IPCC Report.pdf - Adam Curry

138much of the continental United States and every 5 years over mostof Canada; by the end of the century, it would occur every otheryear or more.Meehl et al. (2009c) examined changes in record daily high and lowtemperatures in the United States and show that even with projectedstrong warming resulting in many more record highs than lows, theoccasional record low is still set. For Australia, the CMIP3 ensembleprojected increases in warm nights (15-40% by the end of the 21stcentury) and heat wave duration, together with a decrease in the numberof frost days (Alexander and Arblaster, 2009). Inland regions showgreater warming compared with coastal zones (Suppiah et al., 2007;Alexander and Arblaster, 2009) and large increases in the number ofdays above 35 or 40°C are indicated (Suppiah et al., 2007). For theentire South American region, a study with a single RCM projected morefrequent warm nights and fewer cold nights (Marengo et al., 2009a).Several studies of regional and global model projections of changes inextremes are available for the European continent (see also Table 3-3).Analyses of both global and regional model outputs show majorincreases in warm temperature extremes across the Mediterraneanregion including events such as hot days (Tmax >30°C) and tropicalnights (Tmin>20°C) (Giannakopoulos et al., 2009; Tolika et al., 2009).Comparison of RCM projections using the A1B forcing scenario, withdata for 2007 (the hottest summer in Greece in the instrumental recordwith a record daily Tmax observed value of 44.8°C) indicates that thedistribution for 2007 is closer to the distribution for 2071-2100 than forthe 2021-2050 period, thus 2007 might be considered a ‘normal’ summerof the future (Founda and Giannakopoulos, 2009; Tolika et al., 2009).Beniston et al. (2007) concluded from an analysis of RCM output thatregions such as France and Hungary may experience as many days peryear above 30°C as currently experienced in Spain and Sicily. In thisRCM ensemble, France was the area with the largest projected warmingin the uppermost percentiles of daily summer temperatures althoughthe mean warming was greatest in the Mediterranean region (Fischerand Schär, 2009). New results from an RCM ensemble project increasesin the amplitude, frequency, and duration of health-impacting heat waves,especially in southern Europe (Fischer and Schär, 2010). Overall theseregional assessments are consistent with the global assessments providedabove. It should be noted, however, that the assessed uncertainty is largerat the regional level than at the continental or global level (see Box 3-2).Global-scale trends in a specific extreme may be either more reliable orless reliable than regional-scale trends, depending on the geographicaluniformity of the trends in the specific extreme (Section 3.1.6).Chapter 3Changes in Climate Extremes and their Impacts on the Natural Physical EnvironmentFraction of Warm NightsANN DJF JJA Standard Deviation8 4 0 4 8Fraction of Cold Nights Standard Deviation3 1.5 0 1.5 3Percentage Days with Tmin>20 Percentage of Days28 14 0 14 28Figure 3-4 | Projected annual and seasonal changes in three indices for daily Tmin for 2081-2100 with respect to 1980-1999, based on 14 GCMs contributing to the CMIP3.Left column: fraction of warm nights (days at which Tmin exceeds the 90th percentile of that day of the year, calculated from the 1961-1990 reference period); middle column:fraction of cold nights (days at which Tmin is lower than the 10th percentile of that day of the year, calculated from the 1961-1990 reference period); right column: percentage ofdays with Tmin >20°C. The changes are computed for the annual time scale (top row) and two seasons (December-January-February, DJF, middle row, and June-July-August, JJA,bottom row) as the fractions/percentages in the 2081-2100 period (based on simulations under emission scenario SRES A2) minus the fractions/percentages of the 1980-1999period (from corresponding simulations for the 20th century). Warm night and cold night changes are expressed in units of standard deviations, derived from detrended per yearannual or seasonal estimates, respectively, from the three 20-year periods 1980-1999, 2046-2065, and 2081-2100 pooled together. Tmin >20°C changes are given directly asdifferences of percentage points. Color shading is only applied for areas where at least 66% (i.e., 10 out of 14) of the GCMs agree in the sign of the change; stippling is appliedfor regions where at least 90% (i.e.,13 out of 14) of the GCMs agree in the sign of the change. Adapted from Orlowsky and Seneviratne (2011); updating Tebaldi et al. (2006) foradditional number of indices and CMIP3 models, and including seasonal time frames. For more details, see Appendix 3.A.