T he G lobal P icture
the increase in monthly extremes, as projected for a 4°C world by the end of the century, would be avoided. Although unusual heat extremes (beyond 3-sigma) would still become substantially more common over extended regions, unprecedented extremes (beyond the 5-sigma threshold) would remain essentially absent over most continents. The patterns of change are similar to those described for a 4°C world, but the frequency of threshold-exceeding extremes is strongly reduced. It is only in some localized tropical regions that a strong increase in frequency compared to the present day is expected (see the regional chapters). In these regions, specifically in western tropical Africa (see Chapter 3 on “Regional Patterns of Climate Change”) and South East Asia (see Chapter 5 on “Regional Patterns of Climate Change”), summer months with unusual temperatures become dominant, occurring in about 60–80 percent of years, and extremes of unprecedented temperatures become regular (about 20–30 percent of years) by the end of the century. In parallel with the increase in global mean temperature, in a 2°C world the percentage of land area with unusual temperatures steadily increases until 2050; it then plateaus at around 20 percent, as shown in Figure 2.7. On a global scale, the land area affected by northern hemisphere summer months with unprecedented temperatures remains relatively small (at less than 5 percent). This implies that, in the near term, extremes would increase manifold compared to today even under the low-emissions scenario. In a 4°C world, the land area experiencing extreme heat would continue to increase until the end of the century. This results in unprecedented monthly heat covering approximately 60 percent of the global land area by 2100. Although these analyses are based on a new set of climate models (that is, those used in ISI-MIP—see Appendix 2), the projections for a 4°C world are quantitatively consistent with the results published in the previous report. Under RCP8.5 (or a 4°C world), the annual frequency of warm nights beyond the 90th percentile increases to between 50–95 percent, depending on region, by the end of the century (Sillmann and Kharin 2013a). Under RCP2.6 (or a 2°C world), the frequency of warm nights remains limited to between 20–60 percent, with the highest increases in tropical South East Asia and the Amazon region (Sillmann and Kharin 2013a). Extremes, expressed as an exceedance of a particular percentile threshold derived from natural variability in the base period, show the highest increase in tropical regions, where interannual temperature variability is relatively small. Under RCP8.5, the duration of warm spells, defined as the number of consecutive days beyond the 90th percentile (Sillmann and Kharin 2013b), increases in tropical regions to more than 300, occurring essentially year round (Sillmann and Kharin 2013a).
Precipitation Projections On a global scale, warming of the lower atmosphere strengthens the hydrological cycle, mainly because warmer air can hold more
Figure 2.7: Multi-model mean (thick line) and individual models (thin lines) of the percentage of global land area warmer than 3-sigma (top) and 5-sigma (bottom) during boreal summer months (JJA) for scenarios RCP2.6 and RCP8.5
water vapor (Coumou and Rahmstorf 2012). This strengthening causes dry regions to become drier and wet regions to become wetter (Trenberth 2010). There are other important mechanisms, however, such as changes in circulation patterns and aerosol forcing, which may lead to strong deviations from this general picture. Increased atmospheric water vapor can also amplify extreme precipitation (Sillmann and Kharin 2013a). Although modest improvements have been reported in the precipitation patterns simulated by the state-of-the-art CMIP5 models (Kelley, Ting, Seager, and Kushnir 2012; Jia & DelSole 2012; Zhang and Jin 2012) as compared to the previous generation (CMIP3), substantial uncertainty remains. This report therefore only provides changes in precipitation patterns on annual and seasonal timescales. The ISI-MIP models used were bias-corrected such that they reproduce the observed historical mean and variation in precipitation. The projections might therefore also provide more robust and consistent trends on regional scales. The expected change in annual mean precipitation by 2071– 99 relative to 1951–80 is shown in Figure 2.8 for RCP2.6 (a 2°C 13