose basic shape yields an inverted U-shaped curve. The generalized storyline is that as countries become wealthier, the material intensity of their economy grows sharply and then declines as the economy transitions toward services. Concurrently, new cheaper materials substitute for older materials, and technological change reduces the required amount of material to achieve the same task. There are, of course, problems with projecting this storyline with historical demand data because the results are completely independent of fundamental driving forces. Ayres et al. (14) and Kapur (15) have adopted this technique and storyline specifically for copper and have used the IPCC SRES scenario input variables; thus, a comparison can be made to copper demand calculated from this paper (Figure 4). Ayres et al. (14) calculates a single intensity of use curve per region on the basis of historical data. For simplicity, the resulting world intensity of use curve is shown in Figure 4. In contrast, Kapur (15) calculates intensity of use curves by region as well as by scenarios. A1, B1, and B2 are considered, but A2 is omitted because it is considered an unlikely outcome. His intensity of use curves are parametrized by partially qualitative assumptions about how material substitution and technological change will affect the height and placement of peak intensity of use (dotted boxes) and the
asymptotic value intensity of use approaches as GDP per capita approaches infinity (ranges to right of plot). Plotting historical (from 1950) and the bounding scenario (to 2100) intensity of use from this paper by region yields inverted U-shaped curves much like those hypothesized by Ayres et al. (14), Kapur (15), and van Vuuren et al. (12). However, the reason for the curve shape is at odds with the usual focus on transitioning economies, material substitution, and technological change because this paper makes no assumptions about the shape of intensity of use as regions grow wealthier and does not assume the copper demandreducing effects of material substitution and technological change. In the case of copper, saturation in the penetration of building wiring and piping as well as concavity in the relationship of the power distribution network length as a function of electrical generating capacity (Supporting Information) are the main influences on intensity of use curve shape. Referring to Figure 1 shows that the penetration rate curves become concave before 10000 GDP ppp per capita, roughly the same window at which the intensity of use curves peak. This is an important result with respect to dematerialization for two reasons. First, it shows how dematerialization can be an emergent property of demand for services and stocks. Second, the results shown in Figure 4 are the result of a bounding scenario. Therefore, if it is environmentally advantageous to reduce copper demand, significant opportunity remains for material substitution and technological change. The combined results of this paper offer some preliminary guidance to the long-term resource management of copper and clearly show the benefit of the methodology introduced by this paper. It is likely that in 100 years the amount of copper stock will be significantly larger. The bounding scenario analysis shows that a continuation of current recycling rates over the next 100 years will alleviate absolute supply concerns. However, issues will remain as to the energy and water requirements of mining lower ore grades (18) as well as social and environmental impacts of open pit mining. Furthermore, transformation of the in-use stock results into material intensity of use shows that stock dynamics can have a significant effect on dematerialization. In the case of copper, significant potential for reduction in intensity of use still exists. However, environmental impact must be put in context as further growth of in-use stock is highly dependent on growing economic output, which leads to ever increasing per capita in-use stock in the industrialized and developing world. Further studies of these questions can use the results of this paper as a basis by expanding into further scenario analysis, economic modeling, and life cycle assessment.
Acknowledgments The author thanks T.E. Graedel, R.B. Gordon, D.B. Mu ¨ ller, and S. Baker for their comments and suggestions on earlier drafts of this paper. Research was undertaken while the author was a Ph.D. student at the Center for Industrial Ecology and the School of Engineering and Applied Science at Yale University, with project funding supplied by a grant from the Grainger Foundation.
Supporting Information Available Mmethodology and results for the formulation and calibration of in-use stock models for copper-containing technologies. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Fischer-Kowalski, M. Society’s metabolism: The intellectual history of materials flow analysis. Part I. 1860-1970. J Ind. Ecol. 1998, 2 (1), 61–78. (2) Fischer-Kowalski, M.; Hu ¨ ttler, W. Society’s metabolism: The intellectual history of materials flow analysis, Part II, 19701998. J. Ind. Ecol. 1999, 2 (4), 107–136. (3) Harberger, A. C. Projection of 1975 Materials Demand. In Report of the President’s Material Policy Commission; Resources for Freedom: Washington, DC, 1952. (4) Gordon, R. B.; Koopmans, T. C.; Nordhaus, W. D.; Skinner, B. J. Toward a New Iron Age? Quantitative Modeling of Resource Exhaustion; Harvard University Press: Cambridge, MA, 1987. (5) Mu ¨ ller, D. B. Stock dynamics for forecasting material flows: Case study for housing in The Netherlands. Ecol. Econ. 2006, 59 (1), 142–156. (6) Schipper, L.; Bartlett, S.; Hawk, D.; Vine, E. Linking life-styles and energy use: A matter of time. Annu. Rev. Energy 1989, 14, 273–320. (7) Gerst, M. D.; Graedel, T. E. In-use stocks of metals: Status and implications. Environ. Sci. Technol. 2008, 42 (19), 7038–7045. (8) Economic Analysis of the Copper Industry; Working paper for General Services Administration; Charles River Associates: Washington, DC, 1970. (9) Fisher, J. C.; Cootner, P. H.; Baily, M. N. An econometric model of the world copper industry. Bell J. Econ. 1972, 3 (2), 568–609. (10) Hartman, R. S.; Bozdogan, K.; Nadkarni, R. M. Economic Impact of Environmental Regulations on the U.S. Copper Industry; Arthur D. Little, Inc.: Cambridge, MA, 1978. (11) Bozdogan, K.; Hartman, R. S. U.S. demand for copper: An introduction to theoretical and econometric analysis. In The World Copper Industry; Mikesell, R. F., Ed.; The Johns Hopkins University Press for Resources for the Future: Baltimore, MD, 1979. (12) van Vuuren, D. P.; Stenger, B. J.; de Vries, H. J. M. Long-term perspectives on world metal use: A systems dynamics model. Resour. Policy 1999, 25 (4), 239–255. (13) Zeltner, C.; Bader, H.-P.; Scheidegger, R.; Baccini, P. Sustainable metal management exemplified by copper in the U.S.A. Reg. Environ. Change 1999, 1 (1), 31–46. (14) Ayres, R. U.; Ayres, L. W.; Rade, I. The Life Cycle of Copper, Its Co-Products and Byproducts; Springer: Berlin, 2003. (15) Kapur, A. The future of the red metal: Scenario analysis. Futures 2005, 37 (10), 1067–1094. (16) Greenhouse Gas Initiative Scenario Database; International Institute for Applied Systems Analysis: Laxenburg, Austria, 2008; http://www.iiasa.ac.at/Research/GGI/DB/. (17) Nakicenovic, N.; Swart, R. Special Report on Emissions Scenarios. Cambridge University Press: Cambridge, U.K., 2000; p 612. (18) Gerst, M. D. Revisiting the cumulative grade-tonnage relationships for major copper ore types. Econ. Geol. 2008, 103 (3), 615–628. (19) Kesler, S. E.; Wilkinson, B. H. Earth’s copper resources estimated from tectonic diffusion of porphyry copper deposits. Geology 2008, 36 (3), 255–258. (20) Graedel, T. E.; van Beers, D.; Bertram, M.; Fuse, K.; Gordon, R. B.; Gristsinin, A.; Kapur, A.; Klee, R. J.; Lifset, R. J.; Memon, L.; Rechberger, H.; Spatari, S.; Vexler, D. Multilevel cycle of anthropogenic copper. Environ. Sci. Technol. 2004, 38 (4), 1242– 1252. (21) Cleveland, C.; Ruth, M. Indicators of dematerialization and the materials intensity of use. J. Ind. Ecol. 1998, 2 (3), 15–50.
ES900845V
VOL. 43, NO. 16, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6325
