3.3.1 Climatic drivers
Projections for the future
The most dominant climatic drivers for water availability are precipitation, temperature, and evaporative demand (determined by net radiation at ground level, atmospheric humidity, wind speed, and temperature). Temperature is particularly important in snow-dominated basins and in coastal areas (due to the impact of temperature on sea level).
The following summary of future climate change is taken from the Working Group I Fourth Assessment Report (WGI AR4), Chapter 10 (Meehl et al., 2007). The most likely global average surface temperature increase by the 2020s is around 1°C relative to the pre-industrial period, based on all the IPCC Special Report on Emissions Scenarios (SRES; Naki?enovi? and Swart, 2000) scenarios. By the end of the 21st century, the most likely increases are 3 to 4°C for the A2 emissions scenario and around 2°C for B1 (Figure 10.8). Geographical patterns of projected warming show the greatest temperature increases at high northern latitudes and over land (roughly twice the global average temperature increase) (Chapter 10, Executive summary, see also Figure 10.9). Temperature increases are projected to be stronger in summer than in winter except for Arctic latitudes (Figure 10.9). Evaporative demand is likely to increase almost everywhere (Figures 10.9 and 10.12). Global mean sea-level rise is expected to reach between 14 and 44 cm within this century (Chapter 10, Executive summary). Globally, mean precipitation will increase due to climate change. Current climate models tend to project increasing precipitation at high latitudes and in the tropics (e.g., the south-east monsoon region and over the tropical Pacific) and decreasing precipitation in the sub-tropics (e.g., over much of North Africa and the northern Sahara) (Figure 10.9).
While temperatures are expected to increase during all seasons of the year, although with different increments, precipitation may increase in one season and decrease in another. A robust finding is that precipitation variability will increase in the future (Trenberth et al., 2003). Recent studies of changes in precipitation extremes in Europe (Giorgi et al., 2004; Räisänen et al., 2004) agree that the intensity of daily precipitation events will predominantly increase, also over many areas where means are likely to decrease (Christensen and Christensen, 2003, Kundzewicz et al., 2006). The number of wet days in Europe is projected to decrease (Giorgi et al., 2004), which leads to longer dry periods except in the winters of western and central Europe. An increase in the number of days with intense precipitation has been projected across most of Europe, except for the south (Kundzewicz et al., 2006). Multi-model simulations with nine global climate models for the SRES A1B, A2, and B1 scenarios show precipitation intensity (defined as annual precipitation divided by number of wet days) increasing strongly for A1B and A2, and slightly less strongly for B1, while the annual maximum number of consecutive dry days is expected to increase for A1B and A2 only (WGI AR4, Figure 10.18).
Uncertainties in climate change projections increase with the length of the time horizon. In the near term (e.g., the 2020s), climate model uncertainties play the most important role; while over longer time horizons, uncertainties due to the selection of emissions scenario become increasingly significant (Jenkins and Lowe, 2003).
General Circulation Models (GCMs) are powerful tools accounting for the complex set of processes which will produce future climate change (Karl and Trenberth, 2003). However, GCM projections are currently subject to significant uncertainties in the modelling process (Mearns et al., 2001; Allen and Ingram, 2002; Forest et al., 2002; Stott and Kettleborough, 2002), so that climate projections are not easy to incorporate into hydrological impact studies (Allen and Ingram, 2002). The Coupled Model Intercomparison Project analysed outputs of eighteen GCMs (Covey et al., 2003). Whereas most GCMs had difficulty producing precipitation simulations consistent with observations, the temperature simulations generally agreed well. Such uncertainties produce biases in the simulation of river flows when using direct GCM outputs representative of the current time horizon (Prudhomme, 2006).
For the same emissions scenario, different GCMs produce different geographical patterns of change, particularly with respect to precipitation, which is the most important driver for freshwater resources. As shown by Meehl et al. (2007), the agreement with respect to projected changes of temperature is much higher than with respect to changes in precipitation (WGI AR4, Chapter 10, Figure 10.9). For precipitation changes by the end of the 21st century, the multi-model ensemble mean exceeds the inter-model standard deviation only at high latitudes. Over several regions, models disagree in the sign of the precipitation change (Murphy et al., 2004). To reduce uncertainties, the use of numerous runs from different GCMs with varying model parameters i.e., multi-ensemble runs (see Murphy et al., 2004), or thousands of runs from a single GCM (as from the climateprediction.net experiment; see Stainforth et al., 2005), is often recommended. This allows the construction of conditional probability scenarios of future changes (e.g., Palmer and Räisänen, 2002; Murphy et al., 2004). However, such large ensembles are difficult to use in practice when undertaking an impact study on freshwater resources. Thus, ensemble means are often used instead, despite the failure of such scenarios to accurately reproduce the range of simulated regional changes, particularly for sea-level pressure and precipitation (Murphy et al., 2004). An alternative is to consider a few outputs from several GCMs (e.g. Arnell (2004b) at the global scale, and Jasper et al. (2004) at the river basin scale).
Uncertainties in climate change impacts on water resources are mainly due to the uncertainty in precipitation inputs and less due to the uncertainties in greenhouse gas emissions (Döll et al., 2003; Arnell, 2004b), in climate sensitivities (Prudhomme et al., 2003), or in hydrological models themselves (Kaspar, 2003). The comparison of different sources of uncertainty in flood statistics in two UK catchments (Kay et al., 2006a) led to the conclusion that GCM structure is the largest source of uncertainty, next are the emissions scenarios, and finally hydrological modelling. Similar conclusions were drawn by Prudhomme and Davies (2007) regarding mean monthly flows and low flow statistics in Britain.
Incorporation of changing climatic drivers in freshwater impact studies
Most climate change impact studies for freshwater consider only changes in precipitation and temperature, based on changes in the averages of long-term monthly values, e.g., as available from the IPCC Data Distribution Centre (www.ipcc-data.org). In many impact studies, time series of observed climate values are adjusted with the computed change in climate variables to obtain scenarios that are consistent with present-day conditions. These adjustments aim to minimise the error in GCMs under the assumption that the biases in climate modelling are of similar magnitude for current and future time horizons. This is particularly important for precipitation projections, where differences between the observed values and those computed by climate models for the present day are substantial. Model outputs can be biased, and changes in runoff can be underestimated (e.g., Arnell et al. (2003) in Africa and Prudhomme (2006) in Britain). Changes in interannual or daily variability of climate variables are often not taken into account in hydrological impact studies. This leads to an underestimation of future floods, droughts, and irrigation water requirements.
Another problem in the use of GCM outputs is the mismatch of spatial grid scales between GCMs (typically a few hundred kilometres) and hydrological processes. Moreover, the resolution of global models precludes their simulation of realistic circulation patterns that lead to extreme events (Christensen and Christensen, 2003; Jones et al., 2004). To overcome these problems, techniques that downscale GCM outputs to a finer spatial (and temporal) resolution have been developed (Giorgi et al., 2001). These are: dynamical downscaling techniques, based on physical/dynamical links between the climate at large and at smaller scales (e.g., high resolution Regional Climate Models; RCMs) and statistical downscaling methods using empirical relationships between large-scale atmospheric variables and observed daily local weather variables. The main assumption in statistical downscaling is that the statistical relationships identified for the current climate will remain valid under changes in future conditions. Downscaling techniques may allow modellers to incorporate future changes in daily variability (e.g., Diaz-Nieto and Wilby, 2005) and to apply a probabilistic framework to produce information on future river flows for water resource planning (Wilby and Harris, 2006). These approaches help to quantify the relative significance of different sources of uncertainty affecting water resource projections.