Sea level is projected to rise between the present (1980–1999) and the end of this century (2090–2099) by 0.35 m (0.23 to 0.47 m) for the A1B scenario (see Section 10.6). Due to ocean density and circulation changes, the distribution will not be uniform and Figure 10.32 shows a distribution of local sea level change based on ensemble mean of 14 AOGCMs. A lower-than-average rise in the Southern Ocean can be seen, possibly due to increased wind stress. Also obvious is a narrow band of pronounced sea level rise stretching across the southern Atlantic and Indian Oceans at about 40°S. This is also seen in the southern Pacific at about 30°S. However, large deviations among models make estimates of distribution across the Caribbean, Indian and Pacific Oceans uncertain. Extreme sea level changes, including storm surges, are discussed in Box 11.5 in a broader context. The range of uncertainty cannot be reliably quantified due to the limited set of models addressing the problem.
Box 11.5: Coastal Zone Climate Change
Introduction
Climate change has the potential to interact with the coastal zone in a number of ways including inundation, erosion and salt water intrusion into the water table. Inundation and intrusion will clearly be affected by the relatively slow increases in mean sea level over the next century and beyond. Mean sea level is addressed in Chapter 10; this box concentrates on changes in extreme sea level that have the potential to significantly affect the coastal zone. There is insufficient information on changes in waves or near-coastal currents to provide an assessment of the effects of climate change on erosion.
The characteristics of extreme sea level events are dependent on the atmospheric storm intensity and movement and coastal geometry. In many locations, the risk of extreme sea levels is poorly defined under current climate conditions because of sparse tide gauge networks and relatively short temporal records. This gives a poor baseline for assessing future changes and detecting changes in observed records. Using results from 141 sites worldwide for the last four decades, Woodworth and Blackman (2004) find that at some locations extreme sea levels have increased and that the relative contribution from changes in mean sea level and atmospheric storminess depends on location.
Methods of simulating extreme sea levels
Climate-driven changes in extreme sea level will come about because of the increases in mean sea level and changes in the track, frequency or intensity of atmospheric storms. (From the perspective of coastal flooding, the vertical movement of land, for instance due to post glacial rebound, is also important when considering the contribution from mean sea level change.) To provide the large-scale context for these changes, global climate models are required, although their resolution (typically 150 to 300 km horizontally) is too coarse to represent the details of tropical cyclones or even the extreme winds associated with mid-latitude cyclones. However, some studies have used global climate model forcing to drive storm surge models in order to provide estimates of changes in extreme sea level (e.g., Flather and Williams, 2000). To obtain more realistic simulations from the large-scale drivers, three approaches are used: dynamical and statistical downscaling and a stochastic method (see Section 11.10 for general details).
As few RCMs currently have an ocean component, these are used to provide high-resolution (typically 25 to 50 km horizontally) surface winds and pressure to drive a storm surge model (e.g., Lowe et al., 2001). This sequence of one-way coupled models is usually carried out for a present-day (Debenard et al., 2003) or historical baseline (e.g., Flather et al., 1998) and a period in the future (e.g., Lowe et al., 2001; Debenard et al., 2003). In the statistical approach, relationships between large-scale synoptic conditions and local extreme sea levels are constructed. These relationships can be developed either by using analyses from weather prediction models and observed extreme sea levels, or by using global climate models and present-day simulations of extreme water level generated using the dynamic methods described above. Simulations of future extreme sea level are then derived from applying the statistical relationships to the future large-scale atmospheric synoptic conditions simulated by a global climate model (e.g., von Storch and Reichardt, 1997). The statistical and dynamical approach can be combined, using a statistical model to produce the high-resolution wind fields forcing the wave and storm surge dynamical models (Lionello et al., 2003). Similarly, the stochastic sampling method identifies the key characteristics of synoptic weather events responsible for extreme sea levels (intensity and movement) and represents these by frequency distributions. For each event, simple models are used to generate the surface wind and pressure fields and these are applied to the storm surge model (e.g., Hubbert and McInnes, 1999). Modifications to the frequency distributions of the weather events to represent changes under enhanced greenhouse conditions are derived from global climate models and then used to infer a future storm surge climatology.
Extreme sea level changes – sample projections from three regions
1. Australia
In a study of storm surge impacts in northern Australia, a region with only a few short sea level records, McInnes et al. (2005) used stochastic sampling and dynamical modelling to investigate the implications of climate change on extreme storm surges and inundation. Cyclones occurring in the Cairns region from 1907 to 1997 were used to develop probability distribution functions governing the cyclone characteristics of speed and direction with an extreme value distribution fitted to the cyclone intensity. Cyclone intensity distribution was then modified for enhanced greenhouse conditions based on Walsh and Ryan (2000), in which cyclones off northeast Australia were found to increase in intensity by about 10%. No changes were imposed upon cyclone frequency or direction since no reliable information is available on the future behaviour of the main influences on these, respectively ENSO or mid-level winds. Analysis of the surges resulting from 1,000 randomly selected cyclones with current and future intensities shows that the increased intensity leads to an increase in the height of the 1-in-100 year event from 2.6 m to 2.9 m with the 1-in-100 year event becoming the 1-in-70 year event. This also results in the areal extent of inundation more than doubling (from approximately 32 to 71 km2). Similar increases for Cairns and other coastal locations were found by Hardy et al. (2004). (continued)
2. Europe
Several dynamically downscaled projections of climate-driven changes in extreme water levels in the European shelf region have been carried out. Woth (2005) explored the effect of two different GCMs and their projected climates changes due to two different emissions scenarios (SRES A2 and B2) on storm surges along the North Sea coast. She used data from one RCM downscaling the four GCM simulations (Woth et al., 2006) (using data from four RCMs driven by one GCM produced indistinguishable results) and demonstrates significant increases in the top 1% of events (10 to 20 cm above average sea level change) over the continental European North Sea coast. The changes projected by the different experiments were statistically indistinguishable, although those from the models incorporating the A2 emissions scenario were consistently larger. When including the effects of global mean sea level rise and vertical land movements, Lowe and Gregory (2005) find that increases in extreme sea level are projected for the entire UK coastline, using a storm surge model driven by one of the RCMs analysed by Woth et al. (2006) (Box 11.5, Figure 1). Using a Baltic Sea model driven by data from four RCM simulations, Meier (2006) finds that the changes in storm surges vary strongly between the simulations but with some tendency for larger increases in the 100-year surges than in the mean sea level.
Lionello et al. (2003) estimate the effect of atmospheric CO2 doubling on the frequency and intensity of high wind waves and storm surge events in the Adriatic Sea. The regional surface wind fields were derived from the sea level pressure field in a 30-year long ECHAM4 high-resolution (about 1.5 degrees) time slice experiment by statistical downscaling and then used to force a wave and an ocean model. They find no statistically significant changes in the extreme surge level and a decrease in the extreme wave height with increased atmospheric CO2. An underestimation of the observed wave heights and surge levels calls for caution in the interpretation of these results. Using AOGCM projections, X.L. Wang et al. (2004) infer an increase in winter and autumn seasonal mean and extreme wave heights in the northeast and southwest North Atlantic, but a decrease in the mid-latitudes of the North Atlantic. Not all changes were significant and in some regions (e.g., the North Sea), their sign was found to depend on the emissions scenario.
3. Bay of Bengal
Several dynamic simulations of storm surges have been carried out for the region but these have often involved using results from a small set of historical storms with simple adjustments (such as adding on a mean sea level or increasing wind speeds by 10%) to account for future climate change (e.g., Flather and Khandker, 1993). This technique has the disadvantage that by taking a relatively small and potentially biased set of storms it may lead to a biased distribution of water levels with an unrealistic count of extreme events. In one study using dynamical models driven by RCM simulations of current and future climates, Unnikrishnan et al. (2006) show that despite no significant change in the frequency of cyclones there are large increases in the frequency of the highest storm surges.
Uncertainty
Changes in storm surges and wave heights have been addressed for only a limited set of models. Thus, we cannot reliably quantify the range of uncertainty in estimates of future coastal flooding and can only make crude estimates of the minimum values (Lowe and Gregory, 2005). There is some evidence that the dynamical downscaling step in providing data for storm surge modelling is robust (i.e., does not add to the uncertainty). However, the general low level of confidence in projected circulation changes from AOGCMs implies a substantial uncertainty in these projections.