3.4 Key future impacts and vulnerabilities
3.4.1 Surface waters
Since the TAR, over 100 studies of climate change effects on river flows have been published in scientific journals, and many more have been reported in internal reports. However, studies still tend to be heavily focused on Europe, North America, and Australasia. Virtually all studies use a hydrological model driven by scenarios based on climate model simulations, with a number of them using SRES-based scenarios (e.g., Hayhoe et al., 2004; Zierl and Bugmann, 2005; Kay et al., 2006a). A number of global-scale assessments (e.g., Manabe et al., 2004a, b; Milly et al., 2005, Nohara et al., 2006) directly use climate model simulations of river runoff, but the reliability of estimated changes is dependent on the rather poor ability of the climate model to simulate 20th century runoff reliably.
Methodological advances since the TAR have focused on exploring the effects of different ways of downscaling from the climate model scale to the catchment scale (e.g., Wood et al., 2004), the use of regional climate models to create scenarios or drive hydrological models (e.g., Arnell et al., 2003; Shabalova et al., 2003; Andreasson et al., 2004; Meleshko et al., 2004; Payne et al., 2004; Kay et al., 2006b; Fowler et al., 2007; Graham et al., 2007a, b; Prudhomme and Davies, 2007), ways of applying scenarios to observed climate data (Drogue et al., 2004), and the effect of hydrological model uncertainty on estimated impacts of climate change (Arnell, 2005). In general, these studies have shown that different ways of creating scenarios from the same source (a global-scale climate model) can lead to substantial differences in the estimated effect of climate change, but that hydrological model uncertainty may be smaller than errors in the modelling procedure or differences in climate scenarios (Jha et al., 2004; Arnell, 2005; Wilby, 2005; Kay et al., 2006a, b). However, the largest contribution to uncertainty in future river flows comes from the variations between the GCMs used to derive the scenarios.
Figure 3.3 provides an indication of the effects of future climate change on long-term average annual river runoff by the 2050s, across the world, under the A2 emissions scenario and different climate models used in the TAR (Arnell, 2003a). Obviously, even for large river basins, climate change scenarios from different climate models may result in very different projections of future runoff change (e.g., in Australia, South America, and Southern Africa).
Figure 3.3. Change in average annual runoff by the 2050s under the SRES A2 emissions scenario and different climate models (Arnell, 2003a).
Figure 3.4 shows the mean runoff change until 2050 for the SRES A1B scenario from an ensemble of twenty-four climate model runs (from twelve different GCMs) (Milly et al., 2005). Almost all model runs agree at least with respect to the direction of runoff change in the high latitudes of North America and Eurasia, with increases of 10 to 40%. This is in agreement with results from a similar study of Nohara et al. (2006), which showed that the ensemble mean runoff change until the end of the 21st century (from nineteen GCMs) is smaller than the standard deviation everywhere except at northern high latitudes. With higher uncertainty, runoff can be expected to increase in the wet tropics. Prominent regions, with a rather strong agreement between models, of decreasing runoff (by 10 to 30%) include the Mediterranean, southern Africa, and western USA/northern Mexico. In general, between the late 20th century and 2050, the areas of decreased runoff expand (Milly et al., 2005).
Figure 3.4. Change in annual runoff by 2041-60 relative to 1900-70, in percent, under the SRES A1B emissions scenario and based on an ensemble of 12 climate models. Reprinted by permission from Macmillan Publishers Ltd. [Nature] (Milly et al., 2005), copyright 2005.
A very robust finding of hydrological impact studies is that warming leads to changes in the seasonality of river flows where much winter precipitation currently falls as snow (Barnett et al., 2005). This has been found in projections for the European Alps (Eckhardt and Ulbrich, 2003; Jasper et al., 2004; Zierl and Bugmann, 2005), the Himalayas (Singh, 2003; Singh and Bengtsson, 2004), western North America (Loukas et al., 2002a, b; Christensen et al., 2004; Dettinger et al., 2004; Hayhoe et al., 2004; Knowles and Cayan, 2004; Leung et al., 2004; Payne et al., 2004; Stewart et al., 2004; VanRheenen et al., 2004; Kim, 2005; Maurer and Duffy, 2005), central North America (Stone et al., 2001; Jha et al., 2004), eastern North America (Frei et al., 2002; Chang, 2003; Dibike and Coulibaly, 2005), the entire Russian territory (Shiklomanov and Georgievsky, 2002; Bedritsky et al., 2007), and Scandinavia and Baltic regions (Bergström et al., 2001; Andreasson et al., 2004; Graham, 2004). The effect is greatest at lower elevations (where snowfall is more marginal) (Jasper et al., 2004; Knowles and Cayan, 2004), and in many cases peak flow would occur at least a month earlier. Winter flows increase and summer flows decrease.
Many rivers draining glaciated regions, particularly in the Hindu Kush-Himalaya and the South-American Andes, are sustained by glacier melt during the summer season (Singh and Kumar, 1997; Mark and Seltzer, 2003; Singh, 2003; Barnett et al., 2005). Higher temperatures generate increased glacier melt. Schneeberger et al. (2003) simulated reductions in the mass of a sample of Northern Hemisphere glaciers of up to 60% by 2050. As these glaciers retreat due to global warming (see Chapter 1), river flows are increased in the short term, but the contribution of glacier melt will gradually decrease over the next few decades.
In regions with little or no snowfall, changes in runoff are dependent much more on changes in rainfall than on changes in temperature. A general conclusion from studies in many rain-dominated catchments (Burlando and Rosso, 2002; Evans and Schreider, 2002; Menzel and Burger, 2002; Arnell, 2003b, 2004a; Boorman, 2003a; Booij, 2005) is that flow seasonality increases, with higher flows in the peak flow season and either lower flows during the low flow season or extended dry periods. In most case-studies there is little change in the timing of peak or low flows, although an earlier onset in the East Asian monsoon would bring forward the season of peak flows in China (Bueh et al., 2003).
Changes in lake levels are determined primarily by changes in river inflows and precipitation onto and evaporation from the lake. Impact assessments of the Great Lakes of North America show changes in water levels of between -1.38 m and +0.35 m by the end of the 21st century (Lofgren et al., 2002; Schwartz et al., 2004). Shiklomanov and Vasiliev (2004) suggest that the level of the Caspian Sea will change in the range of 0.5 to 1.0 m. In another study by Elguindi and Giorgi (2006), the levels in the Caspian Sea are estimated to drop by around 9 m by the end of the 21st century, due largely to increases in evaporation. Levels in some lakes represent a changing balance between inputs and outputs and, under one transient scenario, levels in Lake Victoria would initially fall as increases in evaporation offset changes in precipitation, but subsequently rise as the effects of increased precipitation overtake the effects of higher evaporation (Tate et al., 2004).
Increasing winter temperature considerably changes the ice regime of water bodies in northern regions. Studies made at the State Hydrological Institute, Russia, comparing the horizon of 2010 to 2015 with the control period 1950 to 1979, show that ice cover duration on the rivers in Siberia would be shorter by 15 to 27 days and maximum ice cover would be thinner by 20 to 40% (Vuglinsky and Gronskaya, 2005).
Model studies show that land-use changes have a small effect on annual runoff as compared to climate change in the Rhine basin (Pfister et al., 2004), south-east Michigan (Barlage et al., 2002), Pennsylvania (Chang, 2003), and central Ethiopia (Legesse et al., 2003). In other areas, however, such as south-east Australia (Herron et al., 2002) and southern India (Wilk and Hughes, 2002), land-use and climate-change effects may be more similar. In the Australian example, climate change has the potential to exacerbate considerably the reductions in runoff caused by afforestation.
Carbon dioxide enrichment of the atmosphere has two potential competing implications for evapotranspiration, and hence water balance and runoff. First, higher CO2 concentrations can lead to reduced evaporation, as the stomata, through which evaporation from plants takes place, conduct less water. Second, higher CO2 concentrations can lead to increased plant growth and thus leaf area, and hence a greater total evapotranspiration from the area. The relative magnitudes of these two effects, however, vary between plant types and also depend on other influences such as the availability of nutrients and the effects of changes in temperature and water availability. Accounting for the effects of CO2 enrichment on runoff requires the incorporation of a dynamic vegetation model into a hydrological model. A small number of models now do this (Rosenberg et al., 2003; Gerten et al., 2004; Gordon and Famiglietti, 2004; Betts et al., 2007), but are usually at the GCM (and not catchment) scale. Although studies with equilibrium vegetation models suggest that increased leaf area may offset stomatal closure (Betts et al., 1997; Kergoat et al., 2002), studies with dynamic global vegetation models indicate that stomatal responses dominate the effects of leaf area increase. Taking into account CO2-induced changes in vegetation, global mean runoff under a 2*CO2 climate has been simulated to increase by approximately 5% as a result of reduced evapotranspiration due to CO2enrichment alone (‘physiological forcing’) (Betts et al., 2007; Leipprand and Gerten, 2006). This may be compared to (often much larger) changes at the river basin scale (Figures 3.3, 3.4, and 3.7), and global values of runoff change. For example, global mean runoff has been simulated to increase by 5%-17% due to climate change alone in an ensemble of 143 2*CO2 GCM simulations (Betts et al., 2006).