To determine the practical consequences of projections of global sea level rise in particular coastal regions, it is necessary to understand the various components leading to relative sea level changes. These components include local land movements, global eustatic sea level rise, any spatial variability from that global average, local meteorological changes and changes in the frequency of extreme events.
Titus and Narayanan (1996) propose a simple method for computing local projections of mean sea level rise given historical observations at a site and projections of global average sea level (such as Figure 11.12). To allow for local land movements and our current inability to model sea level change accurately (Section 11.4), they propose linearly extrapolating the historical record and adding to this a globally averaged projection. However, they point out that to avoid double counting, it is necessary to correct the global projection for the corresponding modelled trend of sea level rise during the period of the historical observations. Caution is required in applying this method directly to the projections of this chapter for several reasons. First, current model projections indicate substantial spatial variability in sea level rise. This variability has a standard deviation of up to 0.1 m by 2100; some locations experience a sea level rise of more than twice the global-average thermal expansion, while others may have a fall in sea level (Section 11.5.2; Table 11.15). Second, there are uncertainties in the accuracy of the trend from the historical record and in modelling of past sea level changes (Sections 188.8.131.52 and 11.4). Third, as well as changes in mean sea level there may be changes in the local meteorological regime resulting in modified storm surge statistics (Section 184.108.40.206). If the method of Titus and Narayanan is not applied, it is nonetheless important to recognise that in all models and scenarios the rate of local sea level rise in the 21st century is projected to be greater than in the 20th century at the great majority of coastal locations.
The probability of flood risk in coastal areas is generally expressed in terms of extreme sea level distributions. Such distributions are usually computed from observed annual maximum sea levels from several decades of tide gauge data, or from numerical models. While such distributions are readily available for many locations, a worldwide set has never been computed to common standards for studies of impacts of global sea level change.
Changes in the highest sea levels at a given locality could result mainly from two effects. First, if mean sea level rises, the present extreme levels will be attained more frequently, all else being equal. This may imply a significant increase in the area threatened with inundation (e.g., Hubbert and McInnes, 1999) and an increased risk within the existing flood plain. The effect can be estimated from a knowledge of the present day frequency of occurrence of extreme levels (e.g., Flather and Khandker, 1993; Lowe et al., 2001; Figure 11.14).
Second, changes in storm surge heights would result from alterations to the occurrence of strong winds and low pressures. At low-latitude locations, such as the Bay of Bengal, northern Australia and the southern USA, tropical cyclones are the primary cause of storm surges. Changes in frequency and intensity of tropical cyclones could result from alterations to sea surface temperature, large-scale atmospheric circulation and the characteristics of ENSO (Pittock et al., 1996) but no consensus has yet emerged (see Box 10.2). In other places, such as southern Australia and north-west Europe, storm surges are associated with mid-latitude low-pressure systems. For instance, Hubbert and McInnes (1999) showed that increasing the wind speeds in historical storm surge events associated with the passage of cold fronts could lead to greater flooding in Port Phillip Bay, Victoria, Australia. Changes in extratropical storms also cannot be predicted with confidence (Section 220.127.116.11).
Several studies have attempted to quantify the consequences of changes in storm climatology for the north-west European continental shelf using regional models of the atmosphere and ocean. Using five-year integrations of the ECHAM T106 model for present and doubled CO2, Von Storch and Reichardt (1997) and Flather and Smith (1998) did not find any significant changes in extreme events compared with the variability of the control climate (see also WASA Group, 1998). However, Langenberg et al. (1999) reported increases of 0.05 to 0.10 m in five-winter-mean high-water levels around all North Sea coasts, judged to be significant compared with observed natural variability. Lowe et al. (2001) undertook a similar study using multi-decadal integrations of the Hadley Centre regional climate model for the present climate and the end of the 21st century (Figure 11.14), finding statistically significant changes of up to 0.2 m in five-year extremes in the English Channel. Differences between these various results relate to the length of model integration and to systematic uncertainty in the modelling of both the atmospheric forcing and the ocean response.
Changes in wind forcing could result in changes to wave heights, but with the short integrations available, the WASA Group (Rider et al., 1996) were not able to identify any significant changes for the North Atlantic and North Sea for a doubling of CO2. Günther et al. (1998) noted that changes in future wave climate were similar to patterns of past variation.
Other reports in this collection