IPCC Report.pdf - Adam Curry

Chapter 3Changes in Climate Extremes and their Impacts on the Natural Physical Environmentin the 21st century (IPCC, 2007a). The tendency for an increase in heavydaily precipitation events was found in many regions, including someregions in which the total precipitation was projected to decrease.Post-AR4 analyses of climate model simulations partly confirm thisassessment but also highlight fairly large uncertainties and model biasesin projections of changes in heavy precipitation in some regions(Section 3.2.3 and Table 3-3). On the other hand, more GCM and RCMensembles have now been analyzed for some regions (Table 3-3; seealso, e.g., Kharin et al., 2007; Kim et al., 2010). At the time of the AR4,Tebaldi et al. (2006) was the main global study available on projectedchanges in precipitation extremes (e.g., Figure 10.18 of Meehl et al.,2007b). Orlowsky and Seneviratne (2011) extended this analysis to alarger number of GCMs from the CMIP3 ensemble and for seasonal inaddition to annual time frames (see also Section 3.3.1). Figure 3-6 providescorresponding analyses of projected annual and seasonal changes ofthe wet-day intensity, the fraction of days with precipitation above the95% quantile of daily wet-day precipitation, and the fraction of dayswith precipitation above 10 mm day -1 . It should be noted that the10 mm day -1 threshold cannot be considered extreme in several regions,but highlights differences in projections for absolute and relativethresholds (see also discussion in Box 3-1 and beginning of this section).Figure 3-6 indicates that regions with model agreement (at least 66%)with respect to changes in heavy precipitation are mostly found in thehigh latitudes and in the tropics, and in some mid-latitude regions of theNorthern Hemisphere in the boreal winter. Regions with at least 90%model agreement are even more limited and confined to the highlatitudes. Overall, model agreement in projected changes is found to bestronger in boreal winter (DJF) than summer (JJA) for most regions.Kharin et al. (2007) analyzed changes in annual maxima of 24-hourprecipitation in the outputs of 14 CMIP3 models. Figure 3-7a displaysthe projected percentage change in the annual maximum of the 24-hourprecipitation rate from the late 20th-century 20-year return values,while Figure 3-7b displays the corresponding projected return periodsfor late 20th-century 20-year return values of the annual maximum24-hour precipitation rates in the mid-21st century (left) and in late 21stcentury (right) under three different emission scenarios (SRES B1, A1B,and A2). Between the late 20th and the late 21st century, the projectedresponses of extreme precipitation to future emissions show increasedprecipitation rates in most regions, and decreases in return periods inmost regions in the high latitudes and the tropics and in some regionsin the mid-latitudes consistent with projected changes in several indicesrelated to heavy precipitation (see Figure 3-6 and Tebaldi et al., 2006),although there are increases in return periods or only small changesprojected in several regions. Except for these regions, the return periodfor an event of annual maximum 24-hour precipitation with a 20-yearreturn period in the late 20th century is projected to be about 5 to 15years by the end of the 21st century. The greatest projected reductionsin return period are in high latitudes and some tropical regions. Thestronger CO 2 emissions scenarios (A1B and A2) lead to greater projecteddecreases in return period. In some regions with projected decreases intotal precipitation (Christensen et al., 2007) such as southern Africa,west Asia, and the west coast of South America, heavy precipitation isnevertheless projected to increase (Figure 3-7, Table 3-3). In some otherareas with projected decreases in total precipitation (e.g., Central Americaand northern South America), however, heavy precipitation is projectedto decrease or not change. It should be noted that Figure 3-7 addressesvery extreme heavy precipitation events (those expected to occur aboutonce in 20 years) whereas Figure 3-6 addresses less extreme, but stillheavy, precipitation events. Projections of changes for these differentlydefined extreme events may differ.Future precipitation projected by the CMIP3 models has also beenanalyzed in a number of studies for various regions using differentcombinations of the models (see next paragraphs and Table 3-3). Ingeneral these studies confirm the findings of global-scale studies byTebaldi et al. (2006) and Kharin et al. (2007).By analyzing simulations with a single GCM, Khon et al. (2007) reporteda projected general increase in extreme precipitation for the differentregions in northern Eurasia especially for winter. Su et al. (2009) foundthat for the Yangtze River Basin region in 2001-2050, the 50-year heavyprecipitation events become more frequent, with return periods fallingto below 25 years (relative to 1951-2000 behavior). For the Indianregion, the Hadley Centre coupled model HadCM3 projects increases inthe magnitude of the heaviest rainfall with a doubling of atmosphericCO 2 concentration (Turner and Slingo, 2009). Simulations by 12 GCMsprojected an increase in heavy precipitation intensity and meanprecipitation rates in east Africa, more severe precipitation deficits inthe southwest of southern Africa, and enhanced precipitation furthernorth in Zambia, Malawi, and northern Mozambique (Shongwe et al.,2009, 2011). Rocha et al. (2008) evaluated differences in the precipitationregime over southeastern Africa simulated by two GCMs underpresent (1961-1990) and future (2071-2100) conditions as a result ofanthropogenic greenhouse gas forcing. They found that the intensity ofall episode categories of precipitation events is projected to increasepractically over the whole region, whereas the number of episodes isprojected to decrease in most of the region and for most episodecategories. Extreme precipitation is projected to increase over Australia in2080-2099 relative to 1980-1999 in an analysis of the CMIP3 ensemble,although there are inconsistencies between projections from differentmodels (Alexander and Arblaster, 2009).High spatial resolution is important for studies of extreme precipitationbecause the physical processes responsible for extreme precipitationrequire high spatial resolution to resolve them (e.g., Kim et al., 2010).Post-AR4 studies have employed three approaches to obtain high spatialresolution to project precipitation extremes: high-resolution GCMs,dynamical downscaling using RCMs, and statistical downscaling (seealso Section 3.2.3.1). Based on the Meteorological Research Instituteand Japan Meteorological Agency 20-km horizontal grid GCM, heavyprecipitation was projected to increase substantially in south Asia, theAmazon, and west Africa, with increased dry spell persistence projectedin South Africa, southern Australia, and the Amazon at the end of the21st century (Kamiguchi et al., 2006). In the Asian monsoon region,heavy precipitation was projected to increase, notably in Bangladesh147