Introduction

1Since the late 19^th century, global mean sea level has risen by 15 to 20 cm due to ocean warming and glacier melting caused by the anthropogenic climate change. While future sea level projections remain uncertain, the sea level will continue to rise during the 21^st century and beyond. However, with unabated greenhouse gas emissions, the rate of about 3 mm/year observed during the last 20 years will probably accelerate and reach values comparable to the last marine transgression that took place from 21,000 to 6,000 years before present (B.P.). This is a major concern for the future of the majority of low lying coastal areas in the world, especially if extensive melting of the Greenland and Antarctica ice sheets leads to a rise of several meters in global sea levels over the coming centuries (IPCC, WG1, Ch13, 2013; Clark et al. 2016).

2In this general context, the case of the Mediterranean Sea deserves special attention. First, exposure to coastal hazards is a real concern, as illustrated by historical events. Second, this region is particularly vulnerable to climate change: comparisons with other coastal regions of the world have shown that Mediterranean cities, wetlands and dry lands are expected to undergo the most damage according to sea level rise projections for the 21^st century.

3Authors of current impact studies recognize that they are limited by the sparse knowledge regarding sea level changes in this semi-enclosed basin. Indeed, the Mediterranean Sea is not only connected to the global ocean through the Strait of Gibraltar, implying that sea level is controlled by local and remote processes, but it is also affected by vertical ground motions due to active tectonic and volcanic processes or caused by changing water volumes in subsurface sedimentary layers (e.g., due to groundwater extraction).

4This raises the following questions: to what extent could the Mediterranean Sea differ from the global mean? What impacts are to be expected under plausible sea level change projections? Finally, what are the implications of these findings for adaptation and the mitigation of climate change? This article examines these three questions in turn.

Sea level rise in the Mediterranean: why is it so complex?

Present day sea level changes and the role of vertical ground motions

5Sea level observations using satellite altimetry showed an increase in absolute (geocentric) sea level of 2.6±0.2 mm/yr on average across the Mediterranean Sea during the period 1993-2015 (Figure 1). Different features can be observed within the basin (Figure 1a), with linear trends ranging from-4 to 6 mm/yr, although these are associated with ocean circulation variability rather than long term and persistent structures. For the pre-altimetry period, i.e. prior to 1992, sea level observations are provided by tide gauges. The Mediterranean basin is monitored by a large number of tide gauges, and some records span several decades (see Figure 1b for some examples); however, with a very few exceptions, these gauges are concentrated along the European coasts. The longest tide gauge records in the Mediterranean Sea (Marseille, Genova and Trieste) report a rate of around 1.2±0.1 mm/yr in sea level rise in the 20^th century. In contrast to satellite altimetry, tide gauges measure sea level changes relative to the land on which they are grounded. This means that the signal they measure includes vertical land motion in addition to the climate related signal from the ocean (Wöppelmann and Marcos, 2016). The most recent GPS solution (ULR6) produced by the University of La Rochelle Consortium, shows linear rates of vertical land motion varying between 1 and -4 mm/yr (fig. 1a). As in the case of the tide gauges, either continuous or episodic GPS observations are mostly made along the northern coasts of the Mediterranean Sea.

6Rates of vertical land motion can be comparable in magnitude to climate related sea level changes (Wöppelmann and Marcos, 2016), thus mitigating (in the case of uplift) or exacerbating (in the case of subsidence) the impacts of sea level rise such as future flooding risks. A paradigmatic case is the city of Venice (Figure 1b) with a rate of 2.5 mm/yr in sea level rise in the 20^th century, of which about 1.2 mm/yr is due to local subsidence. The city of Alexandria was also considered to be one of the most exposed to coastal flooding due its location on the Nile Delta (Hanson et al. 2011). However, it was subsequently demonstrated that Alexandria is only subject to moderate subsidence (Wöppelmann et al. 2013), implying very different evaluation of the future coastal risks. Therefore, accurate knowledge of rates of land motion is of paramount importance in the assessment of potential damage due to increased exposure to sea level rise.

Future projections of mean sea level in the Mediterranean

7The variability of the Mediterranean Sea level is caused by different mechanisms. Local processes such as changes in the circulation, in the local thermal expansion or in the atmospheric mechanical forcing are combined with remote processes represented as the changes in the northeast (NE) Atlantic sea level. The answer to the question of which mechanism dominates depends on the time scales considered. While atmospheric mechanical forcing dominates at intraseasonal scales (Jordà et al., 2012a), at multidecadal timescales, variations in sea level in the northeast Atlantic are the dominant mechanism. Therefore, in order to obtain reliable projections of Mediterranean Sea level, the numerical models need to consider both local and remote forcings. Realistic simulation of the level of the Mediterranean Sea can be obtained by decomposing the signal into two components: relative variations in the Mediterranean Sea level with respect to the nearby NE Atlantic, and variations in the NE Atlantic sea level. This is an important consideration given that most regional climate models of the Mediterranean Sea do not include variations in the Atlantic sea level in their configurations, while the coarse resolution of global models prevents them from correctly reproducing processes that take place within the basin. For these reasons, here we generate the projections of Mediterranean Sea level combining the results of global climate models (GCMs) for the variations in the NE Atlantic and the results of regional climate models (RCMs) for the relative changes between the sea level of the Mediterranean Sea and of the NE Atlantic.

8The projections of NE Atlantic sea level were obtained using the CMIP3 and CMIP5 simulations for the moderate scenarios SRES-A1b and RCP6.0 and the high emission scenarios SRES-A2 and RCP8.5. The GCM simulations take into account projected changes in thermal expansion and in the circulation but not the contribution of terrestrial ice melt. For the latter component, we used the results of Spada et al. (2012) who solved the sea level equation to investigate the pattern of the gravitationally self-consistent sea level variations corresponding to future scenarios of terrestrial ice melt. Finally, the projected changes in the relative differences between Mediterranean Sea level and NE Atlantic sea level were obtained from an ensemble of regional models. In particular, we used RCMs to simulate the changes related to circulation and local thermal expansion and storm surge models to simulate the changes related to atmospheric mechanical forcing.

9The projections of changes in the NE Atlantic sea level caused by thermal expansion and changes in the circulation (fig. 2a) show an increase that ranges from 16 cm to 57 cm depending on the scenario. The projected mean values are 24, 32, 22 and 41 cm in the SRES-A1b, RCP6.0, SRES-A2 and RCP8.5 scenarios, respectively. These projected changes are a combination of the global thermal expansion (from 10 to 30 cm) with projected changes in the circulation (from -3 to + 37 cm). The latter refers to the projected regional redistribution of water due to changes in the distribution of wind and water masses. It is interesting to note that the latter factor is expected to contribute significantly to the sea level in the NE Atlantic and hence to future changes in the Mediterranean Sea level. If we consider the relative differences between the Mediterranean and the NE Atlantic, we find that all the models show small changes, irrespective of the scenario concerned. The expected evolution in local thermal expansion and circulation only implies a small decrease ranging from -3 to 0 cm with respect to the NE Atlantic. The atmospheric mechanical forcing would also have only a small impact (between -2 and 2 cm with respect to the NE Atlantic). Finally, we can estimate the evolution of Mediterranean Sea level taking into account all the above mentioned factors as well as the terrestrial ice melt (fig. 2c). The results show that the basin average Mediterranean Sea level will increase by between 44 cm and 102 cm by the end of the 21^st century. The mean values are 60, 68, 57 and 76 cm in the SRES-A1b, RCP6.0, SRES-A2 and RCP8.5 scenarios, respectively. In sum, the Mediterranean will basically mirror the evolution of the NE Atlantic, which, in turn, is expected to reach higher levels than the global mean sea level because of the expected changes in the circulation. It is worth noting that these projections refer to the basin average sea level. The regional models show a complex pattern of changes within the Mediterranean including local changes that can differ by + 10 cm from the above mentioned values. However, the results of the different RCMs are not consistent and the ensemble is too small to draw robust conclusions. Therefore, we recommend taking into account that, at local scale, the projected values can differ by up to + 10 cm with respect to the basin average values.

Impacts of sea level rise on shorelines and on coastal flooding in the Mediterranean

10During storms, several physical phenomena cause to water to rise above the predicted levels. In addition to the effects of direct atmospheric forcing (lower pressure and strong winds), the increase in mean water level may be due to the presence of breaking waves, called the wave setup. If the sea level at the coast exceeds the height of coastal defenses or dunes, the result is coastal flooding generated by overflow. Otherwise, the instantaneous effect of each individual wave can still cause overtopping and flooding.

11With the projected increase in sea levels, extreme surge events associated with atmospheric storminess become potentially more hazardous threats to the coastal environment. Although the intensity and frequency of extreme sea levels vary independently of mean sea level, these changes are small compared with rates of mean sea level change. Assessment of projected changes in storm surges in the Mediterranean Sea has revealed that expected variations are towards less frequent/more intense extreme events, possibly associated with a northward shift of the Atlantic storm tracks. However, the projected changes in storminess are relatively modest and would not compensate for the expected increase in the mean sea level.

12Box 1 illustrates all these phenomena with historical storms in Languedoc Roussillon (region on the French Mediterranean coast). For example, in situ and modeling data show that during the 1997 storm, atmospheric forcing caused a sea level rise of 0.4 m with an additional 0.4 m caused by the wave setup in the harbor of Sète. In such cases, the values reached by the wave setup are far from negligible, and depend to a great extent on the slopes of the subseafloor in shallow waters. Hence, extreme water levels will differ from site to site, depending on the local topography, bathymetry and defense systems built. This illustrates the need to take the local context and how it changes over time into account for accurate coastal impact assessments.

13Despite the need to consider the local context, a general conclusion has been drawn from all the studies that examined future coastal flooding under the assumption that sea level follows global or regional IPCC projections (Hallegatte et al. 2013): without adaptation, coastal flooding will not only become more intense (as shown in Box 1), but also more frequent, especially in the Mediterranean region. The latter issue is illustrated in Figure 3 using data from Hallegatte et al. (2013) in the case of Alexandria (Egypt), assuming that the sea level follows IPCC projections, and that vertical ground motions contribute less to relative sea level changes.

14The expected increase in the frequency and intensity of storm surges has raised concerns regarding the possible increase in damage in the future. In large cities, both risks and adaptation capabilities are large, so investments needed to ensure the same level of protection are indispensable (Hallegatte et al. 2013; Hunter et al. 2013). However, the Mediterranean coasts are also characterized by many rural, peri-urban and tourist resorts and small harbors, where funding and local knowledge might not be sufficient to address the issue of coastal adaptation.

Shoreline retreat

15Coastal erosion already affects a large part of the Mediterranean shores. The coastal database Eurosion (www.eurosion.org) contains information on changes in the shorelines of European member states collected from 1980 to 2000. Along the Mediterranean and Black Sea coasts, this dataset shows that for the 6,750 km of sandy beaches for which information exists, 3,400 km were already eroding during that period, while only 580 km were accreting. This raises significant management issues not only on the northern shore of the Mediterranean, but also on the southern shores, as shown by the growing number of studies addressing the issue of shoreline retreat (see section 2.2.2).

16While it is widely acknowledged that future sea level rise will exacerbate coastal erosion, current assessments of future shoreline changes suffer from large uncertainties. Obviously, a rapid extensive sea level rise (typically 2 to 4 m/century) would result in the permanent flooding of the lowest coastal areas. However, as long as sea level continues to rise a few millimeters a year, a dynamic response of the shoreline is to be expected. Figure 4 shows evaluations of future shoreline change rates for typical Mediterranean beaches in the Gulf of Lion, France, which are currently severely affected by erosion.. These simulations provide very different perspectives depending on future greenhouse gas emissions: on the one hand with reduced greenhouse gas emissions (RCP 2.6), shoreline change rates would be little affected by sea-level rise; on the other hand, unabated emissions would result in a significant acceleration of shoreline retreat. These results are based on assumptions that are debated in the scientific literature. However, they support the fact that an acceleration of sea-level rise could result in notable losses of beaches and coastal wetlands on both southern and northern Mediterranean shores.

Conclusion

17While adaptation to current environmental conditions is being increasingly addressed in current risk prevention policies, the possible impacts of future sea level rise are the subject of fewer studies. It is however certain that, particularly in the Mediterranean region, future adaptation will be especially difficult if sea level rise reaches or exceeds rates of 1 cm/year, as predicted by GHG emission scenarios. Therefore, public policies addressing the future challenges of Mediterranean coastal hazards should not only focus on adaptation, but also mitigate climate change by reducing concentrations of GHG in the atmosphere.

Acknowledgements

18We thank IMEDEA and the BRGM for supporting this research and Stephane Hallegatte for providing extreme sea level data.