SCOPE 29 - The Greenhouse Effect, Climatic Change, and Ecosystem

2.1 INTRODUCTION

Numerous gases present in low concentrations in the Earth's atmosphere are transparent to incoming solar radiation but absorb the infrared radiation emitted by the Earth's surface and by certain atmospheric constituents. An increase of the concentration of these 'greenhouse' gases in the Earth's atmosphere could lead to considerable global temperature increases and other climatic changes. One of these gases is carbon dioxide (CO₂), which is emitted to the atmosphere through both human activities and natural processes.

The main anthropogenic source of CO₂ emissions is the combustion of carbon-based fuels. In 1983, approximately five gigatons of carbon per year were emitted globally from fossil fuels (Rotty and Masters, 1984). Other net CO₂ emissions from terrestrial biota, including the clearing and burning of forests, have contributed significantly to the present atmospheric CO₂ concentration. However, these sources are expected to contribute relatively less in the future compared to the combustion of fossil fuels. The role of terrestrial biota in the carbon cycle is discussed more fully in Chapter 3.

Besides CO₂ in the atmosphere, the concentrations of other radiatively active trace substances (greenhouse gases) are observed to be increasing. The accumulation of these other gases may eventually affect global temperature as much as or more than CO₂ does. For this reason, these non-CO₂ gases have received much recent attention, and concern about global warming has increased. This new dimension of the atmospheric greenhouse problem is discussed more fully in Chapter 4.

The purpose of this chapter is to explore the feasible range of future CO₂ emissions from energy-related activities. Numerous recently published scenarios for the future global energy system are analysed to determine how each will affect the rate and cumulative level of CO₂ emissions to the atmosphere. Particular attention is paid to the underlying assumptions and uncertainties in projections of future energy developments and CO₂ emissions.

In Section 2.2, the discussion focuses first on the basic links between energy uses, economic activity, and CO₂, and then on the historical pattern of CO₂ emissions. Section 2.3 presents a detailed discussion of the uncertainties that affect projections of future emissions of CO₂. Section 2.4 provides an overview of methodological and other problems that arise in forecasting future CO₂ emissions, together with a critical review of several recent projections of CO₂ emissions to the year 2050. In Section 2.5, feasible upper and lower bounds (or 'extreme scenarios') for CO₂ emissions between now and 2050 are suggested, based on the review in the preceding sections. Section 2.6 presents key findings and conclusions.

2.2 THE LINK BETWEEN ENERGY USE, ECONOMIC ACTIVITY, AND CO₂ EMISSIONS

Modern economic growth is generally attributed to the spread of industrialization. Historically, growth has depended on the availability of energy and on the combustion of fossil fuels. For numerous countries, energy and economic data for the years between 1860 and 1973 suggest that economic growth and energy use increased together. Yet, the data do not establish that one caused the other. (See Section 2.3.)

Since 1973, the ratio of energy per unit of economic output has declined in many countries belonging to the Organization for Economic Cooperation and Development with attendant increases in the real price of energy. Of course, this ratio also tended to decrease even when the real price of energy was declining, as a result of factors not related solely to energy prices. A recent study of energy use in U .S. business has shown that the ratio of primary energy inputs to the economic value of outputs has declined steadily since 1920 (Schurr, 1984).

Increased energy efficiency has resulted from technological improvements in industrial processes, transportation systems, and building design. In addition, some industrialized countries have shifted economic activity away from the production of basic materials and toward fabrication and finishing activities. At the same time, the relative economic role of the services sector has increased (Marley, 1984).

The minimal level of per capita energy use needed to support modern medical, educational, and leisure services varies somewhat in different cultures and regions. In many developing countries, considerable growth in per capita consumption of high-quality fuels is essential to bring the population above the subsistence level. However, it now appears likely that continued economic growth and an attractive standard of living do not depend on ever greater rates of per capita energy consumption. Williams et al. (1984) have argued that, depending on the energy-using technologies employed, demand of as little as 1 kW per capita is theoretically enough to support the whole world at the standard of living enjoyed by Western Europe in the 1970s. To date, the social and economic transition process for reaching this level of demand has not been addressed in detail. Because developing countries must, in many cases, build up the basic components of industrial infrastructure already existing in industrialized countries, some analysts have argued that these societies will temporarily have to experience levels of energy consumption in the transition period that are significantly in excess of an average level of 1 kW per capita (Sh'pillrain, 1985) even if they were later to reduce energy demand to lower levels. However, referring to analyses of energy use patterns during the 'infrastructure-building' phase of industrialized countries, other analysts suggest that this may not be necessary (Goldemberg et al., 1985).

For many countries, the price of energy in the short term will heavily influence the rate of economic growth. Over the 70-year time horizon considered here, most economies will adjust to higher energy prices by substituting labour, information, or other inputs and perhaps by shifting the balance between production of goods and provision of services. Different energy strategies, e.g. using different fractions of coal or nuclear fuel in the energy system, will also directly influence CO₂ emission rates. In short, the weak correlation between economic growth and energy consumption means that steady economic growth need not entail steadily increasing CO₂ emissions. (This will be discussed in later sections.)

2.2.1 Energy Consumption and CO₂ Emissions

No one seeks energy for its own sake. Industrialized nations use energy primarily for industry, transportation, and space heating and cooling in buildings (see Table 2.1). Many recent studies show that current usage in industrialized countries could be reduced by using more energy-efficient technologies. Mintzer and Miller (1984) showed that if the United States-the world's largest emitter of CO₂-switched to the most energy-efficient technologies now available to produce key commodities and services, its CO₂ emissions would fall approximately 25 percent (0.32 GT/year), with output remaining constant. These changes would reduce global carbon emissions by 7 percent. Similar, but smaller, reductions are possible in other industrialized countries. (A small but growing number of studies assess the economic cost of some of these technological possibilities, e.g. Goldemberg et al., 1984.)

Further reductions occur when the composition of economic activity shifts as the result of more efficient methods of providing goods and services. In studies to assess the global potential for improving energy productivity or displacing energy demand (Goldemberg et al., 1984; Rose et al., 1984; Lovins et al., 1981), a wide range of technologies has been considered-such as improvements in building insulation, automobile design, and industrial processes. Overall, these studies show that energy efficiency is currently far from optimal and technological improvements could significantly reduce demand for fossil fuels. At the same time, however, some of the technical fixes being developed and applied to existing industrial processes to reduce negative environmental impacts other than CO₂ emissions-the use of fluegas desulphurization techniques in coal-electric power plants to reduce acid precipitation, for example-may increase overall fuel consumption and, consequently, emissions. (See Sections 2.4 and 2.5.)

Table 2.1 Primary energy use by sector in selected Western European countries. Reproduced from Jäger (1984)

2.2.2 Energy Supply, Fuel Mix, and CO₂ Emissions

Useful energy is produced from several primary energy forms: fossil, nuclear, and renewable (solar, biomass, hydro, and wind). The approximate total world production from these sources is shown in Table 2.2, together with estimated reserves and resources. Future levels of CO₂ emissions will depend on both the quantity and the mix of fuels burned. These two factors, in turn, will depend as much on the future structure of national economies as on the monetary value of total outputs. Globally, the quantities and mix of fuel consumed are determined partly by national trends in industrialization, urbanization, and consumer preferences.

Current global shares of primary energy supply are illustrated in Figure 2.1. For each form of primary energy, several technologies contribute to commercial and non-commercial energy supplies. Most commercial energy comes from burning fossil fuels, which are reviewed below with other supply technologies.

Table 2.2 Global primary energy

Figure 2.1 Global primary energy shares by source. Redrawn from Williams et al. (1984)

Figure 2.2 Comparisons of the release of carbon (in gigatonnes) for the production of 1 TWyr of energy from gas, oil, and coal (from Marland, 1982)

Fossil Fuels

In the combustion of any fossil fuel, carbon is oxidized and CO₂ is released to the atmosphere. When burned, different fuels produce varying amounts of CO₂ for a given release of thermal energy (see Figure 2.2). Excluding oil shale, coal is the most 'CO₂-intensive' fossil fuel, followed by oil and gas.^l

Recent estimates for proven reserves and total recoverable resources ²

¹ Refining and burning of oil shale produces up to 30% more CO₂ per exajoule released than does combustion of coal (depending on the process of refining used). Most of the excess comes from emissions in the refining process.
² Reserves represent the quantity of ores known to exist as a result of direct exploration, which can be extracted using available technology and economically exploited at current price levels.

Resources represent the total quantity of ore in the ground whose presence can be inferred from the results of direct exploration and may be recoverable in the future.all highly uncertainare shown in Table 2.2 for oil, coal, and natural gas. Estimated coal resources are considered in more detail in Table 2.3, which indicates that estimates of both reserves and resources have been continually revised upwards. In just six years, coal reserves 'increased' by 40 percent and total resources of coal went 'up' by 29 percent, largely because of rising prices. These revisions indicate the uncertainty surrounding estimates of reserves and resources. Moreover, not all known deposits have been extensively explored. Vast areasincluding much of Africa, Central America, eastern Siberia, and northern Chinahave seen comparatively little fossil-fuel exploration (Wood, 1983). Indeed, large deposits have recently been found in Colombia, Sumatra in Indonesia, and Botswana (Rose, 1984a). Such finds cast further doubt on the figures shown in Table 2.3 and suggest that the future of CO₂ emissions may not remain within the control of just a handful of nations.

Besides conventional fossil resources, unconventional formsnotably tar sands and shale oilshould be taken into account. Although the future development of these resources is uncertain, Canada has been exploiting the Athabasca tar sands since the 1960s, Venezuela is expected to develop the huge oil sands deposits in the Orinoco Basin, and the Soviet Union is beginning to develop its oil shale resources. Taken together, global exploitable resources of shale oil and tar sands equal an estimated 70 billion tons of oil (WEC, 1983), as Table 2.2 shows.

Table 2.3 Global coal resource estimates (billion tonnes coal equivalent, tce). From Rose et al. (1983)

Conventional oil and gas resources are limited, but will not be depleted by 2050. If the production of deep or geopressurized gas becomes economical, then gas reserves will increase considerably. Known reserves of coal are more than sufficient to meet expected energy demand for solid fuels well beyond 2050. Tar sands and oil shale represent supplies substantially larger than global resources of conventional oil and gas.

Non-fossil Supplies of Primary Energy

Nuclear energy, solar heat and photovoltaics (PV), hydropower, wind energy, and biomass can power numerous end-use applications of energy without net emissions of CO₂. If any of these sources (or geopressurized 'deep' gas deposits) were deployed widely and rapidly, future global demand for coal would be significantly affected. However, for various economic, technological, and environmental reasons, they provide a much smaller share of global energy today than conventional fossil fuels (see Figure 2.1). Both in industrialized and developing countries, most electricity is still produced by burning fossil fuels (see Figure 2.3).

 Figure 2.3 Global electricity supply by source. Redrawn from Williams et al. (1984)

The nuclear technologies that are economically competitive with fossil fuel systems employ nuclear fission in light-water and heavy-water reactors. Both consume uranium fuel in a 'once-through' conversion cycle. Commercial light-water reactors use artificially 'enriched' uranium while heavy-water reactors can use either natural or man-made isotopes as fuel. The most promising candidates for the future are fission breeder reactors (which can produce fissile plutonium from uranium fuel elements) and fusion reactors (which use natural or man-made isotopes of 'heavy' hydrogen for fuel). Environmental and safety problems in connection with nuclear power have been hotly debated for many years, including environmental contamination during routine disposal or storage of waste material, unauthorized diversion of fuel and the risks of catastrophic accidents. In addition, nuclear power and the proliferation of nuclear weapons are linked-though how strongly is also hotly debated. These issues add to the uncertainty about how widespread nuclear power is to become as a future source of electricity, though nuclear programmes are well under way in the United States, the Soviet Union, France, Great Britain, Japan, Korea, Taiwan, and several other countries. Because nuclear energy would substitute primarily for coal combustion in electrical generation, the problem of future CO₂ emissions to the atmosphere will of course be substantially reduced if nuclear energy becomes the world's main source of electricity.

Although the share of global electricity generated by fission reactors rose rapidly during the 1970s, the growth rate has tapered off recently. In many industrial countries, rising construction and running costs, long lead-times, unanticipated problems during routine operations, and safety-related environmental protection requirements have combined to make the rapid expansion of fission electric power systems seem less likely today than a decade ago. As for the longer term, how and how much commercial nuclear power will develop is difficult to anticipate.

Geothermal energy is another alternative to fossil fuel combustion in some applications. Because high-quality geothermal resources are limited to specific geographic areas, this energy form currently contributes less than one percent to the global energy supply. However, it does provide an alternative in a few locations. Dry steam plants are currently economical and being developed in, e.g., Italy, New Zealand, Mexico, and the United States. Geothermal resources are also being developed in the Soviet Union, Australia, Hawaii, Africa, and Europe. Economic and environmental considerations make it unlikely that, by the year 2000, geothermal energy will be as widely used as commercial nuclear power is now, even if so-called 'hot dry rock' resources prove to be economically attractive. Yet, this variant of geothermal technology, now being tested in prototype units, would allow the heat of deep strata in the bedrock to be recovered, thus making an essentially unlimited source of heat available.

Renewable energy has contributed to global energy supplies for millennia, and a number of those technologies may continue to make significant contributions also in the future. These include hydropower, wind, biomass and solar energy.

Hydropower resources are exploited throughout the world.. Recently, world hydroelectric generation has increased: by 1982, it accounted for approximately 14 percent of electricity generation (about equal to the nuclear contribution) (see Figure 2.3). Today, large dams are being built in, e.g., the Soviet Union, South America, and Canada, and thousands of small-scale hydropower plants (250 kW or less) are under development in other parts of the world. Considerable resources remain to be tapped worldwide, particularly in developing countries. Several industrial countries are now reaching their limits for environmental and economic exploitation of hydropower whereas resources are still plentiful in many developing regions, and could play a very significant role in the development process. The total realizable global resource of available hydropower sites has been estimated to be up to 1.5 TW (IIASA, 1981; Kristoferson, 1977; WEC, 1983).

Wind electric systems, both small scale and large scale, contribute only marginally to the energy supply today. The fastest growing markets for these systems are in California, Australia, and Scandinavia. Although development is still limited by economics and by the resource's geographic distribution, the global potential for recoverable wind energy has been estimated to be of the order of 1 TW (Rose et al., 1984; IIASA, 1981 ).

Biomass has historically contributed most to global energy supplies. Although seldom recorded in commercial transactions or in national income accounts in many developing countries, biomass use (primarily fuelwood and charcoal) accounts for more than 80 percent of annual energy consumption in many developing countries (O'Keefe et al., 1984). It also plays an important role in developed countries. For example, the use of wood for home and office heating is the fastest growing energy supply technology in the United States; in 1981, more than 15 million households used wood to meet some part of home heating demand. The combustion of biomass for industrial process heat and the cogeneration of steam and electricity is also expanding, for example in the United States, Brazil, Sweden, Jamaica, and the Philippines. Today, large quantities of biomass are converted to liquid fuel for transportation in Brazil. In the future, other countries may also enter into ethanol or methanol production on a large scale. Biomass's total potential as a sustainable energy source has been estimated to be about 5 TW (Rose et al., 1984; IIASA, 1981).

Whether biomass combustion adds to the global atmospheric CO₂ concentration depends on how the fuel cycle is managed. If feedstock production equals consumption, the carbon is merely being cycled between terrestrial biota and the atmosphere. However, if forests or other ecosystems are cleared or harvested more rapidly than they can be regrown, the standing biomass is, in effect, being 'mined', and CO₂ emissions from combustion would increase the carbon content of the atmosphere. Current and past rates of increase are matters of controversy. (See Chapter 3.)

Solar energy: although the direct conversion of solar energy to heat or electricity currently contributes negligible amounts to commercial energy consumption on a global scale, it shows promise for large scale applications for the future if economic and other problems can be overcome. In many regions, solar energy cannot currently compete economically with fossil fuels, and its major contribution is such decentralized uses as home heating. However, according to some analysts recent technological developments in photovoltaic power systems show promise for delivering bulk electricity at costs comparable with those of fossil and nuclear fuels by the turn of the century (see Maycock and Stirewalt, 1982; Green, 1983). Also by that time R & D efforts may have brought photovoltaic power systems to such a stage where they may be economically coupled to electrolyzers to produce hydrogen as a fuel for transportation. The timing or success of these systems cannot be predicted and is also subject to debate, but if successful, either could seriously challenge fossil fuel applications (Rose et al., 1983). In part because their environmental impacts are less severe than those of the technologies that they might replace, new or emerging renewable technologies could successfully displace fossil fuel use if their production costs continue to fall. Also a few other renewable energy technologies, such as ocean thermal energy conversion and wave or tidal power, show the theoretical potential of delivering significant amounts of energy in the future. Future contributions are however very hard to estimate, but are likely to be limited.

Over the long term, any one of several advanced technologies (e.g., deep gas, fusion, fission breeders, hot dry rock, or photovoltaics) may become economically attractive as a substitute for fossil-fired electricity production. In addition, improvements in the efficiency of industrial processes, residential and commercial building design, and transportation systems may greatly reduce the amount of fossil fuel consumed per unit of energy service. The 70 or more years time-scale applied in this study is enough to allow substantial technological innovations and changes in behaviour to occur. Thus, it is also necessary to be prepared for technological breakthroughs and substantial improvements, and policy-makers may have more room to manoeuvre than current practices and historical energy trends alone would suggest.

2.2.3 Historical Development of CO₂ Emissions

Since 1860, industrialization has progressed with an increase in the use of fuelsespecially fossil fuelsand a corresponding increase in CO₂ emissions. Keeling (1973) identified a striking increase in global CO₂ emissions from commercial sources of energy between 1860 and 1953 (Figure 2.4). Rotty and Masters (1984) reported that those results cannot be substantially improved. As Figure 2.5 shows, growth has been steady and exponential at 4.2 percent per year, interrupted only by the two World Wars and the Great Crash of 1929. According to Rotty and Masters, this growth 'should not be taken as indicative of the future' since it 'cannot continue indefinitely' (p. 6).

Figure 2.4 Annual carbon dioxide emissions resulting from fossil fuel combustion, 1860-1949. Redrawn from Rotty and Masters (1984)

Figure 2.5 Annual carbon dioxide emissions resulting from fossil fuel combustion, 1950-1982. Redrawn from Rotty and Masters (1984)

Figure 2.6 The changing patterns of global CO₂ emissions. Redrawn from Rotty and Masters (1984). A: North America; B: Western Europe; C: USSR, C.P. Europe; D: Japan, Australia; E: Developing; F: Centrally Planned Asia, USSR; G: Other

The detailed data collected and published each year since 1950 in the United Nations' Energy Statistics Yearbook show growth at about 4.4 percent annually unti11973, when the average growth rate dropped significantly. In the last decade, emission rates grew more slowly and then actually declined (see Figure 2.5). The global economic recession between 1979 and 1982 accounts largely for the decline in CO₂ emissions observed during this period. Because industrialized nations have become more energy-efficient and have turned in part to non-fossil fuel sources, CO₂ emissions have declined since their historic peak in 1979 (approximately 5.3 GT/year; Marland and Rotty, 1984).

On the other hand, developing nations have cut their growth rate in energy consumption only slightly. Thus, their share of global CO₂ emissions has increased. As Figure 2.6 illustrates, CO₂ emissions from Western industrialized countries have declined from 68 percent in 195.0 to 43 percent in 1980, while the share from developing countries has risen sharply. From 1960 to 1973, the rate of growth in commercial energy consumption was 7.3 percent per year in the developing countries, compared with 4.8 percent per year in the industrialized countries, and 4.3 percent per year in the centrally planned economies, Since 1973, growth of energy use has fallen dramatically to 0.5 percent per year in the industrialized countries (Darmstadter, 1984), with smaller reductions in the developing countries (6.2 percent per year) and centrally planned economies (4.0 percent per year).

2.3 UNCERTAINTIES IN FUTURE ENERGY USE AND CO₂ EMISSIONS

2.3.1 Introduction

The only way to project future CO₂ emissions is to estimate the annual amount of primary energy consumed and the mix of fuels used. In turn, this requires estimating economic growth and the dependence of energy consumption on economic activity. Fundamental and irresolvable uncertainty surrounds both the future events affecting energy use and the simulation of future energy systems. Scientists' understanding of the complex interaction of technological, cultural, and political forces that determine the development of national energy economies is limited and whether today's forces and relationships will persist largely unchanged for 50 to 75 years is unknown.

Often, investigations of uncertainty in modelling experiments and policy analysis are confined to structural relationships between variables perceived to influence the system significantly and to estimates of these key parameters at various times, as discussed below. In the exploration of possible outcomes in real systems, however, the element of ignorance itself must be considered. In the case of future estimates of CO₂ emissions, the structures of national economies and energy systems in the distant future are unknown and largely unpredictable. Too often, analysts' ignorance of these factors remains implicit and unacknowledged when short-term engineering and economic models are extrapolated into the more distant future, only to erupt as an unwelcome surprise when the real system is modelled.

As often noted, no existing model could have predicted the development of the global energy system in the 1970s, using data from the 1950s and 1960s. Even the best models could not predict the oil price shocks of 1973 and 1979. Indeed, even if modellers had been given advance notice of these events, the models probably would not have anticipated such other key systemic shifts as the slowdown in the development of commercial nuclear power at the very time it was expected to advance rapidly and edge out increasingly costly fossil fuels.

The presence of uncertainty about economic and political issues does not cripple the analysis of energy systems. Any single-value estimate or scenario of future energy use and CO₂ emissions is of limited value as a foundation for policy formulations. The prudent analyst must instead consider a range of feasible energy futures that might logically result from alternative policy choices.

With this caveat voiced, what are the uncertainties in demographic, economic, technological, and institutional factors that will affect the actual level of future energy demand, the mix of energy supplies consumed, and the associated rates of CO₂ emissions? In the following discussion, these first-order uncertainties are explored along with those characteristic of simulation models and projections of energy economic systems.

2.3.2 Factors Contributing to Uncertainty about Future Energy Use and CO₂ Emissions

Demographic Uncertainties

Today, most of the world scientific community agrees on rates of world population increase over the next century and generally accepts the United Nations' mid-range estimates of six-billion people by the year 2000 and ten-billion by the year 2100. However, Ausubel and Nordhaus (1983) still found estimates for rates of global population growth for the period 1975 to 2025 ranging from 1.2 to 1.7 percent per year, compared to the estimates of 1.6 percent between 1975 and 2000, and 0.5 percent between 2000 and 2100 used in the mid-range estimates of the United Nations.

Many global population estimates do not reflect the effects of regional social, and cultural issues and national population policies. For example, the current campaign for 'one family, one child' in the People's Republic of China has substantially lowered the birthrate in that nation, where nearly one-quarter of humanity lives. The age structure of a population also affects its patterns of energy consumption and its productivity. In countries where much of the population is under 15 or over 65, economic output per capita may be lower than in a region with an equal level of investment but a higher percentage of its population in the more productive 20-to-60 age group.

Urbanization also plays a role in per capita energy demand. In developing countries, urbanization can triple or quadruple per capita demand for primary energy, as charcoal is substituted for fuelwood (Goodman, 1983). However, when kerosene, liquid petroleum gas, electricity, or other high-quality fuels are substituted for traditional fuels, urban households able to afford these fuels can use as little as 30 percent of the energy required for cooking (the largest single household demand) as do their rural counterparts (Williams et al., 1984). Thus, while urbanization helps determine energy demand in developing countries, the direction and magnitude of the effect on energy demand depend not only on the technologies and fuels employed but also on the level of economic development and household buying power.

Global population trends are important for predicting future energy use, but they are not enough. Energy demand cannot be predicted simply by assuming that there will be mit X percent more people who will each use y amount more energy. The rate of population increase in any period interacts with, and may correlate with, the rate of GDP growth, but not necessarily with the rate of GDP growth per capita (Radetski, 1984). As discussed below, other factors contribute to uncertainty in estimates of economic growth.

Uncertainties in Future Economic Variables

The major analyses of future energy use and CO₂ emissions incorporate differing estimates of future rates of economic growth. Even when looking forward for only a decade and at only one or two factor prices, economic experts offer no consensus. To analyse changes in global climate, one must look at 50 to 100 yearsan impossibly long time frame for economic forecasts. When compounded over half a century, even a small difference in the estimated rate of economic growth has a significant effect on the global GDP projected for the final year of the period.

Manne and Schrattenholzer (1984a) have surveyed estimates for future global GDP and oil prices. Looking only 25 years ahead, they requested estimates from 70 governmental and international agencies, corporations, individuals, research institutes, and universities. Estimates of world GDP varied by 50 percent for the year 1990, by a factor of two for the year 2000, and by nearly a factor of four for 2010. Projections of the world price of crude petroleum (in constant economic units) varied by a factor of three for 1990, four in the year 2000, and five in 2010 (Manne and Schrattenholzer , 1984b ). Similar differences have been obtained in other reviews (see Ausubel and Nordhaus, 1983).

Many factors, some not easily quantifiable, contribute to the uncertainty of these estimates. These include:

1. Natural resource endowments.
2. International relations and stability, particularly East-West relations and North-South dynamics.
3. Consumer values and expectations.
4. Technological developments.
5. Annual investments. and the efficiency of national and international investment allocation.
6. The availability of infrastructure to coordinate and distribute goods and services.
7. Economic policies to provide counter-cyclical stimuli to growth, promote infrastructure development, and encourage specialization through international trade.

Cleveland et al. (1984) argue that increasing the thermodynamic quality of energy used in production would spur productivity and increase the GDP . International economic policies to promote stability, free trade and peace may be the most important determinant of the pattern of long-term energy use and CO₂ emissions over the next 50 to 100 years. Few, if any, of these factors can be easily incorporated into computer models.

Technological Factors Affecting Future CO₂ Emissions

The magnitude and composition of future energy demand will be influenced by four technological factors: the difficulty of substituting fossil for non-fossil fuels and vice versa, the rate of general productivity growth, the ease of substitution between energy and non-energy inputs to production (Nordhaus and Yohe, 1983), and trends in the real cost of producing fuel. In addition, technical breakthroughs could well occur between now and 2050 that would change the energy economy.

These four factors are the product of interactions involving four other factors: (1) incremental change in both energy supply technologies and the efficiency of energy use; (2) climatic change and other environmental constraints on the maximum mobilization of various energy technologies; (3) the market for new technologies; and (4) changes in the composition of energy demand.

Predicting the opportunities and timing of technological advances is a gambler's art. Edmonds and Reilly (1983a, see also Edmonds et al., 1984) estimate the aggregate rate of technological change in terms of a single parameter with values between zero and 1 percent per year. Using the same model, Rose et al., (1984) found higher rates of technological innovation to be plausible and consistent with other analyses. Such aggregated rates do not provide a precise description of the dynamics of change, but they are one way of representing the effects.

Environmental constraints on mobilizing energy technologies may take many forms. In some cases, energy activities will deplete limited local resources of high-quality ore, clean water, or the ability to absorb wastes and effluents. Major prolonged damage to lakes and forests from acid deposition, for example, could lead to public protest against a technology that seems risky.

Most analyses of future energy demand ignore the fact that climatic changes caused by a CO₂ buildup can alter energy supply-and-demand patterns. Analysing the results of two general circulation models for five urban sites representative of much of western Europe, Jäger (1984) found that doubling atmospheric CO₂ levels could cut the heating season by one to three months and increase the cooling season significantly. This suggests a decrease of up to 10 percent in the demand for space heating and a reduction of regional energy demand by as much as 2 to 3 percent.

Changes in regional climate can also affect energy supply. Quirk (1981) has shown that droughts can substantially reduce hydroelectric output. Jäger (1984) observed that climatic changes can also significantly affect supplies of both conventional and renewable energy.

An array of economic, institutional, and physical factors may affect the rate of market penetration of new energy technologies. Laurmann (1984), drawing on the work of Marchetti (Marchetti, 1980; Marchetti and Nakicenovic, 1979), states that past evidence on shifts in global primary energy supply implies a characteristic time of 50 years for new technologies to capture half of the market. Conditions for market penetration, however, differ among cultural and economic systems, local institutional and regulatory settings, and technological characteristics. Thus, use of a single-value estimate for all societies and time periods is inappropriate.

Another potential technological development that could affect the rate of CO₂ buildup is the introduction of a process to remove (or 'scrub') CO₂ from other exhaust gases after combustion. Once such a process were developed and its economics became attractive, there would still be a problem of where to dispose of the solid effluent (for example, as CaCO₃). To date, no economically efficient method has been identified and none is within immediate sight. In fact, most approaches use more energy than the original fuel releases. Some recent studies however suggest that the cost of such techniques may be coming down (see Steinberg et al., 1984). If technological breakthroughs occur, they could seriously alter the atmospheric impacts of fossil-fuel use, although the problems with other greenhouse gases than CO₂ would remain to be solved.

Social, Political, and Institutional Factors

Changes in cultural values, lifestyles, political structures, and institutional commitments could substantially affect energy use (and CO₂ emissions) over long periods of time. One of the factors to be taken into account is social resistance to changes in lifestyles and familiar technologies, as well as institutional, professional, and political resistance to innovations. In addition, inherent instability in natural and social systems could limit the range of predictability of future energy needs.

Policy choices regarding efficiency and conservation measures, accelerated development of specific technologies, strict environmental regulations (or absence thereof), or lack of such policies will determine which of many possible energy futures will actually be realized.

However, a theoretically feasible policy may be impossible to implement. Technologies with long lead times could be stopped by the considerable inertia in the global energy/economic system. In addition, a technology (e.g., nuclear power or hydropower sited in a sensitive area) that may look good to scientists, engineers, and economists may prove to be socially unacceptable.

On the other hand, infrastructural and institutional changes (or lack thereof) can enhance or inhibit rapid changes and shifts of emphasis. For example, a national policy to subsidize local biomass production, nuclear power and waste disposal or coal exploitation can make an economically marginal technology much more attractive in a country where it does not yet exist or is under development. Reasons for adopting such national policies may include job creation, national security or environmental considerations. Energy policies are thus often decided by factors completely outside the energy sector.

In addition to the uncertainty about future policy choices is that surrounding implementationan uncertainty that will vary with each policy and the evolution of institutional structures. Because policy implementation does help determine the future, uncertainty as to which future policies bearing on the development of energy technology will be implemented or what constraints on the permissible level of CO₂ emissions will be imposed matters greatly. Moreover, unlike the date of a future technological break-through, policy is one determinant of the future that people can controla self-evident fact that is too often ignored in modelling efforts.

Uncertainties in Models and Projections

The models built to simulate the evolution of the global energy system are themselves subject to uncertainty. Most studies acknowledge uncertainty as a difficulty. But only recently has it been featured prominently in global forecasting. In long-range forecasting, uncertainty grows with time. However, in energy- and economic-modelling, disagreement about the near future is frequently greater than about the distant future (Nordhaus and Yohe, 1983). Indeed, many of the most common assumptions about the future are the least analysedone reason why long-range uncertainties are so frequently underestimated. Authors can influence one another greatly, and when taken together, their 'agreement' creates an unjustified sense of certainty. In several cases, all forecasts turned out to be uniformly incorrect, especially in the energy field (see Greenberger et al., 1983). Moreover, projections that downplay uncertainty frequently carry a false sense of accuracy or inevitability. A model with lots of data and equations can appear to be analytically rigorous, even though most of the data and equations may reflect assumptions with little empirical or theoretical basis.

In forecasting, especially with the use of computer modelling, little attention is usually given to the importance of surprises or atypical events, even though the global energy system is likely to respond to these unpredictable episodes. It matters greatly, for instance, if prices triple in the first year of a twenty-year planning period (due to, say, a national disaster or a war), instead of tripling gradually over 20 years. When looking forward 50 to 75 years, it is worthwhile to return to 1910 and consider how well 1980 could have been forecast then, given the intervening world wars, the Great Depression, and the discovery of nuclear fission or computers among other things. Thus, although the evolution of social and institutional factors cannot be predicted, it must be acknowledged.

Parametric and Structural Uncertainty

In modelling, the two general categories of uncertainty are parametric and structural. Parametric uncertainty relates to such questions as what the GDP or energy demand will be in the future and refers to doubts about a variable's precise value. Structural uncertainty is more fundamental and subtle, involving relationships between variables and the way they affect one another. For example, what is the relationship between GDP and energy demand? Will the relationship hold for the next 50 years? Uncertainties of this kind relate to unknown or poorly understood behavioural and causal relationships between variables.

The concepts of parametric and structural uncertainty are well illustrated by a familiar example of energy economics: the energy/GDP coupling. For many years, aggregate energy consumption and GDP seemed to march in lock-step, and this was empirically well represented by the income elasticity of demand. But since 1973, the GDP grew but energy use did not. This 'breaking of the coefficient' or 'decoupling' has shifted attention from parametric uncertainties to the more fundamental structural issues, which are not well understood.

Handling Uncertainty

Methods for handling uncertainty in models usually focus on parametric uncertainties, including sensitivity analyses, statistics, stochastic processes, and such probabilistic techniques as 'Monte Carlo' simulations. These methods have proved to be powerful tools when carefully applied in energy and economic modelling. But few methods for handling structural uncertainty exist. Most models have one basic structure, which makes exploring structural uncertainty difficult. One solution is to build different models with different structures in order to test the impact of alternative mathematical constructs on the robustness of simulation results.

The growing emphasis on sensitivity analyses and probabilistic treatments is welcome, but not without pitfalls. In sensitivity analyses, the usual approach is to search for output variables that are either greatly affected, or largely unaffected, by varying input data. Such variables, if found, are commonly considered as representative of highly sensitive policy 'levers' or indicative of fundamental changes in structural relationships. But such conclusions should be drawn with caution since sensitivities of the model itself are one thing and the sensitivities of the natural system being modelled are another.

In building a complex dynamic simulation model incorporating many specified formal relationships, the model's overall behaviourhow the most important relationships and parameters interact to influence the outcome of simulation experiments-may not be well understood. Sensitivity analysis should reduce this uncertainty, but even a robust model can be a poor simulation if the correspondence between the behavioural characteristics of the model and real world characteristics is not verified.

Which historical data are the most relevant to test model validity? The analyst's choice may reflect only implicit beliefs and prejudices about the driving forces already built into the central structural relationships in the model. In such cases, validation becomes self-referential and circular, and more fundamental uncertainties about the real systems are concealed.

Often, the implicit assumption in modelling exercises is that uncertainty in analyses arises only from inaccurate measurement of the objective system that exists 'out there' in reality. Yet, the far more subtle, misleading, and unadmitted kind of uncertainty results from ignorance about the dynamics of the real system and from the analyst's undetected prejudices and implicit preconceptions. The impact of these types of uncertainty goes beyond those introduced by the choice of specific parameter values. In short, they introduce a 'hidden', unmanageable element of analytical brittleness.

The increasingly wide use of probabilistic techniques and of extensive sensitivity testing helps uncover and combat built-in biases, but surprises will inevitably occur. Thus, prudent public policy should be formulated on the basis of a range of feasible outcomes and developed using a variety of modelling tools and simulations that identify underlying assumptions, uncertainties, and biases.

2.3.3 Developing Countries-A Special Case

Some 75 percent of the world's population live in developing countries and earn about 20 percent of the global income (Goodman, 1983). Less than 10 percent of the world's industry is located in these regions. In most global energy studies, developing countries are considered 50 to 100 years 'behind' the industrialized countries. Some analysts also believe that the various aid programs operated by the industrialized countries are accelerating developing countries' efforts to 'catch up' and mimic the economies of industrialized states. These and other assumptions are often built into energy and economic models and forecasts, despite mounting evidence that existing aid programs are not always helping to build balanced, self-sustaining economies in developing countries (Cole, 1983). Leaving aside the question of whether developing countries should be measured with an industrialized countries' yardstick, the measurements and projections themselves have often been inaccurate. The economies of many developing countries are deteriorating, and their economic growth falls short of that assumed in early global energy studies.

In addition, in comparisons of energy systems in developing countries with those in the industrialized states, it is not always noted that in many developing countries the largest fractions of primary energy consumption are in the non-commercial sectors (fuelwood, crop residues, and dung, for example). Informal economic activities, or 'shadow economies', often contribute significantly to GNP. Continuing problems of hunger, disease, and inadequate shelter undermine economic growth. The need to address these issues simultaneously, and in an integrated fashion in the process of economic development, makes linear extrapolations of past trends in energy consumption of little value. New systematic methodologies must be developed to simulate the changing conditions in developing countries and to project appropriate patterns of energy use.

Variations in development potential among countries must also be recognized. Brazil, for example, is more likely to develop an economy that resembles that of the U.S. than that of Belgium. In 70 years, Ethiopia may look more like Spain or Portugal than like Britain or the Netherlands. No single model of economic development or energy consumption is well-matched to the range of circumstances facing developing countries.

2.3.4 Conclusions

A host of complex, interrelated factors combine to make it impossible to predict long-term future global energy and CO₂ developments precisely. Thus, if energy development policies are to take environmental and other long-term impacts into account, the full range of feasible outcomes must be explored and the impacts of specific policy choices evaluated in the light of the uncertainties inherent in the processes of formulating long-term projections of energy use and CO₂ emissions. Keeping in mind the dimensions of uncertainty discussed above, the next section presents a review and analysis of a range of recently published energy/economic projections and explores their implications for future CO₂ emissions.

2.4 REVIEW OF FUTURE ENERGY AND CO₂ PROJECTIONS

2.4.1 Introduction

This section provides a critical review of several global energy and CO₂ forecasts. Other reviews have appeared recently, notably those by Ausubel and Nordhaus ( 1983), Perry ( 1982), and Darmstadter ( 1984). Rather than duplicate these efforts, a complementary discussion and analysis is presented here. Earlier reviews have concentrated on final results from various studies and comparisons of these, giving only brief summaries of how the results were obtained. Here, the focus is on the methods employed to obtain results, with brief summaries of the results.

Many studies covered in previous reviews are omitted. Attention is focused on major and/or recent studies. To obtain a general picture of results in the field of energy/CO₂ projections, the reader is also urged to consult the earlier reviews (particularly Ausubel and Nordhaus, 1983).

2.4.2 Modelling and Forecasts

Any forecast is the result of some kind of model, even if only a mental model. In the last decade, formal models, particularly computer models, have been increasingly utilized for investigating the effects of alternative national and international policies. Models vary in complexity from simple accounting schemes to sophisticated systems of behavioural functions involving intricate feedback mechanisms. The process of building, testing, applying, and interpreting models needs to be clarified and demystified in order to make the strengths and limitations of the results more apparent and open for discussion and the models thus more useful as planning and scientific tools.

A model is a representation of a structural relationship or other underlying aspect of reality. In principle, the major advantages of mathematical models are that they require formal representation of the relationships between factors and they demand that assumptions and inputs be made explicit. This helps to control bias or at least make them more transparent. In addition, models permit investigation of different policy options and simulation of unforeseen developments. Thus, they can provide counterintuitive insights that might otherwise go unnoticed.

Characteristics of Models

All models have certain elements in common. Every model is a simplification of the real system it is designed to reflect. Values and prejudices are rarely explicit in the description of the model, but the world view and biases of the analyst often shapes the model structure, the assumptions made about underlying processes, and the range of 'plausible results'. Politics, for example, are a crucial element of energy systems the world over. But they are often ignored or only implicitly incorporated into formal models. Koreisha and Stobaugh (1979) point out that models 'are often modified by personal judgments to make the results correspond more closely to the specialist's understanding of the real world'.

Such modifications can take the form of selection of input data and parameter values in order to steer a model toward a desired result. In other cases, it may be accomplished by adjusting the internal workings of the model so that the input data are transformed in ways more believable to the analyst. This is sometimes regarded as the application of informed judgment in order to fertilize the sterile workings of the mathematical model. Provided that adjustments to the model structure are clearly and fully documented, this is not an improper practice.

An important aspect of modelling is that models are built for different purposes. In global energy forecasting, one researcher may be interested in the future role of nuclear power while another may focus on the interactions between the industrialized countries and the Third World. Both researchers will produce global energy models, but the resulting models will be different. This makes direct comparison difficult, if not meaningless.

There are a number of simplifying assumptions that characterize all models, and these are discussed briefly below.

1. Omissions: Regardless of the degree of sophistication or detail, a model can treat only a rather limited number of relationships. The implicit assumption made by the modeller is that the influence of excluded factors is unimportant. In fact, a model can sometimes mislead its users by inadvertently focusing attention on unimportant aspects of the real system while ignoring critical variables.
2. Approximations: Even a well understood relationship can often only be approximated in a model. This introduces an immediate source of error , the size of which is not always known.
3. Assumptions: When a relationship is not well understood, or is uncertain, or not easily quantified, then either it is omitted, or an assumption must be made to represent it. Such assumptions are often arbitrary and unverifiable.
4. Relevant range of variables: Data used to develop equations and forecasts from historical relationships are typically derived from experiences with one range of values. If the underlying behaviour and circumstances undergo little change, the relationships will probably remain constant. But if conditions change rapidly or radically (as happened to oil prices in 1973), the historic trends and variable ranges may have no relevance. Most analytic forecasting methods are applicable for very short time horizons at best (e.g., five years), and the validity of extending the time horizon to several decades is very questionable.
5. Reversibility: Many models assume that fundamental economic relationships are reversible. They assume that, for example, consumer behaviour patterns typical of a period of declining energy prices will continue in a period of rising prices. But there is no evidence that the relationships will remain true under different conditions.

Model Size

An important issue in model building is how much detail should be included in the structure. Large, disaggregated models offer considerable detail in their pictures of the energy/CO₂ system of the future; they are capable of incorporating a variety of technological, economic, and demographic features. Such projections can highlight the differences between technologies and the effects of changes in lifestyles and values. Regional detail can reflect differences in circumstances, policies, or attitudes.

Models of this type often require much detailed data and many assumptions regarding lifestyles and values. These data are often incomplete or unobtainable, and thus assumptions are often fabricated in an ad hoc manner. Systematic treatment of uncertainty is difficult with such models and accurate documentation is a mammoth task.

Small models offer complementary advantages and drawbacks. They are easier to understand, modify, and document; and they facilitate systematic treatment of uncertainty. Computational simplicity can lend transparency to the analysis. However, small models do not include much detail: important aspects of the real system may be lost in the aggregation, e.g., political and socioeconomic contrasts between developing and developed countries. Also, it can prove difficult to establish realistic interpretations for the data or variables.

Types of Future Projections

The simplest types of future projections are forecasts based on the extrapolation of historical data. The key mathematical relationships in these models are usually either linear functions or exponential growth curves. The relationships that produced the historical data are assumed to continue uninterrupted. Such forecasts cannot easily incorporate radical changes. Many of the projections of energy demand made before 1973 were forecasts of this type, and they were unable to anticipate the results of events in the last decade.

Some recent studies have used the scenario approach, a picture of an unfolding future resulting from specific assumptions. Scenarios usually incorporate both technological and economic elements, and may also address such issues as social values, capital accumulation, and infrastructure development. They offer a richness of description, but not necessarily a forecast.

Backcasts begin with descriptions of a desirable future and work backwards to identify specific actions or limits to action that might make that future possible. One example is an attempt to identify energy futures that hold global atmospheric CO₂ concentrations below some specified level (see Perry, 1982). Backcasts can be used, however, to provide important guidance in formulating policy on issues such as CO₂ buildup in the atmosphere which have uncertain long-term consequences.

2.4.3 Future Energy and CO₂ Projections

Several specific energy/CO₂ projections are now considered in detail. The principal results are summarized in Tables 2.4 and 2.5. The first table presents global primary energy projections for the year 2050; the second table gives the corresponding projections for CO₂ emissions. These results are illustrated graphically in this section's concluding discussion.

Edmonds and Reilly (1983a, b, 1984)

The Edmonds and Reilly (ER) model is a detailed partial equilibrium model, specifically designed to investigate long-term alternative energy policies and their implications for future CO₂ emissions. The world is disaggregated into nine geopolitical regions, as shown in Figure 2.7. A total of nine primary and four secondary energy categories are considered. The major inputs are assumptions about population and economic growth, supply and demand schedules for each type of fuel, and initial conditions. Interregional trade is incorporated into the market clearing calculations for all fuels except electricity. Demand for energy services is driven by population and GNP , the latter being calculated from the labour force and labour productivity.

Table 2.4 Projected primary energy in 2050 (in TWyr/yr)

Report	'Base case'	Range TWyr/yr	Page ref. in source
MIT (1983)		16.332.3	p.56
EPA (1983)	30.0a	Not given	p.44
ER (1983)	52.2	52.2	p.31
ER (1984)	28.4	1762.5	p.49
Goldemberg et al. (1984)	11.2(2020)	11.214.8	b
Lovings et al.(1981)	4.56	4.56	c
IIASA (1981)		22.4-35.7(2030)	p.522
IIASA (1983)	26.3	26.3	p.2d
Legasov et al.(1984)	42	42	e
NY (1983)		9.7109.3	f
Reister (1984)	29.9	24.041.3	p.21
WEC (1983)	29.8	18.929.8	g
a Estimated from Figure 4.1, p. 44.
b From Tables 9 and 14.
c Based on linear interpolation between 2030 and 2080; Krause ( 1983 ). Table 2.
d Based on linear extrapolation from 2030 to 2050.
c Estimated from Figure 3a, p. 1092
f Estimated from Figure 2.15, p. 136
g Calculated from Table 1, p. 252

The GNP is also affected by price via a constant elasticity feedback mechanism. Demand for energy services is converted into actual energy demand by a process involving interfuel substitution elasticities, and efficiency indices (reflecting technological change) which determine the coupling between energy demand and GNP.

The nine sources of primary energy supply are divided into three categories: constrained nonrenewable resources (conventional oil and gas), constrained renewable sources (hydroelectricity and biomass), and unconstrained sources (coal, unconventional oil and gas, nuclear, and solar). The cumulative extraction for resources in the first category is modelled by a logistic function of time (except for the Middle East, which is determined by OPEC policy, specified exogenously). Thus, the extraction rates for these resources are prescribed, and the total supply is offered without regard to market conditions. In the second category, hydroelectric power is also price insensitive, and is phased in by an exogenous logistic function (and exogenous price). Biomass is a price-sensitive resource. Finally, the five supply sources in the unconstrained category (called backstop technologies) are price sensitive, and contribute if production costs exceed exogenously specified 'breakthrough' prices. The total CO₂ emissions are calculated from the fossil supply projections by summing contributions from various source points in the conversion process.

Table 2.5 Projected CO₂ emissions in 2050 (present level is 5.0 Gt/yr)

Report	'Base case' emissions	Range (Gt/yr)	Page ref. in source	Approximate carbon intensity (Gt/TWyr)
MIT(1983)	15a	2.7-15	p.57	0.17-0.46
EPA (1983)	15	10-18	4-25
ER (1983)	26.3	15.7-26.3	p.41	0.50
ER (1984)	14.5	6.8-47.4	p.37	0.40-0.76
Goldemberg et al.
(1984)	4.6(2020)	4.6-5.9	b	0.40-0.43
Lovins et al.
(1981)	<1	<1(2030)	c	0.15
IIASA (1981)		10-17(2030)	d	0.45-0.48
IIASA (1983)	9.4	9.4	e	0.43
Legasov et al.
(1984)	13	13	f	0.31
NY (1983)	15	5-26g	p.94
Reister (1984)	10.7	9.7-27.1	p.30	0.29-0.66
WEC (1983)	14.4	10.0-14.4	h	0.48-0.53
a This assumes no CO2 abatement: it is not a best guess (Rose, 1984b).
b For 2020, calculated from Tables 6 and 11 and Note (11) in Williams et al. (1984).
c From Krause (1983) Figure 9, p.28.
d p.586.
e Approximate value shown in Rotty (1984), p.34.
f Estimated from Figure 3b, p.1092.
g These are the 5th and 95th percentile scenarios.
h Calculated from data on p.252, using carbon conversion factors in Gt/yr of 0.78,
0.63, and 0.48 for solid mineral fuels (SMF), oil, and gas, respectively.

The ER model has certain weaknesses. Reister (1984) reports that the supply function for coal is highly inelastic, thus influencing the equilibrium between supply and demand. His conclusion is that the model is not completely specified and that it needs a more elastic supply function for the backstop technologies. This suggests that the analytic behaviour of the model may be highly sensitive to the arbitrary specification of exogenous supply parameters, which is characteristic of many large-scale energy models. Other criticisms are that conventional oil and gas are not price sensitive, and that capital formation and depreciation are not incorporated. In addition, the ER model is inappropriate for detailed end-use analysis, because end-use efficiency is aggregated into a few parameters. Also, the major focus is on commercial energy technologies of highly developed countries. This makes proper analysis of the energy sectors in many developing countries difficult, since non-commercial energy (fuelwood and charcoal) accounts for a large part of energy consumption (up to 80 percent). Finally, because supply and demand are not determined by price, the appropriateness of using an equilibrium model for analysing centrally planned economies or other economies where governments intervene actively in energy markets can be questioned.

Figure 2.7 Typical geopolitical disaggregation in large energy/CO₂ models. Redrawn from Edmonds and Reilly ( 1983a). 1: USA, 2: OECD West, 3: OECD Asia, 4: Centrally planned Europe, 5: Centrally planned Asia, 6: Middle East, 7: Africa, 8: Latin Americ, 9: South and East Asia

Despite these limitations, the ER model is the only globally disaggregated model sufficiently well documented and tested to be useful. Considerable sensitivity testing has been performed and most researchers report no major vulnerabilities.

Edmonds and Reilly (1983a) developed a 'base case' scenario of the global energy future from 1980 to 2050. This is not a forecast, but a kind of 'best guess' scenario, incorporating what are thought to be likely developments. By the year 2050, global CO₂ emissions reach 26.3 Gt carbon per year (Gt/yr) as shown in Table 2.5. In several CO₂ taxation scenarios, a range of 15.7 to 26.3 Gt/yr is obtained. The conclusion is that the U.S. acting alone can do little to ameliorate the worldwide buildup of CO₂ and even a stringent global tax policy would delay a doubling of atmospheric CO₂ by only a decade or so.

More recently, Edmonds and Reilly (1984) presented a new set of three scenarios: a base case and two extreme cases that were 'developed by varying key parameters within the range of what are currently felt to be likely future values in order to generate extremes in carbon emissions'. In the base case, CO₂ emissions are 14.5 Gt/yr in 2050,45 percent lower than the 1983 base case. This substantial reduction results from the adjustment of unrealistically low coal prices in the earlier runs (Rose, 1984a).

The ER model has been used by several independent research groups to investigate the global CO₂ question in detail. These efforts are briefly reviewed below.

EPA (Seidel and Keyes, 1983)

The U.S. Environmental Protection Agency has applied the ER model to investigate 13 global scenarios to the year 2100. A 'mid-range baseline' scenario was presented that was 'believed to be representative of likely future conditions'. It presumed an annual 0.6 percent increase in end-use efficiency for non-OECD countries, and a corresponding increase of 1 percent/yr in the industrial sector of the OECD countries. Five other baseline and seven policy-specific scenarios are also investigated. The study focused on the effectiveness of reduced fossil fuel consumption in delaying global warming. Thus, one 'high fossil' scenario is included. The other scenarios considered fossil fuel consumption and CO₂ emissions below the midrange baseline scenario (see resulting carbon emissions in Table 2.5).

The baseline scenarios include, besides the midrange case, increased renewable and solar energy, nuclear energy, and reduced energy demand scenarios. For these cases, 'the variation in CO₂ emissions...appears small', ranging between 10 and 18 Gt/yr by 2050 (Table 2.5).These results are reported to be quite robust to moderate changes in GNP growth rates and income elasticity of demand.

The EPA study also included seven scenarios that simulated specific policies. Three scenarios include taxes on fuel use (resulting in CO₂ emissions between 12 and 14 Gt/yr by 2050); two involved bans on shale oil and syn-fuels. These five scenarios resulted in CO₂ emissions between about 10.9 and 14 Gt/yr. Two coal ban scenarios were considered ( with CO₂ emissions of about 4.7 and 9 Gt/yr in 2050). But these were judged economically and politically infeasible.

In conclusion, the 11 feasible scenarios investigated by the EPA, using the ER model, resulted in CO₂ emissions ranging from about 10 to 18 Gt/yr by the middle of the next century. This is between two and four times the current level of about 5 Gt/yr. The EPA analysis concluded that the timing of a two degree global warming is rather insensitive to feasible future energy policies, based on two factors: the feasible scenarios entailed at least a 100 percent increase (over today's values) in CO₂ emissions by 2050 and the assumption that other greenhouse gases will significantly contribute to global warming and that the size of this contribution is not sensitive to any of the energy policies discussed above.

Rose et al., (1983, 1984)

A group of researchers at the Massachusetts Institute of Technology (MIT) applied the ER model to study available energy options relevant to ameliorating the future buildup of CO₂. Results were reported in a detailed report (1983) and a summary journal article (1984). Accounting for technological, economic, and environmental opportunities and constraints, the study explored the energy options for holding down CO₂ emissions. Eleven scenarios were investigated, extending to the year 2050 incorporating increased end-use efficiency and other possibilities, such as increased fossil fuel prices, reduced resource bases, increased nuclear costs, reduced photovoltaic costs, a nuclear moratorium, higher oil prices, a cut-off of Middle East oil, and combinations of these possibilities. Assessments of major energy supply technologies were also provided, and analyses of electricity demand fluctuations and energy storage were provided. All scenarios investigated were judged to be feasible (no fuel bans were considered).

The MIT study differed from other studies that have employed the ER model. First, detailed calculations of major materials requirements are provided and the study investigated realistic possibilities for improved energy efficiency, and found that 'we are far from the limits of what can be achieved by more efficient energy usage'. Substantial evidence was provided to support the specific conclusions that:

'The effectiveness of energy use on a global scale can be increased by about 1 percent per year for decades without any social strain. This seemingly small figure leads to a halving of energy use by the year 2050 and a 50 percent reduction in CO₂ emissions. This result is quite independent of the effect on CO₂ of any shifts to non-fossil sources for primary energy supplies.' It is not yet clear whether the prevailing societal, institutional. and economic circumstances will favour this course of development.

By 2050, the range of CO₂ emissions in the 11 scenarios is from less than 3 to about 15 Gt/yr. Thus, 'the bounds appear to be fairly wide, with a spread of a factor of five in annual carbon emissions by the middle of the next century' (1983). The conclusion is that 'an option space exists in which the CO₂/climate problem is much ameliorated'. Note the discrepancy between this conclusion and the corresponding conclusion reached in the EPA study, and the fact that consideration of the effect of greenhouse gases has not been incorporated explicitly into the MIT analysis.

Reister (1984)

Another study using the ER model was that of Reister, in which two detailed scenarios were developed: a base case and a 'high carbon' case that focus on the contribution of gas (both natural and synthetic) to future CO₂ emissions. Four different gas options were considered, all of which increased gas demand and reduced electricity demand. However, these gas options were found to have minor effects on future CO₂ emissions (the combined effect of all four options is, at most, a 10% reduction in emissions). In addition, sensitivity analysis to variations in GNP in developing countries is included. Resulting CO₂ emissions in 2050 are 10.7 Gt/yr for the base case and 27.1 Gt/yr for the high carbon case.

Comparison of Studies Using the Edmonds-Reilly Model

The results from the ER, EPA, MIT, and Reister studies appear to differ significantly. All four employed the ER model, and so their results may be directly compared. (Since the time horizons differ, results are compared here for the most distant year common to all, 2050.) The results are shown in Table 2.5. The basis for these differing results is not the model or its detailed analytic structure, but rather the informal judgments that determined what the model inputs would be.

As an example, both the EPA and MIT studies look into possible ways to reduce CO₂ emissions. There is a difference of a factor of 3 in the effective lower bound projected for CO₂ emissions into the middle of the next century. Since today's emissions are approximately 5 Gt/yr, MIT finds it possible to postpone the timing of an effective doubling of atmospheric CO₂ emissions by 40% over the next 65 years, while EPA, using the same model, finds it infeasible to delay the doubling for any significant period. Thus, it becomes apparent that the subjective selection of input data sometimes plays a more important role in determining model results than does the internal analytical structure.

To investigate this a bit more thoroughly, it is of interest to examine some of the key input data used in the different studies that employed the ER model. Table 2.6 shows the annual growth rate assumptions for population and GNP (in %/yr). It is clear that in all cases, the same input assumptions were employed for the base case. Moreover, only one study investigated sensitivity to variations in these assumptions. Note in particular that the differences between the MIT and EPA studies are not explained by differences in population or GNP assumptions. To help explain these differences, Table 2.7 displays the range of input assumptions for technological efficiency improvements. The most interesting observation from this table is the wide variation in efficiency improvements in non-OECD countries. This is part of the reason for the different results obtained in the EPA and MIT studies (another principal reason is that the price of coal was much lower in the EPA study than in the MIT study).

Table 2.6 Growth rate assumptions for population and GNP

	Population (%/yr, 1975-2050)		GNP (%/yr, 1975-2050)
	Base	Range	Base	Range
MIT (1983)	1.0		2.6
EPA (1983)	1.0a		2.6a
ER (1984)	1.0b	0.7-1.3b	2.6b	2.1-3.4b
a Based on data from Seidel and Keyes (1983).
b Calculated from Table 2.11 in Edmonds et al. (1984)

Table 2.7 Assumed range of technical efficiency improvements (%/yr, 1975-2050)

The values chosen for key input data such as those shown in Tables 2.6 and 2.7 are crucial in determining model results, and yet this is not always clear in reports. The data in these tables reflect a host of behavioural and institutional assumptions that are represented by a few simple but key parameters. The different conclusions reached in different studies often boil down to basic differences in assumptions such as these that are in fact quite arbitrary.

Most energy/CO₂ studies provide little justification for the selection of input data. In this example, the principal differences between the EPA and MIT inputs relate to energy efficiency improvements: only the latter study presented analysis to support its choice of input data.

IIASA (1981,1983)

The Energy Systems Program at the International Institute for Applied Systems Analysis (IIASA) conducted probably the most ambitious global energy study ever undertaken (Perry , 1982), which is described in Energy in a Finite World (IIASA, 1981). There were at least three distinct energy and/or CO₂ projection efforts within the study: the focus here is on the most prominent, the IIASA global energy scenarios, which have been a benchmark in energy and CO₂ projections. Other modelling efforts within the IIASA program were reviewed in Ausubel and Nordhaus (1983).

The stated purpose of the IIASA study was to provide an objective analysis of the facts and conditions for any energy policy (Häfele, 1980). Two detailed global energy scenarios were presented, labelled 'high' and 'low', to span the conceivable evolutions of the global energy system from 1980 to 2030. The world was disaggregated into seven geopolitical regions similar to the ER regions (Figure 2.7).

The detailed documentation of the IIASA scenarios described an iterative procedure involving three computer models. The iteration began with population and economic growth projections and lifestyle parameters, in an energy demand model. The resulting projections of energy consumption (from 1980 to 2030) were then fed into an energy supply model that computed the least expensive supply strategy that met the specified consumption levels. This was fed into a model that calculated economic and environmental impacts. These variables were fed back to modify the original economic projections (closing the iterative model loop). The entire procedure was repeated until internal consistency was achieved (interfaces between models are not formalized). The IIASA models have been widely considered to be 'the closest existing approach to an appropriate disaggregated technique for forecasting CO₂ emissions' (Ausubel and Nordhaus, 1983). The CO₂ emissions in 2030 were about 10 Gt/yr in the low scenario and about 17 Gt/yr in the high scenario.

The IIASA scenarios have recently been criticized by Keepin and Wynne (1984) who argued that the iterative modelling procedure was never achieved in practice, and because the models were primarily an accounting framework for displaying the subjective projections of the analysts. Moreover , key results were found to depend strongly on uncertain cost and resource estimates. Sensitivity analysis showed that the models' analytic structure was unstable. Thus, although the models included many detailed aspects of the energy system, the variability of future energy costsan issue of overriding importancewas omitted. These findings bring into question the widely publicized policy recommendations drawn from the scenarios.

Recently, IIASA has produced a new scenario (1983), using an improved energy supply model, which foresees an expanded role for natural gas, and is otherwise similar to the 1981 low scenario. The resulting CO₂ emissions would be about 9.4 Gt/yr in 2050 (see Table 2.5).

Nordhaus and Yohe (1983)

Nordhaus and Yohe (NY) employed a compact global model to analyse CO₂ emissions from 1975 to 2100. The work was a principal component of a recent U.S. National Academy of Sciences study, Changing Climate. The principal focus was the systematic treatment of uncertainties in the overall energy/economic/CO₂ nexus. The model employed a rather generalized Cobb-Douglas production function, in which global GNP is basically a product of labour productivity, population, and energy consumption. There was no regional disaggregation. Energy was divided into two categories: fossil and non-fossil, which facilitated the calculation of CO₂ emissions and highlighted the substitution effects between the two. Demand for fossil and non-fossil energy was based on price. Calculations of CO₂ emissions were fed into a simple carbon cycle model to obtain atmospheric CO₂ concentration.

The unique feature of this work was the systematic treatment of uncertainties in the overall system. For this purpose, NY selected ten variables likely to greatly influence the outcome, but whose values were not well known. Each of the variables was assigned a low, medium, and high value, reflecting historical uncertainties or spreads in the published literature. This gave 3102 (or 59,049) possible scenarios. Results were then presented by randomly selecting 1000 different cases. This procedure was called probabilistic scenario analysis.

The results from the NY analysis were not presented simply as forecasts, but in percentiles. An example is given in Figure 2.8, showing the 5th, 25th, 50th, 75th, and 95th percentile scenarios of carbon emissions. The 50th percentile for carbon emissions in 2050 is 15 Gt/yr. The authors drew the following major conclusions:

'Given current knowledge, we find that the odds are even whether the doubling of carbon dioxide will occur in the period 2050-2100 or outside that period. We further find that it is a 1-in-4 possibility that CO₂ doubling will occur before 2050, and a 1-in-20 possibility that doubling will occur before 2035.'

These conclusions were derived from the statistical nature of their analysis. However, since all parameter variations were judgmental, this result is also judgmental. Moreover, a difficulty with probabalistic scenario analysis is that the global energy future is not the same as tossing a coin. Choosing among energy policies can significantly alter the probabilities of occurrence for particular energy futures. Thus, due to policy choices, reality is closer to having the opportunity to load the coin before tossing it, e.g., if we go all out for nuclear power, then a nuclear future is likely. Assigning percentiles to various scenarios implicitly assigns low probabilities to certain policy options which the world might choose to pursue vigorously. For example. nine feasible scenarios in the MIT study (Rose et al., 1983) yielded CO₂ emissions, in 2050,  ranging from 3 to 14 Gt/yr (corresponding to a CO₂ doubling time of up to 'several centuries'). Meanwhile, the NY analysis implies that the odds are 50 percent that emissions would be less than 15 Gt/yr (see Figure 2.8), However, if the world chooses full implementation of the MIT scenarios (particularly, the lower CO₂ emission cases) the chances are certainly greater than 50 percent that CO₂ emissions in 2050 would be less than 15 Gt/yr, assuming that the MIT study was not greatly in error. In fact, such policy decisions would shift all percentiles indicated in Figure 2.8,  downward. Other policy decisions could shift the percentiles upward. In either case, this would alter the conclusions, quoted in the preceding paragraph, regarding the doubling time for atmospheric CO₂ concentration.

 Figure 2.8 Carbon dioxide emissions (gigatons of carbon) from a sample of 100 randomly chosen runs. The 5th, 25th, 50th, 75th and 95th percentile runs for yearly emissions. with emissions for the years 2000, 2025, 2050, and 2100 indicated. Redrawn from Nordhaus and Yohe (1983)

The ten uncertain variables selected for detailed consideration in the NY work were tested for relative importance via sensitivity analysis. They were then ranked from one to ten, according to importance. A parameter representing the ease of substitution between fossil and non-fossil fuels turned out to be the most important uncertain variable. This was described as a surprising finding, because sensitivity to assumptions about substitution had not been previously noted. Such 'sensitivity ranking' is useful but it should be regarded with caution. Such conclusions implicitly require that uncertainties in all variables which have been omitted would not significantly alter the ranking. Furthermore, the ranking itself depends on the analytical structure of the model. As an example, two of the most sensitive parameters in the ER model were effectively excluded or constant in the NY model (see discussion in Section 2.5). It is possible that the sensitivity ranking in the NY analysis would have been significantly altered if these or other uncertain variables had been incorporated into the NY analysis.

Nordhaus and Yohe included a set of experiments involving five scenarios in which taxes on CO₂ were imposed (having five different dynamic variations as a function of time). These were compared with the 50th percentile scenario and the tentative conclusion was that forceful taxation policies would be required to significantly reduce future CO₂ emissions. This Conclusion is consistent with results obtained by both Edmonds and Reilly (1983a) and the EPA study.

In conclusion, the NY work represented an overdue step in energy/economic forecasting because it was the first systematic attempt to handle parametric uncertainties and address model validation issues. The fundamental message is that the range of possible futures is wide.

World Energy Conference (WEC, 1983)

The World Energy Conference (WEC) is an international organization of energy economists, planners, manufacturers, scientists, and politicians. Eighty countries are members and have played a significant role in shaping global energy policy and planning. A number of forecasts have been produced; the most recent (WEC, 1983) is considered here.

The objective of the study was to forecast primary energy balances in 2000 and 2020 for the globe, disaggregated into ten regions. These regions are roughly similar to those of other globally disaggregated studies (see Figure 2.7). A primary motivation for the study was to offer a decentralized alternative to the 'centralized forecasting models which have almost exclusively dominated the minds of researchers'. There was a separate regional working team (RWT) for each of the regions. Each RWT was composed of experts living in that region, and the RWTs were allowed freedom in formulating the regional forecasts. Fifty experts participated and a central team coordinated the effort.

The major advantage of this approach was that a single analytical framework, developed by researchers from a single cultural background, was not imposed on the entire globe. (This is especially advantageous for regions consisting largely of developing countries, which have traditionally been modelled using a framework appropriate for industrialized nations.) The challenge was to obtain results from all regions that were reasonably consistent and meaningful. To facilitate this, the central team provided each RWT with a homogeneous historical reference base for the years 1960 and 1978, a common system of units, and two general forecasting scenarios that defined a general framework. The two scenarios were characterized as optimistic ('rosy') and pessimistic ('grey') clearly reflecting the values and prejudices of the WEC authors. They were based in spirit on scenarios B and C from the Interfutures study (OECD, 1979). Each RWT was supplied with fixed input forecasts for population and economic growth. The principal results generated by each RWT included forecasts for primary energy demand, energy supply (from eight sources), total fossil imports required, and fossil exports available. Forecasts were cross-sections for the years 2000 and 2020. Only primary supplies were analysed. No analyses of useful energy, demand sectors, or secondary conversion were included. The global sum of imports and exports turned out to be roughly in balance without any need for iteration.

CO₂ emissions are not included as part of the study. But they can be calculated from the primary fossil fuel supply data in Table 2.4. The results are 14.4 Gt/yr for the 'rosy' and 10.0 Gt/yr in 2020 for the 'grey' scenarios.

The greatest strength of the study was probably also its greatest weakness: the knitting together of ten regional forecasts, prepared by different experts, using various methods. Although care was taken to ensure overall consistency, it is likely that significant inconsistencies remain embedded.

Goldemberg et al. (1984)

This study was a global end-use analysis, based on extrapolation from four national case studies: India, Sweden, Brazil, and the U.S. These nations have about one-fourth of the world's population, and represent a wide range of cultures, economies, politics, and geography. The study did not include forecasts, but presented a kind of 'existence proof' that outlined how population and economy could grow, while consuming less energy than historical trends would suggest.

With conservation and improvements in energy efficiency, the per capita demand for primary energy in the industrialized nations could be halved. Meanwhile, the per capita demand in developing countries would grow by only about 10 percent during the next 40 years, and the economic activity levels could approach that of Western Europe and Japan today. This surprising result was not presented as a target strategy but rather as a demonstration that 'dramatic improvements in living standards can be achieved without increasing per capita energy use'. A range of technologies were considered. Assumptions about technological performance, economic cost, and feasibility were conservative, rather than optimistic.

A base case was analysed to the year 2020, and various sensitivity analyses were performed. Primary energy consumption remains essentially constant at 10.3 TW in the base case (growing to 14.8 TW in a high demand case). The authors considered this outcome to be readily feasible. The problems ahead were not seen to be primarily technological or financial because all of the technologies now exist. The challenge would be largely that of information transfer and encouraging appropriate investments in capital goods.

The supply side incorporated a range of technologies and resources. In both scenarios, coal consumption declined about 20 percent from today's level, and natural gas consumption increased by 80 percent. Oil consumption declined by 30 percent in the base case (with CO₂ emissions of 4.6 Gt/yr) and grew by 17 percent in the high demand case (with CO₂ emissions of 5.9 Gt/yr; see Table 2.5). Nuclear power, which expanded almost fourfold in both scenarios, was regarded by Goldemberg et al. as a last resort because of the danger of proliferation of nuclear weapons, via recovery of plutonium or other weapons-usable material. In particular, low-cost uranium is sufficiently abundant that advanced technologies requiring fuel reprocessing (such as the breeder reactor) were not needed. Nevertheless, the share of electricity grows with time. Hydropower plays an increasingly important role, and wind and solar PV have a share of 5 percent by 2020.

The difficulty with this study is that it extrapolated from a few detailed case studies to a broader domain. The Soviet Union and Eastern Europe are treated as 'Sweden/U.S.-like' countries, in the sense that opportunities for efficiency improvements are equally large there (or perhaps even larger). Nevertheless, the four countries in the scenarios do span a broad range of socioeconomic and political differences, and the authors have diverse backgrounds representing four countries. Most other globally disaggregated models use a fixed framework for analysing all regions which is equivalent to extrapolation.

This study (along with that of Rose et al.) demonstrates that economic growth can be decoupled from growing energy consumption. In addition, the global energy future was analysed in the context of agriculture and food shortages, nuclear weapons proliferation, and the environment. As such, the study is unique among those reviewed here.

Lovins et al. (1981)

The starting point for this study was the population and economic growth projections for the seven world regions of IIASA's 'low' global scenario. Rather than attempting to forecast the future, the objective was to describe a future that minimizes the total energy consumption required to meet IIASA's population and economic growth projections. To implement the scenario outlined would require a vigorous shift of emphasis for global energy policy, from expanded supply capabilities toward technological improvements in efficiency and increased conservation. For this reason, the scenario is controversial. The authors stated that their analysis projects the least-cost energy future, and resolves the problem of CO₂ emissions by largely eliminating them.

The scenario is based on extrapolation from a case study for the Federal Republic of Germany (FRG), which is examined briefly (Krause, 1983). The core of the analysis was the application of 120 measures to improve energy efficiency and supply in 15 sectors of the West German economy. All technologies discussed exist today. The required investments were compared with the marginal costs of expanding supply within the present energy system. The conclusion was that the alternative scenario could supply the same services at lower cost. The measures would involve replacement, expansion and repair of capital stock, but not accelerated retirement.

In the household sector, space heating requirements for new buildings would be reduced with superinsulation and improved heating systems, including passive solar design. Existing buildings would be retrofitted with superinsulation. The newest electrical appliances and light bulbs would be introduced (up to three or four times more efficient than older ones).

In the commercial sector, office buildings would use existing state-of-the- art technologies. There are Canadian and northern European buildings now which use only 20 to 30 percent of the energy consumed in conventional buildings.

In the transport sector, super efficient automobiles and trucks would be introduced. Trains and airliners would be replaced and retrofitted with more efficient technology.

Major gains in efficiency would be achieved in the steel, chemicals, cement, and aluminium industries. Electric drive technology would be improved, waste heat utilized, and materials used more efficiently.

Taken together, these measures would replace about 70 percent of end-use energy in all sectors combined, with a shift toward greater shares for electricity and liquid fuels. Total energy consumption is reduced to 30 percent of the 1973 level for the FRG.

These measures would be introduced gradually, cutting global energy projections for 2030 in the IIASA low scenario by approximately 50 percent. This would be further reduced by assumed economic structural changes, resulting in 5.23 TW primary energy consumption in 2030 (about 25 percent of IIASA's projection of 24 TW), with CO₂ emissions and acid rain greatly reduced as a consequence.

The validity of such an extrapolation from a single nation to the entire globe is problematical, as discussed above for Goldemberg et al. (1984). In particular, there is little basis for assuming that the economies of the developing countries will evolve to look like the FRG. Furthermore, details on capital and labour requirements were not included. Finally, many institutional changes not considered in the scenario would be required.

The scenario was not presented as a forecast, but as another 'existence proof'. In view of the last 70 years, the potential for tremendous change over a long time horizon should not be underestimated. Even if it were possible to only partially achieve the scenarios described, it indicates a future global energy strategy that would permit continued economic growth along with reduced environmental strain.

Legasov et al. (1984)

This study is the latest of several Soviet analyses of the energy/climate nexus. In an earlier study, Legasov and Kuzmin ( 1981) analysed two long- term global scenarios to study the effects of waste heat discharge. Annual per capita energy consumption was represented by a logistic function of time, whose parameters were estimated, based on historical data.

More recently, Kuzmin et al. (1984) have investigated various CO₂ abatement strategies, concluding that a whole complex of feedback relations must be accounted for in technological risk assessments.

Legasov et al. present two scenarios in which global energy consumption reaches 6 kW/capita and 20 kW/capita, respectively, by 2100. The lower of these scenarios (called the 'minimal variant') is used for the detailed analysis of CO₂ emissions, and thus we focus on this case here.

Assuming a global population of 10 billion by 2100, the minimal variant results in global energy consumption of 60 TW by 2100. This scenario is elaborated in Stirikovich et al. (1981), and leads to approximately 42 TW by 2050. Fossil fuel consumption peaks at less than 20 TW before 2050, and slowly tapers off thereafter (Legasov et al., 1984). Resulting CO₂ emissions follow a bell-shaped curve, reaching a peak of approximately 13.3 Gt/yr in 2050 (and decreasing thereafter), as shown in Table 2.5. The study proceeds to analyse a variety of broader climatic implications (including effects of atmospheric electroconductivity changes due to krypton emissions from nuclear fuel cycle plants).

Discussion

It is clear after reviewing these projections that the global energy/CO₂ future is highly uncertain (for a more detailed review, see Keepin, 1984b). It is important to note that even if future GDP or energy consumption were known with certainty, future CO₂ emissions would still exhibit a range of possible values. To illustrate this, rough calculations of 'carbon intensity' are given for most of the studies in the last column of Table 2.5. These numbers were obtained by computing the ratio of CO₂ emissions to primary energy consumption. Thus, a low value means a relatively 'clean' energy supply system in regard to CO₂; a higher value means a more fossil-based supply system. These ratios are only intended to illustrate that CO₂ emissions can be decoupled from energy consumption. (They are not appropriate for judging the merits of individual projections.)

Figure 2.9 Projections of primary energy consumption in 2050

Figure 2.10 Projections of carbon emissions in 2050. Sources: EPA: Environmental Protection Agency (1983), ER: Edmonds and Reilly (1984), Goldemberg (1984), IIASA: International Institute for Applied Systems Analysis (1981,1983), Legasov et al. (1984), Lovins et al. (1981), MIT: Massachusetts Institute of Technology (1983), NY: Nordhaus and Yohe (1983), Reister (1984), and WEC: World Energy Conference (1983)

Results from the various projections are illustrated in Figures 2.9 and 2.10, showing projections for global primary energy and CO₂ emissions. These figures present various point estimates only, omitting specific trajectories. This unusual format was adopted because the uncertainties in primary energy and CO₂ emissions for the year 2050 are much greater than uncertainties associated with specific trajectories (see Section 2.5 for further discussion).

The uncertainty exhibited in Figures 2.9 and 2.10 is more fundamental than simple lack of knowledge of future parameter values. Indeed, as discussed in Section 2.3, the large variation in future projections is due to a general ignorance about future cultural, political, social, and technological developments. These factors are the most important determinants of the future, but they cannot be anticipated by analytic forecasting methods. Thus specific projections can only focus on second- and third-order effects, and it is unavoidable ignorance of the first-order effects that gives rise to the great uncertainty observed in Figures 2.9 and 2.10.

2.5 BOUNDS ON FUTURE CO₂ EMISSIONS

2.5.1 Introduction

Future anthropogenic CO₂ emission rates are highly uncertain, primarily because they result from a complex set of interactions among a number of uncertain sometimes unknown factors. Thus, it is natural to question whether it is possible to identify upper and lower bounds on likely future CO₂ emissions. A word of caution is needed regarding the term 'bounds'. It is clearly not possible to specify actual upper and lower bounds within which future CO₂ emissions will lie. The term 'bound' is used here for lack of a better word, and it is used informally to suggest an interval or range of projections, outside of which conditions and scenarios are regarded as extreme.

2.5.2 General Approach

No attempt is made here to produce new projections. In fact, the many uncertainties in existing estimates are of greater significance than the individual projections themselves. Rather than make additions to the existing array of forecasts, we attempt to establish extremes for the range of uncertainties in future carbon emissions. The approach will be to consider the past projections and identify tentative upper and lower 'bounds' on future CO₂ emissions. This does not mean that two scenarios will be singled out and labelled 'upper' and 'lower' limits. On the contrary, each bound corresponds to more than one scenario, because there are a multitude of ways to achieve a given level of CO₂ emissions. Moreover, we do not specify a quantitative measure of confidence that future CO₂ emissions will indeed fall within the bounds identified here. To do so would suggest that greater precision inheres in this process than can be justified.

2.5.3 Comparative Assessment of Future CO₂ Projections

To help explain the choice of what seem to be reasonable bounds offered in this section, a comparative assessment and discussion of the projections and forecasts will be reviewed. The purpose is to illustrate why the results of various studies differ and why we should (or should not) believe the results.

Nordhaus and Yohe (1983); Edmonds and Reilly (1983b, 1984)

A significant difference in results between the NY and ER models concerns the sensitivity to the ease of substitution between fossil and nonfossil fuels. Each model has a parameter which represents this factor, the most sensitive parameter in the NY model. The authors reported the significance of this factor to be important and unexpected. Since the NY model estimated the CO₂ concentration with a simple atmospheric accumulation model, this finding implies that the CO₂ emissions themselves are sensitive to variations in the substitution parameter. In the ER study, however, 'carbon emissions are very insensitive to changes in the interfuel substitution parameter' (Reilly and Edmonds, 1984b). Thus, opposite conclusions regarding the importance of this parameter are reached, and it is natural to ask why. Nordhaus and Yohe observed sensitivity to the substitution parameter because it entered their model as an exponent that heavily influences the relative demand for fossil and nonfossil fuels. In the ER model, 'fossil and nonfossil fuel prices are relatively close together and therefore the substitution parameter is less important' (Reilly and Edmonds, 1984b). Thus, one model is sensitive to variations in a particular parameter, while the other is not. Either structural representation is plausible. The point here is not to decide which is 'correct', but to emphasize the importance of inherent structural uncertainties that characterize all energy and CO₂ models. These differences in sensitivity suggest the importance of model validation exercises that attempt to establish a correspondence between model and reality. Behavioural patterns suggested in model runs may either provide genuine insights concerning the underlying systems, or they may be simply artifacts of the model (Keepin, 1984b).

Given the observed lack of sensitivity in the ER model to the most sensitive parameter in the NY model, the question can be reversed: what is the most sensitive parameter in the ER model and what is its effect in the NY model? The two most sensitive parameters in the ER model are income elasticity of demand and the growth rate of non-price-induced efficiency improvements. However, the former of these parameters is explicitly represented in the NY model. Thus, there is no way to assess the importance of these parameters in the NY analysis. This illustrates another problem: one model effectively holds constant a variable whose variation is critical in another model.

When the same conclusion is reached independently, by different researchers, employing different analytic approaches or methods, such a conclusion deserves careful scrutiny. This is the case with the ER and NY models on the issue of fuel tax policies for reducing CO₂ emissions. Both models indicate that such policies would have to be extreme in order to have a significant effect.

IIASA (1981) and Lovins et al. (1981)

Various independent researchers have compared the IlASA and Lovins et al. projections (Rose et al., 1983; Meinl et al., 1984; Wynne, 1984). Robinson (1982b ) finds that the two studies began from quite dissimilar conceptual frameworks, and that the conclusions of each study are rooted in fundamental presuppositions that do not result from the quantitative analysis, but rather guide it. Thus, the 'driving force is either an explicit or implicit view of what constitutes a desirable energy future' (Rose et al., 1983). Lovins et al. are explicit about this, whereas IIASA claims to be objective, particularly because of its use of sophisticated computer models. Robinson ( 1982b ) reported that each study labels the other's scenario as highly undesirable and unaffordable, and that policymakers have tended to ignore these mutual condemnations, while the policies themselves have slowly shifted toward conservation and more efficient end uses.

Gotdemberg et al. (1984) and MIT (Rose et al., 1983)

These two studies reach similar conclusions. As Darmstadter (1984) has observed, 'few pairs of energy analyses are ever totally free of incest'. However, the analysts were unaware of the others' work. They employ different analytic and philosophical approaches although their authors shared the view that the possible consequences of conservation on the greenhouse problem were worthy of systematic exploration. Goldemberg et al. provided a detailed end-use analysis, focusing on low energy futures that would incorporate a variety of energy-saving measures, and technological improvements. The analysis focused on the demand side, included detailed treatment of developing countries, and embedded the energy system in a global context that included major agricultural, environmental, and political factors. The MIT study employed a classical economic equilibrium model that focused on the supply side (commercial technologies), is best suited to developed market economies, and treated the energy sector in isolation. Thus, the two studies may be considered 'analytically orthogonal'. That is, although using very different methods and assumptions, the two studies reach very similar conclusions about the feasibility and implications of a low-demand end-state.

An important conclusion common to both Goldemberg et al. and MIT is that over the next several decades, CO₂ emissions can be kept at or even substantially below today's levels (see Figure 2.10). Both studies find considerable room for improvement in overall energy efficiency, particularly in the developed countries, and recommend that policies be formulated to that end.

Bounds on Future CO₂ Emissions

For the purpose of specifying bounds on the range of future CO₂ emissions, the focus here is on the end point of the time horizon considered in this study: the year 2050. A discussion of emission trajectories to the year 2100 is also included.

In the attempt to identify bounds on future CO₂ emissions, it is necessary to consider the feasibility of different futures. The notion of feasibility occurs frequently in analyses that assess future possibilities, opportunities, and consequences of policy options. Although 'feasibility' appears to be a simple concept, it actually encompasses a broad range of technical, economic, environmental, sociopolitical, and even cultural factors. Thus a truly feasible future is one that is feasible simultaneously in each of several dimensions.

In this work the focus is largely on technical, economic, and environmental feasibility, which can be analysed quantitatively to some extent. However , consideration is also given to subjective and human dimensions that are not quantifiable but can limit what is technically, economically, and environmentally feasible.

To estimate an upper bound for CO₂ emissions in 2050, several different studies are relied upon. High fossil fuel projections are first considered, especially coal projections, since coal is the most 'CO₂-intensive' of the conventional fossil fuels and also, by far, the most abundant. A typical projection is the IIASA ( 1981) high scenario, which projected a five-fold expansion in global coal extraction by 2030. Another example is the world coal study (WOCOL, 1980) which produced coal projections to the year 2000 roughly equivalent to IIASA. Although the IIASA and WOCOL scenarios do not extend to the year 2050, they are of particular interest because they are among the few that have been studied by independent researchers to determine overall feasibility. Alcamo ( 1984a,b ) has carried out a detailed analysis of the coal projections in the IIASA high scenario. He found that for the U.S. and the USSR (which have some of the largest known coal reserves), the projected four- or five-fold expansion is probably not feasible, due to insufficient coal reserves and lack of water. Chadwick and Lindman (1982) provided a detailed investigation of the environmental implications of expanded coal utilization, based on the IIASA coal projections. Perry (1982) investigated the CO₂ implications of the IlASA/WOCOL scenarios and found emissions to be extremely high, due to the huge coal projections. Prior (1982) performed a 'bottom-up' analysis of future coal consumption in Western Europe by considering each country's actual and likely performance in converting industrial and power plants to coal firing. He concluded that demand for coal will fall considerably below the WOCOL and IIASA levels. Long ( 1982) investigated constraints on future coal demand relating to the development of port facilities and shipping terminals, comparative economics of competing coal-exporting regions, and resource distribution issues. His analysis supported the conclusions of Prior, that future coal production will be below the IIASA and WOCOL levels.

More recently, Fischer ( 1984) has found that observed increases in coal demand have been only 40 percent of that predicted in many forecasts. He suggests that this is due to unrealistic assumptions about price differentials between oil and coal, and the ease of substitution (e.g., failure to appreciate the effects of conservation and the slowdown in the world economy since 1973; failure to appreciate the strength of the 'antisocial factor' associated with coal).

Since coal is the most 'CO₂-intensive' fossil fuel (and the IIASA high scenario projects major expansions for conventional oil, shale oil, and natural gas), these findings are interpreted to mean that the IIASA high scenario (and others equivalent to it) was unrealistic. Consequently, the upper bound for carbon emissions in this study will be below these levels. (In Table 2.5, the IIASA high scenario shows carbon emissions of approximately 22.3 Gt/year by 2050, based on a linear extrapolation from 2030.)

For equivalent levels of consumption, high coal scenarios would mean greater carbon emission levels. Several studies have explored such base case scenarios which represent the extrapolation of current trends, without policies to restrict carbon emissions. The results are shown in the final column of Table 2.5 and in Figure 2.10. Note that the projected emissions are at or below 15 Gt/yr in all base cases except one (Edmonds and Reilly, 1983b). (This last case has recently been replaced by the authors with a new base case scenario (Edmonds et al., 1984) that yielded CO₂ emissions of 14.5 Gt/yr by 2050.) Thus, 15 Gt/yr appears to be a plausible upper bound. EPA (1983) explored a 'high fossil' scenario that resulted in approximately 18 Gt/yr by 2050.

Based on these considerations, an upper bound for carbon emissions in 2050 of approximately 20 Gt/yr was selected. Although emissions could exceed this value, it appears to be unlikely. A few studies do include estimates above 20 Gt/yr, but none of these have analysed the realism of such projections. For example, Edmonds et al. (1984) included an unusual scenario in which CO₂ emissions reach 47.4 Gt/yr by 2050, increasing to 91.1 Gt/yr by 2075. This scenario incorporated forecasts of future coal consumption that are higher than the IIASA high scenario (judged to be unrealistic), in addition to a shale oil program of comparable magnitude. While this scenario is perhaps technically feasible, no analysis of the other implications is provided. The scenario appears unrealistic in view of the studies cited above (Long, 1982; Prior, 1982; Alcamo, 1984a,b; Perry, 1982; Chadwick and Lindman, 1982). In fact, the upper bound identified here corresponds more closely to the Edmonds and Reilly (1984) base case.

In the statistical study by Nordhaus and Yohe ( 1983), the 75th and 95th percentile scenarios yielded 17 and 26 Gt/yr by 2050. The NY study was primarily an aggregated statistical analysis rather than an investigation of feasibility. For the latter purpose, a more disaggregated approach would be more appropriate. In terms of the IIASA scenarios, the upper bound chosen here lies between the high and low scenarios (1981), and well above the 1983 scenario.

To estimate a lower bound for CO₂ emissions in 2050, the work of Goldemberg et al. (1984), MIT (Rose et al., 1983), and Lovins et al. (1981) is used. The first two studies reached similar conclusions regarding the extent to which CO₂ emissions can be limited. This consistency is particularly significant because they utilize such different analytic approaches and tools. The lowest figure obtained by MIT is just under 3 Gt/yr for 2050. This is less than half the lowest value, of 6.8 Gt/yr, obtained by Edmonds and Reilly (1984), using the same model. However, MIT developed three different scenarios using the ER model that resulted in lower CO₂ emissions than the Edmonds and Reilly (1984) low case, and also included additional analysis to il1ustrate overall feasibility. Thus, the lower bound obtained by MIT will be used here. This scenario is characterized by a combination of efficiency improvements and large non-fossil energy shares (nuclear 29%, hydro 19%, and solar 12%). This results in CO₂ emissions of 2.7 Gt/yr in 2050. Meanwhile, Goldemberg et al. obtained about 4.3 Gt/yr in 2020 for their base case. (The trajectory between 1980 and 2020 is flat. Thus, it is assumed that 4.3 Gt/yr is maintained through 2050.)

Lovins et al. (1981) found that carbon emissions could be less than 1 Gt/yr by 2030. Recall that this scenario was offered as an 'existence proof' rather than a forecast. While this scenario may be technically feasible, it seems unlikely that global energy/economic policy will be sufficiently stringent and well-coordinated for this scenario to actually be realized. But it suggests that concerted effort could greatly reduce carbon emissions. Based on these considerations, a lower bound for 2050 of approximately 2 Gt/yr was selected.

The upper and lower bounds specified here yield an estimated uncertainty factor of ten for carbon emissions by the middle of the next century. If we include the effects of biological CO₂ sources (which could add another 1 to 2 Gt/yr), the lower bound could increase to 3 Gt/year and the upper bound extend to 22 Gt/yr. The higher emission levels are associated with relatively high fossil fuel consumption. The lower levels are associated with a broader range of possibilities that include displacement of energy demand (conservation, efficiency improvements, etc.) and/or vigorous development of CO₂-benign supply sources (renewable, nuclear, etc.).

The upper bound corresponds to an annual growth rate for carbon emissions of 2.3 percent/yr, which is considerably greater than the average of 1.46 percent/yr since 1973. The corresponding growth rate for the lower bound is 1.4 percent/yr. However, these figures presume an exponential model (either growth or decay) which is not realistic for CO₂ emissions, because it suggests that emissions could increase to infinity or decrease to zero, making the model highly sensitive to the value of the growth rate parameter.

In choosing upper and lower bounds, it is not meant to suggest that a 'best guess' for the future lies halfway between. Choosing a median or best guess scenario is intentionally avoided here. A stringent policy to reduce CO₂ emissions would yield a lower outcome for future carbon emissions than a policy that ignores the greenhouse effect. Since it is not possible to predict policy, it is impossible to make an authoritative 'best guess' for future CO₂ emissions. The prerequisites for arriving at different end-values for CO₂ emissions within the upper and lower bounds will be briefly discussed in the last section of this chapter.

What is actually needed for analysis of future global warming is not a best guess of future CO₂ emissions, but the resulting concentration of CO₂ in the atmosphere. To estimate the latter requires analysis of the global carbon cycle, which is considered in Chapter 3. Using the carbon cycle model developed by Siegenthaler (1983), an idea of the bounds on atmospheric CO₂ concentration that would result from our selected upper and lower bounds on CO₂ emissions has been produced. The resulting atmospheric CO₂ concentrations in 2050 are 531 and 367 ppmv for the upper and lower bounds, respectively (Svenningsson, 1985). These reflect increases over the pre-industrial level (270 ppmv) by factors of 2.0 and 1.4. These results assume a contribution of CO₂ from the terrestrial biota of 1.2 Gt/yr in 1980, decreasing linearly to zero in 2055.

When assessing projections, it is prudent to try to bound the greatest uncertainties first. Thus, while there is more uncertainty related to different emission trajectories, it is of secondary importance compared with the greater uncertainty as to the actual level of CO₂ emissions in the middle of the next century. For this reason, bounding the latter, within a range that seems reasonable, has been the primary focus of the efforts described in this study. Moreover, there is probably no justification for placing confidence in any particular shape of trajectories of long-term energy consumption or carbon emissions. For this reason, the projections reviewed here are represented with points only in Figures 2.9 and 2.10. The issue is considered again in Chapter 3 in connection with limiting the atmospheric concentration of CO₂.

Conclusions

Future emission rates of CO₂ resulting from fossil fuel combustion are highly uncertain, especially when looking into the distant future. Based on a critical examination of the literature and consultations with experts, tentative upper and lower 'bounds' on these emission rates for the year 2050 were selected. These bounds are not rigorous limits, but rather scenarios that are judged to be extreme. The upper bound is 20 Gt/yr, which was arrived at by considering several recent studies of constraints on future coal consumption. The lower bound is 2 Gt/yr, which was obtained from consideration of various recent analyses of the future opportunities for efficiency improvements and non-fossil supply technologies. Both the upper and lower bounds reflect extreme cases that are judged to be very difficult to achieve in practice. In addition, biotic sources of CO₂ could contribute up to 1 or 2 Gt/yr, resulting in a range of 2-22 Gt/yr for total carbon emissions by 2050. It is possible that the actual emissions will fall outside this range, but this is judged to be unlikely.

The 'envelope' of uncertainty identified here is illustrated graphically with dotted lines in Figure 2.10. This wide range of uncertainty is not likely to be significantly narrowed by further research, because this envelope signifies a basic ignorance about future cultural evolution. Thus any particular trajectory implies some combination of political, technological, and social developments that cannot be predicted.

2.6 CONCLUSIONS

Several conclusions emerge from this analysis. First, future emissions of carbon dioxide (CO₂) cannot be predicted with precision. Attempts to forecast energy and CO₂ developments into the distant future are hindered by great uncertainties. The precision and elegance of analytic forecasting methods are overwhelmed by the arbitrary and imprecise nature of the structural assumptions and parameters. For this reason, recent forecasts have generally included a wide range of possible outcomes. It is estimated that CO₂ emissions from fossil fuel combustion in the year 2050 will be between 2 and 20 Gt/yr (compared with today's level of 5 Gt/yr). Coupled with an additional contribution of upto 1 or 2 Gt/yr from terrestrial biota and other sources, the resulting uncertainty factor in CO₂ emissions by the middle of the next century exceeds an order of magnitude.

This does not imply that nothing meaningful can be said about the future, however. One reason for uncertainty is that the energy/CO₂ future depends, in significant measure, on present and future policies. Recent studies have shown that future economic growth can be largely decoupled from growing energy consumption, at least in the industrialized countries, and that future CO₂ emissions can be decoupled in industrialized countries from both economic growth and energy consumption. Thus, the causal relationship between energy and climate is weaker than has hitherto been believed.

Perhaps the most important conclusion in the more recent published studies is that global energy productivity and efficiency is far from optimal. Those studies which have not explicitly considered the 'efficiency factor' are without practical interest. Numerous opportunities exist for improving the efficiency of energy use, and are economically attractive. Global energy efficiency probably can be increased at a sustained rate of at least 1 percent/year for several decades with no adverse social effects. By 2050, this would cut energy consumption and CO₂ emissions in half from what they would otherwise be. In addition, the global energy supply mix may well move away from fossil fuels (especially coal) and toward CO₂-benign technologies. To the extent that energy demand is reduced via increased efficiency (e.g. along the lines suggested in the low energy scenarios discussed), this shift to a cleaner supply system is made easier.

The effects of global warming are potentially serious. Estimates show that substantial warming may occur in the future due to other greenhouse gases as well as CO₂. If the global warming expected to result from a doubling of CO₂ is the upper limit of 'acceptable' climatic changes, then because of the effects of other greenhouse gases, CO₂ accumulation must be limited to approximately 1.5 to 1.7 times the pre-industrial level. According to several studies, there are, however, technically and economically viable strategies for global energy development that are compatible with a high level of concern about the climatic impacts of a continued CO₂ build-up. Some of these strategies allow for a considerable continued growth in the industrialized countries and for bringing developing countries to the economic level of Western European countries in the 1970s.

For global society to reach the 'lower bound' CO₂ emission scenario it would undoubtedly require a cooperative effort based on a common view of the seriousness of the greenhouse and climate issue. In order to reach the 'upper bound', a series of concurrent policy choices based on a disregard of the greenhouse and climate issue would likewise have to be implemented on the global scale. A much more likely outcome of the myriad of energy policy decisions to be taken in the future is that CO₂ emission rates will therefore be in the middle range of the interval.

Actual future emission rates of CO₂ (and other greenhouse gases) will be influenced by many factors. These include policy feedback from scientific studies about the greenhouse and climate issue, economic growth rates and the penetration rates of existing and new technologies for increased energy efficiency and non-fossil energy production. Given the current situation and trends, it seems likely that the net result of all these influences will be to push global CO₂ emissions down, towards the lower parts of the interval between the bounds.

However, monumental uncertainties surrounding all long-term energy projections must not overshadow the fact that it is within our power to control the future concentration of CO₂ through the implementation of systematic policy choices, although these choices must be made in the face of substantial scientific uncertainty.

NOTE ON AUTHORSHIP AND ACKNOWLEDGEMENTS

Although the three authors provided critical discussion and commentary on one another's material, principal responsibility for researching and writing the Sections of this Chapter is as follows: 2.1: all authors, 2.2 and 2.3: I. Mintzer, 2.4 and 2.5: W. Keepin, and 2.6: all authors.

L. Kristoferson co-ordinated, critiqued and supervised the production of the Draft Report. Dr. A. Miller contributed most helpfully to Sections 2.2 and 2.3.

The authors are indebted to Professor Thomas B. Johansson for his helpful commentary on the Draft; also to Academician M. A. Stirikovich for personal discussions and valuable commentary on the Draft. We also wish to thank the following for many helpful suggestions and advice on the content of an earlier draft: B. Bolin; B. R. Döös; P. Gleick; G. T. Goodman; J. Harte; J. Mathews; A. Natkin; R. Repetto; G. Speth; P. Thacher; S. Wennerberg; R. Williams; and, especially, D. Rose.

Additionally, by arrangement with the USSR Academy of Sciences (Moscow), a small Workshop was convened in Tbilisi, USSR (5-7 February, 1985) under the auspices of the Academy of Sciences of the Soviet Socialist Republic of Georgia. We are particularly pleased to acknowledge the support of these two Academies of Science for arranging such a successful Meeting and to the following participants for all their useful proposals, which have now been incorporated in this Final Draft: M. Ja. Antonovsky; B. Bolin; E. Buettner; M. Chadwick; P. Godoy; V. Gomelauri; G. Goodman; J. Hollander; V. Kakabadze; I. Kuzmin; T. Lagidze; M. Lönnroth; and B. Wynne.

We are also thankful for the comments and advice given by the WMO/ICSU/UNEP Review Panel, including W. Böhme, R.E. Dickinson, J. Edmonds, H. W .Ellsaesser, H.L. Ferguson, R.M. Gifford, F .K. Hare, P.M.R. Kiangi, H.E. Landsberg, J.A. Laurmann, W.J. Maunder, A.S. Monin, O. Preining, R.M. Rotty, G.E. Suarez, A. Weinberg, and M.M. Yoshino.

We hope that we have been able to successfully coordinate and as far as possible incorporate all the many valuable suggestions made by all the persons listed above. The Chapter has benefited greatly from their advice. However, any errors that remain are entirely our own.

We also wish to acknowledge the Beijer Institute (Royal Swedish Academy of Sciences) of Stockholm, the Swedish Energy Research Commission (Stockholm) and the World Resources Institute (Washington D.C.) for financial support.

2.7 REFERENCES

Alcamo, J. (1984a) A regional perspective of global coal scenarios, Energy Policy, 12 (2), 213-15.

Alcamo, J. (1984b) Fire and water: Water needs of future coal development in the Soviet Union and the United States, Resources and Energy, 6 (2).

Ausubel, J., and Nordhaus, W. D. (1983) A review of estimates of future carbon dioxide emissions, in Changing Climate, Washington, D. C., National Academy Press.

Caputo, R. (1984) Worlds in collision: Is a rational energy policy possible for Western Europe? Futures, 16 (3), 233-59.

Chadwick, M.J., and lindman, N. (eds) (1982) Environmental Implications of Expanded Coal Utilization. A study by the Beijer Institute, UNEP, and the USSR Academy of Sciences, Oxford, Pergamon.

Cleveland, C. J., Costanza, R., Hall, C. A., and Kaufmann, R. (1984) Energy and the U .S. Economy: A Biophysical Perspective, Science, 225 ( 4665), 890-897. 

Cole, S. (1983) After Cancun: The magic of the market, Futures, 15 (1),2-12. Darmstadter, J. (1984) Energy Patterns, draft, (Washington, D. C., Resources for the Future.

Edmonds, J.,and Reilly, J. (1983a) A long-term global energy-economic model of carbon dioxide release from fossil fuel use, Energy Econ., 5 (2), 74-88.

Edmonds, J.,and Reilly, J. (1983b) Global energy and CO₂ to the year 2050, Energy Journal, 4 (3),21-47.

Edmonds, J. A., ReiIly, J.,Trabalka, J. R., and Reichle, D. E. (1984) An Analysis of Possible Future Atmospheric Retention of Fossil Fuel CO₂, Report No.DOE/OR/21400-1, Washington, D. C., Department of Energy 

EPA (1983), see Seidel and Keyes (1983).

Fischer, W. (1984) Coal Trade Statistics, Vol.11, Llandudno, UK, Robertson Research.

Goldemberg, J., Johansson, T. B., Reddy, A. K. N., and Williams, R. H. (1984) Energy for a Sustainable World (forthcoming).

Goldemberg, J., Johansson, T. B., Reddy, A. K. N., and Williams, R. H. (1985) Basic needs and much more, with 1 kW per capita (to be published).

Goodman, G. T. (1983) Societal problems in meeting current and future energy needs, Ambio, 12 (2),97-101.

Green, M. A. (1983) Solar Cells: Operating Principles, Technology, and Systems Applications, Englewood Cliffs, NJ, Prentice-Hall.

Greenberger, M., et al. (1983) Caught Unawares: The Energy Decade in Retrospect, Cambridge, MA, Ballinger .

Häfele, W. (1980) A global and long range picture of energy developments, Science, 209, 174-182.

Hammond, P. B. (1984) The energy model muddle, Policy Sci., 16, 227-243.

IIASA (1981) Energy in a Finite World. Ed. W. Häfele, Cambridge, MA, Ballinger.

IIASA (1983) IIASA '83 Scenario of Energy Development: Summary. Ed. H.-H. Rogner, Laxenburg, Austria, IIASA.

Jäger, J. (1984) Impacts on the energy sector, in Meinl, H. et al. (1984), Comparison of two energy scenarios. Socioeconomic Impacts of Climatic Changes due to a Doubling of Atmospheric CO₂ Content, Brussels, EEC, 396-418.

Keeling, C. D. (1973) Industrial production of carbon dioxide from fossil fuels and limestone, Tellus, 28, 174-198.

Keepin, B. (1984a) A technical appraisal of the IIASA energy scenarios, Policy Sci., 17 (4), Special issue.

Keepin, B. (1984b) Review of Global Energy Forecasts, Stockholm, Beijer Institute, Royal Swedish Academy of Sciences.

Keepin, B., and Wynne, B. (1984) Technical analysis of IIASA energy scenarios, Nature, 312, 691-695.

Koreisha, S., and Stobaugh, R. (1979) limits to models, in Stobaugh, R., and Yergin, D. (eds) Energy Future, New York, Vintage, 309-342.

Krause, F. (1983) A low cost/low risk global energy future. Paper submitted to Annu. Rev. Energy.

Kristoferson, L. ( 1977) Waterpower, Ambio, 6, 44-45.

Kuzmin, I. I., Romanov, S. V., and Chernoplyokov, A. I. (1984) On a Quantitative Approach to the Safety Assessment, Moscow, Kurchatov Institute of Atomic Energy.

Laurmann, J. A. (1984) Market Penetration as an Impediment to Replacement of Fossil Fuels in the CO₂ Environmental Problem. Paper prepared for the H. H. Humphrey Institute of Public Affairs Symposium on Greenhouse Problem Policy Options, Minneapolis, MN, May 29-31,1984 (Proceedings in preparation).

Legasov, V. A., and Kuzmin, I.I. (1981) The problem of energy production, Prioroda, 2, 8-23.

Legasov, V. A., Kuzmin, I. I., and Chernoplyokov, A. I. (1984) The influence of energetics on climate, Fizika Almospheri i Okeana, 11, 1089-1103, USSR Academy of Sciences.

Long, R. (1982) Constraints on International Trade in Coal, London, IEA Coal Research.

Lovins, A. B., Lovins, L. H., Krause, F., and Bach, W. (1981) Least Cost Energy: Solving the CO₂ Problem, Andover, MA, Brick House Press.

Manne, A., and Schrattenholzer, L. (1984a) International Energy Workshop Part II: Individual Poll Responses, Laxenburg, Austria, IIASA.

Manne, A., and Schrattenholzer, L. (1984b) International Energy Workshop: A summary of the 1983 poll responses, Energy Journal, 5, (1),45-65.

Marchetti, C. (1980) On Energy Systems in Historical Perspective, Working Paper WP-80, Laxenburg, Austria, IIASA.

Marchetti, C., and Nakicenovic, N. (1979) The Dynamics of Energy Systems and the Logistic Substitution Model, Research Report RR-79-l3, Laxenburg, Austria, IIASA.

Marland, G. (1982) The impact of synthetic fuels on global CO₂ emission, in W. Clark (ed.) CO₂ Review 1982, Oxford, Clarendon Press, 406-410.

Marland, G., and Rotty, R. M. (1984) Carbon dioxide emissions from fossil fuels: A procedure for estimation and results for 1950-82, Tellus (forthcoming).

Marley, R. C. (1984) Trends in industrial use of energy, Science 226, 1277-1283. 

Maycock, P ., and Stirewalt, E. (1982) Photovoltaics: Sunlight to Electricity in One Step, Andover, MA, Brick House Press.

Meinl, H., et al. (1984) Comparison of Two Energy Scenarios. Socioeconomic Impacts of Climatic Changes due to a Doubling of Atmospheric CO₂ Content, Brussels, F-EC, 25-30.

Mintzer, I., and Miller, A. (1984) Impacts of Energy Conservation Measures and Renewable Energy Technologies on Global Emissions of Carbon Dioxide, Washington. D. C., World Resources Institute.

MIT (1983), see Rose et al. (1983).

Nader et al. (1979) Impacts of Consumption, Location, and Occupational Patterns on Energy Demand: A Report to the Demand Panel of National Academy of Sciences Committee on Nuclear and Alternative Energy Systems, Washington. DC, National Research Council.

Nordhaus, W. D., and Yohe, G. (1983) Future paths of energy and carbon dioxide emissions, in Changing Climate, Washington, D.C., National Academy Press.

OECD (1979) OECD Interfutures Project 'Facing the Future', Paris, OECD. O'Keefe, P., Raskin, P., and Bernow, S. (eds) (1984) Energy and development in Kenya: opportunities and constraints. Energy, Environment and Development in Africa, Vol. 1, Stockholm, Beijer Institute and Scandinavian Institute of African Studies.

Peng. T. H., et al. (1983) A deconvolution of the tree ring based on the ¹⁴C record, J. Geophys. Res., 88. 3609-3620.

Perry. A. N. (1982) CO₂ production scenarios: An assessment of alternative futures, in Clark. W.C. (ed.) Carbon Dioxide Review:1982. New York, Oxford University Press.

Prior, M. (1982) Steam Coal and Energy Needs in Western Europe Beyond 1985, EIU Special Report No.134, London, Economist Intelligence Unit.

Quirk, W. J. (1981) Climate and Energy Emergencies, Bull. Am. Meteorol. Soc., 62, 623-632.

Radetski, M. (1984) The Art of Forecasting Long Run Economic Growth. Internal Memorandum, Beijer Institute, Royal Swedish Academy of Sciences.

Reilly, J., and Edmonds, J. (1984a) Time and Uncertainty: Analytic Paradigms and Policy Requirements, Oak Ridge, TN, Institute for Energy Analysis.

Reilly. J., and Edmonds, J. (1984b) 'Changing climate' and energy modeling: A review, Energy Journal (forthcoming).

Reister, D. B. (1984) An Assessment of the Contribution of Gas to the Global Emissions of Carbon Dioxide. Final Report GRI-84/003, Chicago, IL, Gas Research Institute.

Robinson, J. B. (1982a) Backing into the future: On the methodological and institutional biases embedded in energy supply and demand forecasting. Technological Forecasting and Social Change, 21, 229-240.

Robinson. J. B. (1982b) Apples and horned toads: On the framework-determined nature of the energy debate, Policy Sci.. 15, 23-45.

Rose. D. J. (1984a) Private communication.

Rose. D. J. (1984b) Interpreting the Rose-Miller-Agnew Global Energy Scenarios (unpublished memorandum).

Rose, D. J., Miller, M. M., and Agnew, C. (1983) Global Energy Futures and CO₂- Induced Climate Change. MITEL 83-015, Cambridge MA, MIT Energy Laboratory.

Rose. D. J., Miller, M. M., and Agnew, C. (1984) Reducing the problem of global warming, Technol. Rev., May/june, 49-58.

Rotty. R. M., and Masters, C. (1984) Past and Future Releases of CO₂ from Fossil Fuel Combustion. Oak Ridge, TN, Institute for Energy Analysis (forthcoming).

Schurr. S. H. (1984) Energy use, technological change, and productive efficiency: An economic-historial interpretation, Annu. Rev. Energy,. 9. 409-425.

Seidel, S., and Keyes, D. (1983) Can We Delay a Greenhouse Warming? Washington, D.C.. Environmental Protection Agency.

Sh'pillrain, E. (1985) Personal communication.

Siegenthaler, U. (1983) Uptake of excess CO₂ by an outcrop-diffusion model of the ocean, J. Geophys. Res., 88, 3598-608.

Skalkin, F. V., Kanaev. A. A., and Kopp, I. Z. (1981) Energetics and the Environment, Leningrad.

Steinberg et al. (1984) A Systems Study for the Removal, Recovery, and Disposal of Carbon Dioxide from fossil Fuel Power Plants in the US, Report No.DOE/CH/00016-2., Upton; NY, Brookhaven National Laboatory.

Stirikovich, M. A., Sinyak. Y. V., and Chernyavskii. S. Y. (1981) Long term global energy developments-achievements and perspectives, Energitika, 14, Toplivo, No.3.

Svenningsson, P. ( 1985) Energy Strategies and Atmospheric CO₂ Concentrations. Report. Stockholm, Beijer Institute, Royal Swedish Academy of Sciences.

Thompson, M. (1984) Among the energy tribes: A cultural framework for the analysis and design of policy. Policy Sci., 17 (4) Special issue.

WEC (1983) Energy 2000-2020: World Prospects and Regional Stresses, Ed. J. R. Frisch, World Energy Conference Conservation Commission; and Oil Substitution: World Outlook to 2020, London, Graham and Trotman and Oxford University Press.

Williams, R. H., Goldemberg, J., Johansson, T. B., Reddy, A. K. N., and Larson, E. (1984) Overview of An End-Use Oriented Global Energy Strategy. Paper presented at the Greenhouse Problem: Policy Option Symposium, University of Minnesota, May 1984.

WOCOL (1980) CoalBridge to the Future: Report of the World Coal Study, Ed. C. Wilson, Cambridge, MA, Ballinger.

Wood, G. V. (1983) Private communication to D. J. Rose.

Wynne, B. (1983) Redefining the issues of risk and social acceptance. Futures, 20, 1-21.

Wynne, B. (1984) The institutional context of science, models and policy: The IIASA energy study, Policy Sci., 17 (4) Special issue.