The following sections assess future climate projections for several regions, and relate them, where possible, to projected changes in the major climate phenomena assessed in Sections 14.2 to 14.7. The regional climate change assessments are mainly of mean surface air temperature and mean precipitation based primarily on multi-model ensemble projections from general circulation models. Reference is made to the appropriate projection maps from CMIP5 (Taylor et al., 2011c) presented in Annex I: Atlas of Global and Regional Climate Projections. Annex I uses smaller sub-regions similar to those introduced by SREX (Seneviratne et al., 2012). Table 14.1 presents a quantitative summary of the regional area averages over three projection periods (2016–2035, 2046–2065 and 2081–2100 with respect to the reference period 1986– 2005, representing near future, middle century and end of century) for the RCP4.5 scenario. The 26 land regions assessed here are presented in Seneviratne et al., 2012, page 12 and the coordinates can be found from their online Appendix 3.A. Added to this are six additional regions containing the two polar regions, the Caribbean, Indian Ocean and Pacific Island States (see Annex I for further details). Table 14.1 identifies the smaller sub-domains grouped within the somewhat large regions that are discussed in Sections 14.8.2 to 14.8.15. Tables for RCP2.6, RCP6.0 and RCP8.5 scenarios are presented in Supplementary Material Tables 14.SM.1a to 14.SM.1c. For continental-scale regions, projected changes in mean precipitation between (2081–2100) and (1986–2005) are compared in two generations of models forced under two comparable emission scenarios: RCP4.5 in CMIP5 versus A1B in CMIP3. In contrast to the Annex, the seasons here are chosen differently for each region so as to best capture the regional features such as monsoons. Downscaling issues are illustrated in panels showing results from an ensemble of high-resolution time-slice experiments with the Meteorological Research Institute (MRI) model (Endo et al., 2012; Mizuta et al., 2012). To facilitate a direct comparison across the scenarios, the precipitation changes are normalized by the global annual mean surface air temperature changes in each scenario. Published results using other downscaling methods are also assessed when found essential to illustrate issues related to regional climate change.

Regional climate projections are generally more uncertain than projections of global mean temperature but the sources of uncertainty are similar (see Chapters 8, 11, and 12) yet differ in relative importance. For example, natural variability (Deser et al., 2012), aerosol forcing (Chapter 7) and land use/cover changes (DeFries et al., 2002; Moss et al., 2010) all become more important sources of uncertainty on a regional scale. Regional climate assessments incur additional uncertainty due to the cascade of uncertainty through the hierarchy of models needed to generate local information (cf. downscaling in Section 9.6). Calibration (bias correction) of model output to match local observations is an additional important source of uncertainty in regional climate projections (e.g., Ho et al., 2012), which should be considered when interpreting the regional projections. Therefore, the model spread shown in Annex 1 should not be interpreted as the final uncertainty in the observable regional climate change response.

Table 14.2 summarizes the assessed confidence in the ability of CMIP5 models to represent regional scale present-day climate (temperature and precipitation, based on Chapter 9), the main controlling phenomena for weather and climate in that region and the assessed resulting confidence in the future projections. There is generally less confidence in projections of precipitation than of temperature. For example, in Annex I, the temperature projections for 2081–2100 are almost always above the model estimates of natural variability, whereas the precipitation projections less frequently rise above natural variability. Although some projections are robust for reasons that are well understood (e.g., the projected increase in precipitation at high latitudes), many other regions have precipitation projections that vary in sign and magnitude across the models. These issues are further discussed in Section Details on how the confidence table is constructed are found in the Supplementary Material.

Credibility in regional climate change projections is increased if it is possible to find key drivers of the change that are known to be well-simulated and well-projected by climate models. Table 14.3 summarizes the assessment of how major climate phenomena might be relevant for future regional climate change. For each entry in the table, the relevance is based on an assessment of confidence in future change in the phenomenon and the confidence in how the phenomenon influences regional climate. For example, NAO is assigned high relevance (red) for the Arctic region because NAO is known to influence the Arctic and there is high confidence that the NAO index will increase in response to anthropogenic forcing. If there is low confidence in how a phenomenon might change (e.g., ENSO) but high confidence that it has a strong regional impact, then the cell in the table is assigned medium relevance (yellow). It can be seen from the table that there are many cases where major phenomena are assessed to have high (red) or medium (yellow) relevance for future regional climate change. See Supplementary Material Section 14.SM.6.1 for more details on how this relevance table was constructed.