3.1.1.1 Types of scenarios

The scenario literature can be split into two largely non-overlapping streams – quantitative modelling and qualitative narratives ( Morita et al., 2001 [NPR, SRC] ). This dualism mirrors the twin challenges of providing systematic and replicable quantitative representation, on the one hand, and contrasting social visions and non-quantifiable descriptors, on the other ( Raskin et al., 2005 [NPR, SRC] ). It is particularly noteworthy that recent developments in scenario analysis are beginning to bridge this difficult gap ( Nakicenovic et al., 2000 [NPR, SRC] ; Morita et al., 2001 [NPR, SRC] ; and Carpenter et al., 2005 ).

Catastrophic futures feature prominently in the narrative scenarios literature. They typically involve large-scale environmental or economic collapse, extrapolating current unfavourable conditions and trends in many regions. ^[1] Many of these scenarios suggest that catastrophic developments may draw the world into a state of chaos within one or two decades. Greenhouse-gas emissions might be low in such scenarios because of low or negative economic growth, but seem unlikely to receive much attention in any case, in the light of more immediate problems. This report does not analyze such futures, except where cases provide emissions pathways.

3.1.1.2 Narrative storylines and modelling

The literature based on narrative storylines that describe futures is rich going back to the first global studies of the 1970 s (e.g. Kahn et al., 1976 [NPR] ; Kahn and Weiner, 1967 [NPR] ) and is also well represented in more recent literature (e.g. ( Peterson and Peterson, 1994; ) Gallopin et al., 1997 [NPR, ARC] ; Raskin et al., 1998 [NPR, ARC] ; Glenn and Gordon, 1997 [NPR, MoS] ). Well known are the Shell scenarios that are principally based on narrative stories with illustrative quantification of salient driving forces and scenario outcomes ( (Wack, 1985a, ) 1985b; Schwartz, 1991 [NPR] ; Shell, 2005 ).

3.1.1.3 Global futures scenarios

Global futures scenarios are deeply rooted in the long history of narrative scenarios ( Carpenter et al., 2005 [NPR] ; UNEP, 2002 [NPR, MoS] ). The direct antecedents of contemporary scenarios lie with the future studies of the 1970 s ( Raskin et al., 2005 [NPR, SRC] ). These responded to emerging concerns about the long-term sufficiency of natural resources to support expanding global populations and economies. This first wave of global scenarios included ambitious mathematical simulation models ( Meadows et al., 1972 [NPR] ; Mesarovic and Pestel, 1974 [NPR] ) as well as speculative narrative ( Kahn et al., 1976 [NPR] ). At this time, scenario analysis was first used at Royal Dutch/Shell as a strategic management technique ( (Wack, 1985a, ) 1985b; Schwartz, 1991 [NPR] ).

A second round of integrated global analysis began in the late 1980 s and 1990 s, prompted by concerns with climate change and sustainable development. These included narratives of alternative futures ranging from ‘optimistic’ and ‘pessimistic’ worlds to consideration of ‘surprising’ futures ( Burrows et al., 1991 [NPR, MoS] ; the Central Planning Bureau of the Netherlands, 1992; ( Kaplan, 1994; ) Svedin and Aniansson, 1987 [NPR, MoS] ; Toth et al., 1989 [NPR, SRC] ). The long-term nature of the climate change issue introduced a new dimension and has resulted in a rich new literature of global emissions scenarios, starting from the IPCC IS92 scenarios ( Pepper et al., 1992 [NPR, ARC] ; Leggett et al., 1992 [NPR, ARC] ) and most recent scenario comparisons projects (e.g. EMF and IMCP). The first decades of scenario assessment paved the way by showing the power – and limits – of both deterministic modelling and descriptive future analyses. A central challenge of global scenario exercises today is to unify these two aspects by blending the objectivity and clarity of quantification with the richness of narrative ( Raskin et al., 2005 [NPR, SRC] ).

3.2.1.1 Population projections

Current population projections reflect less global population growth than was expected at the time the TAR was published. Since the early 1990 s demographers have revised their outlook on future population downward, based mainly on new data indicating that birth rates in many parts of the world have fallen sharply.

Recent projections indicate a small downward revision to the medium (or ‘best guess’) outlook and to the high end of the uncertainty range, and a larger downward revision to the low end of the uncertainty range ( Van Vuuren and O’Neill, 2006 [JoC, MoS, SRC] ). This global result is driven primarily by changes in outlook for the Asia and the Africa-Latin America-Middle East (ALM) region. On a more detailed level, trends are driven by changes in the outlook for Sub-Saharan Africa, the Middle East and North Africa region, and the East Asia region, where recent data show lower than expected fertility rates, as well as a much more pessimistic view on the extent and duration of the HIV/AIDS crisis in Sub-Saharan Africa. In contrast, in the OECD region, updated projections are somewhat higher than previous estimates. This comes from changes in assumptions regarding migration (in the case of the UN projections), or to a more optimistic projection of future life expectancy (in the case of International Institute for Applied Systems Analysis (IIASA) projections). In the Eastern Europe and Central Asia (Reforming Economic, REF) region, projections have been revised downward, especially by the UN, driven mainly by recent data showing very low fertility levels and mortality rates that are quite high relative to other industrialized countries.

Lutz et al. 2004 [NPR] UN 2004 [NPR] and Fisher et al. 2006 [MoS, SRC] ) have produced updated projections for the world that extend to 2100 . The most recent central projections for global population are 1.4–2.0 billion (13–19%) lower than the medium population scenario of 10.4 billion used in the SRES B2 scenarios. As was the case with the outlook for 2050, the long-term changes at the global level are driven by the developing-country regions (Asia and ALM), with the changes particularly large in China, the Middle East and North Africa, and Sub-Saharan Africa.

Most of the SRES scenarios still fall within the plausible range of population outcomes, according to more recent literature (see Figure 3.1 ). However, the high end of the SRES population range now falls above the range of recent projections from IIASA and the UN. This is a particular problem for population projections in East Asia, the Middle East, North Africa and the Former Soviet Union, where the differences are large enough to strain credibility ( Van Vuuren and O’Neill, 2006 [JoC, MoS, SRC] ). In addition, the population assumptions in SRES and the vast majority of more recent emissions scenarios do not cover the low end of the current range of population projections well. New scenario exercises will need to take the lower population projections into account. All other factors being equal, lower population projections are likely to result in lower emissions. However, a small number of recent studies that have used updated and lower population projections ( Carpenter et al., 2005 [NPR] ; van Vuuren et al., 2007 [JoC, SRC, 2007] ; Riahi et al., 2006 [NPR, SRC] ) indicate that changes in other drivers of emissions might partly offset the impact of lower population assumptions, thus leading to no significant changes in emissions.

Figure 3.1: Comparison of population assumptions in post-SRES emissions scenarios with those used in previous scenarios. Blue shaded areas span the range of 84 population scenarios used in SRES or pre-SRES emissions scenarios; individual curves show population assumptions in 117 emissions scenarios in the literature since 2000 . The two vertical bars on the right extend from the minimum to maximum of the distribution of scenarios by 2100 . The horizontal bars indicate the 5^th, 25^th, 50^th, 75^th and the 95^th percentiles of the distributions.

Data source: After Nakicenovic et al., 2006 [SRC] .

3.2.1.2 Economic development

Economic activity is a dominant driver of energy demand and thus of greenhouse gas emissions. This activity is usually reported as gross domestic product (GDP), often measured in per-person (per-capita) terms. To derive meaningful comparisons over time, changes in price levels must be taken into account and corrected by reporting activities as constant prices taken from a base year. One way of reducing the effects of different base years employed across various studies is to report real growth rates for changes in economic output. Therefore, the focus below is on real growth rates rather than on absolute numbers.

Given that countries and regions use particular currencies, another difficulty arises in aggregating and comparing economic output across countries and world regions. There are two main approaches: using an observed market exchange rate (MER) in a fixed year or using a purchasing power parity rate (PPP) (see Box 3.1 ). GDP trajectories in the large majority of long-term scenarios in the literature are calibrated in MER. A few dozen scenarios exist that use PPP exchange rates, but most of them are shorter-term, generally running until the year 2030 .

3.2.1.3 GDP growth rates in the new literature

Many of the long-term economic projections in the literature have been specifically developed for climate-related scenario work. Figure 3.2 compares the global GDP range of 153 baseline scenarios from the pre-SRES and SRES literature with 130 new scenarios developed since SRES (post-SRES). There is a considerable overlap in the GDP numbers published, with a slight downward shift of the median in the new scenarios (by about 7%) compared to the median in the pre-SRES scenario literature. The data suggests no appreciable change in the distribution of GDP projections.

Figure 3.2: Comparison of GDP projections in post-SRES emissions scenarios with those used in previous scenarios. The median of the new scenarios is about 7% below the median of the pre-SRES and SRES scenario literature. The two vertical bars on the right extend from the minimum to maximum of the distribution of scenarios by 2100 . The horizontal bars indicate the 5^th, 25^th, 50^th, 75^th and the 95^th percentiles of the distributions.

A comparison of some recent shorter-term global GDP projections using the SRES scenarios is illustrated in Figure 3.3 . The SRES scenarios project a very wide range of global economic per-person growth rates from 1% (A2) to 3.1% (A1) to 2030, both based on MER. This range is somewhat wider than that covered by the USDOE 2004 [NPR, MoS] ) high and low scenarios (1.2–2.5%). The central projections of USDOE, IEA and the World Bank all contain growth rates of around 1.5–1.9%, thus occurring in the middle of the range of the SRES scenarios. Other medium-term energy scenarios are also reported to have growth rates in this range ( IEA, 2004 [NPR] ).

Figure 3.3: Comparison of global GDP growth per person in the SRES scenarios and more recent projections.

Notes: SRES = (Nakicenovic et al., 2000 [NotFound] ), WB = World Bank (World Bank, 2004 [NPR] ), DoE = assumptions used by US Department of Energy ( USDOE, 2004 [NPR, MoS] ), IEA assumptions used by IEA ( IEA, 2002 [NPR] and 2004 ( Van Vuuren and O’Neill, 2006 [JoC, MoS, SRC] ).

Regionally, for the OECD, Eastern Europe and Central Asia (REF) regions, the correspondence between SRES outcomes and recent scenarios is relatively good, although the SRES GDP growth rates are somewhat conservative. In the ASIA region, the SRES range and its median value are just above that of recent studies. The differences between the SRES outcomes and more recent projections are largest in the ALM region (covering Africa, Latin America and the Middle East). Here, the A1 and B1 scenarios clearly lie above the upper end of the range of current projections (4%–5%), while A2 and B2 fall near the centre of the range (1.4–1.7%). The recent short-term projections reported here contain an assumption that current barriers to economic growth in these regions will slow growth, at least until 2015 .

3.2.1.4 The use of MER in economic and emissions scenarios modelling

The uses of MER-based economic projections in SRES have recently been criticized ( (Castles and Henderson, 2003a, ) 2003b; ( Henderson, 2005 ) ). The vast majority of scenarios published in the literature use MER-based economic projections. Some exceptions exist, for example, MESSAGE in SRES, and more recent scenarios using the MERGE model ( Manne and Richels, 2003 [NPR, SRC] ), along with shorter term scenarios to 2030, including the G-Cubed model ( (McKibbin et al., 2004a, ) 2004b ), the International Energy Outlook ( USDOE, 2004 [NPR, MoS] ), the IEA World Energy Outlook ( IEA, 2004 [NPR] ) and the POLES model used by the European Commission ( 2003 ). The main criticism of the MER-based models is that GDP data for world regions are not corrected with respect to purchasing power parities (PPP) in most of the model runs. The implied consequence is that the economic activity levels in non-OECD countries generally appear to be lower than they actually are when measured in PPP units. In addition, the high growth SRES scenarios (A1 and B1 families) assume that regions tend to conditionally converge in terms of relative per-person income across regions (see Section 3.1.4 ). According to the critics, the use of MER, together with the assumption of conditional convergence, lead to overstated economic growth in the poorer regions and excessive growth in energy demand and emission levels.

A team of SRES researchers responded to this criticism, indicating that the use of MER or PPP data does not in itself lead to different emission projections outside the range of the literature. In addition, they stated that the use of PPP data in most scenarios models was (and still is) infeasible, due to lack of required data in PPP terms, for example price elasticities and social accounting matrices ( Nakicenovic et al., 2003 [Ambiguous] ; Grübler et al., 2004 [SRC] ). A growing number of other researchers have also indicated different opinions on this issue or explored it in a more quantitative sense (e.g. Dixon and Rimmer, 2005 [NPR, MoS] ; Nordhaus, 2006b [NPR] ; Manne and Richels, 2003 [NPR, SRC] ;( McKibbin et al., 2004a, ) 2004b; Holtsmark and Alfsen, 2004a [JoC, SRC] , 2004b; van Vuuren and Alfsen, 2006 [JoC, SRC] ).

There are at least three strands to this debate. The first is whether economic projections based on MER are appropriate, and thus whether the economic growth rates reported in the SRES and other MER-based scenarios are reasonable and robust. The second is whether the choice of the exchange rate matters when it comes to emission scenarios. The third is whether it is possible, or practical, to develop robust scenarios given the sparseness of relevant and required PPP data. While the GDP data are available in PPP, other economic scenario characteristics, such as capital and operational cost of energy facilities, are usually available either in domestic currencies or MER. Full model calibration in PPP for regional and global models is still difficult due to the lack of underlying data. This could be one of the reasons why a vast majority of long-term emissions scenarios continues to be calibrated in MER.

On the question of whether PPP or MER should be employed in economic scenarios, the general recommendations are to use PPP where practical. ^[3] This is certainly necessary when comparisons of income levels across regions are of concern. On the other hand, models that analyse international trade and include trade as part of their economic projections, are better served by MER data given that trade takes place between countries in actual market prices. Thus, the choice of conversion factor depends on the type of analysis or comparison being undertaken.

For principle and practical reasons, Nordhaus 2005 [NPR, MoS] ) recommends that economic growth scenarios should be constructed by using regional or national accounting figures (including growth rates) for each region, but using PPP exchange rates for aggregating regions and updating over time by use of a superlative price index. In contrast, Timmer 2005 [NPR] ) actually prefers the use of MER data in long-term modelling, as such data are more readily available, and many international relations within the model are based on MER. Others (e.g. van Vuuren and Alfsen, 2006 [JoC, SRC] ) also argue that the use of MER data in long-term modelling is often preferable, given that model parameters are usually estimated on MER data and international trade within the models is based on MER. The real economic consequences of the choice of conversion rates will obviously depend on how the scenarios are constructed, as well as on the type of model used for quantifying the scenarios. In some of the short-term scenarios (with a horizon to 2030 ) a bottom-up approach is taken where assumptions about productivity growth and investment/saving decisions are the main drivers of growth in the models (e.g. ( McKibbin et al., 2004a, ) 2004b ). In long-term scenario models, a top-down approach is more commonly used where the actual growth rates are prescribed more directly, based on convergence or other assumptions about long-term growth potentials.

When it comes to emission projections, it is important to note that in a fully disaggregated (by country) multi-sector economic model of the global economy, aggregate index numbers play no role and the choice between PPP and MER conversion of income levels does not arise. However, in an aggregated model with consistent specifications (i.e. where model parameter estimation and model calibrations are all carried out based on consistent use of conversion factors), the effects of the choice of conversion measure on emissions should approximately cancel out. The reason can be illustrated by using the Kaya identity, which decomposes the emissions as follows:

GHG = Population x GDP per person x Emissions per GDP

or:

where GHG stands for greenhouse gas emissions, GDP stands for economic output, and POP stands for population size. ^[4]

Given this relationship, emission scenarios can be represented, explicitly based on estimates of population development, economic growth, and development of emission intensity.

Population is often projected to grow along a pre-described (exogenous) path, while economic activity and emission intensities are projected based on differing assumptions from scenario to scenario. The economic growth path can be based on historical growth rates, convergence assumptions, or on fundamental growth factors, such as saving and investment behaviour, productivity changes, etc. Similarly, future emission intensities can be projected based on historical experience, economic factors, such as labour productivity or other key factors determining structural changes in an economy, or technological development. The numerical expression of GDP clearly depends on conversion measures; thus GDP expressed in PPP will deviate from GDP expressed in MER, particularly for developing countries. However, when it comes to calculating emissions (or other physical measures such as energy), the Kaya identity shows that the choice between MER-based or PPP-based representations of GDP will not matter, since emission intensity will change (in a compensating manner) when the GDP numbers change. While using PPP values necessitates using lower economic growth rates for developing countries under the convergence assumption, it is also necessary to adjust the relationship between income and demand for energy with lower economic growth, leading to slower improvements in energy intensities. Thus, if a consistent set of metrics is employed, the choice of metric should not appreciably affect the final emission level.

In their modelling work, Manne and Richels 2003 [NPR, SRC] ( and McKibbin et al. 2004a, ) 2004b ) find some differences in emission levels between using PPP-based and MER-based estimates. Analysis of their work indicates that these results critically depend on, among other things, the combination of convergence assumptions and the mathematical approximation used between MER-GDP and PPP-GDP. In the Manne and Richels work for instance, autonomous efficiency improvement (AEI) is determined as a percentage of economic growth and estimated on the basis of MER data. In going from MER to PPP, the economic growth rate declines as expected, leading to a decline in the autonomous efficiency improvement. However, it is not clear whether it is realistic not to change the AEI rate when changing conversion measure. On the other hand, Holtsmark and Alfsen 2004a [JoC, SRC] , 2004b ), showed that in their simple model consistent replacement of the metric (PPP for MER) – for income levels as well as for underlying technology relationships – leads to a full cancellation of the impact of choice of metric on projected emission levels.

To summarize: available evidence indicates that the differences between projected emissions using MER exchange rates and PPP exchange rates are small in comparison to the uncertainties represented by the range of scenarios and the likely impacts of other parameters and assumptions made in developing scenarios, for example, technological change. However, the debate clearly shows the need for modellers to be more transparent in explaining conversion factors, as well as taking care in determining exogenous factors used for their economic and emission scenarios.

Box 3.1 Market Exchange Rates and Purchasing Power Parity

To aggregate or compare economic output from various countries, GDP data must be converted into a common unit. This conversion can be based on observed market exchange (MER) rates or purchasing power parity (PPP) rates where, in the latter, a correction is made for differences in price levels between countries. The PPP approach is considered to be the better alternative if data is used for welfare or income comparisons across countries or regions. Market exchange rates usually undervalue the purchasing power of currencies in developing countries, see Figure 3.4 .

Figure 3.4 : Regional GDP per person, expressed in MER and PPP on the basis of World Bank data aggregated to 17 global regions.

Note: The left y-axis and columns compare absolute data, while the right y-axis and line graph compare the ratio between PPP and MER data. EECCA = countries of Eastern Europe, the Caucasus and Central Asia.

Source: van Vuuren and Alfsen, 2006 [JoC, SRC] .

Clearly, deriving PPP exchange rates requires analysis of a relatively large amount of data. Hence, methods have been devised to derive PPP rates for new years on the basis of price indices. Unfortunately, there is currently no single method or price index favoured for doing this, resulting in different sets of PPP rates (e.g. from the OECD, Eurostat, World Bank and Penn World Tables) although the differences tend to be small.

3.2.1.5 Energy use

Future evolution of energy systems is a fundamental determinant of GHG emissions. In most models, energy demand growth is a function of key driving forces such as demographic change and the level and nature of human activities such as mobility, information processing, and industry. The type of energy consumed is also important. While Chapters 4 through 11 report on medium-term projections for different parts of the energy system, long-term energy projections are reported here. Figure 3.5 compares the range of the 153 SRES and pre-SRES scenarios with 133 new, post-SRES, long-term energy scenarios in the literature. The ranges are comparable, with small changes on the lower and upper boundaries, and a shift downwards with respect to the median development. In general, the energy growth observed in the newer scenarios does not deviate significantly from the previous ranges as reported in the SRES report. However, most of the scenarios reported here have not adapted the lower population levels discussed in Section 3.2.1.1 .

Figure 3.5: Comparison of 153 SRES and pre-SRES baseline energy scenarios in the literature compared with the 133 more recent, post-SRES scenarios. The ranges are comparable, with small changes on the lower and upper boundaries.

Note: The two vertical bars on the right extend from the minimum to maximum of the distribution of scenarios by 2100 . The horizontal bars indicate the 5^th, 25^th, 50^th, 75^th and the 95^th percentiles of the distributions.

In general, this situation also exists for underlying trends as represented by changes in energy intensity, expressed as gigajoule (GJ)/GDP, and change in the carbon intensity of the energy system (CO₂/GJ) as shown in Figure 3.6 . In all scenarios, energy intensity improves significantly across the century – with a mean annual intensity improvement of 1%. The 90% range of the annual average intensity improvement is between 0.5% and 1.9% (which is fairly consistent with historic variation in this factor). Actually, this range implies a difference in total energy consumption in 2100 of more than 300% – indicating the importance of the uncertainty associated with this ratio. The carbon intensity is more constant in scenarios without climate policy. The mean annual long-term improvement rate over the course of the 21^st century is 0.4%, while the uncertainty range is again relatively large (from -0.2 to 1.5%). At the high end of this range, some scenarios assume that energy technologies without CO₂ emissions become competitive without climate policy as a result of increasing fossil fuel prices and rapid technology progress for carbon-free technologies. Scenarios with a low carbon-intensity improvement coincide with scenarios with a large fossil fuel base, less resistance to coal consumption or lower technology development rates for fossil-free energy technologies. The long-term historical trend is one of declining carbon intensities. However, since 2000, carbon intensities are increasing slightly, primarily due to the increasing use of coal. Only a few scenarios assume the continuation of the present trend of increasing carbon intensities. One of the reasons for this may be that just a few of the recent scenarios include the effects of high oil prices.

Figure 3.6: Development of carbon intensity of energy (left) and primary energy intensity of GDP (right). Historical development and projections from SRES and pre-SRES scenarios compared to post-SRES scenarios.

Note: The blue coloured range illustrates the range of 142 carbon intensity and 114 energy intensity – SRES and pre-SRES non-intervention scenarios.

Source: After Nakicenovic et al., 2006 [SRC] .

3.2.1.6 Land-use change and land-use management

Understanding land-use and land-cover changes is crucial to understanding climate change. Even if land activities are not considered as subject to mitigation policy, the impact of land-use change on emissions, sequestration, and albedo plays an important role in radiative forcing and the carbon cycle.

Over the past several centuries, human intervention has markedly changed land surface characteristics, in particular through large-scale land conversion for cultivation ( Vitousek et al., 1997 [JoC] ). Land-cover changes have an impact on atmospheric composition and climate via two mechanisms: biogeophysical and biogeochemical. Biogeophysical mechanisms include the effects of changes in surface roughness, transpiration, and albedo that, over the past millennium, are thought to have had a global cooling effect ( Brovkin et al., 1999 [MoS] ). Biogeochemical effects result from direct emissions of CO₂ into the atmosphere from deforestation. Cumulative emissions from historical land-cover conversion for the period 1920 – 1992 have been estimated to be between 206 and 333 Pg CO₂ ( McGuire et al., 2001 [PoC, JoC, MoS] ), and as much as 572 Pg CO₂ for the entire industrial period 1850 – 2000, roughly one-third of total anthropogenic carbon emissions over this period ( Houghton, 2003 [NPR, MoS] ). In addition, land management activities (e.g. cropland fertilization and water management, manure management and forest rotation lengths) also affect land-based emissions of CO₂ and non-CO₂ GHGs, where agricultural land management activities are estimated to be responsible for the majority of global anthropogenic methane (CH₄) and nitrous oxide (N₂O) emissions. For example, USEPA 2006a [NPR] ) estimated that agricultural activities were responsible for approximately 52% and 84% of global anthropogenic CH₄ or N₂O emissions respectively in the year 2000, with a net contribution from non-CO₂ GHGs of 14% of all anthropogenic greenhouse gas emissions in that year.

Projected changes in land use were not explicitly represented in carbon cycle studies until recently. Previous studies into the effects of future land-use changes on the global carbon cycle employed trend extrapolations ( (Cramer et al., 2004 ) ), extreme assumptions about future land-use changes ( House et al., 2002 [NPR] ), or derived trends of land-use change from the SRES storylines ( Levy et al., 2004 [JoC] ). However, recent studies (e.g. Brovkin et al., 2006 [MoS] ; Matthews et al., 2003 [MoS] ; Gitz and Ciais, 2004 [MoS] ) have shown that land use, as well as feedbacks in the society-biosphere-atmosphere system (e.g. Strengers et al., 2004 [MoS, SRC] ), must be considered in order to achieve realistic estimates of the future development of the carbon cycle; thereby providing further motivation for ongoing development to explicitly model land and land-use drivers in global integrated assessment and climate economic frameworks. For example, in a model comparison study of six climate models of intermediate complexity, Brovkin et al. 2006 [MoS] ) concluded that land-use changes contributed to a decrease in global mean annual temperature in the range of 0.13–0.25°C, mainly during the 19th century and the first half of the 20^th century, which is in line with conclusions from other studies, such as Matthews et al. 2003 [MoS] ).

In general, land-use drivers influence either the demand for land-based products and services (e.g. food, timber, bio-energy crops, and ecosystem services) or land-use production possibilities and opportunity costs (e.g. yield-improving technologies, temperature and precipitation changes, and CO₂ fertilization). Non-market values – both use and non-use such as environmental services and species existence values respectively – will also shape land-use outcomes.

Food demand is a dominant land-use driver, and population and economic growth are the most significant food demand drivers through per person consumption. Total world food consumption is expected to increase by over 50% by 2030 ( Bruinsma, 2003 [NPR] ). Moreover, economic growth is expected to generate significant structural change in consumption patterns, with diets shifting to include more livestock products and fewer staples such as roots and tubers. As a result, per person meat consumption is expected to show a strong global increase, in the order of 25% by 2030, with faster growth in developing and transitional countries of more than 40% and 30%, respectively ( Bruinsma, 2003 [NPR] ; Cassman et al., 2003 [ARC] ). The Millennium Ecosystem Assessment (MEA) scenarios projected that global average meat consumption would increase from 36 kg/person in 1997 to 41–70 kg/person by 2050, with corresponding increases in overall food and livestock feed demands ( Carpenter et al., 2005 [NPR] ). Additional cropland is expected to be required to support these projected increases in demand. Beyond 2050, food demand is expected to level off with slow-down of population growth.

Technological change is also a critical driver of land use, and a critical assumption in land-use projections. For example, Sands and Leimbach 2003 [JoC, MoS] ) suggest that, globally, 800 million hectares of cropland expansion could be avoided with a 1% annual growth in crop yields. Similarly, Kurosawa 2006 [NPR] ) estimates decreased cropland requirements of 18% by 2050, relative to 2000, with 2% annual growth in global average crop yields. Alternatively, the MEA scenarios implement a more complex representation of yield growth projections that, in addition to autonomous technological change, reflect the changes in production practices, investments, technology transfer, environmental degradation, and climate change. The net effect is positive, but shows declining productivity growth over time for some commodities, due in large part to diminishing marginal technical productivity gains and environmental degradation. In all these studies, increasing (decreasing) net productivity per hectare results in reduced (increased) cropland demand.

Also important to land-use projections are potential changes in climate. For instance, rising temperatures and CO₂ fertilization may improve regional crop yields in the short term, thereby reducing pressure for additional cropland and resulting in increased afforestation. However, modelling the beneficial impacts of CO₂ fertilization is not as straightforward as once thought. Recent results suggest: lower crop productivity improvements in the field than shown previously with laboratory results (e.g. Ainsworth and Long, 2005 [NPR] ); likely increases in tropospheric ozone and smog associated with higher temperatures that will depress plant growth and partially offset CO₂ fertilization; expected increases in the variability of annual yields; CO₂ effects favouring C3 plants (e.g. wheat, barley, potatoes, rice) over C4 plants (e.g. maize, sugar cane, sorghum, millet) while temperature increases favour C4 over C3 plants; potential decreased nutritional content in plants subjected to CO₂ fertilization and increased frequency of temperature extremes; and increases in forest disturbance frequency and intensity. See IPCC 2007b [NPR, 2007] , Chapter 5 ) for an overall discussion of these issues and this literature. Long-term projections need to consider these issues, as well as examining the potential limitations or saturation points of plant responses. However, to date, long-term scenarios from integrated assessment models are only just beginning to represent climate feedbacks on terrestrial ecosystems, much less fully account for the many effects. Current integrated assessment representations only consider CO₂ fertilization and changes in yearly average temperature, if they consider climate change effects at all (e.g. USCCSP, 2006 [NPR] ; van Vuuren et al., 2007 [JoC, SRC, 2007] ).

Only a few global studies have focused on long-term (century) land-use projections. The most comprehensive studies, in terms of sector and land-type coverage, are the SRES ( Nakicenovic et al., 2000 [NPR, SRC] ), the SRES implementation with the IMAGE model ( Strengers et al., 2004 [MoS, SRC] ), the scenarios from the Global Scenarios Group ( Raskin et al., 2002 [NPR, ARC] ), UNEP’s Global Environment Outlook ( UNEP, 2002 [NPR, MoS] ), the Millennium Ecosystem Assessment ( Carpenter et al., 2005 [NPR] ), and some of the EMF-21 Study models ( Kurosawa, 2006 [NPR] ; van Vuuren et al., 2006a [NPR, MoS, SRC] ; Rao and Riahi, 2006 [SRC] ; Jakeman and Fisher, 2006 [NPR, SRC] ; Riahi et al., 2006 [NPR, SRC] ; van Vuuren et al., 2007 [JoC, SRC, 2007] ). Recent sector-specific economic studies have also contributed global land-use projections for climate analysis, especially for forestry ( Sands and Leimbach, 2003 [JoC, MoS] ; Sohngen and Mendelsohn, 2003 [MoS, ARC] , 2007; Sathaye et al., 2006 [NPR, ARC] ; Sohngen and Sedjo, 2006 [NPR, ARC] ). In general, the post-SRES scenarios, though scarce in number for agricultural land use, have projected increasing global cropland areas, smaller forest-land areas, and mixed results for changes in global grassland ( Figure 3.7 ). Unlike the SRES land-use scenarios that span a broader range while representing diverse storylines, the post-SRES scenarios, for forestry in particular, illustrate greater convergence across models on projected land-use change.

Most post-SRES global scenarios project significant changes in agricultural land caused primarily by regional changes in food demand and production technology. Scenarios with larger amounts of land used for agriculture result from assumptions about higher population growth rates, higher food demands, and lower rates of technological improvement that generate negligible increases in crop yields. Combined, these effects are projected to lead to a sizeable expansion (up to 40%) of agricultural land between 1995 and 2100 ( Figure 3.7 ). Conversely, lower population growth and food demand, and more rapid technological change, are projected to result in lower demand for agricultural land (as much as 20% less global agricultural acreage by the end of the century). In the short-term, almost all scenarios suggest an increase in cropland acreage and decline in forest land to meet projected increases in food, feed, and livestock grazing demands over the next few decades. Cropland changes range from -18% to +69% by 2050 relative to 2000 (from -123 to +, (from -123 to +, million hectares) and forest-land changes range from -18% to +3% (from -680 to +94 million hectares) by 2050 . The changes in global forest generally mirror the agricultural scenarios; thereby, illustrating both the positive and negative aspects of some existing global land modelling. Most of the long-term scenarios assume that forest trends are driven almost exclusively by cropland expansion or contraction, and only deal superficially with driving forces, such as global trade in agricultural and forest products and conservation demands.

Figure 3.7: Global cropland (a), forest land (b) and grassland (c) projections.

Notes: shaded areas indicate SRES scenario ranges, post-SRES scenarios denoted with solid lines. IMAGE-EMF21 = van Vuuren et al. 2006a [NPR, MoS, SRC] ) scenario from EMF-21 Study; IMAGE-MA-xx = Millennium Ecosystem Assessment ( Carpenter et al., 2005 [NPR] ) scenarios from the IMAGE model for four storylines (GO = Global Orchestration, OS = Order from Strength, AM = Adapting Mosaic, TG = TechnoGarden); AgLU-x.x% = Sands and Leimbach 2003 [JoC, MoS] ) scenarios with x.x% annual growth in crop yield; GTM- 2003 = Sohngen and Mendelsohn 2003 [MoS, ARC] ) global forest scenario; GTM-EMF21 = Sohngen and Sedjo 2006 [NPR, ARC] ) global forest scenario from EMF-21 Study; GCOMAP-EMF21 = Sathaye et al. 2006 [NPR, ARC] ) global forest scenario from EMF-21 Study; GRAPE-EMF21 = Kurosawa 2006 [NPR] ) scenario from EMF-21 Study.

Without incentives or technological innovation, biomass crops are currently not projected to assume a large share of global business as usual land cover – no more than about 4% by 2100 . Until long-run energy price expectations rise (due to a carbon price, economic scarcity, or other force), biomass and other less economical energy supply technologies (some with higher greenhouse gas emission characteristics than biomass), are not expected to assume more significant baseline roles.

3.2.2.1 CO₂ emissions from energy and industry

This category of emissions encompasses CO₂ emissions from burning fossil fuels, and industrial emissions from cement production and sometimes feedstocks. ^[5] Figure 3.8 compares the range of the pre-SRES and SRES baseline scenarios with the post-SRES baseline scenarios. The figure shows that the scenario range has remained almost the same since the SRES. There seems to have been an upwards shift on the high and low end, but careful consideration of the data shows that this is caused by only very few scenarios and the change is therefore not significant. The median of the recent scenario distribution has shifted downwards slightly, from 75 GtCO₂ by 2100 (pre-SRES and SRES) to about 60 GtCO₂ (post SRES). The median of the recent literature therefore corresponds roughly to emissions levels of the intermediate SRES-B2 scenarios. The majority of scenarios, both pre-SRES and post-SRES, indicate an increase in emissions across most of the century, resulting in a range of 2100 emissions of 17–135 GtCO₂ emissions from energy and industry (90^th percentile of the full scenario distribution). Also the range of emissions depicted by the SRES scenarios is consistent with the range of other emission scenarios reported in the literature; both in the short and long term (see van Vuuren and O’Neill, 2006 [JoC, MoS, SRC] ).

Figure 3.8: Comparison of the SRES and pre-SRES energy-related and industrial CO₂ emissions scenarios in the literature with the post-SRES scenarios.

Note: The two vertical bars on the right extend from the minimum to maximum of the distribution of scenarios and indicate the 5^th, 25^th, 50^th, 75^th and the 95^th percentiles of the distributions by 2100 .

Source: After Nakicenovic et al., 2006 [SRC]

Several reasons may contribute to the fact that emissions have not declined in spite of somewhat lower projections for population and GDP. An important reason is that the lower demographic projections are only recently being integrated into emission scenario literature. Second, indirect impacts in the models are likely to offset part of the direct impacts. For instance, lower energy demand leads to lower fossil fuel depletion, thus allowing for a higher share of fossil fuels in the total energy mix over a longer period of time. Finally, in recent years there has been increasing attention to the interpretation of fossil fuel reserves reported in the literature. Some models may have decreased oil and gas use in this context, leading to higher coal use (and thus higher emissions).

Analysis of scenario literature using the Kaya identity shows that pre-SRES and post-SRES baseline scenarios indicate a continuous decline of the primary energy intensity (EJ/GDP), while the change in carbon intensity (CO₂/E) is much slower – or even stable (see Figure 3.6 and Section 3.2.1.5 ) in the post-SRES scenarios. In other words, in the absence of climate policy, structural change and energy efficiency improvement do contribute to lower emissions, but changes in the energy mix have a much smaller (or even zero) contribution. This conclusion is true for both the pre-SRES, SRES, as well as the post-SRES scenario literature.

Baseline or reference emissions projections generally come from three types of studies:

1. Studies meant to represent a ‘best-guess’ of what might happen if present-day trends and behaviour continue.

2. Studies with multiple baseline scenarios under comprehensively different assumptions (storylines).

3. Studies based on a probabilistic approach.

In literature, since the TAR, there has been some discussion of the purpose of these approaches (see Schneider, 2001 [PoC, JoC] ; Grübler et al., 2002 [ARC] ; Webster et al., 2002 [MoS, SRC] ). Figure 3.9 (left panel) shows a comparison of the outcomes of some prominent examples of these approaches by comparing the outcome of baselines scenarios reported in the set of EMF-21 scenarios, representing the ‘best-guess’ approach, to the outcomes of the SRES scenarios, representing the storyline approach. In the right panel the SRES range is compared to the probabilistic approach (see Webster et al., 2002 [MoS, SRC] ; Richels et al., 2004 [NPR, SRC] , for the probability studies).

Figure 3.9: Comparison of various long-term scenario studies for CO₂ emissions. Left panel: IPCC SRES, EMF-21 range (grey area), indicating the range of the lowest and highest reported values in the EMF-21 study ( Weyant et al., 2006 [NPR, SRC] ). Right panel: Webster et al. 2002 [MoS, SRC] and Richels et al. 2004 [NPR, SRC] ), indicating the mean (markers) and 95% intervals of the reported ranges of these studies (for the latter, showing the 95% interval of the combined range for optimistic and pessimistic technology).

The figure shows that the range of different models participating in the EMF-21 study is somewhat smaller than those from SRES and the probabilistic approach. The range of EMF-21 scenarios result from different modelling approaches and from modeller’s insights into ‘the mostly likely values’ for driving forces. The two probabilistic studies and SRES explicitly assume more radical developments, but the number of studies involved is smaller. This leads to the low end of scenarios for the second category having very specific assumptions on development that may lead to low greenhouse gas emissions. The range of scenarios in the probabilistic studies tends to be between these extremes. Overall, the three different approaches seem to lead to consistent results, confirming the range of emissions reported in Figure 3.8 and confirming the emission range of scenarios used for the TAR.

3.2.2.2 Anthropogenic land emissions and sequestration

Some of the first global integrated assessment scenario analyses to account for land-use-related emissions were the IS92 scenario set ( Leggett et al., 1992 [NPR, ARC] ) and the SRES scenarios ( Nakicenovic et al., 2000 [NPR, SRC] ). However, out of the six SRES models, only four dealt with land use specifically (MiniCAM, MARIA, IMAGE 2.1, AIM), of which MiniCAM and MARIA used more simplified land-use modules. ASF and MESSAGE also simulated land-use emissions, however ASF did not have a specific land-use module and MESSAGE incorporated land-use results from the AIM model ( Nakicenovic et al., 2000 [NPR, SRC] ). Although SRES was a seminal contribution to scenario development, the treatment of land-use emissions was not the focus of this assessment; and, therefore, neither was the modelling of land-use drivers, land management alternatives, and the many emissions sources, sinks, and GHGs associated with land.

While some recent assessments, such as UNEP’s Third Global Environment Outlook ( UNEP, 2002 [NPR, MoS] ) and the Millennium Ecosystem Assessment ( Carpenter et al., 2005 [NPR] ), have evaluated land-based environmental outcomes (global environment and ecosystem goods and services respectively), the Energy Modelling Forum’s 21^st Study (EMF-21) was the first large-scale exercise with a special focus on land as a climate issue. In EMF-21, the integrated assessment models incorporated non-CO₂ greenhouse gases, such as those from agriculture, and carbon sequestration in managed terrestrial ecosystems ( Kurosawa, 2006 [NPR] ; van Vuuren et al., 2006a [NPR, MoS, SRC] ; Rao and Riahi, 2006 [SRC] ; Jakeman and Fisher, 2006 [NPR, SRC] ). A few additional papers have subsequently improved upon their EMF-21 work ( Riahi et al., 2006 [NPR, SRC] ; van Vuuren et al., 2007 [JoC, SRC, 2007] ). In general, the land-use change carbon emissions scenarios since SRES project high global annual net releases of carbon in the near future that decline over time, leading to net sequestration by the end of the century in some scenarios (see Figure 3.10 ). The clustering of the non-harmonized post-SRES scenarios in Figure 3.10 suggests a degree of expert agreement that the decline in annual land-use change carbon emissions over time will be less dramatic (slower) than suggested by many of the SRES scenarios. Many of the post-SRES scenarios project a decrease in net deforestation pressure over time, as population growth slows and crop and livestock productivity increase; and, despite continued projected loss of forest area in some scenarios ( Figure 3.7 ), carbon uptake from afforestation and reforestation result in net sequestration.

Figure 3.10: Baseline land-use change and forestry carbon net emissions.

Notes: MESSAGE-EMF21 = Rao and Riahi 2006 [SRC] ) scenario from EMF-21 Study; GTEM-EMF21 = Jakeman and Fisher 2006 [NPR, SRC] ) scenario from EMF-21 Study; MESSAGE-A2r = Riahi et al. 2006 [NPR, SRC] ) scenario with revised SRES-A2 baseline; IMAGE 2.3 = van Vuuren et al. 2007 [JoC, SRC, 2007] ) scenario; see Figure 3.7 notes for additional scenario references. The IMAGE 2.3 LUCF baseline scenario also emits non-CO₂ emissions (CH₄ and N₂O) of 0.26, 0.30, 0.16 GtCO₂-eq in 2030, 2050, and 2100, respectively.

There also seems to be a consensus in recent non-CO₂ GHG emission baseline scenarios that agricultural CH₄ and N₂O emissions will increase until the end of this century, potentially doubling in some baselines (see Table 3.1 Kurosawa, 2006 [NPR] ; van Vuuren et al., 2006a [NPR, MoS, SRC] ; Rao and Riahi, 2006 [SRC] ; Jakeman and Fisher, 2006 [NPR, SRC] ; Riahi et al., 2006 [NPR, SRC] ; van Vuuren et al., 2007 [JoC, SRC, 2007] ). The modelling of agricultural emission sources varies across scenarios, with livestock and rice paddy methane and crop soil nitrous oxide emissions consistently represented. However, the handling of emissions from biomass burning and fossil fuel combustion are inconsistent across models; and cropland soil carbon fluxes are generally not reported, probably due to the fact that soil carbon sequestration mitigation options are not currently represented in these models.

Table 3.1: Baseline global agricultural non-CO₂ greenhouse gas emissions from various long-term stabilization scenarios (GtCO₂-eq).

As noted in Section 3.2.1.6 climate change feedbacks could have a significant influence on long-term land use and, to date, are only partially represented in long-term modelling of land scenarios. Similarly, climate feedbacks can also affect land-based emissions. For instance, rising temperatures and CO₂ fertilization can influence the amount of carbon that can be sequestered by land and may also lead to increased afforestation due to higher crop yields. Climate feedbacks in the carbon cycle could be extremely important. For instance, Leemans et al. 2002 [SRC] ) showed that CO₂ fertilization and soil respiration could be as important as the socio-economic drivers in determining the land-use emissions range.

In addition, potentially important additional climate feed-backs in the carbon-climate system are currently not accounted for in integrated assessment scenarios. Specifically, new in-sights suggest that soil drying and forest dieback may naturally reduce terrestrial carbon sequestration ( Cox et al., 2000 [JoC, MoS] ). However, these studies, as well as studies that try to capture changes in climate due to land-use change ( Sitch et al., 2005 [JoC, MoS, SRC] ) have thus far not been able to provide definitive guidance. A modelling system that fully couples land use change scenarios with a dynamic climate-carbon system is required in the future for such an assessment.

3.2.2.3 Non-CO₂ greenhouse gas emissions

The emissions scenario chapter in the TAR ( Morita et al., 2001 [NPR, SRC] ) recommended that future research should include GHGs other than CO₂ in new scenarios work. The reason was that, at that time, certainly regarding mitigation, most of the scenarios literature was still primarily focused on CO₂ emissions from energy. Nevertheless, some multi-gas scenario work existed, including the SRES baseline scenarios, but also some other modelling efforts ( Manne and Richels, 2001 [JoC, SRC] ; Babiker et al., 2001 [NPR, MoS, SRC] ;( Tol, 1999 ) ). The most important non-CO₂ gases include: methane (CH₄), nitrous oxide (N₂O), and a group of fluorinated compounds (F-gases, i.e., HFCs, PFCs, and SF₆). Since the TAR, the number of modelling groups producing long-term emission scenarios for non-CO₂ gases has dramatically increased. As a result, the quantity and quality of non-CO₂ emissions scenarios has improved appreciably.

Unlike CO₂ where the main emissions-related sectors are few (i.e. energy, industry, and land use), non-CO₂ emissions originate from a larger and more diverse set of economic sectors. Table 3.2 provides a list of the major GHG emitting sectors and their corresponding emissions, estimated for 2000 . Note that there is significant uncertainty concerning emissions from some sources of the non-CO₂ gases, and the table summarizes the central values from Weyant et al. 2006 [NPR, SRC] ) which has been used in long-term multi-gas scenario studies of the EMF-21. To make the non-CO₂ emissions comparable to those of CO₂, the common practice is to compare and aggregate emissions by using global warming potentials (GWPs).

Table 3.2: Global Anthropogenic GHG Emissions for 2000 at sector level, as used in EMF-21 studies (MtCO₂-eq/yr).

The most important work on non-CO₂ GHG emissions scenarios has been done in the context of EMF-21 ( De la Chesnaye and Weyant, 2006 [SRC] ). The EMF-21 study updated the capability of long-term integrated assessment models for modelling non-CO₂ GHG emissions. The results of the study are illustrated in Figure 3.11 .

Figure 3.11: Development of baseline emissions in the EMF-21 scenarios (left) and comparison between EMF21 and SRES scenarios (right).

Source: de la Chesnaye and Weyant 2006 [SRC] ; see also van Vuuren et al., 2006b [MoS, SRC]

Evaluating the long-term projections of anthropogenic methane emissions from the EMF-21 data shows a significant range in the estimates, but this range is consistent with that found in the SRES. The methane emission differences in the SRES are due to the different storylines. The differences in the EMF-21 reference cases are mainly due to changes in the economic activity level projected in key sectors by each of the models ^[6] . This could include, for example, increased agriculture production or increased supply of natural gas and below-ground coal in the energy sector. In addition, different modelling groups employed various methods of representing methane emissions in their models and also made different assumptions about how specific methane emission factors for each economic sector change over time. Finally, the degree to which agricultural activities are represented in the models differs substantially. For example, some models represent all agricultural output as one large commodity, ‘agriculture’, while others have considerable disaggregation. Interestingly, the latter group of models tend to find slower emissions growth rates (see van Vuuren et al., 2006b [MoS, SRC] ).

The range of long-term projections of anthropogenic nitrous oxide emissions is wider than for methane in the EMF-21 data. Note that for N₂O, base year emissions of the different models differ substantially. Two factors may contribute to this. First, different definitions exist as to what should be regarded as human-induced and natural emissions in the case of N₂O emissions from soils. Second, some models do not include all emission sources.

The last group of non-CO₂ gases are fluorinated compounds, which include hydrofluorocarbons (HFCs), perfluorocarbons (PFCs), and sulphur hexafluoride (SF₆). The total global emissions of these gases are almost 450 MtCO₂-eq, or slightly over 1% of all GHG for 2000 . While the emissions of some fluorinated compounds are projected to decrease, many are expected to grow substantially because of the rapid growth rate of some emitting industries (e.g. semiconductor manufacture and magnesium production and processing), and the replacement of ozone-depleting substances (ODSs) with HFCs. Long-term projections of these fluorinated GHGs are generated by a fewer number of models, but still show a wide range in the results over the century. Total emissions of non-CO₂ GHGs are projected to increase, but somewhat less rapidly than CO₂ emissions, due to agricultural activities growing less than energy use.

3.2.2.4 Scenarios for air pollutants and other radiative substances

Sulphur dioxide emission scenarios

Sulphur emissions are relevant for climate change modelling as they contribute to the formation of aerosols, which affect precipitation patterns and, taken together, reduce radiative forcing. Sulphur emissions also contribute to regional and local air pollution. Global sulphur dioxide emissions have grown approximately in parallel with the increase in fossil fuel use ( Smith et al., 2001 [SRC] , 2004; ( Stern, 2005 ) ). However, since around the late 1970 s, the growth in emissions has slowed considerably ( Grübler, 2002 [ARC] ). Implementation of emissions controls, a shift to lower sulphur fuels in most industrialized countries, and the economic transition process in Eastern Europe and the Former Soviet Union have contributed to the lowering of global sulphur emissions ( Smith et al., 2001 [SRC] ). Conversely, with accelerated economic development, the growth of sulphur emissions in many parts of Asia has been high in recent decades, although growth rates have moderated recently ( (Streets et al., 2000; ) ( Stern, 2005; ) Cofala et al., 2006 [NPR, ARC] ; Smith et al., 2004 [NPR] ). A review of the recent literature indicates that there is some uncertainty concerning present global anthropogenic sulphur emissions, with estimates for the year 2000 ranging between 55.2 MtS ( (Stern, 2005 ) ), 57.5 MtS ( Cofala et al., 2006 [NPR, ARC] ) and 62 MtS ( Smith et al., 2004 [NPR] ). ^[7]

Many empirical studies have explored the relationship between sulphur emissions and related drivers, such as economic development (see for example, Smith et al., 2004 [NPR] ). The main driving factors that have been identified are increasing income, changes in the energy mix, and a greater focus on air pollution abatement (as a consequence of increasing affluence). Together, these factors may result in an inverted U-shaped pattern of SO₂ emissions, where emissions increase during early stages of industrialization, peak and then fall at higher levels of income, following a Kuznets curve (World Bank, 1992 [NPR] ). This general trend is also apparent in most of the recent emissions scenarios in the literature.

Over time, new scenarios have generally produced lower SO₂ emissions projections. A comprehensive comparison of the SRES and more recent sulphur-emission scenarios is given in van Vuuren and O’Neill 2006 [JoC, MoS, SRC] ). Figure 3.12 illustrates that the resulting spread of sulphur emissions over the medium term (up to the year 2050 ) is predominantly due to the varying assumptions about the timing of future emissions control, particularly in developing countries ^[8] . Scenarios at the lower boundary assume the rapid introduction of sulphur-control technologies on a global scale, and hence, a reversal of historical trends and declining emissions in the initial years of the scenario. Conversely, the upper boundaries of emissions are characterized by a rapid increase over coming decades, primarily driven by the increasing use of coal and oil at relatively low levels of sulphur control (SRES A1 and A2).

Figure 3.12: Sulphur dioxide emission scenarios.

Notes: Thick coloured lines depict the four SRES marker scenarios and the black dashed lines show the median, 5^th and 95^th percentile of the frequency distribution for the full ensemble of all 40 SRES scenarios. The blue area (and the thin dashed lines in blue) illustrates individual scenarios and the range of Smith et al. 2004 [NPR] ). Dotted lines indicate the minimum and maximum of sulphur emissions scenarios developed pre-SRES.

The comparison shows that overall the SRES scenarios are fairly consistent with recent projections concerning the long-term uncertainty range ( Smith et al., 2004 [NPR] ; see Figure 3.12 ). However, the emissions peak over the short-term of some high emissions scenarios in SRES, which lie above the upper boundary estimates of the recent scenarios. There are two main reasons for this difference. First, recent sulphur inventories for the year 2000 have shifted downward. Second, and perhaps more importantly, new information on present and planned sulphur legislation in some developing countries, such as India ( Carmichael et al., 2002 [ARC] ) and China ( Streets et al., 2001 [JoC, SRC] ) has become available. Anticipating this change in legislation, recent scenarios project sulphur emissions to peak earlier and at lower levels compared to the SRES. Also the lower boundary projections of the recent literature have shifted downward slightly compared to the SRES scenario.

NO_x emission scenarios

The most important sources of NO_x emissions are fossil fuel combustion and industrial processes, which combined with other sources such as natural and anthropogenic soil release, biomass burning, lightning, and atmospheric processes, amount to around 25 MtN per year. Considerable uncertainties exist, particularly around the natural sources ( Prather et al., 1995 [NPR, MoS] ; Olivier et al., 1998 [NPR, ARC] ; Olivier and Berdowski, 2001 [NPR, ARC] ; Cofala et al. 2006 [NPR, ARC] ). Fossil fuel combustion in the electric power and transport sectors is the largest source of NO_x, with emissions largely being related to the combustion practice. In recent years, emissions from fossil fuel use in North America and Europe are either constant or declining. Emissions have been increasing in most parts of Asia and other developing parts of the world, mainly due to the growing transport sector ( Cofala et al., 2006 [NPR, ARC] ; Smith, 2005 [MoS] ; WBCSD, 2004 [NPR] ). However in the longer term, most studies project that NO_x emissions in developing countries will saturate and eventually decline, following the trend in the developed world. However, the pace of this trend is uncertain. Emissions are projected to peak in the developing world as early as 2015 ( WBCSD, 2004 [NPR] , focusing on the transport sector) and, in worst cases, around the end of this century (see the high emissions projection of Smith, 2005 [MoS] ).

There have been very few global scenarios for NO_x emissions since the earlier IS92 scenarios and the SRES. An important characteristic of these (baseline) scenarios is that they consider air pollution legislation (in the absence of any climate policy). Some scenarios, such as those by Bouwman and van Vuuren 1999 [NPR, SRC] ( and Collins et al. 2000 ) ) often use IS92a as a ‘loose’ baseline, with new abatement policies added. Many scenarios report rising NO_x emissions up to the 2020 s ( Figure 3.13 ), with the lower boundary given by the short-term Cofala et al. 2006 [NPR, ARC] ) reference scenario, projecting emissions to stay at about present levels for the next two to three decades. In the most recent longer-term scenarios ( Smith, 2005 [MoS] ), NO_x emissions range between 32 MtN and 47 MtN by 2020, which corresponds to an increase in emissions of around 6–50% compared to 2000 . The long-term spread is considerably larger, ranging from 9 MtN to 74 MtN by 2100 (see Figure 3.13 ). The majority of the SRES scenarios (70%) lie within the range of the new Smith 2005 [MoS] ) scenarios. However, the upper and lower boundaries of the range of the recent projections have shifted downward compared to the SRES.

Figure 3.13: NO_x emission scenarios.

Notes: Thick coloured lines depict the four SRES marker scenarios and the black dashed lines show the median, 5^th and 95^th percentile of the frequency distribution for the full ensemble of all 40 SRES scenarios. The blue area illus-trates the range of the recent Smith 2005 [MoS] ) projections.

Emission scenarios for black and organic carbon

Black and organic carbon emissions (BC and OC) are mainly formed by incomplete combustion, as well as from gaseous precursors ( (Penner et al., 1993; ) ( Gray and Cass, 1998 ) ). The main sources of BC and OC emissions include fossil fuel combustion in industry, power generation, traffic and residential sectors, as well as biomass and agriculture waste burning. Natural sources, such as forest fires and savannah burning, are other major contributors. There has recently been some research suggesting that carbonaceous aerosols may contribute to global warming ( Hansen et al., 2000 [JoC] ; Andrae, 2001 [JoC] ; Jacobson, 2001 [JoC] ; Ramaswamy et al., 2001 [NPR, PoC, MoS] ). However, the uncertainty concerning the effects of BC and OC on the change in radiative forcing and hence global warming is still high (see Jacobson, 2001 [JoC] ; and Penner et al., 2004 [JoC, MoS] ).

In the past, BC and OC emissions have been poorly represented in economic and systems engineering models due to unavailability of data. For example, in the IPCC’s Third Assessment Report, BC and OC estimates were developed by using CO emissions ( IPCC, 2001b [NPR] ). One of the main reasons for this has been the lack of adequate global inventories for different emission sources. However, some detailed global and regional emission inventories of BC and OC have recently become available ( Table 3.3 ). In addition, some detailed regional inventories are also available including Streets et al. 2003 [JoC] and Kupiainen and Klimont 2004 [NPR] ). While many of these are comprehensive with regard to detail, considerable uncertainty still exists in the inventories, mainly due to the variety in combustion techniques for different fuels as well as measurement techniques. In order to represent these uncertainties, some studies, such as Bond et al. 2004 [JoC] ), provide high, low and ‘best-guess’ values.

Table 3.3: Emission inventories for black and organic carbon (Tg/yr).

The development in the inventories has resulted in the possibility of estimating future BC and OC emissions. Streets et al. 2004 [JoC, MoS] ) use the fuel-use information and technological change in the SRES scenarios to develop estimates of BC and OC emissions from both contained combustion as well as natural sources for all the SRES scenarios until 2050 Rao et al. 2005 [SRC] and Smith and Wigley 2006 [NPR] ) estimate BC and OC emissions until 2100 for two IPCC SRES scenarios, with an assumption of increasing affluence leading to an additional premium on local air quality. Liousse et al. 2005 [NPR] ) use the fuel-mix and other detail in various energy scenarios and obtain corresponding BC and OC emissions.

The inclusion of technological development is an important factor in estimating future BC and OC emissions because, even though absolute fossil fuel use may increase, a combination of economic growth, increased environmental consciousness, technology development and legislation could imply decreased pollutant emissions ( Figure 3.14 Liousse et al. 2005 [NPR] ) neglect the effects of technological change leading to much higher emission estimates for BC emissions in the long-term in some cases, as compared to other studies such as Streets et al. 2004 [JoC, MoS] Rao et al. 2005 [SRC] and Smith and Wigley 2006 [NPR] ), all of which show declining emissions in the long-term. Another important factor that Rao et al. 2005 [SRC] ) also account for is current and proposed environmental legislation. This suggests the necessity for technology-rich frameworks that capture structural and technological change, as well as policy dynamics in the energy system in order to estimate future BC and OC emissions.

Figure 3.14: Total black carbon (left panel) and organic carbon (right panel) emission estimates in scenarios from different studies.

Notes: Rao et al., 2005 [SRC] ) include emissions from contained combustion only. Liousse et al. 2005 [NPR] ), A2 not included as 2100 value of 100 Tg lies way above range.

Streets et al. 2004 [JoC, MoS] and Rao et al. 2005 [SRC] ) show a general decline in BC and OC emissions in developed countries, as well as in regions such as East Asia (including China). In other developing regions, such as Africa and South Asia, slower technology penetration rates lead to much lower emission reductions. There is a large decline in emissions from the residential sector in the developing countries, due to the gradual replacement of traditional fuels and technologies with more efficient ones. Transport-related emissions in both industrialized and developing countries decline in the long-term due to stringent regulations, technology improvements and fuel switching.

To summarize, an important feature of the recent scenario literature is the long-term decline in BC/OC emission intensities per unit of energy use (or economic activity). The majority of the above studies thus indicate that the long-term BC and OC emissions might be decoupled from the trajectory of CO₂ emissions.

3.3.5.1 Emission reductions and timing

Figure 3.17 shows the projected CO₂ emissions associated with the new mitigation scenarios. In addition, the figure depicts the range of the TAR stabilization scenarios (more than 80 scenarios) ( Morita et al., 2001 [NPR, SRC] ). Independent of the stabilization level, scenarios show that the scale of the emissions reductions, relative to the reference scenario, increases over time. Higher stabilization targets do push back the timing of most reductions, even beyond 2100 .

An increasing body of literature assesses the attainability of very low targets of below 450 ppmv CO₂ (e.g. van Vuuren et al., 2007 [JoC, SRC, 2007] ; Riahi et al., 2006 [NPR, SRC] ). These scenarios from class I and II extend the lower boundary beyond the range of the TAR stabilization scenarios of 450 ppmv CO₂ (see upper panels of Figure 3.17 ). The attainability of such low targets is shown to depend on: 1) using a wide range of different reduction options; and 2) the technology ‘readiness’ of advanced technologies, in particular the combination of bio-energy, carbon capture and geologic storage (BECCS). If biomass is grown sustainably, this combination may lead to negative emissions ( Williams, 1998 [NPR] ; Herzog et al., 2005 [NPR, SRC] Rao and Riahi 2006 [SRC] Azar et al. 2006 [JoC] and van Vuuren et al. 2007 [JoC, SRC, 2007] ) all find that such negative emissions technologies might be essential for achieving very stringent targets.

Figure 3.17: Emissions pathways of mitigation scenarios for alternative groups of stabilization targets (Categories I to VI, see Table 3.5). The pink area gives the projected CO₂ emissions for the recent mitigation scenarios developed post-TAR. Green shaded areas depict the range of more than 80 TAR stabilization scenarios ( Morita et al., 2001 [NPR, SRC] ). Category I and II scenarios explore stabilization targets below the lowest target of the TAR.

Source: After Nakicenovic et al., 2006 [SRC] , and Hanaoka et al., 2006 [NPR, SRC]

The emission range for the scenarios with low and intermediate targets between 3.5 and 5 W/m² (scenarios in categories III and IV) are consistent with the range of the 450 and 550 ppmv CO₂ scenarios in the TAR. Emissions in this category tend to show peak emissions around 2040 – with emissions in 2100 similar to, or slightly below, emissions today. Although for these categories less rapid and forceful reductions are required than for the more stringent targets, studies focusing on these stabilization categories find that a wide portfolio of reduction measures would be needed to achieve such emission pathways in a cost-effective way.

The two highest categories of stabilization scenarios (V and VI) overlap with low-medium category baseline scenarios (see Section 3.2 ). This partly explains the relatively small number of new studies on these categories. The emission profiles of these scenarios are found to be consistent with the emissions ranges as published in the TAR.

There is a relatively strong relationship between the cumulative CO₂ emissions in the 2000 – 2100 period and the stringency of climate targets (see Figure 3.18 ). The uncertainties associated with individual stabilization levels (shown by the different percentiles ^[9] ) are primarily due to the ranges associated with individual stabilization categories, substitutability of CO₂ and non-CO₂-emissions, different model parameterizations of the carbon cycle, but they are also partly due to differences in emissions pathways (delayed reduction pathways can allow for somewhat higher cumulative emissions). In general, scenarios aiming for targets below 3 W/m² require cumulative CO₂ emissions of around 1100 GtCO₂ (range of 800– 1500 GtCO₂). The cumulative emissions increase for subsequently less stringent targets. The middle category (4–5 W/m²) requires emissions to be in the order of 3000 GtCO₂ (range of 2270 –3920 GtCO₂). The highest category (>6 W/m²) exhibits emissions, on average, around 5020 GtCO₂ (range of 4400–6600 GtCO₂).

Figure 3.18: Relationship between the scenario’s cumulative carbon dioxide emissions ( 2000 – 2100 ) and the stabilization target (stabilization categories I to VI, of Table 3.5).

Data source: After Nakicenovic et al., 2006 [SRC] , and Hanaoka et al., 2006 [NPR, SRC]

The timing of emission reductions also depends on the stringency of the stabilization target. Timing of climate policy has always been an important topic in the scenario literature. While some studies argue for early action for smooth transitions and stimulating technology development (e.g. ( Azar and Dowlatabadi, 1999, ) van Vuuren and de Vries, 2001 [JoC, SRC] ), others emphasize delayed response to benefit from better technology and higher CO₂ fertilization rates from natural systems at later points in time (e.g. Wigley et al., 1996 [JoC, SRC] ;( Tol, 2000; ) for a more elaborate discussion on timing see also Section 3.6 ). This implies that a given stabilization target can be consistent with a range of interim targets. Nevertheless, stringent targets require an earlier peak of CO₂ emissions (see Figure 3.19 and Table 3.5 ). In the majority of the scenarios concerning the most stringent group (< 3 W/m²), emissions start to decline before 2015, and are further reduced to less than 50% of today’s emissions by 2050 ( Table 3.5 ). The emissions profiles of these scenarios indicate the need for short-term infrastructure investments for a comparatively early decarbonization of the energy system. Achieving these low-emission trajectories requires a comprehensive global mitigation effort, including a further tightening of existing climate policies in Annex I countries, and simultaneous emission mitigation in developing countries, where most of the increase in emissions is expected in the coming decades. For the medium stringency group (4-5 W/m²) the peak of global emissions generally occurs around 2010 to 2030; followed by a return to 2000 levels, on average, around 2040 (with the majority of these scenarios returning to 2000 emissions levels between 2020 and 2060 ). For targets between 5–6 W/m², the median emissions peak around 2070 . The figure also indicates that the uncertainty range is relatively small for the more stringent targets, illustrating the reduced flexibility of the emissions path and the requirement for early mitigation. The less stringent categories allow more flexibility in timing. Most of the stringent stabilization scenarios of category I (and some II scenarios) assume a temporal overshoot of the stabilization target (GHG concentration, radiative forcing, or temperature change) before the eventual date of stabilization between 2100 and 2150 . Recent studies indicate that while such ‘overshoot’ strategies might be inevitable for very low targets (given the climate system and socio-economic inertia), they might also provide important economic benefits. At the same time, however, studies note that the associated rate of warming from large overshoots might significantly increase the risk of exceeding critical climate thresholds. (For further discussion, see Section 3.3.4 .)

The right-hand panel of Figure 3.19 illustrates the time at which CO₂ emissions will have to return to present levels. For stringent stabilization targets (below 4 W/m²; category I, II and III) emissions return to present levels, on average, before the middle of this century, that is about one to two decades after the year in which emissions peak. In most of the scenarios for the highest stabilization category (above 6 W/m²; category VI) emissions could stay above present levels throughout the century.

Figure 3.19: Relationship between the stringency of the stabilization target (category I to VI) and 1) the time at which CO₂ emissions have to peak (left-hand panel), and 2) the year when emissions return to present ( 2000 ) levels.

Data source: After Nakicenovic et al., 2006 [SRC] , and Hanaoka et al., 2006 [NPR, SRC] .

The absolute level of the required emissions reduction does not only depend on the stabilization target, but also on the baseline emissions (see Hourcade and Shukla, 2001 [NPR, ARC] ). This is clearly shown in the right-hand panel of Figure 3.20 , which illustrates the relationship between the cumulative baseline emissions and the cumulative emissions reductions for the stabilization scenarios (by 2100 ). In general, scenarios with high baseline emissions require a higher reduction rate to reach the same reduction target: this implies that the different reduction categories need to show up as diagonals in figure 3.20 . This is indeed the case for the range of studies and the ‘category averages’ (large triangles). As indicated in the figure, a scenario with high baseline emissions requires much deeper emission reduction in order to reach a medium stabilization target (sometimes more than 3600 GtCO₂) than a scenario with low baseline emissions to reach the most stringent targets (in some cases less than 1800 GtCO₂). For the same target (e.g. category IV) reduction may differ from 370 to 5500 GtCO₂. This comes from the large spread of emissions in the baseline scenarios. While scenarios for both stringent and less-stringent targets have been developed from low and high baseline scenarios, the data suggests that, on average, mitigation scenarios aimed at the most stringent targets start from the lowest baseline scenarios.

In the short-term ( 2030 ), the relationship between emission reduction and baseline is less clear, given the flexibility in the timing of emission reductions (left-hand panel in Figure 3.20 ). While the averages of the various stabilization categories are aligned in a similar way to those discussed for 2100 (with exception of category I, for which the scenario sample is smaller than for the other categories); the uncertainty ranges here are very large.

Figure 3.20: Relationship between required cumulative emissions reduction and carbon emissions in the baseline by 2030 (left-hand panel) and 2100 (right-hand panel).

Notes: Coloured rectangles denote individual scenarios for alternative stabilization targets (categories I to VI). The large triangles indicate the averages for each category.

Data source: After Nakicenovic et al., 2006 [SRC] , and Hanaoka et al., 2006 [NPR, SRC]

3.3.5.2 GHG abatement measures

The abatement of GHG emissions can be achieved through a wide portfolio of measures in the energy, industry, agricultural and forest sectors (see also Edmonds et al., 2004b [MoS, ARC] ; Pacala and Socolow, 2004 [JoC] ; Metz and van Vuuren, 2006 [NPR, SRC] ). Measures for reducing CO₂ emissions range from structural changes in the energy system and replacement of carbon-intensive fossil fuels by cleaner alternatives (such as a switch from coal to natural gas, or the enhanced use of nuclear and renewable energy), to demand-side measures geared towards energy conservation and efficiency improvements. In addition, capturing carbon during energy conversion processes with subsequent storage in geological formations (CCS) provides an approach for reducing emissions. Another important option for CO₂ emission reduction encompasses the enhancement of forest sinks through afforestation, reforestation activities and avoided deforestation.

In the energy sector the aforementioned options can be grouped into two principal measures for achieving CO₂ reductions:

1. Improving the efficiency of energy use (or measures geared towards energy conservation).

2. Reducing the emissions per unit of energy consumption.

The latter comprises the aggregated effect of structural changes in the energy systems and the application of CCS. A response index has been calculated (based on the full set of stabilization scenarios from the database) in order to explore the importance of these two strategies. This index is equal to the ratio of the reductions achieved through energy efficiency over those achieved by carbon-intensity improvements ( Figure 3.21 ). Similar to Morita et al. 2001 [NPR, SRC] ), it was discovered that the mitigation response to reduce CO₂ emissions would shift over time, from initially focusing on energy efficiency reductions in the beginning of the 21^st century to more carbon-intensity reduction in the latter half of the century ( Figure 3.21 ). The amount of reductions coming from carbon-intensity improvement is more important for the most stringent scenarios. The main reason is that, in the second half of the century, increasing costs of further energy efficiency improvements and decreasing costs of low-carbon or carbon-free energy sources make the latter category relatively more attractive. This trend is also visible in the scenario results of model comparison studies ( Weyant, 2004 [SRC] ; Edenhofer et al., 2006 [NPR, ARC] ).

Figure 3.21: Response index to assess priority setting in energy-intensity reduction (more than 1) or in carbon-intensity reduction (less than 1) for post-TAR stabilization scenarios.

Note: The panels show the development of the index for the years 2020, 2050, and 2100 (66, 77, and 59 scenarios, respectively, for which data on energy, GDP and carbon emissions were available.

Data source: After Nakicenovic et al., 2006 [SRC] , and Hanaoka et al., 2006 [NPR, SRC] )

In addition to measures for reducing CO₂ emissions from energy and industry, emission reductions can also be achieved from other gases and sources. Figure 3.22 illustrates the relative contribution of measures towards achieving climate stabilization from three main sources:

1. CO₂ from energy and industry.

2. CO₂ from land-use change.

3. The full basket of non-CO₂ emissions from all relevant sources.

Figure 3.22: Cumulative contribution of alternative mitigation measures by source ( 2000 – 2100 )

Note: Contributions for a wide range of stabilization targets (2.6–5.3 W/m²), indicated after the model name) and alternative baseline scenarios. Mitigation scenarios using the same baseline are indicated for each model as (a) and (b) reprectively.

Data source: EMF-21, Smith and Wigley., 2006 [NPR] ; van Vuuren et al., 2007 [JoC, SRC, 2007] ; Riahi et al., 2006 [NPR, SRC]

The figure compares the contribution of these measures towards achieving stabilization for a wide range of targets (between 2.6 and 5.3 W/m² by 2100 ) and baseline scenarios. An important conclusion across all stabilization levels and baseline scenarios is the central role of emissions reductions in the energy and industry sectors. All stabilization studies are consistent in that (independent of the baseline or target uncertainty) more than 65% of total emissions reduction would occur in this sector. The non-CO₂ gases and land-use-related CO₂ emissions (including forests) are seen to contribute together up to 35% of total emissions reductions. ^[10] However, as noted further above, the majority of recent studies indicate the relative importance of the latter two sectors for the cost-effectiveness of integrated multi-gas GHG abatement strategies (see also Section 3.3.5.4 on CO₂-only versus multi-gas mitigation and 3.3.5.5 on land-use).

The strongest divergence across the scenarios concerns the contribution of land-use-related mitigation. The results range from negative contributions of land-use change to potential emissions savings of more than 1100 GtCO₂ over the course of the century ( Figure 3.22 ). The primary reason for this is the considerable uncertainty with respect to future competition for land between dedicated bio-energy plantations and potential gains from carbon savings in terrestrial sinks. Some scenarios, for example, project massive expansion of dedicated bio-energy plantations, leading to an increase in emissions due to net deforestation (compared to the baseline).

An illustrative example for the further breakdown of mitigation options is shown in Figure 3.23 . The figure shows stabilization scenarios for a range of targets (about 3–4.5 W/m²) based on four illustrative models (IMAGE, MESSAGE, AIM and IPAC) for which sufficient data were available. The scenarios share similar stabilization targets, but differ with respect to salient assumptions for technological change, long-term abatement potentials, as well as model methodology and structure. The scenarios are also based on different baseline scenarios. For example, cumulative baseline emissions over the course of the century range between 6000 GtCO₂-eq in MESSAGE and IPAC scenarios to more than 7000 GtCO₂-eq in the IMAGE and AIM scenarios. Figure 3.24 shows the primary energy mix of the baseline and the mitigation scenarios.

It should be noted that the figure shows reduction on top of the baseline (e.g. other renewables may already make a large baseline contribution). Above all, Figure 3.23 illustrates the importance of using a wide portfolio of reduction measures, with many categories of measures, showing contributions of more than a few hundred GtCO₂ over the course of the century. In terms of the contribution of different options, there is agreement for some options, while there is disagreement for others. The category types that have a large potential over the long term ( 2000 – 2100 ) in at least one model include energy conservation, carbon capture and storage, renewables, nuclear and non-CO₂ gases. These options could thus constitute an important part of the mitigation portfolio. However, the differences between the models also emphasize the impact of different assumptions and the associated uncertainty (e.g. for renewables, results can vary strongly depending on whether they are already used in the baseline, and how this category competes against other zero or low-emission options in the power sector, such as nuclear and CCS). The figure also illustrates that the limitations of the mitigation portfolio with respect to CCS or forest sinks (AIM and IPAC) would lead to relatively higher contributions of other options, in particular nuclear (IPAC) and renewables (AIM).

Figure 3.23 also illustrates the increase in emissions reductions necessary to strengthen the target from 4.5 to about 3–3.6 W/m². Most of the mitigation options increase their contribution significantly by up to a factor of more than two. This effect is particularly strong over the short term ( 2000 – 2030 ), indicating the need for early abatement in meeting stringent stabilization targets. Another important conclusion from the figure is that CCS and forest sink options are playing a relatively modest role in the short-term mitigation portfolio, particularly for the intermediate stabilization target (4.5 W/m²). The results thus indicate that the widespread deployment of these options might require relatively more time compared to the other options and also relatively higher carbon prices (see also Figure 3.25 on increasing carbon prices over time).

Figure 3.23: Cumulative emissions reductions for alternative mitigation measures for 2000 to 2030 (left-hand panel) and for 2000 - 2100 (right-hand panel). The figure shows illustrative scenarios from four models (AIM, IMAGE, IPAC and MESSAGE) for stabilization levels of 490-540 ppmv CO₂-eq and levels of 650 ppmv CO₂-eq, respectively. Dark bars denote reductions for a target of 650 ppmv CO₂-eq and light bars the additional reductions to achieve 490-540 ppmv CO₂-eq. Note that some models do not consider mitigation through forest sink enhancement (AIM and IPAC) or CCS (AIM) and that the share of low-carbon energy options in total energy supply is also determined by inclusion of these options in the baseline. CCS includes carbon capture and storage from biomass. Forest sinks include reducing emissions from deforestation.

Data source: van Vuuren et al. 2007 [JoC, SRC, 2007] Riahi et al. 2006 [NPR, SRC] ( Hijioka, et al. 2006 ) Masui et al. 2006 [SRC] Jiang et al. 2006 [NPR, SRC] ).

As noted above, assumptions with regards to the baseline can have significant implications for the contribution of individual mitigation options in achieving stabilization. Figure 3.24 clearly shows that the baseline assumptions of the four models differ, and that these differences play a role in explaining some of the results. For instance, the MESSAGE model already includes a large amount of renewables in its baseline and further expansion is relatively costly. Nevertheless, some common trends among the models may also be observed. First of all, almost all cases show a clear reduction in primary energy use. Second, in all models coal use is significantly reduced under the climate policy scenarios, compared to the baseline. It should be noted that in those models that consider CCS, the remaining fossil fuel use is mostly in combination with carbon capture and storage. In 2030, oil use is only modestly reduced by climate policies – this also applies to natural gas use. In 2100, both oil and gas are reduced compared to the baseline in most models. Finally, renewable energy and nuclear power increase in all models – although the distribution across these two options differs.

Figure 3.24: Primary energy mix for the years 2030 and 2100 . Illustrative scenarios aim at stabilizing radiative forcing at low (3–3.6 W/m²) and intermediate levels (4.5 W/m2) respectively.

Note: BL= Baseline. For the corresponding contribution of individual mitigation measures in (in GtCO₂) see also Figure 3.23 .

3.3.5.3 Stabilization costs

Models use different metrics to report the costs of emission reductions. Top-down general equilibrium models tend to report GDP losses, while system-engineering partial equilibrium models usually report the increase in energy system costs or the net present value (NPV) of the abatement costs. A common cost indicator is also the marginal cost/price of emissions reduction (US$/tC or US$/tCO₂).

Figure 3.25 shows the relationship between stabilization targets and alternative measures of mitigation costs, comprising GDP losses, net present value of abatement, and carbon price in terms of US$ /tCO₂-eq.

Figure 3.25: Relationship between the cost of mitigation and long-term stabilization targets (radiative forcing compared to pre-industrial level, W/m² and CO₂-eq concentrations).

Notes: These panels show costs measured as a % loss of GDP (top), net present value of cumulative abatement costs (middle), and carbon price (bottom). The left-hand panels give costs for 2030, the middle panel for 2050, and the right-hand panel for 2100 repectively. Individual coloured lines denote selected studies with representative cost dynamics from very high to very low cost estimates. Scenarios from models sharing similar baseline assumptions are shown in the same colour. The grey-shaded range represents the 80th percentile of the TAR and post-TAR scenarios. NPV calculations are based on a discount rate of 5%. Solid lines show representative scenarios considering all radiatively active gases. CO₂ stabilization scenarios are added based on the relationship between CO₂ concentration and the radiative forcing targets shown in Figure 3.16 . Dashed lines represent multi-gas scenarios, where the target is defined by the six Kyoto gases (other multi-gas scenarios consider all radiatively active gases).

Data sources: CCSP scenarios ( USCCSP, 2006 [NPR] ); IMCP scenarios ( Edenhofer et al., 2006 [NPR, ARC] ); Post-SRES (PS) scenarios ( Morita et al., 2001 [NPR, SRC] Azar et al., 2006 [JoC] ; Riahi et al., 2006 [NPR, SRC] ; van Vuuren et al., 2007 [JoC, SRC, 2007] .

It is important to note that for the following reported cost estimates, the vast majority of the models assume transparent markets, no transaction costs, and thus perfect implementation of policy measures throughout the 21^st century, leading to the universal adoption of cost-effective mitigation measures, such as carbon taxes or universal cap and trade programmes. These assumptions generally result in equal carbon prices across all regions and countries equivalent to global, least-cost estimates. Relaxation of these modelling assumptions, alone or in combination (e.g. mitigation-only in Annex I countries, no emissions trading, or CO₂-only mitigation), will lead to an appreciable increase in all cost categories.

The grey shaded area in Figure 3.25 illustrates the 10^th–90^th percentile of the mitigation cost ranges of recent studies, including the TAR. The area includes only those recent scenarios in the literature that report cost estimates based on a comprehensive mitigation analysis, defined as those that have a sufficiently wide portfolio of mitigation measures. ^[11] The selection was made on a case-by-case basis for each scenario considered in this assessment. The Figure also shows results from selected illustrative studies (coloured lines). These studies report costs for a range of stabilization targets and are representative of the overall cost dynamics of the full set of scenarios. They show cases with high-, intermediate- and low-cost estimates (sometimes exceeding the 80^th (i.e. 10^th–90^th) percentile range on the upper and lower boundaries of the grey-shaded area). The colour coding is used to distinguish between individual mitigation studies that are based on similar baseline assumptions. Generally, mitigation costs (for comparable stabilization targets) are higher from baseline scenarios with relatively high baseline emissions (brown and red lines). By the same token, intermediate or low baseline assumptions result in relatively lower cost estimates (blue and green lines).

Figure 3.25 a shows that the majority of studies find that GDP losses increase with the stringency of the target, even though there is considerable uncertainty with respect to the range of losses. Barker et al. 2006 [NPR, SRC] ) found that, after allowing for baseline emissions, the differences can be explained by:

• The spread of assumptions in modelling-induced technical change.
• The use of revenues from taxes and permit auctions.
• The use of flexibility mechanisms (i.e. emissions trading, multi-gas mitigation, and banking).
• The use of backstop technologies.
• Allowing for climate policy related co-benefits.
• Other specific modelling assumptions.

Weyant 2000 [NPR, SRC] ) lists similar factors but also includes the number and type of technologies covered, and the possible substitution between cost factors (elasticities). A limited set of studies finds negative GDP losses (economic gains) that arise from the assumption that a model’s baseline is assumed to be a non-optimal pathway and incorporates market imperfections. In these models, climate policies steer economies in the direction of reducing these imperfections, for example by promoting more investment into research and development and thus achieving higher productivity, promoting higher employment rates, or removing distortionary taxes.

The left-hand side panel of Figure 3.25 a shows that for 2030, GDP losses in the vast majority of the studies (more than 90% of the scenarios) are generally below 1% for the target categories V and VI. Also in the majority of the category III and IV scenarios (70% of the scenarios) GDP losses are below 1%. However, it is important to note that for categories III and IV costs are higher, on average, and show a wider range than those for categories V and VI. For instance, for category IV the interval lying between the 10^th and 90^th percentile varies from about 0.6% gain to about 1.2% loss. For category III, this range is shifted upwards (0.2–2.5%). This is also indicated by the median GDP losses by 2030, which increases from below 0.2% for categories V and VI, to about 0.2% for the category IV scenarios, and to about 0.6% for category III scenarios. GDP losses of the lowest stabilization categories (I & II) are generally below 3% by 2030, however the number of studies are relatively limited and in these scenarios stabilization is achieved predominantly from low baselines. The absolute GDP losses by 2030 correspond on average to a reduction of the annual GDP growth rate of less than 0.06 percentage points for the scenarios of category IV, and less than 0.1 and 0.12 percentage points for the categories III and I&II, respectively.

GDP losses by 2050 (middle panel of Figure 3.25 a) are comparatively higher than the estimates for 2030 . For example, for category IV scenarios the range is between -1% and 2% GDP loss compared to baseline (median 0.5%), and for category III scenarios the range is from slightly negative to 4% (median 1.3%). The Stern review ( 2006 ), looking at the costs of stabilization in 2050 for a comparable category (500–550 CO₂-eq) found a similar range of between -2% and +5%. For the studies that also explore different baselines (in addition to multiple stabilization levels), Figure 3.25 a also shows that high emission baselines (e.g. high SRES-A1 or A2 baselines) tend to lead to higher costs. However, the uncertainty range across the models is at least of a similar magnitude. Generally, models that combine assumptions of very slow or incremental technological change with high baseline emissions (e.g. IGSM-CCSP) tend to show the relatively highest costs ( Figure 3.25 a). GDP losses of the lowest stabilization categories (I & II) are generally below 5.5% by 2050, however the number of studies are relatively limited and in these scenarios stabilization is achieved predominantly from low baselines. The absolute GDP losses numbers for 2050 reported above correspond on average to a reduction of the annual GDP growth rate of less than 0.05 percentage points for the scenarios of category IV, and less than 0.1 and 0.12 percentage points for the categories III and I&II, respectively.

Finally, the most right-hand side panel of Figure 3.25 a shows that GDP losses show a bigger spread and tend to be somewhat higher by 2100 . GDP losses are between 0.3% and 3% for category V scenarios and -1.6% to about 5% for category IV scenarios. Highest costs are given by category III (from slightly negative costs up to 6.5%). The sample size for category I is not large enough for a statistical analysis. Similarly, for category II scenarios, the range is not shown as the stabilization scenarios of category II are predominantly based on low or intermediate baselines, and thus the resulting range would not be comparable to those from the other stabilization categories. However, individual studies indicate that costs become higher for more stringent targets (see, for example, studies highlighted in green and blue for the lowest stabilization categories in Figure 3.25 a). ^[12]

The results for the net present value of cumulative abatement costs show a similar picture ( Figure 3.25 b). However, given the fact that abatement costs only capture direct costs, this cost estimate is by definition more certain. ^[13] The interval from the 10^th to the 90^th percentile in 2100 ranges from nearly zero to about 11 trillion US$. The highest level corresponds to around 2–3% of the NPV of global GDP over the same period. Again, on the basis of comparison across models, it is clear that costs depend both on the stabilization level and baseline emissions. In general, the spread of costs for each stabilization category seems to be of a similar order to the differences across stabilization scenarios from different baselines. In 2030, the interval covering 80% of the NPV estimates runs from around 0–0.3 trillion for category IV scenarios. The majority of the more stringent (category III) scenarios range between 0.2 to about 1.6 trillion US$. In 2050, typical numbers for category IV are around 0.1–1.2 trillion US$ and, for category III, this is 1–5 trillion US$ (or below about 1% of the NPV of GDP). By 2100 the NPV estimates increase further, with the range up to 5 trillion for category IV scenarios and up to 11 trillion for category III scenarios, respectively. The results of these studies, published since the TAR, are consistent with the numbers presented in the TAR, although the new studies extend results to substantially lower stabilization levels.

Finally, a similar trend is found for carbon price estimates. In 2030, typical carbon prices across the range of models and baselines for a 4.5 W/m² stabilization target (category IV) range from around 1–24 US$/tCO₂ (80% of estimates), with the median of about 11 US$/tCO₂. For category III, the corresponding prices are somewhat higher and range from 18–79 US$/tCO₂ (with the median of the scenarios around 45 US$/tCO₂). Most individual studies for the most stringent category cluster around prices of about 100 US$/tCO₂. ^[14] Carbon prices by 2050 are comparatively higher than those in 2030 . For example, costs of category IV scenarios by 2050 range between 5 and 65 US$/tCO₂, and those for category III range between 30 and 155 US$/tCO₂. Carbon prices in 2100 vary over a much wider range – mostly reflecting uncertainty in baseline emissions and technology development. For the medium target of 4.5 W/m², typical carbon prices in 2100 range from 25–200 US$/tCO₂ (80% of estimates). This is primarily a consequence of the nature of this metric, which often represents costs at the margin. Costs tend to slowly increase for more stringent targets – with a range between the 10^th and 90^th percentile of more than 35 to about 350 US$/tCO₂ for category III.

3.3.5.4 The role of non-CO₂ GHGs

As also illustrated by the scenario assessment in the previous sections, more and more attention has been paid since the TAR to incorporating non-CO₂ gases into climate mitigation and stabilization analyses. As a result, there is now a body of literature (e.g. van Vuuren et al., 2006b [MoS, SRC] ; de la Chesnaye and Weyant, 2006 [SRC] ; de la Chesnaye et al., 2007 [NPR, MoS, SRC, 2007] ) showing that mitigation costs for these sectors can be lower than for energy-related CO₂ sectors. As a result, when all these options are employed in a multi-gas mitigation policy, there is a significant potential for reduced costs, for a given climate policy objective, versus the same policy when CO₂ is the only GHG directly mitigated. These cost savings can be especially important where carbon dioxide is not the dominant gas, on a percentage basis, for a particular economic sector and even for a particular region. While the previous sections have focused on the joint assessment of CO₂ and multi-gas mitigation scenarios, this section explores the specific role of non-CO₂ emitting sectors. ^[15]

A number of parallel numerical experiments have been carried out by the Energy Modelling Forum (EMF-21; de la Chesnaye and Weyant, 2006 [SRC] ). The overall conclusion is that economic benefits of multi-gas strategies are robust across all models. This is even true, despite the fact that different methods were used in the study to compare the relative contribution of these gases in climate forcing (see Section 3.3.3 ). The EMF-21 study specifically focused on comparing stabilization scenarios aiming for 4.5 W/m² compared to pre-industrial levels. There were two cases employed to achieve the mitigation target:

1. Directly mitigate CO₂ emissions from the energy sector (with some indirect reduction in non-CO₂ gases).

2. Mitigate all available GHG in costs-effective approaches using full ‘what’ flexibility.

In the CO₂-only mitigation scenario, all models significantly reduced CO₂ emissions, on average by about 75% in 2100 compared to baseline scenarios. Models still indicated some emission reductions for CH₄ and N₂O as a result of systemic changes in the energy system. Emissions of CH₄ were reduced by about 20% and N₂O by about 10% ( Figure 3.26 ).

Figure 3.26: Reduction of emissions in the CO₂-only versus multi-gas strategies.

Source: de la Chesnaye and Weyant, 2006 [SRC] (see also van Vuuren et al., 2006b [MoS, SRC] )

In the multi-gas mitigation scenario, all models found that an appreciable percentage of the emission reductions occur through reductions of non-CO₂ gases, which then results in smaller required reductions of CO₂. The emission reduction for CO₂ in 2100 therefore drops (on average) from 75% to 67%. This percentage is still rather high, caused by the large share of CO₂ in total emissions (on average, 60% in 2100 ) and partly due to the exhaustion of reduction options for non-CO₂ gases. The reductions of CH₄ across the different models averages around 50%, with remaining emissions coming from sources for which no reduction options were identified, such as CH₄ emissions from enteric fermentation. For N₂O, the increased reduction in the multi-gas strategy is not as large as for CH₄ (almost 40%). The main reason is that the identified potential for emission reductions for the main sources of N₂O emissions, fertilizer use and animal manure, is still limited. Finally, for the fluorinated gases, high reduction rates (about 75%) are found across the different models.

Although the contributions of different gases change sharply over time, there is a considerable spread among the different models. Many models project relatively early reductions of both CH₄ and the fluorinated gases under the multi-gas case. However, the subset of models that does not use GWPs as the substitution metric for the relative contributions of the different gases to the overall target, but does assume inter-temporal optimization in minimizing abatement costs, do not start to reduce CH₄ emissions substantially until the end of the period. The increased flexibility of a multi-gas mitigation strategy is seen to have significant implications for the costs of stabilization across all models participating in the EMF-21. These scenarios concur that multi-gas mitigation is significantly cheaper than CO₂-only. The potential reductions of the GHG price ranges in the majority of the studies between 30% and 85% (See Figure 3.27 ).

Figure 3.27: Reduction in GHG abatement price (%) in multi-gas stabilization scenarios compared to CO₂-only cases. Ranges correspond to alternative scenarios for a stabilization target of 4.5 W/m².

Data source: de la Chesnaye and Weyant, 2006 [SRC]

Finally, the EMF-21 research also showed that, for some sources of non-CO₂ gases, the identified reduction potential is still very limited (e.g. most agricultural sources for N₂O emissions). For long-term scenarios (and more stringent targets) in particular, identifying how this potential may develop in time is a crucial research question. Attempts to estimate the maximum feasible reductions (and the development of potential over time) have been made in van Vuuren et al. 2007 [JoC, SRC, 2007] ).

3.3.5.5 Land use

Changes in land-use practices are regarded as an important component of long-term strategies to mitigate climate change. Modifications to land-use activities can reduce emissions of both CO₂ and non-CO₂ gases (CH₄ and N₂O), increase sequestration of atmospheric CO₂ into plant biomass and soils, and produce biomass fuel substitutes for fossil fuels (see Chapters 4 , 8 , and 9 of this report for discussions of detailed land-related mitigation alternatives). Available information before the TAR suggested that land has the technical potential to sequester up to an additional 319 billion tonnes of CO₂ (GtCO₂) by 2050 in global forests alone ( IPCC, 1996a [NPR] ; IPCC, 2000 [NPR] ; IPCC, 2001a [NPR] ). In addition, current technologies are capable of substantially reducing CH₄ and N₂O emissions from agriculture (see Chapter 8 ). A number of global biomass energy potential assessments have also been conducted (see Berndes et al. 2003 [MoS, ARC] for an overview). ^[16]

The explicit modelling of land-based climate change mitigation in long-term global scenarios is relatively new and rapidly developing. As a result, assessment of the long-term role of global land-based mitigation was not formally addressed by the Special Report on Land use, Land-use Change, and Forestry ( IPCC, 2000 [NPR] ) or the TAR. This section assesses the modelling of land in long-term climate stabilization and the relationship to detailed global forestry mitigation estimates from partial equilibrium sectoral models that model 100-year carbon price trajectories.

Development of, among other things, global sectoral land mitigation models (e.g. Sohngen and Sedjo, 2006 [NPR, ARC] ), bottom-up agricultural mitigation costs for specific technologies (e.g. USEPA, 2006b [NPR] ), and biomass technical potential studies (e.g. Hoogwijk et al., 2005 [SRC] ) has facilitated the formal incorpo-ration of land mitigation in long-term integrated assessment of climate change stabilization strategies. Hoogwijk et al. 2005 [SRC] ), for example, estimated the potential of abandoned agricultural lands for providing biomass for primary energy demand and identified the technical biomass supply limits of this land type (e.g. under the SRES A2 scenario, abandoned agricultural lands could provide for 20% of 2001 total energy demand). Sands and Leimbach 2003 [JoC, MoS] ) conducted one of the first studies to explicitly explore land-based mitigation in stabilization, suggesting that the total cost of stabilization could be reduced by including land strategies in the set of eligible mitigation options (energy crops in this case). The Energy Modelling Forum Study-21 (EMF-21; de la Chesnaye and Weyant, 2006 [SRC] ) was the first coordinated stabilization modelling effort to include an explicit evaluation of the relative role of land in stabilization; however, only a few models participated. Building on their EMF-21 efforts, some modelling teams have also generated even more recent stabilization scenarios with revised land modelling. These studies are conspicuously different in the specifics of their modelling of land and land-based mitigation ( Rose et al., 2007 [NPR, MoS, SRC, 2007] ). Differences in the types of land considered, emissions sources, and mitigation alternatives and implementation imply different opportunities and opportunity costs for land-related mitigation; and, therefore, different outcomes.

Four of the modelling teams in the EMF-21 study directly explored the question of the cost-effectiveness of including land-based mitigation in stabilization solutions and found that including these options (both non-CO₂ and CO₂) provided greater flexibility and was cost-effective for stabilizing radiative forcing at 4.5 W/m² ( Kurosawa, 2006 [NPR] ; van Vuuren et al., 2006a [NPR, MoS, SRC] ; Rao and Riahi, 2006 [SRC] ; Jakeman and Fisher, 2006 [NPR, SRC] Jakeman and Fisher 2006 [NPR, SRC] ), for example, found that including land-use change and forestry mitigation options reduced the emissions reduction burden on all other emissions sources such that the projected decline in global real GDP associated with achieving stabilization was reduced to 2.3% at 2050 (3.4 trillion US$), versus losses of around 7.1% (10.6 trillion US$) and 3.3% (4.9 trillion US$) for the CO₂-only and multi-gas scenarios, respectively. ^[17] Unfortunately, none of the EMF-21 papers isolated the GDP effects associated with biomass fuel substitution or agricultural non-CO₂ abatement. However, given agriculture’s small estimated share of total abatement (discussed below), the GDP savings associated with agricultural non-CO₂ abatement could be expected to be modest overall, though potentially strategically significant to the dynamics of mitigation portfolios. Biomass, on the other hand, may have a substantial abatement role and therefore a large effect on the economic cost of stabilization. Notably, strategies for increasing cropland soil carbon have not been incorporated to date into this class of models (see Chapter 8 for an estimate of the short-term potential for enhancing agricultural soil carbon).

Figure 3.28 presents the projected mitigation from forestry, agriculture, and biomass for the EMF-21 4.5 W/m² stabilization scenarios, as well as additional scenarios produced by the MESSAGE and IMAGE models – an approximate 3 W/m² scenario from Rao and Riahi 2006 [SRC] ), a 4.5 W/m² scenario from Riahi et al. 2006 [NPR, SRC] ), and approximately 4.5, 3.7, and 2.9 W/m² scenarios from van Vuuren et al. 2007 [JoC, SRC, 2007] ) (see Rose et al., 2007 [NPR, MoS, SRC, 2007] , for a synthesis). While there are clearly different land-based mitigation pathways being taken across models for the same stabilization target, and across targets with the same model and assumptions, some general observations can be made. First, forestry, agriculture, and biomass are called upon to provide significant cost-effective mitigation contributions ( Rose et al., 2007 [NPR, MoS, SRC, 2007] ). In the short-term ( 2000 – 2030 ), forest, agriculture, and biomass together could account for cumulative abatement of 10–65 GtCO₂-eq, with 15–60% of the total abatement considered by the available studies, and forest/agricultural non-CO₂ abatement providing at least three quarters of total land abatement. ^[18] Over the entire century ( 2000 – 2100 ), cumulative land-based abatement of approximately 345– 1260 GtCO₂-eq is estimated to be cost-effective, accounting for 15–40% of total cumulative abatement. Forestry, agriculture, and biomass abatement levels are each projected to grow annually with relatively stable annual increases in agricultural mitigation and gradual deployment of biomass mitigation, which accelerates dramatically in the last half of the century to become the dominant land-mitigation strategy.

Figures 3.28 and 3.29 show that additional land-based abatement is expected to be cost-effective with tighter stabilization targets and/or higher baseline emissions (e.g. see the IMAGE 2.3 results for various stabilization targets and the MESSAGE 4.5 W/m² stabilization results with B2 (EMF-21) and A2r baselines). Biomass is largely responsible for the additional abatement; however, agricultural and forestry abatement are also expected to increase. How they might increase is model and time dependent. In general, the overall mitigation role of agricultural abatement of rice methane, livestock methane, nitrous oxide (enteric and manure) and soil nitrous oxide is projected to be modest throughout the time horizon, with some suggestion of increased importance in early decades.

Figure 3.28: Cost-effective agriculture, forest, and commercial biomass annual greenhouse gas emissions abatement from baselines from various 2100 stabilization scenarios (note y-axes have different ranges).

Notes: The colour of the line indicates the 2100 stabilization target modelled: green < 3.25 W/m² (< 420 CO₂ concentration, < 510 CO₂-eq concentration), pink 3.25–4 (42–490, 510–590), and dark blue 4–5 (490–570, 59–710). The IMAGE-EMF21 and IMAGE 2.3 forest results are net of deforestation carbon losses induced by bio-energy crop extensification. These carbon losses are accounted for under forestry by the other scenarios. The MESSAGE-EMF21 results are taken from the sensitivity analysis of Rao and Riahi 2006 [SRC] ). The GTEM-EMF21 scenarios ran through 2050 and the GTEM agriculture mitigation results include fossil fuel emissions reductions in agriculture (5-7% of the annual agricultural abatement). Scenario references: IMAGE-EMF21 ( Van Vuuren et al., 2006a [NPR, MoS, SRC] ); MESSAGE-EMF21 ( Rao and Riahi, 2006 [SRC] ); MESSAGE-A2r ( Riahi et al., 2006 [NPR, SRC] ); GRAPE-EMF21 ( Kurosawa, 2006 [NPR] ); GTEM-EMF21 ( Jakeman and Fisher, 2006 [NPR, SRC] ); and IMAGE 2.3 ( Van Vuuren et al., 2007 [Ambiguous] ).

Source: Rose et al. 2007 [NPR, MoS, SRC, 2007] )

However, there are substantial uncertainties. There is little agreement about the magnitudes of abatement (Figures 3.28 and 3.29 ). The scenarios disagree about the role of agricultural strategies targeting CH₄ versus N₂O, as well as the timing and annual growth of forestry abatement, with some scenarios suggesting substantial early deployment of forest abatement, while others suggest gradual annual growth or increasing annual growth.

A number of the recent scenarios suggest that biomass energy alternatives could be essential for stabilization, especially as a mitigation strategy that combines the terrestrial sequestration mitigation benefits associated with bio-energy CO₂ capture and storage (BECCS), where CO₂ emissions are captured during biomass energy combustion for storage in geologic formations (e.g. Rao and Riahi, 2006 [SRC] ; Riahi et al., 2006 [NPR, SRC] ; Kurosawa, 2006 [NPR] ; van Vuuren et al., 2007 [JoC, SRC, 2007] ; USCCSP, 2006 [NPR] ). BECCS has also been suggested as a potential rapid-response prevention strategy for abrupt climate change. Across stabilization scenarios, absolute emissions reductions from biomass are projected to grow slowly in the first half of the century, and then rapidly in the second half, as new biomass processing and mitigation technologies become available. Figure 3.28 suggests biomass mitigation of up to 7 GtCO₂/yr in 2050 and 27 GtCO₂/yr in 2100, for cumulative abatement over the century of 115–749 GtCO₂ ( Figure 3.29 ). Figure 3.30 shows the amount of commercial biomass primary energy utilized in various stabilization scenarios. For example, in 2050, the additional biomass energy provides approximately 5–55 EJ for a 2100 stabilization target of 4–5 W/m² and approximately 40–115 EJ for 3.25–4 W/m², accounting for about 0–10 and 5–20% of 2050 total primary energy respectively ( USCCSP, 2006 [NPR] ; Rose et al., 2007 [NPR, MoS, SRC, 2007] ). Over the century, the additional bio-energy accounts for 500–6,700 EJ for targets of 4–5 W/m² and 6100–8000 EJ for targets of 3.25–4 W/m² (1–9% and 9–13% of total primary energy, respectively).

Figure 3.30: Commercial biomass primary energy scenarios above baseline from various 2100 stabilization scenarios.

Notes: The colour of the line indicates the 2100 stabilization target modelled: green < 3.25 W/m² (< 420 CO₂ concentration, < 510 CO₂-eq concentration), pink 3.25–4 (420–490, 510–590), dark blue 4–5 (490–570, 590–710), and light blue 5–6 (570–660, 710–860). Scenario references: IMAGE-EMF21 ( Van Vuuren et al., 2006a [NPR, MoS, SRC] ); MESSAGE-EMF21 ( Rao and Riahi, 2006 [SRC] ); MESSAGE-A2r ( Riahi et al., 2006 [NPR, SRC] ); IMAGE 2.3 ( Van Vuuren et al., 2007 [Ambiguous] ); IGSM and MiniCAM ( USCCSP, 2006 [NPR] ).

Source: Rose et al. 2007 [NPR, MoS, SRC, 2007] USCCSP 2006 [NPR] )

More biomass energy is supplied with tighter stabilization targets, but how much is required for any particular target depends on the confluence of the many different modelling assumptions. Modelled demands for biomass include electric power and end-use sectors (transportation, buildings, industry, and non-energy uses). Current scenarios suggest that electric power is projected to dominate biomass demand in the initial decades and, in general, with less stringent stabilization targets. Later in the century (and for more stringent targets) transportation is projected to dominate biomass use. When biomass is combined with BECCS, biomass mitigation shifts to the power sector late in the century, to take advantage of the net negative emissions from the combined abatement option, such that BECCS could represent a signifant share of cumulative biomass abatement over the century (e.g. 30–50% of total biomass abatement from MESSAGE in Figure 3.29 ).

Figure 3.29: Cumulative cost-effective agricultural, forestry, and biomass abatement 2000 – 2100 from various 2100 stabilization scenarios.

Source: Rose et al. 2007 [NPR, MoS, SRC, 2007] )

To date, detailed analyses of large-scale biomass conversion with CO₂ capture and storage is scarce. As a result, current integrated assessment BECCS scenarios are based on a limited and uncertain understanding of the technology. In general, further research is necessary to characterize biomass’ long-term mitigation potential, especially in terms of land area and water requirements, constraints, and opportunity costs, infrastructure possibilities, cost estimates (collection, transportation, and processing), conversion and end-use technologies, and ecosystem externalities. In particular, present studies are relatively poor in representing land competition with food supply and timber production, which has a significant influence on the economic potential of bio-energy crops (an exception is Sands and Leimbach, 2003 [JoC, MoS] ).

Terrestrial mitigation projections are expected to be regionally unique, while still linked across time and space by changes in global physical and economic forces. For example, Rao and Riahi 2006 [SRC] ) offer intuitive results on the potential role of agricultural methane and nitrous oxide mitigation across industrialized and developing country groups, finding that agriculture is expected to form a larger share of the developing countries’ total mitigation portfolio; and, developing countries are likely to provide the vast majority of global agricultural mitigation. Some aggregate regional forest mitigation results also are discussed below. However, given the paucity of published regional results from integrated assessment models, it is currently not possible to assess the regional land-use abatement potential in stabilization. Future research should direct attention to this issue in order to more fully characterize mitigation potential.

In addition to the stabilization scenarios discussed thus far from integrated assessment and climate economic models, the literature includes long-term mitigation scenarios from global land sector economic models (e.g. Sohngen and Sedjo, 2006 [NPR, ARC] ; Sathaye et al., 2006 [NPR, ARC] ; Sands and Leimbach, 2003 [JoC, MoS] ). Therefore, a comparison is prudent. The sectoral models use exogenous carbon price paths to simulate different climate policies and assumptions. It is possible to compare the stabilization and sectoral scenarios using these carbon price paths. Stabilization (e.g. EMF-21, discussed above) and ‘optimal’ (e.g. Sohngen and Mendelsohn, 2003 [MoS, ARC] ) climate abatement policies suggest that carbon prices will rise over time. ^[19] Table 3.6 compares the forest mitigation outcomes from stabilization and sectoral scenarios that have similar carbon price trajectories ( Rose et al., 2007 [NPR, MoS, SRC, 2007] ). ^[20] Rising carbon prices will provide incentives for additional forest area, longer rotations, and more intensive management to increase carbon storage. Higher effective energy prices might also encourage shorter rotations for joint production of forest bioenergy feedstocks.

Table 3.6 shows that the vast majority of forest mitigation is projected to occur in the second half of the century, with tropical regions in all but one scenario in Table 3.6 assuming a larger share of global forest sequestration/mitigation than temperate regions. The IMAGE results from EMF-21 are discussed separately below. Lower initial carbon prices shift early period mitigation to the temperate regions since, at that time, carbon incentives are inadequate for arresting deforestation. The sectoral models project that tropical forest mitigation activities are expected to be heavily dominated by land-use change activities (reduced deforestation and afforestation), while land management activities (increasing inputs, changing rotation length, adjusting age or species composition) are expected to be the slightly dominant strategies in temperate regions. The current stabilization scenarios model more limited and aggregated forestry GHG abatement technologies that do not distinguish the detailed responses seen in the sectoral models.

The sectoral models, in particular, Sohngen and Sedjo 2006 [NPR, ARC] ), suggest substantially more mitigation in the second half of the century compared to the stabilization scenarios. A number of factors are likely to be contributing to this deviation from the integrated assessment model results. First and foremost, is that Sohngen and Sedjo explicitly model future markets, which none of the integrated assessment models are currently capable of doing. Therefore, a low carbon price that is expected to increase rapidly results in a postponement of additional sequestration actions in Sohngen and Sedjo until the price (benefit) of sequestration is greater. Endogenously modelling forest biophysical and economic dynamics will be a significant future challenge for integrated assessment models. Conversely, the integrated assessment models may be producing a somewhat more muted forest sequestration response given:

(i) Their explicit consideration of competing mitigation alternatives across all sectors and regions, and, in some cases, land-use alternatives.

(ii) Their more limited set of forest-related abatement options, with all integrated assessment models modelling afforestation strategies, but only some considering avoided deforestation, and none modelling forest management options at this point.

(iii) Some integrated assessment models (including those in Table 3.6 ) sequentially allocate land, satisfying population food and feed-demand growth requirements first.

(iv) Climate feedbacks in integrated assessment models can lead to terrestrial carbon losses relative to the baseline.

The IMAGE results in Table 3.6 provide a dramatic illustration of the potential implications and importance of some of these counterbalancing effects. Despite the planting of additional forest plantations in the IMAGE scenario, net tropical forest carbon stocks decline (relative to the baseline) due to deforestation induced by bioenergy crop extensification, as well as reduced CO₂ fertilization that affects forest carbon uptake, especially in tropical forests, and decreases crop productivity, where the latter effect induces greater expansion of food crops onto fallow lands, thereby displacing stored carbon.

Table 3.6: Cumulative forest carbon stock gains above baseline by 2020, 2050 and 2100, from long-term global forestry and stabilization scenarios (GtCO₂).

In addition to reducing uncertainty about the maginitude and timing of land-based mitigation, biomass potential, and regional potential, there are a number of other important outcomes from changes in land that should be tracked and reported in order to properly evaluate long-term land mitigation. Of particular importance to climate stabilization are the albedo implications of land-use change, which can offset emissions reducing land-use change ( Betts, 2000 [JoC] ; Schaeffer et al., 2006 [JoC, SRC] ), as well as the potential climate-driven changes in forest disturbance frequency and intensity that could affect the effectiveness of forest mitigation strategies. Non-climate implications should also be considered. As shown in the Millennium Ecosystem Assessment ( Carpenter et al., 2005 [NPR] ), land use has implications for social welfare (e.g. food security, clean water access), environmental services (water quality, soil retention), and economic welfare (output prices and production).

A number of relevant key baseline land modelling challenges have already been discussed in Sections 3.2.1.6 and 3.2.2.2 . Central to future long-term land mitigation modelling are improvements in the dynamic modelling of regional land use and land-use competition and mitigation cost estimates, as well as modelling of the implications of climate change for land-use and land mitigation opportunities. The total cost of any land-based mitigation strategy should include the opportunity costs of land, which are dynamic and regionally unique functions of changing regional biophysical and economic circumstances. In addition, the results presented in this section do not consider climate shifts that could dramatically alter land-use conditions, such as a permanent El-Nino-like state in tropical regions ( Cox et al., 1999 [MoS] ).

To summarize, recent stabilization studies have found that land-use mitigation options (both non-CO₂ and CO₂) provide cost-effective abatement flexibility in achieving 2100 stabilization targets, in the order of 345– 1260 GtCO₂-eq (15–40%) of cumulative abatement over the century. In some scenarios, increased commercial biomass energy (solid and liquid fuel) is significant in stabilization, providing 115–749 GtCO₂-eq (5–30%) of cumulative abatement and 500–9500 EJ of additional bio-energy above the baseline over the century (potentially 1–15% of total primary energy), especially as a net negative emissions strategy that combines biomass energy with CO₂ capture and storage. Agriculture and forestry mitigation options are projected to be cost effective short-term and long-term abatement strategies. Global forestry models project greater additional forest sequestration than found in stabilization scenarios, a result attributable in part to differences in the modelling of forest dynamics and general economic feedbacks. Overall, the explicit modelling of land-based climate change mitigation in long-term global scenarios is relatively immature, with significant opportunities for improving baseline and mitigation land-use scenarios.

3.3.5.6 Air pollutants, including co-benefits

Quantitative analysis on a global scale on the implications of climate mitigation for air pollutants such as SO₂, NO_x, CO, VOC (volatile organic compounds), BC (black carbon) and OC (organic carbon), are relatively scarce. Air pollutants and greenhouse gases are often emitted by the same sources, and changes in the activity of these sources affect both types of emissions. Previous studies have focused on purely ancillary benefits to air pollution that accrue from a climate mitigation objective, but recently there is a focus on integrating air quality and climate concerns, thus analyzing the co-benefits of such policies. Several recent reviews have summarized the issues related to such benefits ( OECD 2000 [NPR] , 2003 ). They cover absolute air pollutant emission reductions, monetary value of reduced pollution, the climatic impacts of such reductions and the improved health effects due to reduced pollution.

The magnitude of such benefits largely depends on the assumptions of future policies and technological change in the baseline against which they are measured, as discussed in Morgenstern 2000 [MoS] ). For example, Smith et al. 2005 [MoS] and Rao et al. 2005 [SRC] ) assume an overall growth in environmental awareness and formulation of new environmental policies with increased affluence in the baseline scenario, and thus reduced air pollution, even in the absence of any climate policies. The pace of this trend differs significantly across pollutants and baseline scenarios, and may or may not have an obvious effect on greenhouse gases. An added aspect of ancillary benefit measurement is the representation of technological options. Some emission-control technologies reduce both air pollutants and greenhouse gases, such as selective catalytic reduction (SCR) on gas boilers, which reduces not only NO_x, but also N₂O, CO and CH₄ ( IPCC, 1997 [NPR] ). But there are also examples where, at least in principle, emission-control technologies aimed at a certain pollutant could increase emissions of other pollutants. For example, substituting more fuel-efficient diesel engines for petrol engines might lead to higher PM/black carbon emissions ( Kupiainen and Klimont 2004 [NPR] ). Thus estimating co-benefits of climate mitigation should include adequate sectoral representation of emission sources, a wide range of substitution possibilities, assumptions on technological change and a clear representation of current environmental legislation.

Only a few studies have explored the longer-term ancillary benefits of climate policies. ( Alcamo 2002 ) and Mayerhofer et al. 2002 [SRC] ) assess in detail the linkages between regional air pollution and climate change in Europe. They emphasize important co-benefits between climate policy and air pollution control but also indicate that, depending on assumptions, air pollution policies in Europe will play a greater role in air pollutant reductions than climate policy. Smith and Wigley 2006 [NPR] ) suggest that there will be a slight reduction in global sulphur aerosols as a result of long-term multi-gas climate stabilization. Rao et al. 2005 [SRC] and Smith and Wigley 2006 [NPR] ) find that climate policies can reduce cumulative BC and OC emissions by providing the impetus for adoption of cleaner fuels and advanced technologies. In addition, the inclusion of co-benefits for air pollution can have significant impacts on the cost effectiveness of both the climate policy and air pollution policy under consideration. van Harmelen et al. 2002 [SRC] ) find that to comply with agreed upon or future policies to reduce regional air pollution in Europe, mitigation costs are implied, but these are reduced by 50–70% for SO₂ and around 50% for NO_x when combined with GHG policies. Similarly, in the shorter-term, van Vuuren et al. 2006c [SRC] ) find that for the Kyoto Protocol, about half the costs of climate policy might be recovered from reduced air pollution control costs. The exact benefits, however, critically depend on how the Kyoto Protocol is implemented.

The different spatial and temporal scale of greenhouse gases and air pollutants is a major difficulty in evaluating ancillary benefits. Swart et al. 2004 [JoC, ARC] ) stress the need for new analytical bridges between these different spatial and temporal scales. Rypdal et al. 2005 [ARC] ) suggest the possibility of including some local pollutants, such as CO and VOCs, in global climate agreement with others (e.g. NO_x) and aerosols being regulated by regional agreements. It should be noted that some air pollutants, such as sulphate and carbonaceous aerosols, exert radiative forcing and thus global warming, but their contribution is uncertain. Smith and Wigley 2006 [NPR] ) find that the attendant reduced aerosol cooling from sulphates can more than offset the reduction in warming that accrues from reduced GHGs. On the other hand, air pollutants such as NO_x, CO and VOCs act as indirect greenhouse gases having an influence for example via their impact on OH (hydroxil) radicals and therefore the lifetime of direct greenhouse gases (e.g. methane and HFCs). Further, the climatic effects of some pollutants, such as BC and OC aerosols, remain unclear.

While there has been a lot of recent research in estimating co-benefits of joint GHG and air pollution policies, most current studies do not have a comprehensive treatment of co-benefits in terms of reduction costs and the related health and climate impacts in the long-term, thus indicating the need for more research in this area.

3.3.6.1 Costs of mitigation in regional and country scenarios

Figure 3.31 shows the relationship between carbon prices and the CO₂ mitigation rates from the baseline in 2050 in some major countries and regions such as the USA, Japan, EU-15, India, China, Former Soviet Union (FSU) and Eastern Europe, taken from the literature since the TAR ( Hanaoka et al., 2006 [NPR, SRC] ). In the developing countries there are many scenarios where relatively high CO₂ reductions are projected even with low carbon prices. With high prices in the range of 100-150 US$/tCO₂ (in 2000 US dollars) more CO₂ reductions are expected in China and India than in developed countries when the same level of carbon price is applied.

Figure 3.31: Relationship between carbon prices and CO₂ reduction from baseline in 2050 in selected countries taken from the literature published since the TAR.

Note: The red box shows the range between the 25^th and 75^th percentile of the scenarios for each price range. EECCA= Countries of Eastern Europe, the Caucasus and Central Asia.

3.4.1.1 Decarbonization trends

Decarbonization denotes the declining average carbon intensity of primary energy over time. Although decarbonization of the world’s energy system is comparatively slow (0.3% per year), the trend has persisted throughout the past two centuries ( Nakicenovic, 1996 [SRC] ). The overall tendency towards lower carbon intensities is due to the continuous replacement of fuels with high carbon content by those with low carbon content; however, intensities are currently increasing in some developing regions. In short- to medium-term scenarios such a declining tendency for carbon intensity may not be as discernable as across the longer-term literature, e.g. in the World Energy Outlook 2004 ( IEA, 2004 [NPR] ), the reference scenario to 2030 shows the replacement of gas for other fossil fuels as well as cleaner fuels due to limited growth of nuclear and bioenergy.

Another effect contributing towards reduced carbon intensity of the economy is the declining energy requirements per unit of GDP, or energy intensity of GDP. Globally, energy intensity has been declining more rapidly than carbon intensity of energy (0.9% per year) during the past two centuries ( Nakicenovic, 1996 [SRC] ). Consequently, carbon intensity of GDP declined globally at about 1.2% per year.

The carbon intensity of energy and energy intensities of GDP were shown in Section 3.2 of this chapter, Figure 3.6 , for the full scenario sample in the scenario database compared to the newer (developed after 2001 ) non-intervention scenarios. As in Sections 3.2 and 3.3 , the range of the scenarios in the literature until 2001 is compared with recent projections from scenarios developed after 2001 ( Nakicenovic et al., 2005 [NPR, SRC] ).

The majority of the scenarios in the literature portray a similar and persistent decarbonization trend as observed in the past. In particular, the medians of the scenario sets indicate energy decarbonization rates of about 0.9% (pre- 2001 literature median) and 0.6% (post- 2001 median) per year, which is a significantly more rapid decrease compared to the historical rates of about 0.3% per year. Decarbonization of GDP is also more rapid (about 2.5% per year for both pre- and post- 2001 literature medians) compared with the historical rates of about 1.2% per year. As expected, the intervention and stabilization scenarios have significantly higher decarbonization rates and the post- 2001 scenarios include a few with significantly more rapid decarbonization of energy, even extending into the negative range. This means that towards the end of the century these more extreme decarbonization scenarios foresee net carbon removal from the atmosphere, e.g. through carbon capture and storage in conjunction with large amounts of biomass energy. Such developments represent a radical paradigm shift compared to the current and more short-term energy systems, implying significant and radical technological changes.

In contrast, the scenarios that are most intensive in the use of fossil fuels lead to practically no reduction in carbon intensity of energy, while all scenarios portray decarbonization of GDP. For example, the upper boundary of the recent scenarios developed after 2001 depict slightly increasing (about 0.3% per year) carbon intensities of energy (A2 reference scenario, ( Mori 2003 ) ), see Figure 3.8 , comparing carbon emissions across scenarios in the literature presented in Section 3.2 ). Most notably, a few scenarios developed before 2001 follow an opposite path compared to other scenarios: decarbonization of primary energy with decreasing energy efficiency until 2040, followed by rapidly increasing ratios of CO₂ per unit of primary energy after 2040 – in other words, recarbonization. In the long term, these scenarios lie well above the range spanned by the new scenarios, indicating a shift towards more rapid CO₂ intensity improvements in the recent literature ( Nakicenovic et al., 2006 [SRC] ). In contrast, there are just a very few scenarios in the post- 2100 literature that envisage increases in carbon intensity of energy.

The highest rates of decarbonization of energy (up to 2.5% per year for the recent scenarios) are from scenarios that include a complete transition in the energy system away from carbon-intensive fossil fuels. Clearly, the majority of these scenarios are intervention scenarios, although some non-intervention scenarios show drastic reductions in CO₂ intensities due to reasons other than climate policies (e.g. the combination of sustainable development policies and technology push measures to promote renewable hydrogen systems). The relatively fast decarbonization rate of intervention scenarios is also illustrated by the median of the post- 2001 intervention scenarios, which depict an average rate of improvement of 1.1% per year over the course of the century, compared to just 0.3% for the non-intervention scenarios. Note, nevertheless, that the modest increase in carbon intensity of energy improvements in the intervention scenarios above the 75^th percentile of the distribution of the recent scenarios. The vast majority of these scenarios represent sensitivity analysis; have climate policies for mitigation of non-CO₂ greenhouse gas emissions (methane emissions policies: Reilly et al., 2006 [NPR] ); or have comparatively modest CO₂ reductions measures, such as the implementation of a relatively minor carbon tax of 10 US$/tC (about 2.7 US$/tCO₂) over the course of the century (e.g. ( Kurosawa, 2004 ) ). Although these scenarios are categorized according to our definition as intervention scenarios, they do not necessarily lead to the stabilization of atmospheric CO₂ concentrations.

3.4.1.2 Key factors for carbon-free energy and decarbonization development

All of the technological options assumed to contribute towards further decarbonization and reduction of future GHG emissions require further research and development (R&D) to improve their technical performance, reduce costs and achieve social acceptability. In addition, deployment of carbon-saving technologies needs to be applied at ever-larger scales in order to benefit from potentials of technological learning that can result in further improved costs and economic characteristics of new technologies. Most importantly, appropriate institutional and policy inducements are required to enhance widespread diffusion and transfer of these technologies.

Full replacement of dominant technologies in the energy systems is generally a long process. In the past, the major energy technology transitions have lasted more than half a century, such as the transition from coal as the dominant energy source in the world some 80 years ago, to the dominance of crude oil during the 1970 s. Achieving such a transition in the future towards lower GHG intensities is one of the major technological challenges addressed in mitigation and stabilization scenarios.

Figures 3.34 and 3.35 show the ranges of energy technology deployment across scenarios by 2030 and 2100 for baseline (non-intervention) and intervention (including stabilization) scenarios, respectively. The deployment of energy technologies in general, and of new technologies in particular, is significant indeed, even through the 2030 period, but especially by 2100 . The deployment ranges should be compared with the current total global primary energy requirements of some 440 EJ in 2000 . Coal, oil and gas reach median deployment levels ranging from some 150 to 250 EJ by 2030 . The variation is significantly higher by 2100, but even medians reach levels of close to 600 EJ for coal in reference scenarios, thereby exceeding by 50% the current deployment of all primary energy technologies in the world. Deployment of nuclear and biomass is comparatively lower, in the range of about 50–100 EJ by 2030 and up to ten times as much by 2100 . This all indicates that radical technological changes occur across the range of scenarios.

Figure 3.34: Deployment of primary energy technologies across pre- 2001 scenarios by 2030 and 2100 : Left-side ‘error’ bars show baseline (non-intervention) scenarios and right-side ones show intervention and stabilization scenarios. The full ranges of the distributions (full vertical line with two extreme tic marks), the 25^th and 75^th percentiles (blue area) and the median (middle tic mark) are also shown.

Figure 3.35: Deployment of primary energy technologies across post- 2001 scenarios by 2030 and 2100 : Left-side ‘error’ bars show baseline (non-intervention) scenarios and right-side ones show intervention and stabilization scenarios. The full ranges of the distributions (full vertical line with two extreme tic marks), the 25^th and 75^th percentiles (blue area) and the median (middle tic mark) are also shown.

The deployment ranges are large for each of the technologies but do not differ much when comparing the pre- 2001 with post- 2001 scenarios over both time periods, up to 2030 and 2100 . Thus, while technology deployments are large in the mean and variance, the patterns have changed little in the new (compared with the older) scenarios. What is significant in both sets of literature is the radically different structure and portfolio of technologies between baseline and stabilization scenarios. Mitigation generally means significantly less coal, somewhat less natural gas and consistently more nuclear and biomass. What cannot be seen from this comparison, due to the lack of data and information about the scenarios, is the extent to which carbon capture and storage is deployed in mitigation scenarios. However, it is very likely that most of the coal and much of the natural gas deployment across stabilization scenarios occurs in conjunction with carbon capture and storage. The overall conclusion is that mitigation and stabilization in emissions scenarios have a significant inducement on diffusion rates of carbon-saving and zero-carbon energy technologies.

3.4.3.1 Summary from the TAR

The IPCC-TAR concluded that reduction of greenhouse gas emissions is highly dependent on both technological innovation and implementation of technologies (a conclusion broadly confirmed in Chapter 2 , Section 2.7.1 ). However, the rate of introduction of new technologies, and the drivers for adoption are different across different parts of the world, particularly in industrial market economies, economies in transition and developing countries. To some extent this is reflected in global emissions scenarios as they often involve technological change at a level that includes a dozen or so world regions. This usually involves making more region-specific assumptions about future performance, costs and investment needs for new and low-carbon technologies.

There are multiple policy approaches to encourage technological innovation and change. Through regulation of energy markets, environmental regulations, energy efficiency standards, financial and other market-based incentives, such as energy and emission taxes, governments can induce technology changes and influence the level of innovations. In emissions scenarios, this is reflected in assumptions about policy instruments such as taxes, emissions permits, technology standards, costs, and lower and upper boundaries of technology diffusion.

3.4.3.2 Dynamics of technology

R&D, technological learning, and spillovers are the three broad categories of drivers behind technological change. These are discussed in Chapter 2 , Section 2.7 , and Chapter 11 , Section 11.5 . The main conclusion is that, on the whole, all three of the sources of induced technological change (ITC) play important roles in technological advance. Here, we focus on the dynamics of technology and ITC in emissions and stabilization scenarios.

Emissions scenarios generally treat technological change as an exogenous assumption about costs, market penetration and other technology characteristics, with some notable exceptions such as in Gritsevskyi and Nakicenovic 2000 [MoS, SRC] Hourcade and Shukla 2001 [NPR, ARC] ), in their review of scenarios from top-down models, indicate that technology assumptions are a critical factor that affects the timing and cost of emission abatement in the models. They identify widely differing costs of stabilization at 550 ppmv by 2050, of between 0.2 and 1.75% of GDP, mainly influenced by the size of the emissions in the baseline.

The International Modelling Comparison Project (IMCP) ( Edenhofer et al., 2006 [NPR, ARC] ) compared the treatment relating to technological change in many models covering a wide range of approaches. The economies for technological change were simulated in three groups: effects through R&D expenditures, learning-by-doing (LBD) or specialization and scale. IMCP finds that ITC reduces costs of stabilization, but in a wide range, depending on the flexibility of the investment decisions and the range of mitigation options in the models. It should be noted, however, that induced technological change is not a ‘free lunch’, as it requires higher upfront investment and deployment of new technologies in order to achieve cost-reductions thereafter. This can lead to lower overall mitigation costs.

All models indicate that real carbon prices for stabilization targets rise with time in the early years, with some models showing a decline in the optimal price after 2050 due to the accumulated effects of LBD and positive spillovers on economic growth. Another robust result is that ITC can reduce costs when models include low carbon energy sources (such as renewables, nuclear, and carbon capture and sequestration), as well as energy efficiency and energy savings. Finally, policy uncertainty is seen as an issue. Long-term and credible abatement targets and policies will reduce some of the uncertainties around the investment decisions and are crucial to the transformation of the energy system.

ITC broadens the scope of technology-related policies and usually increases the benefits of early action, which accelerates deployment and cost-reductions of low-carbon technologies ( Barker et al., 2006 [NPR, SRC] ; Sijm, 2004 [NPR, MoS] ; Gritsevskyi and Nakicenovic, 2000 [MoS, SRC] ). This is due to the cumulative nature of ITC as treated in the new modelling approaches. Early deployment of costly technologies leads to learning benefits and lower costs as diffusion progresses. In contrast, scenarios with exogenous technology assumptions imply waiting for better technologies to arrive in the future, though this too may result in reduced costs of emission reduction (European Commission, 2003 ).

Other recent work also confirms these findings. For example, Manne and Richels 2004 [SRC] and Goulder 2004 [NPR] ) also found that ITC lowers mitigation costs and that more extensive reductions in GHGs are justified than with exogenous technical change. Nakicenovic and Riahi 2003 [NPR, MoS, SRC] ) noted how the assumption about the availability of future technologies was a strong driver of stabilization costs. Edmonds et al. 2004a [ARC] ) studied stabilization at 550 ppmv CO₂ in the SRES B2 world using the MiniCAM model and showed a reduction in costs of a factor of 2.5 in 2100 using a baseline incorporating technical change. Edmonds et al. consider advanced technology development to be far more important as a driver of emission reductions than carbon taxes. Weyant 2004 [SRC] ) concluded that stabilization will require the large-scale development of new energy technologies, and that costs would be reduced if many technologies are developed in parallel and there is early adoption of policies to encourage technology development.

The results from the bottom-up and more technology-specific modelling approaches give a different perspective. Following the work of the IIASA in particular, models investigating induced technical change emerged during the mid- and late- 1990 s. These models show that ITC can alter results in many ways. In the previous sections of this chapter the authors have also illustrated that the baseline choice is crucial in determining the nature (and by implication also the cost) of stabilization. However, this influence is itself largely due to the different assumptions made about technological change in the baseline scenarios. Gritsevskyi and Nakicenovic 2000 [MoS, SRC] ) identified some 53 clusters of least-cost technologies, allowing for endogenous technological learning with uncertainty. This suggests that a decarbonized economy may not cost any more than a carbon-intensive one, if technology learning curves are taken into account. Other key findings are that there is a large diversity across alternative energy technology strategies, a finding that was confirmed in IMCP ( Edenhofer et al., 2006 [NPR, ARC] ). These results suggest that it is not possible to choose an ‘optimal’ direction for energy system development. Some modelling reported in the TAR suggests that a reduction (up to 5 GtC a year) by 2020 (some 50% of baseline projections) might be achieved by current technologies, half of the reduction at no direct cost, the other half at direct costs of less than 100 US$/tC-equivalent (27 US$/tCO₂-eq).

3.4.3.3 Barriers to technology transfer, diffusion and deployment for long-term mitigation

Chapter 2 , Section 2.7.2 includes a discussion of the barriers to development and commercialization of technologies. Barriers to technology transfer vary according to the specific context from sector to sector and can manifest themselves differently in developed and developing countries, and in economies-in-transition (EITs). These barriers range from a lack of information; insufficient human capabilities; political and economic barriers (such as the lack of capital, high transaction costs, lack of full cost pricing, and trade and policy barriers); institutional and structural barriers; lack of understanding of local needs; business limitations (such as risk aversion in financial institutions); institutional limitations (such as insufficient legal protection); and inadequate environmental codes and standards.

In the long term, the key drivers of technological change in developing countries will depend on three ‘changes’ that are simultaneous and inseparable within the context of development: exogenous behavioural changes or changes in social infrastructure; endogenous policies driven by ‘development goals’; and any induced change from climate policies ( Shukla et al., 2006 [MoS, SRC] ).

3.4.3.4 Dynamics in developing countries and timing of technology deployment

National policies in developing countries necessarily focus on more fundamental priorities of development, such as poverty alleviation and providing basic living conditions for their populations, and it is unlikely that short-term national policies would be driven by environmental concerns. National policies driven by energy security concerns can, however, have strong alignment with climate goals. The success of policies that address short-term development concerns will determine the pace at which the quality of life in the developing and the developed world converges over the long term.

3.5.1.1 An iterative risk-management framework to articulate options

Previous IPCC reports conclude that climate change decision-making is not a once-and-for-all event, but an iterative risk-management process that is likely to take place over decades, where there will be opportunities for learning and mid-course corrections in the light of new information ( Lempert et al., 1994 [JoC] ; Keller et al., 2006 [JoC] ).

This iterative process can be described using a decision tree ( Figure 3.37 ), where the square nodes represent decisions, the circles represent the reduction of uncertainty and the arrows indicate the range of decisions and outcomes. Some nodes summarize today’s options – how much should be invested in mitigation, in adaptation, in expanding mitigative and adaptive capacity, or in research to reduce uncertainty? Other nodes represent opportunities to learn and make mid-course corrections. This picture is a caricature of real decision processes, which are continuous, overlapping and iterative. However, it is useful to conceptually put the many determinants of any short-term strategy in a context of progressive resolution of uncertainty.

Figure 3.37: The Iterative Nature of the Climate Policy Process.

3.5.1.2 Qualitative insights into interactions between mitigaton, adaptation and development

Until recently, a main focus in the policy and integrated assessment literature has been on comparing mitigation costs and avoided damages. Since the TAR, attention has shifted towards the interaction between mitigation and adaptation in reducing damages in a risk-management framework. This has accompanied a growing realization that some climate change in the coming decades is inevitable.

Limited treatment of adaptation in climate policy assessments is still a problem and a number of reasons explain this. First, the focus of the international climate change negotiations has largely been on mitigation (perhaps because attention to adaptation could be viewed as ‘giving up’ on mitigation) even though the importance of adaptation is underlined in Article 4 of the UNFCCC and Article 10 of the Kyoto Protocol. Second, adaptation is largely undertaken at the local scale, by individual households, farmers, companies or local governments; it is thus difficult to target through coordinated international incentives, and is more complicated to handle quantitatively by models in global scenarios. Third, it is difficult to generalize the ways that individuals or communities are likely to adapt to specific impacts. However, the literature is evolving quickly and recent work is available in a number of regions; for example, in Finland ( Carter et al., 2005 [NPR, MoS] ), the UK ( West and Gawith, 2005 [NPR] ), Canada ( Cohen et al., 2004 [NPR] ) and the USA (e.g. California, Hayhoe et al., 2004 [PoC, JoC, ARC] ).

Despite the scarcity of global systematic assessments ( (Tol, 2005a ) ), some interesting insights into the interaction between adaptation and mitigation emerge from recent regional-scale studies. Some adaptation measures are ‘no-regret’ measures and should be undertaken anyway ( Agrawala, 2005 [NPR] ), such as preservation of mangroves in coastal zones, which provide a buffer for increased coastal flood risk due to climate change and help to maintain healthy marine ecosystems ( Nicholls et al., 2006 [NPR] ). A few may be synergistic with mitigation ( Bosello, 2005 [NPR] ) such as investing in more efficient buildings that will limit human vulnerability to increasingly frequent heatwaves and also reduce energy use, hence emissions. But many adaptation options involve net costs with a risk of committing to irreversible and misplaced investment given the considerable uncertainty about climate change at a local scale. Given this uncertainty, and the fact that learning about adaptation to climate change imposes some costs and takes time ( Kelly et al., 2005 [ARC] ), mis-allocation of investments may occur, or the rate of long-term investment in adaptation strategies may slow ( Kokic et al., 2005 [MoS, SRC] ; Kelly et al., 2005 [ARC] ).

Finally, the interactions between adaptation and mitigation are intertwined with development pathways. A key issue is to understand at what point (over)investment in mitigation or adaptation might limit funds available for development, and thus reduce future adaptive capacity ( Sachs, 2004 [JoC] ;( Tol, 2005a; ) Tol and Yohe, 2006 [NPR] ). Another issue concerns the point at which climate change damages, and the associated investment in adaptation, could crowd out more productive investments later and harm development ( (Kemfert, 2002; ) Bosello and Zhang, 2005 [NPR] ; Kemfert and Schumacher, 2005 [NPR] ). The answer to these questions depends upon modelling assumptions that drive repercussions in other sectors of the economy and other regions and the potential impacts on economic growth. These are ‘higher-order’ social costs of climate change from a series of climate-change-induced shocks; they include the relative influence of: a) the cross-sectoral interactions across all major sectors and regions; b) a crowding out effect that slows down capital accumulation and technical progress, especially if technical change is endogenous. These indirect impacts reduce development and adaptive capacity and may be in the same order of magnitude, or greater than, the direct impact of climate change ( Fankhauser and Tol, 2005 [JoC] ; Roson and Tol, 2006 [MoS] ; Kemfert, 2006 [MoS] ).

Both the magnitude and the sign of the indirect macro-economic impacts of climate change are conditional upon the growth dynamics of the countries concerned. When confronted by the same mitigation policies and the same climate change impacts, economies experiencing strong disequilibrium (including ‘poverty traps’) and large market and institutional imperfections will not react in the same way as countries that are on a steady and high economic growth pathway. The latter are near what economists call their ‘production frontier’ (the maximum of production attainable at a given point in time); the former are more vulnerable to any climatic shock or badly calibrated mitigation policies, but symmetrically offer more opportunities for synergies between mitigation, adaptation and development policies ( Shukla et al., 2006 [MoS, SRC] ). On the adaptation side for example, ( Tol and Dowlatabadi 2001 ) ) demonstrate that there is significant potential to reduce vulnerability to the spread of malaria in Africa. In some circumstances, mitigation measures can be aligned with development policies and alleviate important sources of vulnerability in these countries, such as dependency on oil imports or local pollution. But this involves transition costs over the coming 10–20 years (higher domestic energy prices, higher investments in the energy sector), which in turn suggests opportunities for international cooperative mechanisms to minimize these costs.

Bosello 2005 [NPR] ) shows complementarity between adaptation, mitigation and investment in R&D, whilst others consider these as substitutes ( (Tol, 2005a ) Schneider and Lane 2006 [NPR, PoC] ) consider that mitigation and adaptation only trade off for small temperature increments where adaptation might be cheaper, whereas for larger temperature increases mitigation is always the cheaper option. ( Goklany 2003 ) ) promotes the view that the contribution by climate change to hunger, malaria, coastal flooding, and water stress (as measured by populations at risk) is small compared to that of non-climate-change-related factors, and that through the 2080 s, efforts to reduce vulnerability would be more cost-effective in reducing these problems than mitigation. This analysis neglects critical thresholds at the regional level (such as the temperature ceiling on feasibility of regional crop growth) and at the global level (such as the onset of ice sheet melting or release of methane from permafrost), and, like many others studies, it neglects the impacts of extreme weather events. It also promotes a very optimistic view of adaptive capacity, which is increasingly challenged in the literature ( (Tompkins and Adger, 2005 ) ). An adaptation-only policy scenario in the coming decades leads to an even greater challenge for adaptation in decades to follow, owing to the inertia of the climate system. In the absence of mitigation, temperature rises will be much greater than would otherwise occur with pusuant impacts on economic development ( IPCC, 2007b [NPR, 2007] , Chapter 19.3.7; Stern 2006 [NPR] ). Hence adaptation alone is insufficient to avoid the serious risks due to climate change (see Table 3.11 ; also IPCC, 2007b [NPR, 2007] , Chapter 19, Table 19.1).

To summarize, adaptation and mitigation are thus increasingly viewed as complementary (on the global scale), whilst locally there are examples of both synergies and conflicts between the two ( IPCC, 2007b [NPR, 2007] , Chapter 18). Less action on mitigation raises the risk of greater climate-change-induced damages to economic development and natural systems and implies a greater need for adaptation. Some authors maintain that adaptation and mitigation are substitutes, because of competition for funds, whilst others claim that such tradeoffs occur only at the margin when considering incremental temperature change and incremental policy action, because for large temperature changes mitigation is always cheaper than adaptation.

Table 3.9: Global mean temperature increase at equilibrium, greenhouse gas concentration and radiative forcing. Equilibrium temperatures here are calculated using estimates of climate sensitivity and do not take into account the full range of bio-geophysical feedbacks that may occur.

3.5.3.1 Scenario and sensitivity analysis of climate targets

Probabilistic scenario analysis can be used to assess the risk of overshooting some climate target or to produce probabilistic projections that quantify the likelihood of a particular outcome. Targets for such analysis can be expressed in several different ways: absolute global mean temperature rise by 2100, rate of climate change, other thresholds beyond which dangerous anthropogenic interference (DAI) may occur, or additional numbers of people at risk to various stresses. For example, Arnell et al. 2002 [JoC] ) show that such stresses (conversion of forests to grasslands, coastal flood risk, water stress) are far less at 550 ppmv than at 750 ppmv.

Recent Integrated Assessment Models (IAM) literature reflects a renewed attention to climate sensitivity as a key driver of climate dynamics (Den Elzen and Meinshausen, 2006 [NotFound] ; Hare and Meinshausen, 2006 [JoC, ARC] ; Harvey, 2006 [JoC] ; Keller et al., 2006 [JoC] ; Mastrandrea and Schneider, 2004 [PoC, JoC, SRC] ; Meehl et al., 2005 [JoC] ; Meinshausen et al., 2006 [NPR, MoS] , Meinshausen, 2006 [NPR, MoS] ; O’Neill and Oppeinheimer, 2002, 2004; Schneider and Lane, 2004 [NPR, PoC] ; Wigley, 2005 [JoC] ). The consideration of a full range of possible climate sensitivity increases the probability of exceeding thresholds for specific DAI. It also magnifies the consequence of delaying mitigation efforts. Hare and Meinshausen 2006 [JoC, ARC] ) estimate that each 10-year delay in mitigation implies an additional 0.2°C–0.3°C warming over a 100–400 year time horizon. For a climate sensitivity of 3°C, Harvey 2006 [JoC] ) shows that immediate mitigation is required to constrain temperature rise to roughly 2°C above pre-industrial levels. Only in the unlikely situation where climate sensitivity is 1°C or lower would immediate mitigation not be necessary. ^[26] Harvey also points out that, even in the case of a 2°C threshold (above pre-industrial levels), acidification of the ocean would still occur and that this might not be considered safe.

Another focus of sensitivity analysis is on mitigation scenarios that overshoot and eventually return to a given stabilization or temperature target ( Kheshgi, 2004 [ARC] ; Wigley, 2005 [JoC] ; Harvey, 2004 [JoC] ; Izrael and Semenov, 2005 [MoS] ; Kheshgi et al., 2005 [ARC] ; Meinshausen et al., 2006 [NPR, MoS] Schneider and Mastrandrea 2005 [PoC, JoC, SRC] ) find that this risk of exceeding a threshold of 2ºC above pre-industrial levels is increased by 70% for an overshoot scenario stabilizing at 500 ppmv CO₂-eq (as compared to a scenario stabilizing at 500 ppmv CO₂-eq). Such overshoot scenarios are likely to be necessary if there is a decision to achieve stablization of GHG concentrations close to (or at) today’s levels. They are indeed likely to lower the costs of mitigation but, in turn, raise the risk of exceeding such thresholds ( Keller et al., 2006 [JoC] ; Schneider and Lane, 2004 [NPR, PoC] ) and may limit the ability to adapt by increasing the rate of climate change, at least temporarily ( Hare and Meinshausen, 2006 [JoC, ARC] O’Neill and Oppenheimer 2004 [SRC] ) find that the transient temperature up to 2100 is equally, or more, controlled by the pathway to stabilization than by the stabilization target, and that overshooting can lead to a peak temperature increase that is higher than in the long-term (equilibrium) warming.

The last and important contribution of this approach is to test the sensitivity of results to carbon cycle and climate change feedbacks ( Cox et al., 2000 [JoC, MoS] ; Friedlingstein et al., 2001 [MoS] ;( Matthews, 2005 ) ) and other factors that may affect carbon cycle dynamics, such as deforestation ( Gitz and Ciais, 2003 [NPR, MoS] ). For example, carbon cycle feedbacks amplify warming ( Meehl et al., 2007 [NPR, MoS, 2007] ) and are omitted from most other studies that thus underestimate the risks of exceeding (or overshooting) temperature targets for a given effort of mitigation in the energy sector only. This could increase warming by up to 1°C in 2100, according to a simple model ( Meehl et al., 2007 [NPR, MoS, 2007] ). The amplification, together with further potential amplification due to feedbacks of uncertain magnitude, such as the potential release of methane from permafrost, peat bogs and seafloor clathrates ( Meehl et al., 2007 [NPR, MoS, 2007] ) are also not included in the analysis presented in Figure 3.38 and Table 3.10 . This analysis reflects only known feedbacks for which the magnitude can be estimated and are included in General Circulation Models (GCMs). Hence, scenario and sensitivity analysis shows that the risks of exceeding a given temperature threshold for a given temperature target may be higher than that shown in Table 3.10 and Figure 3.38 .

3.5.3.2 Inverse modelling and guardrail analysis

Inverse modelling approaches such as Safe Landing Analysis ( Swart et al., 1998 [NPR, MoS] ) and Tolerable Windows Approach ( Toth, 2003 [JoC, SRC] ), aim to define a guardrail of allowable emissions for sets of unacceptable impacts or intolerable mitigation costs. They explore how the set of viable emissions pathways is constrained by parameters such as the starting date, the rate of emission reductions, or the environmental constraints. They provide insights into the influence of short-term decisions on long-term targets by delineating allowable emissions corridor, but they do not prescribe unique emissions pathways, as per cost-effectiveness or costs-benefit analysis.

For example, Toth et al. 2002 [SRC] ) draw on climate impact response functions (CIRFs) by ( Füssel and van Minnen 2001 ) ) that use detailed biophysical models to estimate regionally specific, non-monetized impacts for different sectors (i.e. agricultural production, forestry, water runoff and biome changes). They show that the business-as-usual scenario of GHG emissions (which resembles the SRES A2 scenario) to 2040 precludes the possibility of limiting the worldwide transformation of ecosystems to 30% or less, even with very high willingness to pay for the mitigation of GHG emissions afterwards. Some applications of guardrail analyses assess the relationship between emission pathways and abrupt change such as thermohaline circulation (THC) collapse ( Rahmstorf and Zickfeld, 2005 [PoC, JoC] ). The latter study concludes that stringent mitigation policy reduces the probability of THC collapse but cannot entirely avoid the risk of shutdown.

Corfee-Morlot and Höhne 2003 [JoC, ARC] ) conclude that only low stabilization targets (e.g. 450 ppmv CO₂ or 550 ppmv CO₂-eq) significantly reduce the likelihood of climate change impacts. They use an inverse analysis to conclude that more than half of the SRES (baseline) emission scenarios leave this objective virtually out of reach as of 2020 .

More generally, referring to Table 3.10 , if the peaking of global emissions is postponed beyond the next 15 years to a time period somewhere between the next 15–55 years, then constraining global temperature rise to below 2°C above 1990 (2.6°C above pre-industrial levels) becomes unlikely (using ‘best estimate’ assumptions of climate sensitivity), resulting in increased risks of the impacts listed in Table 3.11 and discussed in Section 3.5.2 .

3.5.3.3 Cost-benefit analysis, damage cost estimates and social costs of carbon

The above analysis provides a means of eliminating those emissions scenarios that are outside sets of pre-determined guardrails for climate protection and provides the raw material for cost-effectiveness analysis of optimal pathways for GHG emissions. If one wants to determine these pathways through a cost-benefit analysis it is necessary to assess the trade-off between mitigation, adaptation and damages, and consequently, to measure damages in the same monetary metric as mitigation and adaptation expenditures. Such assessment can be carried out directly in the form of ‘willingness to pay for’ avoiding certain physical consequences.

Some argue that it is necessary to specify more precisely why certain impacts are undesirable and to comprehensively itemize the economic consequences of climate change in monetary terms. The credibility of such efforts has often been questioned, given the uncertainty surrounding climate impacts and the efficacy of societal responses to them, plus the controversial meaning of a monetary metric across different regions and generations ( Jacoby, 2004 [JoC, MoS] ). This explains why few economists have taken the step of monetizing global climate impacts. At the time of the TAR, only three such comprehensive studies had been published ( Mendelsohn et al., 2000 [JoC] ; Nordhaus and Boyer, 2000 [NPR, MoS] ; and Tol, 2002a [MoS] , 2002b ). Their estimates ranged from negligible to 1.5% of the GDP for a global mean temperature rise of +2.5°C and Nordhaus and Boyer carefully warned: ‘Along the economically efficient emission path, the long-run global average temperature after 500 years is projected to increase 6.2°C over the 1900 global climate. While we have only the foggiest idea of what this would imply in terms of ecological, economic, and social outcomes, it would make the most thoughtful people, even economists, nervous to induce such a large environmental change. Given the potential for unintended and potentially disastrous consequences….’

Progress has been made since the TAR in assessing the impacts of climate change. Nonetheless, as noted in Watkiss et al. 2005 [NPR, MoS] ), estimates of the social costs of carbon (SCC) in the recent literature still reflect an incomplete subset of relevant impacts; many significant impacts have not yet been monetized (see also IPCC, 2007b [NPR, 2007] ; for SCC see IPCC 2007b [NPR, 2007] , Section 20.6) and others are calibrated in numeraires that may defy monetization for some time to come. Existing reviews of available SCC estimates show that they span several orders of magnitude – ranges that reflect uncertainties in climate sensitivity, response lags, discount rates, the treatment of equity, the valuation of economic and non-economic impacts, and the treatment of possible catastrophic losses ( IPCC, 2007b [NPR, 2007] , Chapter 20). The majority of available estimates in the literature also capture only impacts driven by lower levels of climate change (e.g. 3°C above 1990 levels). IPCC 2007b [NPR, 2007] ) highlights available estimates of SCC that run from -3 to 95 US$ /tCO₂ from one survey, but also note that another survey includes a few estimates as high as 400 US$/tCO₂ ( IPCC, 2007b [NPR, 2007] , Chapter 20, ES and Section 20.6.1). However the lower boundary of this range includes studies where climate change is presumed to be low and aggregate benefits accrue. Moreover, none of the aggregate estimates reflect the significant differences in impacts that will be felt across different regions; nor do they capture any of the social costs of other greenhouse gases. A more recent estimate by Stern 2006 [NPR] ) is at the high end of these estimates (at 85 US$/tCO₂) because an extremely low discount rate (of 1.4%) is used in calculating damages that include additional costs attributed to abrupt change and increases in global mean temperature for some scenarios in excess of 7°C ( Nordhaus, 2006a [JoC] ; Yohe, 2006 [MoS] ; Tol and Yohe, 2006 [NPR] ). The long-term high-temperature scenarios are due to inclusion of feedback processes. IPCC 2007b [NPR, 2007] ) also highlights the fact that the social costs of carbon and other greenhouse gases could increase over time by 2–4% per year ( IPCC, 2007b [NPR, 2007] ; Chapter 20, ES and Section 20.6.1).

For a given level of climate change, the discrepancies in estimates of the social costs of carbon can be explained by a number of parameters highlighted in Figure 3.39 . These stem from two different types of questions: normative and empirical. Key normative parameters include the inter-temporal aggregation of damages through discount rates and aggregation methods for impacts across diverse populations within the same time period ( Azar and Lindgren, 2003 [JoC] ; Howarth, 2003 [JoC] ; Mastrandrea and Schneider, 2004 [PoC, JoC, SRC] ) and are responsible for much of the variation.

Figure 3.39: Factors influencing the social costs of carbon.

Source: Downing et al., 2005 [NPR, ARC]

The other parameters relate to the empirical validity of their assessment, given the poor quality of data and the difficulty of predicting how society will react to climate impacts in a given sector, at a given scale in future decades. ( Pearce 2003 ) ) suggests that climate damages and SCC may be over-estimated due to the omission of possible amenity benefits in warmer climates or high-latitude regions ( Maddison 2001 [NPR] ) and possible agricultural benefits. However, overall, it is likely that current SCC estimates are understated due to the omission of significant impacts that have not yet been monetized ( IPCC, 2007b [NPR, 2007] , Chapters 19 and 20; Watkiss et al., 2005 [NPR, MoS] ).

Key empirical parameters that increase the social value of damages include:

• Climate sensitivity and response lag. Equilibrium temperature rise for a doubling of CO₂, and the modelled response time of climate to such a change in forcing. Hope (2006) in his PAGE 2002 model found that, as climate sensitivity was varied from 1.5°–5°C, the model identified a strong correlation with SCC.
• Coverage of abrupt or catastrophic changes, such as the crossing of the THC threshold (Keller et al., 2000 and 2004; Mastrandrea and Schneider, 2001; Hall and Behl, 2006) or the release of methane from permafrost and the weakening of carbon sinks. The Stern Review (2006) finds that such abrupt changes may more than double the market damages (e.g. from 2.1% to 5% of global GDP) if temperatures were to rise by 7.4°C in 2200.
• Inclusion and social value of non-market impacts: what value will future generations place on impacts, such as the quality of landscape or biodiversity?
• Valuation methods for market impacts such as the value of life.
• Adaptative capacity: social costs will be magnified if climate change impacts fall on fragile economies.
• Predictive capacity: studies finding efficient adaptation assume that actors decide using perfect foresight (after a learning process; see Mendelsohn and Williams, 2004). Higher costs are found if one considers the volatility of climate signals and transaction costs. For agriculture, Parry et al. (2004) shows the costs of a mismatch between expectations and real climate change (sunk costs, value of real estates, and of capital stock).
• Geographic downscaling: using a geographic-economic cross-sectional (1990) database, Nordhaus (2006a) concludes that this downscaling leads to increased damage costs, from previous 0.7% estimates to 3% of world output for a 3°C increase in global mean temperature.
• The propagation of local economic and social shocks: this blurs the distinction between winners and losers. The magnitude of this type of indirect impact depends on the existence of compensation mechanisms, including direct assistance and insurance as well as on how the cross-sectoral interdependences and transition costs are captured by models (see Section 3.5.1 ).

The influence of this set of parameters, which is set differently in various studies, explains the wide range of estimates for the SCC.

In an economically-efficient mitigation response, the marginal costs of mitigation should be equated to the marginal benefits of emission reduction. The marginal benefits are the avoided damages for an additional tonne of carbon abated within a given emission pathway, also known as the SCC. As discussed in Section 3.6 , both sides of this equation are uncertain, which is why a sequential or iterative decision-making framework, with progressive resolution of information, is needed. Despite a paucity of analytical results in this area, it is possible to draw on today’s literature to make a first comparison between the range of SCC estimates and the range of marginal costs of mitigation across different scenarios. IPCC 2007b [NPR, 2007] , Chapter 20) reviews ranges of SCC from available literature. Allowing for a range of SCC between 4–95 US$/tCO₂ (14–350 US$/tC from ( Tol 2005b ) ) median and 95^th percentile estimates) and assuming a 2.4% per year increase ( IPCC, 2007b [NPR, 2007] , Chapter 20), produces a range of estimates for 2030 of 8–189 US$/tCO₂. The mitigation studies in this chapter suggest carbon prices in 2030 of 1–24 US$/tCO₂-eq for category IV scenarios, 18–79 US$/tCO₂-eq for category III scenarios, and 31–121 US$/tCO₂-eq for category I and II scenarios (see Sections 3.3 and 3.6 ).

3.6.1.1 Influence of passing from concentration targets to temperature targets in a cost-effectiveness framework

New studies such as den Elzen et al. 2006 [Ambiguous] ) confirm previous results. They establish that reaching a concentration target as low as 450 ppmv CO₂-eq, under even optimistic assumptions of full participation, poses significant challenges in the 2030 – 2040 timeframe, with rapidly increasing emission reduction rates and rising costs. In a stochastic cost-effectiveness framework, reaching such targets requires a significant and early emissions reduction with respect to respective baselines.

But concentration ceilings are a poor surrogate for climate change risks: they bypass many links from atmospheric chemistry to ultimate damages and they only refer to long-term implications of global warming. A better proxy of climate change impacts can be found in global mean temperature: every regional assessment of climate change impacts refers to this parameter, making it easier for stakeholders to grasp the stakes of global warming for their region; one can also take into account the rate of climate change, a major determinant of impacts and damages.

Therefore, with a noticeable acceleration in the last few years, the scientific community has concentrated on assessing climate policies in the context of climate stabilization around various temperature targets. These contributions have mainly examined the influence of the uncertainty about climate sensitivity on the allowable (short-term) GHGs emissions budget and on the corresponding stringency of the climatic constraints, either through sensitivity analyses ( Böhringer et al., 2006 [NPR] ; Caldeira et al., 2003 [JoC] ; den Elzen and Meinshausen, 2006 [NPR, ARC] ; Richels et al., 2004 [NPR, SRC] ) or within an optimal control frame-work ( Ambrosi et al., 2003 [MoS, SRC] ; Yohe et al., 2004 [JoC, MoS] ).

On the whole, these studies reach similar conclusions, outlining the significance of uncertainty about climate sensitivity. Ambrosi et al. 2003 [MoS, SRC] ) demonstrates the information value of climate sensitivity before 2030, given the significant economic regrets from a precautionary climate policy in the presence of uncertainty about this parameter. Such information might not be available soon (i.e. at least 50 years could be necessary – Kelly et al., 2000 [NPR, ARC] Yohe et al. 2004 [JoC, MoS] ) thus conclude: ‘uncertainty (about climate sensitivity) is the reason for acting in the near term and uncertainty cannot be used as a justification for doing nothing’.

A few authors analyze the trade-off between a costly acceleration of mitigation costs and a (temporary) overshoot of targets, and the climate impacts of this overshoot. Ambrosi et al. 2003 [MoS, SRC] ) did so through a willingness to pay for not interfering with the climate system. They show that allowing for overshoot of an ex-ante target significantly decreases the required acceleration of decarbonization and the peak of abatement costs, but does not drastically change the level of abatement in the first period. However, the overshoot may significantly increase climate change damages as discussed above (see Section 3.5 ). Another result is that higher climate sensitivity magnifies the rate of warming, which in turn exacerbates adaptation difficulties, and leads to stringent abatement policy recommendations for the coming decades ( Ambrosi, 2007 [NPR, SRC, 2007] ). This result is robust for the choice of discount rate; uncertainty about the rate constraint is proven to be more important for short-term decisions than uncertainty about the magnitude of warming. Therefore, research should be aimed at better characterizing early climate change risks with a view to helping decision-makers in agreeing on a safe guardrail to limit the rate of global warming.

3.6.1.2 Implications of assumptions concerning damage functions in cost-benefit analysis

What is remarkable in cost-benefit studies of the optimal timing of mitigation is that the shape (or curvature) of the damage function matters even more than the ultimate level of damages – a fact long established by ( Peck and Teisberg 1995 ) ). With damage functions exhibiting smooth and regular damages (such as power functions with integer exponents or polynomial functions), GHG abatement is postponed. This is because, for several decades, the temporal rate of increase in marginal climate change damage remains low enough to conclude that investments to accelerate the rate of economic growth are more socially profitable that investing in abatement.

This result changes if singularities in the damage curve represent non-linear events. Including even small probabilities of catastrophic ‘nasty surprises’ may substantially alter optimal short-term carbon taxes (Mastrandea and Schneider, 2004 [NPR, PoC] ; Azar and Lindgren, 2003 [JoC] ). Many other authors report similar findings ( Azar and Schneider, 2001 [NPR, PoC] ; Howarth, 2003 [JoC] ; Dumas and Ha-Duong, 2005 [NPR] ;( Baranzini et al., 2003 ) ), whilst ( Hall and Behl 2006 ) ) suggest a damage function reflecting climate instability needs to include discontinuities in capital stock and the rate of return on capital, and hysteresis with respect to heating and cooling – resulting in a non-convex optimization function such that economic optimization models can provide no solution. But these surprises may be caused by forces other than large catastrophic events. They may also be triggered by smooth climate changes that exceed a vulnerability threshold (e.g. shocks to agricultural systems in developing countries leading to starvation) or by policies that lead to maladaptations to climate change.

In the case of an irreversible THC collapse, ( Keller et al. 2004 ) ) point out another seemingly paradoxical result: if a climate catastrophe seems very likely within a short-term time horizon, it might be economically sound to accept its consequences instead of investing in expensive mitigation to avoid the inevitable. This shows that temporary overshoot of a pre-determined target may be preferable to bearing the social costs of an exaggerated reduction in emission, as well as the need to be attentive to ‘windows of opportunity’ for abatement action. The converse argument is that timely abatement measures, especially in the case of ITC, can reduce long-term mitigation costs and avoid some of the catastrophic events. In this respect, limited differences in GMT curves for different emissions pathways within coming decades are often misinterpreted. It does not imply that early mitigation activities would make no material difference to long-term warming. On the contrary, if the social value of the damages is high enough to justify deep emission cuts decades from now, then early action is necessary due to inertia in socio-economic systems. For example, one challenge is to avoid further build-up of carbon-intensive capital stock.

3.6.2.1 Studies reporting short-term sectoral reduction levels

While there are many potential emissions pathways to a particular stabilization target from a specific year, it is possible to define emissions trajectories based on short-term mitigation opportunities that are consistent with a given stabilization target. This section assesses scenario results (by sector) from top-down models for the year 2030, to evaluate the range of short-term mitigation opportunities in long-term stabilization scenarios. To put these identified mitigation opportunities in context, Chapter 11 , Section 11.3 compares the short-term mitigation estimates across all of the economic sectors.

Many of the modelling scenarios represented in this section were an outcome from the Energy Modelling Forum Study 21 (EMF-21), which focused specifically on multi-gas strategies to address climate change stabilization (see de la Chesnaye and Weyant, 2006 [SRC] ). Models that were evaluated in this assessment are listed in Table 3.12 .

Table 3.12: Top-down models assessed for mitigation opportunities in 2030

For each model, the resulting emissions in the mitigation case for each economic sector in 2030 were compared to projected emissions in a reference case. Results were compared across a range of stabilization targets. For more detail on the relationship between stabilization targets defined in concentrations, radiative forcing and temperature, see Section 3.3.2 .

Key assumptions and attributes vary across the models evaluated, thus having an impact on the results. Most of the top-down models evaluated have a time horizon beyond 2050 such as AIM, IPAC, IMAGE, GRAPE, MiniCAM, MERGE, MESSAGE, and WIAGEM. Top-down models with a time horizon up to 2050, such as POLES and SGM, were also evaluated. The models also vary in their solution concept. Some models provide a solution based on inter-temporal optimization, allowing mitigation options to be adopted with perfect foresight as to what the future carbon price will be. Other models are based on a recursive dynamic, allowing mitigation options to be adopted based only on today’s carbon price. Recursive dynamic models tend to show higher carbon prices to achieve the same emission reductions as in inter-temporal optimization models, because emitters do not have the foresight to take early mitigation actions that may have been cheaper (for more discussion on modelling approaches, refer to Section 3.3.3 ).

Three important considerations need to be remembered with regard to the reported carbon prices. First, these mitigation scenarios assume complete ‘what’ and ‘where’ flexibility (i.e. there is full substitution among GHGs and reductions take place anywhere in the world, according to the principle of least cost). Limiting the degree of flexibility in these mitigation scenarios, such as limiting mitigation only to CO₂, removing major countries or regions from undertaking mitigation, or both, will increase carbon prices, all else being equal. Second, the carbon prices of realizing these levels of mitigation increase in the time horizon beyond 2030 . See Figure 3.25 for an illustration of carbon prices across longer time horizons from top-down scenarios. Third, at the economic sector level, estimated emission reduction for all greenhouse gases varies significantly across the different model scenarios, in part because each model uses sector definitions specific to that type of model.

Across all the models, the long-term target in the stabilization scenarios could be met through the mitigation of multiple greenhouse gases (CO₂, CH₄, N₂O and high-GWP gases). However the specific mitigation options and the treatment of technological progress vary across the models. For example, only some of the models include carbon capture and storage as a mitigation option (GRAPE, IMAGE, IPAC, MiniCAM, and MESSAGE). Some models also include forest sinks as a mitigation option. The model results shown in Table 3.13 do not include forest sinks as a mitigation option, while the results shown in Table 3.14 do include forest sinks, as described in further detail below.

Table 3.13: Global emission reductions from top-down models in 2030 by sector for multi-gas scenarios.

Table 3.13 illustrates the amount of global GHG mitigation reported by sector for the year 2030 across a range of multi-gas stabilization targets. Across the higher Category IV stabilization target scenarios, emission reductions of 3–31% from the reference case emissions across all greenhouse gases can be achieved for a carbon price of 2–57 US$/tCO₂-eq. The results from the POLES models fall into the higher end of the price range, in part due to the recursive dynamic nature of the model, and also due to its shorter time horizon over which to plan. The results from the GRAPE model fall into the lower end of the price range, which is the only inter-temporally optimizing model shown in the higher stabilization scenarios. In the GRAPE results, only 3% of the emissions are reduced by 2030, implying that the majority of the mitigation necessary to meet the target is undertaken beyond 2030 . In scenarios with lower Category I and II stabilization targets, higher levels of short-term mitigation are required to achieve the target in the long run, resulting in a higher range of prices. Emission reductions of approximately 35% can be achieved at a price of 9–92 US$/tCO₂-eq.

Several of the models included in the EMF-21 study also ran multi-gas scenarios that included forest sinks as a mitigation option. Table 3.14 shows the 2030 mitigation estimates for these scenarios that model net land-use change (including forest carbon sinks) as a mitigation option. When terrestrial sinks are modelled as a mitigation option, it can lessen the pressure to mitigate in other sectors. Further discussion of forest sequestration as a mitigation option is presented in Section 3.3.5.5 . Across the higher Category IV stabilization target scenarios, emission reductions of 4–24% from the reference case emissions across all greenhouse gases can be achieved at a price of 2–21 US$/tCO₂-eq. In scenarios with lower Category I and II stabilization targets, emission reductions of 26–40% can be achieved at a price of 31–121 US$/tCO₂-eq.

Table 3.14: Global emission reductions from top-down models in 2030 (by sector) for multi-gas plus sinks scenarios.

3.6.2.2 Assessment of reduction levels at different marginal prices

To put these identified mitigation opportunities into context they will be compared with mitigation estimates from bottom-up models. Chapters 4 through 10 describe mitigation technologies available in specific economic sectors. Chapter 11 , Section 11.3 compares the short-term mitigation estimates across all of the economic sectors for selected marginal costs levels (20, 50 and 100 US$/tCO₂-eq). For that purpose, we have plotted the permit price and (sectoral) reduction levels of the different studies. These plots have been used to explore whether the combination of the studies suggests certain likely reduction levels at the three target levels of 20, 50 and 100 US$/tCO₂-eq. As far more studies were available that reported economy-wide reduction levels than the ones that provided sectoral information, we were able to use a formal statistical method for the former. For the latter, a statistical method was also applied, but outcomes have been used with more care.

Economy-wide reduction levels

Figure 3.40 shows the available data from studies that report economy-wide reduction levels (multi-gas) and permit prices. The data has been taken from the emission scenario database ( Hanaoka et al., 2006 [NPR, SRC] ; Nakicenovic et al., 2006 [SRC] ) – and information directly reported in the context of EMF-21 ( De la Chesnaye and Weyant, 2006 [SRC] ) and IMCP ( Edenhofer et al., 2006 [NPR, ARC] ). The total sets suggest some form of a relationship with studies reporting higher permit prices: also, in general, reporting higher reduction levels.

Figure 3.40: Permit price versus level of emission reduction – total economy in 2030 (the natural logarithm of the permit price is used for the x-axis). The uncertainty range indicated is the 68% interval.

Obviously, a considerable range of results is also found – this is a function of factors such as:

• Model uncertainties, including technology assumptions and inertia.
• Assumed baseline developments.
• The trajectory of the permit price prior to 2030.

The suggested relationship across the total is linear if permit prices are plotted on a logarithmic scale as shown in Figure 3.40 . In other words, the relationship between the two variables is logarithmic, which is a form that is consistent with the general form of marginal abatement curves reported in literature: increasing reduction levels for higher prices, but diminishing returns at higher prices as the reduction tends to reach a theoretical maximum. The figure not only shows the best-guess regression line, but also 68% confidence interval. The latter can be used to derive the 68 percentile interval of the reduction potential for the 20 and 100 US$/tCO₂-eq price levels, which are 13.3 ± 4.6 GtCO₂-eq/yr and 21.5 ± 4.7 GtCO₂-eq/yr, respectively.

Figure 3.41: Permit price versus emission reduction level – several sectors in 2030 (vertical lines indicate levels at 20, 50 and 100 US$/tCO₂).

Sectoral estimates

A more limited set of studies reported sectoral reduction levels. The same plot as Figure 3.40 has been made for the sectoral data (see Figure 3.41 ), again plotting the logarithm of the permit price against emission reduction levels. The data here are directly taken from Table 3.13 and Table 3.14 . As less data are available, the statistical analysis becomes less robust. Nevertheless, for most sectors, a similarly formed relationship was found across the set of studies as for the economy-wide potential (logarithmic relationship showing increasing reduction levels at relatively low prices, and a much slower increase at higher prices). As expected, in several sectors, the spread across models in the 2030 set is larger than in the economy-wide estimates.

In general, a relatively strong relationship is found in the sectors for energy supply, transport, and industrial energy consumption. The relationship between the price and emission reduction level is less clear in other sectors – and more-or-less absent for the limited reported data on the forestry sector. It should be noted here that definitions across studies may be less well-defined – and also, forest sector emissions may actually increase in mitigation scenarios as a result of net deforestation due to bio-energy production.

It should be noted that emission data (and thus also reduction levels) are reported on a ‘point of emission basis’ (emissions are reported for the sectors in which the emissions occur). For example, the efficiency improvements in end-use sectors for electricity lead to reductions in the energy supply sector. Likewise, using bio-energy leads to emission reductions in the end-use sectors, but at the same time (in some models) may lead to increases in emissions for forestry, due to associated land-use changes. The latter may explain differences in the way that data from top-down models are represented elsewhere in this report, as here (in most cases) only the emission changes from mitigation measures in the forestry sector itself are reported. It also explains why the potential in some of the end-use sectors is relatively small, as emission reductions from electricity savings are reported elsewhere.

Reported estimates

On the basis of the available data, the following ranges have been estimated for the reduction potential at a 20, 50 and 100 US$/tCO₂-eq price ( Table 3.15 ). As estimates have been made independently, the total of the different sectors does not add up to the overall range (as expected, the sum of the sectors gives a slightly wider range).

Table 3.15: Reduction potential at various marginal prices, averages across different models (low and high indicate one standard deviation variation).

The largest potential is found in energy supply – covering both the electricity sector and energy supply – with a relatively high capability of responding to permit prices. Relatively high reduction levels are also found for the industry sector. Relatively small reduction levels are reported for the forestry sector and the waste management sector.