IPCC Report.pdf - Adam Curry

Chapter 3Changes in Climate Extremes and their Impacts on the Natural Physical EnvironmentIn climates where seasonal snow storage and melting play a significantrole in annual runoff, the hydrologic regime is affected by changes intemperature. In a warmer world, a smaller portion of precipitation fallsas snow (Hirabayashi et al., 2008a) and the melting of winter snowoccurs earlier in spring, resulting in a shift in peak river runoff to winterand early spring. This has been observed in the western United States(Regonda et al., 2005; Clow, 2010), in Canada (Zhang et al., 2001), andin other cold regions (Rosenzweig et al., 2007; Shiklomanov et al.,2007), along with an earlier breakup of river ice in Arctic rivers (Smith,2000; Beltaos and Prowse, 2009). The observed trends toward earliertiming of snowmelt-driven streamflows in the western United Statessince 1950 are detectably different from natural variability (Barnett etal., 2008; Hidalgo et al., 2009). Thus, observed warming over severaldecades that is attributable to anthropogenic forcing has likely beenlinked to earlier spring peak flows in snowmelt- and glacier-fed rivers. Itis unclear if observed warming over several decades has affected themagnitude of the snowmelt peak flows, but warming may result eitherin an increase in spring peak flows where winter snow depth increases(Meehl et al., 2007b) or a decrease in spring peak flows because ofdecreased snow cover and amounts (Hirabayashi et al., 2008b; Dankersand Feyen, 2009).There is still a lack of studies identifying an influence of anthropogenicclimate change over the past several decades on rain-generated peakstreamflow trends because of availability and uncertainty in theobserved streamflow data and low signal-to-noise ratio. Evidence hasrecently emerged that anthropogenic climate change could haveincreased the risk of rainfall-dominated flood occurrence in some riverbasins in the United Kingdom in autumn 2000 (Pall et al., 2011). Overall,there is low confidence (due to limited evidence) that anthropogenicclimate change has affected the magnitude and frequency of floods,though it has detectably influenced several components of thehydrological cycle, such as precipitation and snowmelt, that may impactflood trends. The assessment of causes behind the changes in floods isinherently complex and difficult.The number of studies that investigated projected flood changes inrivers especially at a regional or a continental scale was limited whenthe AR4 was published. Projections of flood changes at the catchment/river-basin scale were also not abundantly cited in the AR4. Nevertheless,Kundzewicz et al. (2007) and Bates et al. (2008) argued that morefrequent heavy precipitation events projected over most regions wouldaffect the risk of rain-generated floods (e.g., flash flooding and urbanflooding).The number of regional- or continental-scale studies of projectedchanges in floods is still limited. Recently, a few studies for Europe(Lehner et al., 2006; Dankers and Feyen, 2008, 2009) and a study for theglobe (Hirabayashi et al., 2008b) have indicated changes in the frequencyand/or magnitude of floods in the 21st century at large scale using dailyriver discharge calculated from RCM or GCM outputs and hydrologicalmodels. A notable change is projected to occur in northeastern Europein the late 21st century because of a reduction in snow accumulation(Dankers and Feyen, 2008, 2009; Hirabayashi et al., 2008b), that is, adecrease in the probability of floods, that generally corresponds tolower flood peaks. For other parts of the world, Hirabayashi et al.(2008b) show an increase in the risk of floods in most humid Asianmonsoon regions, tropical Africa, and tropical South America with adecrease in the risk of floods in non-negligible areas of the world suchas most parts of northern North America.Projections of flood changes at the catchment/river-basin scale are alsonot abundant in the scientific literature. Several studies have beenundertaken for UK catchments (Cameron, 2006; Kay et al., 2009;Prudhomme and Davies, 2009) and catchments in continental Europeand North America (Graham et al., 2007; Thodsen, 2007; Leander et al.,2008; Raff et al., 2009; van Pelt et al., 2009). However, projections forcatchments in other regions such as Asia (Asokan and Dutta, 2008;Dairaku et al., 2008), the Middle East (Fujihara et al., 2008), SouthAmerica (Nakaegawa and Vergara, 2010; Kitoh et al., 2011), and Africa(Taye et al., 2011) are rare.Uncertainty is still large in the projected changes in the magnitude andfrequency of floods. It has been recently recognized that the choice ofGCMs is the largest source of uncertainties in hydrological projectionsat the catchment/river-basin scale, and that uncertainties from emissionscenarios and downscaling methods are also relevant but less important(Graham et al., 2007; Leander et al., 2008; Kay et al., 2009; Prudhommeand Davies, 2009), although, in general, hydrological projections requiredownscaling and/or bias-correction of GCM outputs (e.g., precipitationand temperature). Also the choice of hydrological models was found tobe relevant but less important (Kay et al., 2009; Taye et al., 2011).However, the relative importance of downscaling, bias-correction, andthe choice of hydrological models may depend on the selected region/catchment, the selected downscaling and bias-correction methods, andthe selected hydrological models (Wilby et al., 2008). For example, thesign of the above-mentioned flood changes in northeastern Europe isaffected by differences in temporal downscaling and bias-correctionmethods applied in the different studies (Dankers and Feyen, 2009).Chen et al. (2011) demonstrated considerable uncertainty caused byseveral downscaling methods in a hydrological projection for asnowmelt-dominated Canadian catchment. Downscaling (see Section3.2.3) and bias-correction are also a major source of uncertainty in raindominatedcatchments (van Pelt et al., 2009). We also note that biascorrectionand statistical downscaling tend to ignore the energy closureof the climate system, which could be a non-negligible source ofuncertainty in hydrological projections (Milly and Dunne, 2011).The number of projections of flood magnitude and frequency changes isstill limited at regional and continental scales. Projections at thecatchment/river-basin scale are also not abundant in the peer-reviewedscientific literature, especially for regions outside Europe and NorthAmerica. In addition, considerable uncertainty remains in the projectionsof flood changes, especially regarding their magnitude and frequency.Therefore, our assessment is that there is low confidence (due to limitedevidence) in future changes in flood magnitude and frequency derived177