The RCPs were selected from the existing literature on the basis of their emissions and associated concentration levels. This implies that the socio-economic assumptions of the different modeling teams were based on individual model assumptions made within the context of the original publication, and that there is no consistent design behind the position of the different RCPs relative to each other for these parameters. Scenario development after the RCP phase (Section 1) will focus on developing a new set of socio-economic scenarios. Therefore, socio-economic parameters have not been included in the RCP information available for download. Still, this information does form part of the underlying individual scenario development, and thus provides useful information on internal logic and the plausibility of each of the individual RCPs. Here, their primary characteristics are discussed only in this context.
The population and GDP pathways underlying the four RCPs are shown in Fig. 2. The figure also shows, as reference, the UN population projections and the 90th percentile range of GDP scenarios in the literature on greenhouse gas emission scenarios. Figure 2 shows the RCPs to be consistent with these two references. It should be noted that, with one exception (RCP8.5), the modeling teams deliberately made intermediate assumptions about the main driving forces (as illustrated by their position in Fig. 2) (see the relevant papers elsewhere in this Special Issue). In contrast, the RCP8.5 was based on a revised version of the SRES A2 scenario; here, the storyline emphasizes high population growth and lower incomes in developing countries.
For energy use, the scenarios underlying the RCPs are consistent with the literature—with the RCP2.6, RCP4.5 and RCP6 again being representative of intermediate scenarios in the literature (resulting in a primary energy use of 750 to 900 EJ in 2100, or about double the level of today)Footnote 9 (Fig. 3). The RCP8.5, in contrast, is a highly energy-intensive scenario as a result of high population growth and a lower rate of technology development. In terms of the mix of energy carriers, there is a clear distinction across the RCPs given the influence of the climate target (for details, see the papers elsewhere in this Special Issue). Total fossil-fuel use basically follows the radiative forcing level of the scenarios; however, due to the use of carbon capture and storage (CCS) technologies (in particular in the power sector), all scenarios, by 2100, still use a greater amount of coal and/or natural gas than in the year 2000. The use of oil stays fairly constant in most scenarios, but declines in the RCP2.6 (as a result of depletion and climate policy). The use of non-fossil fuels increases in all scenarios, especially renewable resources (e.g. wind, solar), bio-energy and nuclear power. The main driving forces are increasing energy demand, rising fossil-fuel prices and climate policy. An important element of the RCP2.6 is the use of bio-energy and CCS, resulting in negative emissions (and allowing some fossil fuel without CCS by the end of the century).
The Kaya identity describes future emission levels as a simple multiplicative function of population, income per capita, energy per unit of income (energy intensity) and emissions per unit of primary energy (carbon factor) (Kaya 1989; Ehrlich 1971) (Fig. 4). These factors are often used to provide insight into scenario trends. The figure shows all RCPs to be above the mean values in the literature for energy intensity which is caused mostly by the inclusion of traditional fuels. Analysis of the Kaya factors shows the influence of the radiative forcing targets, and indicates that the scenarios underlying the RCPs cover the full range of possible values reasonably well. RCP2.6 achieves most of its emission reductions by reducing the carbon factor (changes in supply mix) but is also the lowest scenario in terms of energy intensity, although much lower values are found in the literature. RCP6 and RCP8.5 both show a rather constant trend for the carbon factor (heavy reliance on fossil fuels), but are very different in terms of the development of energy intensity (high for RCP8.5 and intermediate for RCP6). Finally, RCP4.5 shows trends that are very similar to those in RCP2.6, but far less extreme.
A crucial element of the new scenarios is land use. Land use influences the climate system in many different ways including direct emissions from land-use change, hydrological impacts, biogeophysical impacts (such as changes in albedo and surface roughness), and the size of the remaining vegetation stock (influencing CO2 removal from the atmosphere). Historically, cropland and anthropogenic use of grassland have both been increasing, driven by rising population and changing dietary patterns. There are far fewer land-use scenarios published in the literature than emission or energy-use scenarios. Moreover, far less experience exists with scenario projections (Rose et al. 2011; Smith et al. 2010). Most projections focus on a shorter time period (up to 2030 or 2050) and show an increasing demand for cropland and pasture.
The limited experience in global land-use modeling as part of integrated assessment work is also reflected in the RCP development process. Compared to emission modeling, definitions of relevant variables and base year data differ more greatly across the IAMs for the land use components. This, along with the importance of retaining continuity at grid cell level with historical data, required more extensive harmonization activities (i.e. minimizing the difference between historical reconstruction and future projections, and preserving as much information from IAMs as possible). As a first step, general agreement on the 2005 global land-use definitions and values was reached. Prior to harmonization, inconsistencies in definitions of cropland, pasture, and wood harvest resulted in significant discrepancies between IAM values for their initial year (2005), and the HYDE 3.1 or FAO final year values (also 2005). With consistent definitions and reanalysis, these inconsistencies were reduced to <12% for 2005 between HYDE 3.1, FAO and the IAMs, except for RCP8.5 (MESSAGE) pasture (Hurtt et al. 2011). Subsequently, the IAM decadal changes in land use were aggregated over a 2° × 2° grid, and these changes were applied sequentially to the 2005 land-use distribution of HYDE3.1. As a basic rule, future land use for the RCPs was based on the absolute changes in the IAM output, combined with 2005 historical data. The resulting 2° × 2° grids were then disaggregated into 0.5° × 0.5° grids.
The RCPs cover a very wide-range of land-use scenario projections. This is illustrated by the trends shown in Fig. 5 (i.e. after harmonization). The use of cropland and grasslands increases in RCP8.5, mostly driven by an increasing global population. Cropland also increases in the RCP2.6, but largely as a result of bio-energy production. The use of grassland is more-or-less constant in the RCP2.6, as the increase in production of animal products is met through a shift from extensive to more intensive animal husbandry. The RCP6 shows an increasing use of cropland but a decline in pasture. This decline is caused by a similar trend as noted for RCP2.6, but with a much stronger implementation. Finally, the RCP4.5 shows a clear turning point in global land use based on the assumption that carbon in natural vegetation will be valued as part of global climate policy. As a result of reforestation programs, the use of cropland and grassland decreases, following considerable yield increases and dietary changes. In comparison with the general scenario literature, the range covered by the RCPs is wider as it not only includes pathways of continuing expansion of agricultural land use, but also those that show a contraction of agricultural land.
The four different RCPs also produced different patterns of future land use. By 2100, in RCP8.5, areas of high-density cropland are evident in the United States, Europe, and South-East Asia. High-density pasture areas are evident in the Western United States, Eurasia, South Africa, and Australia. Primary forest is most concentrated in northern high latitudes, and parts of Amazonia, while secondary vegetation is common in the United States, Africa, South America and Eurasia. Patterns from RCP6 are broadly similar, but clearly with less pasture generally and especially in the United States, Africa, Eurasia and Australia. RCP4.5 has less cropland overall than either of the previous RCPs, more land with no fractional cropland, and high-density areas of secondary vegetation in the United States, Africa and Eurasia. Spatial patterns from RCP2.6 are broadly similar to those of RCP4.5.
Greenhouse gas emissions
Emission and concentrations were harmonized to available historical data for the 2000–2005 period. For CO2 emissions from land-use change, in contrast, the average of the four RCP models was used as the 2005 harmonization value. On an aggregate scale, the difference between the original data and the final harmonized data are generally small. For the RCP2.6, RCP4.5 and RCP8.5 scenarios, the difference in total CO2 equivalent greenhouse gas emissions of 2005 was 2 to 4%, with 10% difference for the RCP6 scenario. The difference between the harmonized and unharmonized scenarios for cumulative emissions over the 2000–2050 period in total CO2 equivalent emissions is expected to be 1 to 2%, except for the RCP6 scenario, which has a difference of 5% (Meinshausen et al. 2011b).
The CO2 emissions of the four RCPs correspond well with the literature range, which was part of their selection criterion (Fig. 6). The RCP8.5 is representative of the high range of non-climate policy scenarios. Most non-climate policy scenarios, in fact, predict emissions of the order of 15 to 20 GtC by the end of the century, which is close to the emission level of the RCP6. The forcing pathway of the RCP4.5 scenario is comparable to a number of climate policy scenarios and several low-emissions reference scenarios in the literature, such as the SRES B1 scenario. The RCP2.6 represents the range of lowest scenarios, which requires stringent climate policies to limit emissions.
The trends in CH4 and N2O emissions are largely due to differences in the assumed climate policy along with differences in model assumptions (Fig. 6). Emissions of both CH4 and N2O show a rapidly increasing trend for the RCP8.5 (no climate policy and high population). For RCP6 and RCP4.5, CH4 emissions are more-or-less stable throughout the century, while for RCP2.6, these emissions are reduced by around 40%. The low emission trajectories for CH4 are a net result of low cost emission options for some sources (e.g. from energy production and transport), and a limited reduction for others (e.g. from livestock). Introduction of climate policy, thus, may lead to significant emission reductions, even in the short term, but will not eliminate emissions altogether. While the RCP CH4 emissions are within the ranges from the literature, there is a significant gap between RCP2.6, RCP4.5 and RCP6 on the one hand and the high-emission RCP8.5 scenario on the other. For N2O, the scenarios are placed in similar order, although here the emissions for RCP4.5 remain stable while those for RCP6 increase over time. In this case, the RCPs do not cover the full range in the literature, but only the more representative range. One may, however, question the studies that indicate very rapidly increasing and decreasing N2O emissions, given the main sources of N2O (these are mostly agricultural and will grow at a modest rate, in the future, but to some degree are also difficult to abate). It is important to recognize that there is substantial uncertainty in base-year emissions for many substances (Granier et al. 2011). The RCP scenarios, due to the design of the harmonization process, do not fully represent this uncertainty.
Emissions of atmospheric air pollutants
The RCPs generally exhibit a declining trend of air polluting emissions. The emission trends for air pollutants are determined by three factors: the change in driving forces (fossil-fuel use, fertilizer use), the assumed air pollution control policy, and the assumed climate policy (as the last induces changes in energy consumption leading to changes (generally reductions) in air polluting emissions). We have illustrated the trends in air pollutants by looking at SO2 and NOx (Fig. 7). In general, similar trends can be seen for other air pollutants.
All RCPs include the assumption that air pollution control becomes more stringent, over time, as a result of rising income levels. Globally, this would cause emissions to decrease, over time—although trends can be different for specific regions or at particular moments in time. A second factor that influences the results across the RCPs is climate policy. In general, the lowest emissions are found for the scenario with the most stringent climate policy (RCP2.6) and the highest for the scenario without climate policy (RCP8.5), although this does not apply to all regions, at all times. The overall correlation is a result of the fact that climate policy induces systemic changes in the energy system, away from technologies with high greenhouse gas emission levels, which also have high emissions of air pollutants (e.g. coal use without CCS has high emission levels of CO2, but also of SO2). In contrast, the application of energy efficiency or use of renewables reduces both greenhouse gas emissions and air pollutants. The range of air pollution projections, generally, is smaller than that found in the literature. This is mostly due to the RCPs’ shared assumption of stringent air pollution policies increasing proportionally with income (van Ruijven et al. 2008). As such, one may conclude that the RCPs show a range of plausible development pathways for air pollutants and policy interventions, but they are not fully representative of the literature on air polluting emissions, as the set does not include scenarios which assume that very little or no reduction of emissions will be achieved. This may limit the use of the RCPs for specific air pollution applications.
The emissions in the RCPs have been downscaled to 0.5° × 0.5° grids per sector (Masui et al. 2011; Riahi et al. 2011; Thomson et al. 2011; Van Vuuren et al. 2011a)—allowing their use in atmospheric climate and chemistry models (Fig. 8). The results show that for most gases, emissions are concentrated in specific areas (e.g. Eastern United States, Western Europe, Eastern China and India). Moreover, a general trend can be noted across all RCPs and gases, indicating that emissions tend to become relatively more concentrated in currently low-income regions.
Concentrations of greenhouse gases
The greenhouse gas concentrations in the RCPs closely correspond to the emissions trends discussed earlier (Fig. 9). For CO2, RCP8.5 follows the upper range in the literature (rapidly increasing concentrations). RCP6 and RCP4.5 show a stabilizing CO2 concentration (close to the median range in the literature). Finally, RCP2.6 has a peak in CO2 concentrations around 2050, followed by a modest decline to around 400 ppm CO2, by the end of the century. For CH4 and N2O, the order in which the RCPs can be placed are also a direct result of the assumed level of climate policy. The trends in CH4 concentrations are more pronounced, as a result of the relatively short lifetime of CH4. Emission reductions, as in the RCP2.6 and RCP4.5, therefore, may lead to an emission peak much earlier in the century. For N2O, in contrast, a relatively long lifetime and a modest reduction potential imply an increase in concentrations, in all RCPs. For both CH4 and N2O, the concentration levels correspond well with the range in the literature. Further information on the calculations of concentration can be found in Meinshausen et al. (2011b)
The combination of trends in greenhouse gases and those in atmospheric pollutants translate to changes in concentrations affecting the overall development of radiative forcing. As shown in Fig. 10, the RCPs, as specified in the original selection criteria, cover the trends and level of radiative forcing values of scenarios in the literature very well. Total radiative forcing is determined by both positive forcing from greenhouse gases and negative forcing from aerosols. The most dominant factor, by far, is the forcing from CO2. As a result, both for the RCPs and in the overall literature, 2100 radiative forcing levels are correlated with cumulative 21st century CO2 emissions (see middle panel of Fig. 10). Thus, it is not surprising that the RCPs are consistent with the literature, both in terms of total forcing and cumulative CO2 emissions (over the course of the century).
Concentration of air pollutants
For tropospheric ozone (driven by the changes in NOx, VOC, OC and methane emissions, along with changes in climate conditions), there is a clear difference between the RCPs. For RCP8.5, radiative forcing from tropospheric ozone, according to the CAM3.5 calculations, increases by an additional 0.2 W/m2 by 2100 (Lamarque et al. 2011). In contrast, there is a decrease in radiative forcing, for RCP4.5 and RCP2.6, of 0.07 and 0.2 W/m2, respectively (again CAM3.5). This is the result of assumed trends in air pollution control and climate policy.
Aerosol concentrations eventually decrease in all RCPs, following the strong decrease in emissions, especially those of anthropogenic SO2. This is very different from some of the SRES scenarios. However, the new insights into implementation of air pollution control measures were developed more recently, which were not comprehensively included in the SRES (Smith and Wigley 2006). Moreover, the SRES scenarios did not include climate policy measures. While there is a reduction in the impact of aerosols, at a global level, for some tropical regions, a shift towards higher concentrations is also reported. Finally, for nitrogen deposition, a decrease can be observed across the RCPs for most high-income regions. However, in many developing regions, an increase in nitrogen deposition is projected for the end of the 21st century, mostly related to the projected increases in NH3 emissions due to agricultural activities.
The MAGICC model used for calculating greenhouse gas concentrations results in tropospheric and stratospheric forcing levels that are slightly different from those of the more complex model used for the atmospheric chemistry calculations (Lamarque et al. 2011). For stratospheric ozone, this is related to the MAGICC model assumption of stratospheric ozone being solely driven by the amount of ozone-depleting substances. Full chemistry-climate model simulations (Lamarque et al. 2011) indicate that climate change is an important additional component in the evolution of stratospheric ozone. These small differences in ozone forcing, however, are only a very small fraction of total forcing in the RCP scenarios.
Extending the RCPs to 2300
Figure 11 shows the CO2 emissions and radiative forcing trajectories for each of the four extensions of the RCPs (ECPs). As explained in the method sections, these have not been based on integrated assessment modeling, but on simple extension rules consistent with the rationale of each of the RCPs to which they connect (see Table 3). This has resulted in a set of extended concentration pathways to be used for climate model runs. Still, it is useful to examine the implied changes in emissions. For CO2, these are also shown in Fig. 11. The figure indicates that the simple extension rules (stabilization of RF for ECP8.5, ECP6 and ECP4.5 at 12, 6 and 4.5 W/m2, respectively) imply considerable reductions in CO2 emission beyond 2100. For the last two ECPs, this can be seen as a continuation of the trends of before 2100. For ECP8.5, this, in fact, implies a major trend break and an emission reduction, between 2150 and 2250, which reduces emissions at a similar rate as in RCP2.6 before 2100, but over 2 to 3 times the total emission volume. For ECP3PD, the assumed continuation of negative emissions implies that sufficient storage capacity will be found to store CO2 from bio-energy, CCS use, or other technologies that may remove CO2 from the atmosphere. Storage before 2100 (of emissions from bio-energy and fossil fuels) equals about 600 GtC. Assuming that after 2100 the storage potential only will be used for bio-energy and CCS (BECCS), the continuation of the scenario would at least require another 200 GtC. Optimistic estimates of storage potential are consistent with these numbers. By 2300, this scenario would result in a radiative forcing at roughly the same level as in 2000. Finally, a special extension was added to explore the difference in impacts from direct stabilization at 4.5 W/m2 and an initial overshoot to 6.0 W/m2 (SCP6to4.5). The extension shows that such an overshoot scenario would be possible, but would require a very abrupt emission reduction from the 6.0 W/m2 profile and a long period of negative CO2 emissions. In other words, this would be a scenario that would be relatively hard to achieve (Meinshausen et al. 2011b).