Vol50,6,2007 741 ANNALS OF GEOPHYSICS, VOL. 50, N. 6, December 2007 fied the geological environment of the Mediter- ranean. Given the low tidal excursion and the presence of geological indicators and archaeo- logical remains, the Mediterranean coasts are particularly suitable for the reconstruction of past sea-levels. A considerable number of studies on the Rel- ative Sea-Level (RSL) variations in this region have been published, based on regional analyses aiming at the interpretation of tectonic, eustatic, and glacio-hydro-isostatic contributions to sea- level change. For reviews of regional Holocene observations, the reader is referred to Flemming (1972) and Pirazzoli (1991, 1996), while quanti- tative interpretations based upon models of glacio-isostatic adjustment have been recently proposed by Lambeck and Purcell (2005) and Pi- razzoli (2005). Due to the relevance held by sea- level variations in the context of the present-day Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets Paolo Stocchi (1) (2) and Giorgio Spada (1) (1) Istituto di Fisica, Università degli Studi di Urbino «Carlo Bo», Urbino, Italy (2) Delft University of Technology, Delft, The Netherlands Abstract Solving the sea-level equation for a spherically symmetric Earth we study the relative sea-level curves in the Mediterranean Sea in terms of Clark’s zones and we explore their sensitivity to the time-history of Late-Pleis- tocene ice aggregates. Since the Mediterranean is an intermediate field region with respect to the former ice sheets, glacio- and hydro-isostasy both contribute to sea-level variations throughout the Holocene. In the bulk of the basin, subsidence of the sea floor results in a monotonous sea-level rise, whereas along continental margins water loading produces the effect of «continental levering», which locally originates marked highstands followed by a sea-level fall. To describe such peculiar pattern of relative sea-level in this and other mid-latitude closed basins we introduce a new Clark’s zone (namely, Clark’s zone VII). Using a suite of publicly available ice sheet chronologies, we identify for the first time a distinct sensitivity of predictions to the Antarctic ice sheet. In par- ticular, we show that the history of mid to Late Holocene sea-level variations along the coasts of SE Tunisia may mainly reflect the melting of Antarctica, by a consequence of a mutual cancellation of the effects from the North- ern Hemisphere ice-sheets at this specific site. Ice models incorporating a delayed melting of Antarctica may ac- count for the observations across the Mediterranean, but fail to reproduce the SE Tunisia highstand. Mailing address: Dr. Giorgio Spada, Istituto di Fisica, Università degli Studi di Urbino «Carlo Bo», Via S. Chiara 27, 61029 Urbino (PU), Italy; e-mail: giorgio.spada@gmail.com Key words sea-level variations – glacial isostasy – Mediterranean Sea 1. Introduction The Mediterranean sea-level variations re- sult from complex geodynamical, geological, and metereological processes that span a wide range of time-scales. It is nowadays accepted that the melting of the Holocene remote ice sheets and the subsequent global glacio- and hydro-isostatic readjustment profoundly modi- 742 Paolo Stocchi and Giorgio Spada climate change, new geological investigations have been recently complemented by geomor- phological and archaeological evidence on a re- gional scale (Lambeck et al., 2004a). The latter have allowed us to better constrain the age of the paleo-sea-level indicators available in the Tyrrhenian and to address the influence of tec- tonic motions upon the observed uplift. While sea-level signals caused by tectonic forces and other local mechanisms may exhibit a complex spatial and temporal variability (see e.g., Carminati and Di Donato, 1999; for a re- gional case study), those associated with glacial isostasy are characterized by a smooth, long- wavelength pattern that discloses various re- gions (or zones) sharing the same RSL signa- tures and named after Clark (Clark et al., 1978; Clark and Lingle, 1979). The mechanisms that determine the shape of the Clark’s zones can be identified solving the sea-level equation that de- scribes simultaneously the effects of ice and melt water loads, the gravitational interactions between the solid Earth and the oceans, and the time-dependent Earth’s viscoelastic response (Farrell and Clark, 1976). Due to the limited computing resources, early determinations of the Clark’s zones failed to resolve important de- tails of their global pattern, including their con- tours and their possible fine structure in the Mediterranean Basin. The availability of high- resolution solutions of the sea-level equation (e.g., Lambeck, 1993, 1995) now allows us to scrutinize details of the shapes of the Clark’s zones; in particular, the recent high-resolution ap- proach of Mitrovica and Milne (2002) has shed new light on the mechanisms that determine their shape on a global scale. Using the ice sheet chronology ICE3G (Tushingham and Peltier, 1991) and assuming a moderate viscosity in- crease across the mantle, Mitrovica and Milne have shown that the bulk of the Mediterranean and of other mid-latitude basins is presently sub- ject to a sea-level rise of glacio-isostatic origin. Taking advantage of the recognized sensitivity of sea-level predictions in the Mediterranean to model parameters (e.g., Lambeck and Purcell, 2005), here we investigate the details of the Mediterranean Clark’s zones for various avail- able ice-sheets chronologies, focusing on the role of remote ice sheets. The chronology of the Antarctic ice sheet since the Last Glacial Max- imum (LGM) is still affected by large uncer- tainties. During the last two decades evidence supported by glaciological and glacio-isostatic adjustment modeling and ice core analysis have provided sometimes divergent estimates of its contribution to global sea-level (see e.g., Kauf- mann, 2002 and references therein). In the fol- lowing, we will consider several plausible mod- els for the Holocene time-history of Antarctica and we will evaluate the consequences on the RSL observations in the Mediterranean. This is motivated by the preliminary results of Stocchi et al. (2005b), who first noticed the sensitivity of RSL predictions for SE Tunisia and Gulf of Sirte to the time-history of remote ice sheets and particularly of Antarctica. After a presentation of the methods, we will discuss the shape of Clark’s zones in the Mediterranean for a suite of ice models and we will qualitatively compare predictions to obser- vations obtained from a public domain global database. To gain insight into the results pre- sented, we will separately consider the effects of glacio- and hydro-isostasy. In the results sec- tion, we will consider more recent and reliable RSL observations to address the role of remote ice sheets on firmer grounds. 2. Methods 2.1. Theory To define the contours of the Clark’s zones for the Mediterranean, we solve the sea-level equation (hereafter SLE), the linear integral equation that describes the space and time varia- tion of the geoid due to the mass exchange be- tween the hydro-and the cryo-sphere and to the delayed response of a visco-elastic Earth. Para- phrasing Farrell and Clark (1976), the SLE reads (2.1) where O is the ocean function (O=1 over the oceans and O= 0 elsewhere), Z = OS is sea- ( Z)ρ+ − ( ) O G G G G Z I I S Z i s s EUS s s ) ) ) ρ= − + + ω )7 A 743 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets The SLE is solved according the «pseudo- spectral» approach of Mitrovica and Peltier (1991), by publicly available numerical tools (Spada et al., 2004; Spada and Stocchi, 2006, 2007) based on the spatial discretization scheme introduced by Tegmark (2002). Here we neglect the effects of rotation on GIA-induced sea-level variations and we assume a constant ocean func- tion. However, for most regions the rigorous and accurate solution of the Australian National Uni- versity, which incorporates time-dependent shorelines and accounts for floating ice is recom- mended (Lambeck et al., 2003). In the study re- gion, the inclusion of a time-dependent ocean function may me particularly appropriate to model the sea-level rise in Northern Adriatic, as done by Lambeck et al. (2004a). In our computations, the viscosity profile of the mantle is that implied in the «standard» deglaciation chronology ICE3G of Tushingham and Peltier (1991, 1992), with a shallow upper mantle with viscosity ηSM = 1021 Pa·s, a transi- tion zone viscosity ηTZ = ηSM , and a lower mantle viscosity ηLM = 2 ×1021 Pa·s. Following Peltier (2004), we will refer to this viscosity profile to as VM1. The shear modulus and density pro- files are the same as in Cianetti et al. (2002). 2.2. Ice sheet models In the following we will consider three dis- tinct global ice sheets chronologies: ICE1 (Peltier and Andrews, 1976), ICE3G (Tushing- ham and Peltier, 1991), and ICE5G (Peltier, 2004). Frames (a), (b), and (c) of fig. 1a-f show ice thickness for these models at time t = 18 kyrs BP. Using ice sheets with increasing com- plexity and spatial resolutions will allow us to assess more robustly the sensitivity of the Me- diterranean data to the Antarctic ice sheet and to review some previous results in the litera- ture. Minor constituents of the global ice sheets distribution, such as the Alpine aggregate (Lambeck and Purcell, 2005; Stocchi et al., 2005a), will not be considered here. Frames (d), (e), and (f) of fig. 1a-f show Equivalent Sea-Level (ESL) as a function of time for the ice distributions considered in this study. Here ESL is defined as level change restricted to oceanic regions, S = = S(θ, λ, t) is sea-level change at colatitude θ, longitude λ and time t, I = I (θ, λ, t) is ice thick- ness variation, ρi and ρω are the density of ice and water, respectively, * denotes a convolution over time and over the whole surface of the Earth, Gs=Gs /γ where Gs is the (rheology- dependent) sea-level Green function, γ is mean surface gravity, and the over bar represents the average over the surface of the oceans. In eq. (2.1) SEUS represents the uniform sea-level change for a rigid, non-gravitating Earth (Gs≡ ≡ 0), with (2.2) where mICE and Ao represent the ice mass varia- tion and the area of the oceans, respectively. In the analyses of Section 3.3 it will be in- structive to consider separately ice- and ocean- load induced sea-level variations. Thus, follow- ing e.g., Mitrovica and Milne (2002), we write (2.3) where SICE, SEUS, and SOCE account for glacio-iso- stasy, eustasy, and hydro-isostasy, respectively, with (2.4) and (2.5) where Z represents the solution of eq. (2.1). Due to the integral nature of the SLE, howev- er, SOCE depends from SICE through Z, so that the decoupling given by eq. (2.3) is artificial (Mitrovica and Milne, 2002; Lambeck and Purcell, 2005) and is used here for illustrative purposes. Once the SLE has been solved, the RSL variations at coordinates (θ, λ) are obtained by (2.6) where tBP and tp are time before present and present time, respectively. ( , , ) ( , , ) ( , , )RSL t S t S tBP BP pθ λ θ λ θ λ= − ( )G GS Z ZOCE i s s) )ρ= − ( )G GS I IICE i s s) )ρ= − S S S SICE OCEEUS= ++ S A mEUS ICE oρ = − ω 744 Paolo Stocchi and Giorgio Spada (2.7) where V is melt water volume. Model ICE1 describes the retreat of the Holocene Northern Hemisphere ice sheets ac- cording to geological and geomorphological evidence, and stores ∼78 m of ESL at the LGM (see fig. 1a, solid). It is characterized by a rela- tively low spatial resolution, being specified on a grid of 5°×5° spherical quadrilaterals; the thickness of each element varies at steps of 2 kyrs. Details on how this chronology is incor- porated within the pseudo-spectral method is given e.g., by Spada and Stocchi (2007). ESL A Vi oρ ρ = ω Differently from ICE1, model ICE3G (fig. 1b) has been built to improve the fit to the global RSL data in the near field assuming the VM1 rhe- ological profile and a 120 km thick elastic litho- sphere (Tushingham and Peltier, 1991, 1992). While ICE1 assumes stationary Southern Hemi- sphere ice masses, ICE3G accounts for the melt- ing of Antarctica and South America, and conse- quently stores a larger water volume at the LGM, with an ESL of ∼113.5 m (see fig. 1c). ICE3G has an improved resolution compared to ICE1, being composed by a set of disc-shaped elements with half-amplitude of ∼1°, with thickness varying at steps of 1 kyr. Antarctica stores ∼27 m of ESL at the LGM and its volume is constant until 9 kyrs Fig. 1a-f. Global ice thickness distribution according to ICE1 (a), ICE3G (b) and ICE5G (VM1) (c) at 18 kyrs BP. This epoch corresponds to the LGM for ICE1 and ICE3G, while in ICE5G the LGM is reached 21 kyrs BP. The contribution to ESL of major constituents of the ice models are shown in the right frames as a function of time, where FEN, and NAM stand for Fennoscandia and Laurentide, respectively. With A3 and A5 we label the Antarctic components of ICE3G and ICE5G, respectively. a b c d e f 745 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets BP; its deglaciation occurs between 9 and 5 kyrs BP, when the Northern Hemisphere ice sheets had already lost most of their mass. To fit the constraints of the glaciological studies of Huybrechts (2002) and Denton and Hughes (2002), in the recent ICE5G model of Peltier (2004) the total ESL of Antarctica has been reduced (by ∼ 50%) with respect to ICE3G (ICE5G can be downloaded from http://www.sbl.statkart.no/projects/pgs/). Such reduction is accompanied by an increased ESL for the North American ice sheet to fit constraints from VLBI and gravity observations (Argus et al., 1999). The spatial resolution of ICE5G, whose ESL is shown in fig. 1f as a function of time, is improved relative to ICE35, being de- fined on a gaussian grid with a spacing of ∼ 0.7°. Differently from ICE3G, ICE5G describes the whole history of ice thickness during the last gla- cial phase (i.e., since 122.0 kyrs ago), and as- sumes that the LGM occurred 21 kyrs BP. Here ICE5G has been implemented assuming the VM1 viscosity profile described above. With the purpose of evaluating the sensitiv- ity of Mediterranean Clark zones to the time- history of remote Holocene ice aggregates, in the computations of Section 4.1 we will also consider alternative chronologies for the Ant- arctic ice sheet, which differ from those implic- it in ICE3G and ICE5G. 3. Clark zones Here we will first qualitatively characterize the RSL variations in the Mediterranean Sea studying the shape of the Clark zones and then we will compare results based on ICE1, ICE3G, and ICE5G with observations available from the database of Tushingham and Peltier (1993). The pattern of sea-level change in the Mediter- ranean will also be discussed by considering separately the ocean and the ice-load compo- nents SOCE and SICE (see eq. (2.3)), and evaluat- ing the role of individual regional components of the two ice sheet distributions. An in-depth analysis of the sensitivity of the Mediterranean RSL variations to the ice sheets chronology will be presented in the results section with the aid of further sea-level indicators. 3.1. Pattern of Clark zones for the Mediterranean Sea As shown in fig. 2a-e, the solution of the SLE discloses complex patterns of RSL change across the Mediterranean that strongly depend upon the assumptions about the chronology of the far-field ice sheets. Model ICE3G (frame a) implies a Late-Holocene highstand that marks the end of deglaciation (5 kyrs BP) along most of the Mediterranean coasts, with the exception of Southern Italy, Greece and part of the coasts of Algeria, Lybia and of the Southern Levant, while submergence is predicted in the bulk of the basin. The regions of emergence and submer- gence are separated by narrow transition zones in which the sea-level nearly follows the eustatic curve. The pattern of sea-level change shown in fig. 2a might be mistakenly interpreted as a match of Clark’s zones VI and II, which charac- terize the far-field continental shorelines as a consequence of «continental levering» and the collapsing forebulge regions, respectively (Wal- cott, 1972; Clark et al., 1978). However, as shown by Mitrovica and Milne (2002), Lambeck and Purcell (2005) and Lambeck et al. (2004b) (and independently confirmed by our computa- tions), submergence in the Central Mediter- ranean Sea mainly stems from hydro-isostasy, the direct glacio-isostatic effects of the Northern Hemisphere aggregates, characteristic of zone II, being confined to the coasts of France and Northern Italy. In general, Clark zone VI shows up as bands of offshore sea-level rise and onshore sea-level fall, with the size of the submerging ar- eas that show significant spatial variability with the tendency to increase for concave coast lines (this is clearly visible in the map of fig. 1a-f of Mitrovica and Milne 2002). When the shorelines close on themselves to define a relatively small basin, as in the Mediterranean, the area of sub- mergence tends to cover the central portion of the basin possibly leaving narrow regions of highstand on shore (this is also observed for the Black Sea and in other mid-latitude basins in the global map of Mitrovica and Milne). To charac- terize the peculiar RSL pattern predicted for mid-latitude basins, where zone VI is manifest as a central submergence region contoured by a narrow highstand zone, we propose the name of 746 Paolo Stocchi and Giorgio Spada Clark’s zone VII. The dominance of the influence of hydro-isostasy in the Mediterranean region when all the ice aggregates of ICE3G are consid- ered is apparent from inspection of fig. 1a-f of Mitrovica and Milne (2002), in which the pattern of ṠOCE is broadly similar that of the total signal Ṡ, where the dot indicates the time-derivative computed at present time. Such pattern has re- mained qualitatively stable in the last 5 kyrs, af- ter the end of deglaciation of major ice sheets (Mitrovica and Milne, 2002). When model ICE1 is considered (fig. 2b) the portion of zone VII characterized by high- stands is narrowed significantly, being only present along the coasts of Spain, Northern Mo- rocco, Tunisia and Lybia. Furthermore, the two transition regions of fig. 2a are not observed. Differently from in ICE3G, in ICE1 it is as- sumed that the Antarctic ice sheet is stationary during the Holocene, but the two models also differ for details of the melting chronologies of the North American and Fennoscandian ice sheets (see fig. 1a-f). To understand the origin of the distinct patterns in fig. 2a,b, fig. 2c we consider the ICE3G-A3 chronology, in which ICE3G is deprived of its Antarctic component. The pattern obtained is strikingly similar to that of ICE1 (b), indicating that the existence of Late- Holocene highstand in the Mediterranean mainly result from the melting of the Antarctic ice sheet implemented in model ICE3G. When Antarctica is built into model ICE1 (fig. 2d, model ICE1+ +A3), the configuration of Clark zones closely matches that of ICE3G (a). Significant difference between the results obtained for ICE3G and ICE1+A3, visible along the Northern Adriatic and Tyrrhenian coasts, can be attributed to differ- ences in the time-histories of the Northern Hemi- sphere components of ICE3G and ICE1, with the relatively contiguous Fennoscandian ice sheet that is likely to play a major role. In terms of Clark zones, for model ICE1+A3 zone II counter- acts the highstand of zone VII and merges with its core approximately North of the 42°N parallel. In Fig. 2a-e. Shape of Mediterranean Clark zones for models ICE3G (a), ICE1 (b), ICE3G-A3 (c), ICE1+A3 (d), and A3 (e). RSL variations expected within each zone, qualitatively shown in the legend, include a monotonous submergence, a Late-Holocene emergence with a highstand of a few meters, and possibly two narrow transition zones. In all computations, we have adopted rheological profile VM1. a c e b d 747 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets fig. 2e we only account for the effects of A3, the Antarctic component of ICE3G. In this case, zone VII disappears leaving an ubiquitous high- stand that, as we have verified, extends north of the 45°S parallel. Up to now, we have discussed the shape of Clark zones in the Mediterranean using quite dat- ed models for the global history of deglaciation. For instance, ICE1 is known to overestimate the extent and thickness of ice in the Baltic region (e.g., Nakada and Lambeck, 1988), while ICE3G does not match some well constrained RSL curves, such those relative to Southern France (see our discussion below). In addition, neither ICE1 or ICE3G satisfy the far-field sea-levels for the entire period since the Last Glacial Maximum (Nakada and Lambeck, 1988). In spite of their limitations, these two dated ice models have al- lowed us to appreciate how the shapes of the Clark zones evolve with increasing spatial resolu- tion of the surface load. However, the question of the validity of our results with respect to more re- cent models remains open. To address this point, we implemented ICE5G (Peltier, 2004) within our code, and repeated all the above computa- tions. The results, not shown here, indicate that Clark zones of ICE5G match those of ICE1+A3 very closely. A more detailed comparison of the outcomes of ICE5G with those of ICE3G is con- sidered, in terms of RSL curves, in the following section. 3.2. Synthetic RSL curves for the Mediterranean Figure 3a shows the location of Mediter- ranean sites pertaining to the publicly available Fig. 3a,b. a) Location of the RSL sites considered in the analysis of Section 3.2, namely Roussillon (1), Marseilles (2), Civitavecchia (3), Termoli (4), Catania (5), Messenia (6), SW Turkey (7), Cyprus (8), Beirut (9), Jaffa (10), Djer- ba (11), and Algiers (12). With the exception of Djerba (filled triangle), all the sites shown here belong to the RSL database of Tushingham and Peltier (1993). b) Full set of RSL observations available for the sites in (a) during the last 10 kyrs, where filled symbols denote observations from North Africa and the Levant Sea. a b 748 Paolo Stocchi and Giorgio Spada database (see ftp://ftp.ncdc.noaa.gov/pub/data/ paleo/paleocean/relative_sea_level/) of Tushin- ham and Peltier (1993) (henceafter TP), while RSL indicators are displayed in fig. 3b. Recent and more reliable observations, for which infor- mation on the exact meaning of geomorpholog- ical indicators used to derive sea-level curves can be more easily obtained from the literature, will be considered in Section 4.2 below. Fo- cussing on the TP observations at this stage is useful since they are available on a Mediter- ranean scale, and thus are useful for a global (although qualitative) comparison with predic- tions across the whole region. The TP sites shown in fig. 3a are quite even- ly distributed across the Mediterranean, but since only one indicator is available for North Africa (i.e., Algiers, site 12), here we also con- sider the site of Djerba (Tunisia), based on the work of Jedoui et al. (1998). According to fig. 3b, the only indications of a sea-level highstand in the Late-Holocene come from Beirut (9), Djerba (11) and Algiers (12). While RSL indi- cations from Beirut and Algiers are possibly in- fluenced by local tectonic deformations (Me- ghraoui et al., 2004), those from Tunisia main- ly reflect glacio-isostatic RSL variations (Mo- rhange and Pirazzoli, 2005). We observe that sites indicating an highstand are only situated along the southern and eastern continental Mediterranean coasts: this is qualitatively con- sistent with the pattern of Clark’s zones ob- tained for models ICE1, ICE1+A3, and ICE5G while it appears to be at variance with predic- tions based upon ICE3G (see fig. 2a-e). To assess the role of far field ice sheets, in fig. 4a-i we compare individual RSL indicators from the TP database with model calculations based on ICE3G (solid lines), ICE1 (dashed), ICE1+A3 (dotted), A3 (dash-dotted), and ICE5G (grey solid). Consistently with the qualitative study of fig. 2e, RSL variations driven by the melting of A3 show a clear highstand at 5 kyrs BP in all of the sites considered, with a peak am- plitude varying between ∼2 m in Roussillon and ∼1 m in Messenia. For models ICE3G and ICE1+A3, the curves are characterized by a knee at 5 kyrs BP caused by their Antarctic compo- nents that with the exception of the Aegean sites of Messenia (d) and SW Turkey (e) corresponds to a sea-level highstand. When ICE1 is consid- ered instead (dashed), RSL predictions clearly show a monotonous sea-level rise that broadly agrees with the trend of TP indicators from France, Italy and SW Turkey. In Roussillon (a), Marseilles (b) and Civitavecchia (c), the offset between ICE1+A3 and ICE3G results reflects differences in the time-history of the Northern Hemisphere ice sheets. At lower latitudes, with increasing distance from the former margins of Fennoscandia, the gap between solid and dotted curves is reduced. In the case of Djerba (h), ICE1 implies a highstand of a few centimeters at 5.0 kyrs BP, and the predictions by A3, ICE1+A3, and ICE3G almost overlap, to indicate that RSL observations at this site are virtually sensitive to the sole Antarctic component of the ice sheets distributions. Although evidence in favor of a Late-Holocene highstand at Djerba is weak due to the large (and statistically dubious) data un- certainties, the sensitivity of North Africa RSL observations to the chronology of Antarctica merits further investigations in Section 4.2.3. Predictions based on ICE5G (grey solid curves) are generally intermediate between those of ICE1 and ICE3G, as a direct consequence of the significant reduction of the total ESL of Antarc- tica relative to ICE3G. Even a cursory inspection of fig. 4a-i reveals that ICE3G provides globally a poor fit to the TP RSL indicators across the Mediterranean. Con- versely, with the sole exception of Djerba (h) and of sites belonging to tectonically unstable re- gions such as Beirut (Mahamoud et al., 2005; Morhange and Pirazzoli, 2005) and Algiers (12) (Meghraoui et al., 2004), ICE1 and ICE5G broadly match the RSL trends, but this should be tested by a more rigorous misfit study. The re- sults suggest that the Antarctic component is the main responsible of the failure of ICE3G, but it is also possible that the rheological profile VM1 is not fully suitable to describe the RSL varia- tions in this region. We will return to these issues in Sections 4.2 and in our future investigations. 3.3. Ocean and ice-induced RSL variations Since the Mediterranean is moderately dis- tant from the former Late-Pleistocene ice sheets, 749 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets Fig. 4a-i. Observed and synthetic RSL for some of the sites of fig. 3a. Predictions are obtained by VM1 and models ICE3G, ICE1+A3, ICE1, A3, and ICE5G. RSL data and error bars are taken from the TP database with the exception of Djerba (Jedoui et al., 1998). Since the TP RSL indicators are not calibrated, this comparison has mainly a qualitative character. In Djerba the melting of Antarctica provides virtually the whole signal dur- ing the last 6 kyrs (compare solid with dash-dotted curve in frame h). the ice-induced RSL variations in this region are not expected to dominate the ocean components, as it is the case in the near-field Clark’s zones I and II (Lambeck and Purcell, 2005). To address this point, using the rheological model VM1, in fig. 5 we separately consider the RSL variations of glacio- and hydro-isostatic origin at Roussil- lon, Jaffa and Djerba, representative of the northern, eastern, and southern coasts of the Mediterranean Basin, respectively (see fig. 3a,b). Dashed and dotted lines show ocean- and ice-induced RSL components, obtained by eq. (2.6) with S = S OCE and S = S ICE, respectively. Since SEUS= 0 after the end of deglaciation, in the time window considered RSL variations are only given by ocean- and ice-load effects. a b c d e f ihg 750 Paolo Stocchi and Giorgio Spada For both ICE1 and ICE3G, the trend of the ice-induced component of RSL (dotted) is con- stant across the Mediterranean, being mostly determined by long-wavelength deformations driven by distant sources. However, they are clearly sensitive to the assumed time-history, with a sea-level fall for ICE3G (top frames) and a sea-level rise for ICE1 (bottom). As discussed below, this diametrically opposite trend is to be attributed to the sea-level changes driven by the melting of Antarctica in ICE3G. Differently from the ice term, the ocean term in fig. 5 (dashed) shows a significant spatial variability, with a clear sea-level fall in Roussillon and Djer- ba, and a moderate sea-level rise in Jaffa. Such variability is caused by the sensitivity of this component of RSL to local effects related to the irregular shape of the shorelines (e.g., Mitrovica and Milne, 2002; Lambeck and Purcell, 2005). The trend of the hydro-isostatic term is the same for both ice sheets, but its amplitude tends to be larger for ICE3G, due to the effect of the extra ocean load driven by Antarctica. In fig. 6 we have repeated the computations above for the site Fig. 5. Ice- and ocean-load induced components of RSL for the sites of Roussillon, Jaffa, and Djerba (the sites location is shown in fig. 3a), using ICE3G (top) and ICE1 (bottom). Since in ICE1 the end of deglaciation oc- curs 8 kyrs BP (see fig. 1a-f), its eustatic component vanishes in this time window (not shown in figure). Fig. 6. Ice- and ocean-load induced components of RSL for the site of Djerba (see map of fig. 3a) ob- tained using model ICE5G (VM1). 751 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets of Djerba using the recently published deglacia- tion chronology ICE5G (Peltier, 2004). The re- duced ESL of Antarctica in ICE5G reduces the highstand amplitude relative to ICE3G (compare with fig. 5c); the emergence following the peak is now controlled by the ocean component of RSL, to indicate a dominating role of continental levering at this site and for this specific surface ice load. In fig. 7 we consider individual contribu- tions from major Late-Pleistocene ice sheets (North America, Fennoscandia, and Antarcti- ca), neglecting other minor constituents. The results illustrate how the distance from the for- mer ice sheets affects RSL curves. For Roussil- lon, the sea-level rise produced by the relative- ly nearby Fennoscandian component of ICE3G is counteracted by Antarctica and North Amer- ica to produce a marked Late-Holocene high- stand. This is only barely visible in Jaffa, since during the last 5 kyrs the melting of the North- ern Hemisphere aggregates almost exactly compensates the effects from the Southern Hemisphere. Finally, for Djerba, in the last 6 Fig. 8. Contributions to RSL due to individual com- ponents of ICE5G (VM1) at the site of Djerba (see fig. 3a). Fig. 7. Contributions to RSL due to individual components of the ICE3G (top) and ICE1 (bottom) global ice sheets distributions at the sites of Roussillon, Jaffa, and Djerba (see fig. 3a). 752 Paolo Stocchi and Giorgio Spada kyrs, the North American and Fennoscandian ice sheets have produced equal but opposite trends for both ICE1 and ICE3G that make the RSL observations from the coasts of Tunisia particularly sensitive to the deglaciation of Antarctica, as anticipated in Section 3.2. In fig. 8 the computations of fig. 7c are repeated using ICE5G. In Djerba we still observe a substantial cancellation of the effects of North America and Fennoscandia that shows the robustness of the ICE3G results and motivates further data acquisition from Tunisia, which is underway. 4. Results Assuming ICE3G and VM1 as a priori mod- els for the Late-Pleistocene ice sheets and for the Earth’s rheological profile, respectively, in the previous sections we have confirmed that Antarctica significantly affects the Holocene sea-level variations in the Mediterranean. How- ever, from the results obtained (see in particular fig. 4a-i) it is apparent that ICE3G qualitatively provides a poor fit to the field observations in this area, which are better reproduced by ICE1, which assumes a stationary Antarctic ice sheet throughout the Holocene, and by ICE5G, which implies a sizeable reduction of the ESL of Antarctica relative to ICE3G. To evaluate to what extent recent Mediterranean data might be useful to constrain the history of far-field ice sheets, here we will first review some existing deglaciation histories of Antarctica and then we will implement them in model ICE3G. Our pre- dictions will be compared with field observa- tions from various locations digitalized from the compilation of Pirazzoli (1991, 1996) and from more recent sources. With respect to the TB data- base, employed so far in our study to describe qualitatively the RSL trends across the Mediter- ranean, the sea-level indicators used in Section 4.2 below provide a better spatial coverage and are more reliable in a number of aspects. 4.1. Three ice models for Antarctica In the following, we will modify the origi- nal ice model ICE3G by including the three dis- tinct melting chronologies for Antarctica that in fig. 9 are denoted by S, G, and D, respectively. All of them contribute 14 m of total ESL (i.e., one half of the ICE3G value). Considering the distance of the Mediterranean from Antarctica, this ice sheet is modeled by a disc load of rectan- gular cross-section having an half-amplitude of 20 and a thickness of ∼355 m at the LGM. To fit the ∼113 m lowstand predicted by ICE3G in the far-field sites at the LGM (see fig. 1d), we have increased the volume of the Northern Hemisphere ICE3G components at this epoch keeping their isochrons unaltered. The global chronologies so obtained are named ICE3G(S), ICE3G(G), and ICE3G(D), respectively. Evidence in support of the chosen ESL for Antarctica comes from a number of sources. In his review on the volume of Antarctica at the LGM, Bentley (1999) pro- posed an ESL in the range of 6.1-13.1 m that matches the 12 m estimate of Huybrechts (1992) based on glaciological modeling. Through a 3D thermo-mechanical modeling, recently Huy- brechts (2002) has derived an ESL in the range of 14-18 m, consistent with the value of 14 m ob- tained from geologic constraints by Denton and Hughes (2002), whereas from an ice-dynamical Fig. 9. Time-histories of ESL for the Antarctic component of ICE3G according to the S, G, and D models described in the text. They are constituted by a simple disc-shaped ice element that stores an ESL of 14 m before the beginning of melting (12 kyrs BP). Predictions for modified ICE3G chronologies including these Antarctic components are shown in fig. 11 and 12. 753 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets approach Bintanja et al. (2002) estimated that Antarctica has contributed ∼5 m of ESL since the LGM. These estimates are consistently smaller (by at least a factor of 2) than former values based on the classical reconstruction of Denton and Hughes (1981), and of the ICE3G figure (∼27 m). As we will show in the following, a re- duced ESL for Antarctica improves the fit with the RSL observations in the Mediterranean, mainly because the amplitude of the Late- Holocene highstand is significantly reduced. The simplest of the three time-histories con- sidered for Antarctica (solid curve labeled by S in fig. 9) is characterized by a constant rate of melting between 12 and 5 kyrs BP, when geo- logical evidence indicates that deglaciation was complete (Goodwin, 1996). However, it has been suggested that some additional melt water was still added to the ocean during the last 6 kyrs. Since the main source of this Late- Holocene additional global sea-level rise of ∼3 m is assumed to be the Antarctic ice sheet (Nakada and Lambeck, 1988), we have also im- plemented a delayed (D) melting phase (dot- ted). This second chronology approximately follows S until 7 kyrs BP but subsequently the rate of deglaciation decreases until the ESL vanishes 1 kyr BP. The D chronology provides a Late-Holocene water release sufficient to in- crease sea-level by about 3 m since 6 kyrs BP (Lambeck and Bard, 2000), while its contribu- tion since 3 kyrs BP is less than 1 m (Fleming et al., 1998). Support for a delayed model of deglaciation also comes from the work of Stone et al. (2003), who analyzed surface exposure ages of glacial deposits in the West Antarctic ice sheet. Various pieces of evidence indicate a mid- to Late-Holocene sea-level highstand in the South Pacific, Indian Ocean and in parts of the Northern Atlantic and Pacific Oceans. The sub- sequent sea-level fall is generally attributed to «ocean siphoning» (Mitrovica and Milne, 2002). However, based on glaciological and geological evidence, Goodwin (1998) suggest- ed that a Late-Holocene increase of the Antarc- tic ice volume may be partly responsible for this sea-level fall. According to this hypothesis, the expansion of mountain glaciers, ice sheet margins and the thickening of the ice sheet in- terior could account for ∼1.0 m of the sea-lev- el fall on mid-oceanic islands. Hence, differ- ently from D, the G chronology, shown by a dashed line in fig. 9, is characterized by a Late- Holocene re-advance causing a general eustat- ic sea-level fall of ∼1.0 m since 5 kyrs BP, with ∼0.7 m of sea-level fall between 4 and 2 kyrs BP (Goodwin, 1998). 4.2. RSL data and predictions 4.2.1. French coasts From the collection of sea-level field obser- vations of Pirazzoli (1991), we have borrowed data from nine sites along the Golfe du Lion and the coasts of Corsica, ranging from Rous- sillon (1) to South Corsica (9); sites locations are shown by filled circles in fig. 10. In the original compilation of Pirazzoli (1991), only two sites show some indication of a Late- Holocene highstand, with an emergence of ∼2.0 m near Cap Romarin between 5 and 4 kyrs BP (Aloisi et al., 1978), and a highstand in the range of 2 to 4 m ∼4 kyrs BP deduced from un- dated beach deposits in the Nice area and in the surroundings (Dubar, 1987). From a reexami- nation of the available record, Lambeck and Bard (2000) concluded that along the French coast Holocene sea-levels have never exceeded the present level, the highstand being marked in the erosional notch at cap Romarin of pleis- tocenic age (Laborel et al., 1998). Some further evidence against a Late-Holocene highstand comes from archaeological excavations of the ancient harbor of Marseilles, which have pro- vided a new set of high-precision data for the past 4 kyrs (Morhange et al., 2001), and from the preservation of half-submerged Paleolithic paintings on a wall of the Cosquer cave near Marseilles (Vouve et al., 1996), showing that during the Holocene the sea-level never never exceeded its present-day level. In our previous calculations of fig. 2a-e, ice models ICE3G and ICE1+A3 predicted a Late- Holocene emergence along the Mediterranean coast of France. As shown in fig. 11, different re- sults are obtained when the three modified Antarctic chronologies of fig. 9 are implemented 754 Paolo Stocchi and Giorgio Spada in ICE3G. The solution for ICE3G(S) (solid lines) show a sea-level highstand of ∼1 m at 5 kyrs BP both in Roussillon and in the Rhone Delta region. Due to the reduced ESL of ICE3G(S) and to the 2 kyrs anticipation of the melting inception of its Antarctic component (see fig. 9), the maximum transgression is dimin- ished by about a factor of 2 with respect to ICE3G (compare with fig. 4a-i). The amplitude of the highstand diminishes eastward until it van- ishes at Port Cros (site 6); in Corsica (sites 7-9) the highstand is canceled by the submergence that characterizes Clark zone VII in the bulk of the Mediterranean. Model ICE3G(S) generally provides a poor fit to the data. The Late- Holocene re-advance of the Antarctic ice sheet, implied in the ICE3G(G) chronology (dashed lines), enhances the development of highstands. Those with largest amplitude (∼1.5 m) are ob- served at Roussillon (1) and in the Rhone Delta (2). The predictions for this hypothetical chronology of Antarctica systematically stand above those based upon ICE3G(S) (solid) and are in clear disagreement with observations from Southern France. ICE3G(D) (dotted) implies a smooth sea-level rise through the Late Holocene, similar to that predicted by ICE1 and ICE3G-A3 (see fig. 2c,d) and a subsequent barely visible sea-level fall ending between 2 and 1 kyrs BP. The general agreement of ICE3G(D) predictions with sea-level observations from this region is apparent. Fig. 10. Sites from Southern France and Corsica (sites 1-9), the Tyrrhenian coast of Italy (sites 10-16), North- ern Adriatic (17), and Tunisia (18) for which field observations of past sea-level variations are available from the compilation of Pirazzoli (1991) (filled circles), and Lambeck et al. (2004a) (open). Predictions for these RSL sites are shown in figs. 11 and 12, respectively. 755 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets F ig . 11 . R S L o bs er va ti on s fr om S ou th er n F ra nc e an d C or si ca a cc or di ng t o th e co m pi la ti on o f P ir az zo li ( 19 91 ), to w hi ch t he r ea de r is r ef er re d fo r th e or ig in al s ou rc es . D at a un ce rt ai nt ie s ar e ta ke n fr om s ou rc es w it h th e ex ce pt io n of R ou ss il lo n (1 ) an d th e R ho ne D el ta ( 2) , fo r w hi ch a q ua li ta ti ve er ro r ba nd i s gi ve n. S ol id , da sh ed , an d do tt ed c ur ve s pe rt ai n to m od el s IC E 3G (S ), IC E 3G (G ), an d IC E 3G (D ), re sp ec ti ve ly . F ig . 12 . R S L o bs er va ti on s an d un ce rt ai nt ie s fr om t he T yr rh en ia n co as ts o f It al y (s it es 1 0- 16 ) an d fr om t he N or th er n A dr ia ti c (1 7) a cc or di ng t o L am - be ck e t a l. (2 00 4a ). P re di ct io ns a re b as ed o n th e th re e va ri an ts o f m od el I C E 3G a lr ea dy c on si de re d in f ig . 9. 11 1 2 756 Paolo Stocchi and Giorgio Spada 4.2.2. Tyrrhenian and Northern Adriatic coasts of Italy The pioneering investigations of Alfieri and Caputo (Schmiedt, 1972) have shown the im- portance of the archaeological remains as sea- level indicators of past sea-levels along the Tyrrhenian coasts of Italy. For this region, they estimated a sea-level rise of ∼1.7 mm yr−1 be- tween 600 BC and 100 AD; in particular, dating Roman fish tanks and submerged harbors, they showed that between 100 years BC and 100 years AD the sea-level was ∼1.0 m lower than present. Subsequently, the radiocarbon-based RSL curve of Antonioli and Frezzotti (1989) evidenced, for the southern coasts of Lazio, a sea-level similar to the present one between 7 and 5.4 kyrs BP, followed by a slight oscillation below the present level (Pirazzoli, 1991), while the cumulative Tyrrhenian sea-level curve of Alessio et al. (1994) confirmed that during the Holocene the sea-level was never above the present level. From the recently published high- quality data of Lambeck et al. (2004a) we se- lected those relative to the tectonically stable Tyrrhenian sites marked by open circles in fig. 10. RSL observations in fig. 12 clearly indicate a monotonous submergence since 8 kyrs BP with no evidence of highstands. With the excep- tion of the Northern Adriatic (site 17), ICE3G(D) (dotted) systematically overestimates RSL observations, but correctly reproduce their trends. In qualitative agreement with the results obtained for Southern France (see fig. 11), chronologies ICE3G(S) and ICE3G(G) do not improve the fit with the observations. It should be remarked, however, that since the ice models scrutinized here do not account for the effects of melting of the Alpine ice-sheet, a bias may be present in our computations (see Stocchi et al., 2005a). In addition, in the case of Northern Adriatic, assuming a constant ocean function may provide a poor approximation due to the shallow water environment. More refined mod- elizations can account for time-evolving shore- lines (e.g., Lambeck et al., 2004a). 4.2.3. Gulf of Gabes, Tunisia Paskoff and Sanlaville (1983) proposed a tentative sea-level curve for South Tunisia in the last 8000 years (see also Pirazzoli 1991). The fluctuating RSL curve of Paskoff and San- laville, relative to Djerba (site 18) and repro- duced in fig. 13 by filled squares, shows a max- imum emergence of reproduce ∼2.0 m between 6 and 5 kyrs BP and a second minor highst and ∼3 kyrs BP. The subsequent study of Jedoui et al. (1998) evidenced two fossilized bioclastic beaches at different elevations (see the diamonds in fig. 13a). The older palaeobeach deposit (6.4- 4.3 years BP) is found at an elevation of about 40 to 100 cm above the present sea-level, while the younger, dated 1850 years BP, at present sea-lev- el. Uncertainties on the interpretation of past sea- level indicators from this area have been dis- cussed by Pirazzoli (1987) and Lambeck et al. (2004a). Recently Morhange and Pirazzoli (2005) published a tentative sea-level curve for Southern Tunisia based on new indicators col- lected between the Gulf of Gabes and the Libyan border. The curve, reproduced by open circles in Fig. 13. RSL observations for Djerba according to various sources, and predictions using ice models ICE3G(S) (solid line), ICE3G(G) (dashed), ICE3G(D) (dotted), and ICE5G (solid grey). A dash-dotted curve shows results obtained assuming that the Antarctic ice sheet model D of fig. 9 is the only active load since the LGM. 757 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets fig. 13, shows a transgression peak of 1.86±0.11 m between 6 and 5 14C kyrs BP. In fig. 13 we compare the RSL observations from Djerba with predictions based on the three modified ICE3G chronologies described in Sec- tion 4.1. While ICE3G(S) (solid lines) and ICE3G(G) (dashed) support the Late-Holocene submergence suggested by the data of Mor- hange and Pirazzoli (2005) and Paskoff and San- laville (1983) in the last ∼5 kyrs, ICE3G(D) shows a monotonous emergence in contrast with observations. However, since reliable confidence bands for dotted curves are lacking, a rigorous evaluation of misfit is not possible. Provided that uncertainties reported by Jedoui et al. (1998) are reliable, fig. 13 shows that none of the three time-histories considered is indeed in contrast with the observations. While ICE5G (grey curve) is broadly consistent with Jedoui et al. (1998), it does not account for the neat highstand suggest- ed by Paskoff and Sanlaville (1983) and Morhange and Pirazzoli (2005), nor with its tim- ing. In every instance, from the results obtained for this region it is clear that improved RSL ob- servations for the coasts of Tunisia could signif- icantly contribute to constrain the time-history of Antarctica in the last 6 kyrs. The cancellation of the effects from the Northern Hemisphere ice sheets that occur here once again is manifest ob- serving that ICE3G(D) (dotted) basically repro- duces the effect from the sole Antarctic compo- nent of this ice sheet (D, dash-dotted). 4.2.4. The Levant Sea, Israel Evidence for Late-Holocene sea-level high- er than the present in Israel has been reported by Sneh and Klein (1984) and by Raban and Galili (1985) who derived two similar sea-level curves for the site of Dor, showing sea-level fluctuations of over 2.0 m of amplitude (Piraz- zoli, 1991) as portrayed in fig. 14b. In a subse- quent study, Nir and Eldar (1987) proposed a Fig. 14a. Sites along the coasts of Israel, ranging between Haifa (site 21) and Yavne Yam (24). 758 Paolo Stocchi and Giorgio Spada curve characterized by small oscillations for the last 2.5 kyrs. From the investigation of sub- merged archaeological remains along the conti- nental shelf of Israel, between Haifa and Atlit, Galili et al. (1988) obtained the RSL curve shown by a dashed line in fig. 14b, indicating a monotonous and smooth sea-level rise until ∼1 kyrs BP (Pirazzoli, 1991). As shown in frame (c), archaeological evidence from the sites of Tel Nami (20), Dor (21), Michmoret (23) and Yavne Yam (24) indicate that 6 kyrs BP sea-lev- el was ∼4 m lower than today and that it reached the present-day level between 3 and 2 kyrs BP (Sivan et al., 2001). The observational limits derived from the coastal water wells in Caesarea Maritima (site 22) have extended the record of the Late-Holocene sea-level change to 1300 AD (Sivan et al., 2004) and suggest that during the Byzantine period sea-level was high- er by ∼30 cm than today. The RSL curve ob- tained using ICE3G(G) (dashed curve in fig. 14c) shows a highstand between 5 and 4 kyrs BP, followed by emergence and lastly by a neg- ative oscillation. When the ICE3G(S) and the ICE3G(D) ice chronologies are considered (solid and dotted lines, respectively), the pre- dicted Late-Holocene sea-level curves define a narrow band broadly consistent with the ar- chaeological evidence until ∼2 kyrs BP. These latter predictions do not match the more recent sea-level data and thus appear to be inconsistent with the higher than present sea-level between 2 and 1 kyr BP observed at Caesarea Maritima according to Sivan et al. (2004). 5. Conclusions The main results can be summarized as fol- lows: i) The pattern of sea-level change in the Mediterranean can be described by means of two types of RSL curves. The first denotes a rising sea-level through the Late-Holocene, while the second exhibits a Late-Holocene highstand that implies a sea-level fall in the last ∼5 kyrs. The ex- tent of the so-called Clark’s regions characterized by these distinct sea-level patterns is strongly de- pendent upon the assumptions about the time-his- tory of the Late-Pleistocene ice sheets surround- ing the Mediterranean. When the ICE3G chronol- ogy is employed, a well developed highstand re- gion is expected along the Mediterranean coasts, while the submergence region covers the bulk of the basin. Such peculiar pattern for closed basins has been named here as «Clark’s zone VII». The assumption of a stationary Antarctic ice sheet and the enhanced effect of the melting of Fennoscan- dia disrupt zone VII when ICE1 is employed, leaving highstand zones along the indented coast lines of the Alboran Sea and SE Tunisia. This clearly shows the significant role played by the melting of Antarctica upon the Holocene sea-lev- el variations in the Mediterranean. Fig. 14b,c. In (b) we have reproduced tentative RSL curves from the existing literature, while in (c) we compare the data of Sivan et al. (2001, 2004) with results based on the same models previously considered in fig. 13. b c 759 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets ii) Using models ICE3G and ICE5G, we have shown that along the coasts of SE Tunisia (i.e., at Djerba) the effects due to the melting of Fennoscandia are almost exactly counterbal- anced by that of North America. Such fortu- itous cancellation makes the highstand of ∼2 m predicted in this region only sensitive to the ef- fects of the remote Antarctica ice sheet. A high- stand is indeed suggested by both available ob- servations and tentative RSL curves at Djerba, but their resolving power is not sufficient to constrain its amplitude unequivocally. Incorpo- rating within ICE3G a suite of plausible models for the melting of Antarctica during the last 6 kyrs, all characterized by a sensibly reduced ESL at the LGM, it has been possible to fully enlight- en the sensibility of RSL observations from Southern Tunisia to the details of the time-histo- ry. A Late-Holocene highstand of ∼ 1 m in this region is predicted when Antarctica is assumed to melt at a constant rate between 12 and 5 kyrs BP, and its amplitude is enhanced (nearly dou- bled) if a late ice re-advance is assumed. On the contrary, a delayed melting of Antarctica until 1 kyr BP is responsible for a regular sea-level rise since 6 kyrs BP. Only an improved spatial cover- age and sampling frequency of the RSL data, made however difficult by the large tidal excur- sions in the Gulf of Gabes (Sammari et al., 2006), could help to put tight bounds on the de- tails of the melting chronology of Antarctica. Model ICE5G(VM1) is unable to account for the amplitude and timing of the highstand suggested by the recent observations of Morhange and Pi- razzoli (2005). This motivates further data acqui- sition in SE Tunisia and a refinement of existing ice sheets models. iii) An assessment of the effects of mantle viscosity is beyond our purposes here. Howev- er, we have verified that an increase in ηLM from 2×1021 to 1022, a value more appropriate accord- ing to a number of authors (see e.g., Nakada and Lambeck, 1989) has a significant influence on the predicted RSL curves for the Mediter- ranean. In particular, this implies the disappear- ance of the Late-Holocene highstand from the coasts of Southern France, that would improve the agreement with RSL observations from this region. In general, an increase in lower mantle viscosity limits the variability of the predicted RSL curves corresponding to different assump- tions about the chronology of Antarctica. As- suming a high-viscosity lower mantle, the con- firmation of the existence of a Late-Holocene highstand in Djerba would indicate for Antarc- tica an ESL consistent with the value implicit in ICE3G, whereas an essentially stationary sea- level would be suggestive of a sensibly reduced ESL. Acknowledgements We thank Fabrizio Antonioli for a constructive review and Maurizio Bonafede for suggestions and encouragement. Florence Colleoni, Kurt Lam- beck, Jeffrey Donnelly, and Liviu Giosan are ac- knowledged for discussions and encouragement. Anna Radi is acknowledged for help in the compi- lation of bibliography. PS has benefited from an exchange program at the Department of Geologi- cal Sciences of Brown University at Providence (RI), that we acknowledge for having provided support and a stimulating environment useful for the preparation of this work. The code for solving the SLE (SELEN, see http://flocolleoni.free.fr/ SELEN.html/) is freely available (Spada and Stoc- chi, 2007) and can be requested to GS (e-mail: giorgio.spada@gmail.com). The figures were drawn using the GMT public domain software (Wessel and Smith, 1998). Research funded by MIUR (Ministero dell’Università, dell’Istruzione e della Ricerca) by the PRIN2006 grant «Il ruolo del riaggiustamento isostatico postglaciale nelle variazioni del livello marino globale e mediterra- neo: nuovi vincoli geofisici, geologici, ed archeo- logici». REFERENCES ALESSIO, M., L. ALLEGRI, F. ANTONIOLI, G. BELLUOMINI, S. IMPROTA, L. MANFRA and M. PREITE MARTINEZ (1994): La curva di risalita del mare Tirreno negli ultimi 43 ka ricavata da datazioni su speleotemi sommersi e dati archeologici, Mem. Descr. Carta Geol. d’Italia, LII, 261-276. ALOISI, J.C., A. MONACO, N. PLANCHAIS, J. THOMMERET and Y. THOMMERET (1978): The Holocene transgression in the Golfe du Lion, southwestern France: paleogeo- graphic and paleobotanical evolution, Geogr. Phys. Quat., 32 (2), 145-162. ANTONIOLI, F. and M. FREZZOTTI (1989): I sedimenti tardo- 760 Paolo Stocchi and Giorgio Spada pleistocenici e olocenici compresi nella fascia costiera tra Sabaudia e Sperlonga, Mem. Soc. Geol. Ital., 42, 321-334. ARGUS, D.F., W.R. PELTIER and M.M. WATKINS (1999): Glacial isostatic adjustment observed using very long baseline interferometry and satellite laser ranging geo- desy, J. Geophys. Res., 104 (B12), 29077-29094. BENTLEY, M.J. (1999): Volume of Antarctic Ice at the Last Glacial Maximum, and its impact on global sea-level change, Quat. Sci. Rev., 18, 1569-1595. BINTANJA, R., R.S.W. VAN DE WAL and J. OERLEMANS (2002): Global ice volume variations through the last glacial cicle simulated by a 3D ice-dynamical model, Quat. Int., 95 (96), 11-23. CARMINATI, E. and G. DI DONATO (1999): Separating natu- ral and anthropogenic vertical movements in fast sub- siding areas: Po Plain (North Italy) case, Geophys. Res. Lett., 26, 2291-2294. CIANETTI, S., C. GIUNCHI and G. SPADA (2002): Mantle vis- cosity beneath the Hudson Bay: an inversion based on the Metropolis algorithm, J. Geophys. Res., 107, doi: 10.1029/2001JB000585, 2352. CLARK, J.A. and C.S. LINGLE (1979): Predicted relative sea- level changes (18 000 years BP to present) caused by late-glacial retreat of Antarctic ice sheet, Quat. Res., 11, 279-298. CLARK, J.A., W.E. FARRELL and W.R. PELTIER (1978): Glob- al changes in postglacial sea-level: a numerical calcu- lation, Quat. Res., 9, 265-287. DENTON, G.H. and T.H. HUGHES (1981): The Last Greast Ice Sheets (John Wiley, New York), pp. 484. DENTON, G.H. and T.H. HUGHES (2002): Reconstructing the Antarctic ice sheet at the last glacial maximum, Quat. Sci. Rev., 21, 193-202. DUBAR, M. (1987): Donnees nouvelles sur la transgression holocène dans la region de Nice (France), Bull. Soc. Géol. France (Paris), Serie 8, 3 (1), 195-198. FARRELL, W.E. and J.A. CLARK (1976): On postglacial sea- level, Geophys. J.R. Astr. Soc., 46, 647-667. FLEMING, K., P. JOHNSTON, D. ZWARTZ, Y. YOKOAMA, K. LAMBECK and J. CHAPPELL (1998): Refining the eustat- ic sea-level curve since the Last Glacial Maximum us- ing far- and intermediate- field sites, Quat. Res., 9, 265-287. FLEMMING, N.C. (1972): Eustatic and tectonic factors in the relative vertical displacement of the Aegean coast, in The Mediterranean Sea: a Natural Sedimentation Lab- oratory, edited by D.J. STANLEY (Dowden, Hutchinson and Ross, Inc., Stroudsburg), 180-201. GALILI, E., M. WEINSTEIN-EVRON and A. RONEN (1988): Holo-cene sea-level changes based on submerged ar- chaeological sites of the Northern Carmel coast in Is- rael, Quat. Res., 29, 36-42. GOODWIN, I.D. (1996): A mid to Late-Holocene readvance of the Law Dome ice margin, Budd Coast, East Antarc- tica, Antarctic Sci., 8, 395-406. GOODWIN, I.D. (1998): Did changes in Antarctic ice volume influence the Late-Holocene sea-level lowering?, Quat. Sci. Rev., 17, 319-332. HUYBRECHTS, P. (1992): The Antarctic ice sheet and envi- ronmental change: a three-dimensional modelling study, Rep. Polar Res. (Alfred Wegener Institute, Bre- hemerhaven), 99, pp. 242. HUYBRECHTS, P. (2002): Sea-level changes at the LGM from ice-dynamic reconstructions of the Greenland and Antarctic ice sheets during the glacial cicles, Quat. Sci. Rev., 21, 203-231. JEDOUI, Y., N. KALLEL, M. FUNTUGNE, H.B. ISMAIL, A. M’RABET and M. MONTACER (1998): A high relative sea-level stand in the middle Holocene in southeastern Tunisia, Mar. Geol., 147, 123-130. KAUFMANN, G. (2002): Predictions of secular geoid changes from the Late-Pleistocene and Holocene Antarctic ice- ocean mass imbalance, Geophys. J. Int., 148 (2), 340- 347. LABOREL, J., M. AARNOLD, F. LABOREL-DEGUEN, C. MORHAN- GE and N. TISNERAT-LABARD (1998): Confirmation of a Pleistocene age for the marine notch at Cap Romarin (Port-La-Nouvelle, Languedoc, France), Geomorfol. Re- lief. Process. Environ., 2, 125-130. LAMBECK, K. (1993): Glacial rebound of the British Isles- II. A high-resolution, high-precision model, Geophys. J. Int., 115, 960-990. LAMBECK, K. (1995): Late-Pleistocene and Holocene sea- level change in Greece and south-western Turkey: a separation of eustatic, isostatic and tectonic contribu- tions, Geophys. J. Int., 122, 1022-1044. LAMBECK, K. and E. BARD (2000): sea-level changes along the French Mediterranean coast from the past 30 000 years, Earth Planet. Sci. Lett., 175, 203-222. LAMBECK, K. and A. PURCELL (2005): Sea-level change in the Mediterranean Sea since th LGM: model predic- tions for tectonically stable areas, Quat. Sci. Rev., 24, 1969-1988. LAMBECK, K., A. PURCELL, P. JOHNSTON, M. NAKADA and Y. YOKOYAMA (2003): Water-load definition in the glacio- hydro-isostatic sea-level equation, Quat. Sci. Rev., 22, 309-318. LAMBECK, K., F. ANTONIOLI, A. PURCELL and S. SILENZI (2004a): Sea-level change along the Italian coast from the past 10 000 years, Quat. Sci. Rev., 23, 1567-1598. LAMBECK, K., M. ANZIDEI, F. ANTONIOLI, A. BENINI and A. ESPOSITO (2004b): Sea-level in Roman time in the Cen- tral Mediterranean and implications for recent change, Earth Planet. Sci. Lett., 224, 563-575. MAHAMOUD, S., R. REILINGER, S. MCCLUSKY, P. VERNANT and A. TEALEB (2005): GPS evidence for northward motion of the Sinai Block: implications for E. Mediter- ranean tectonics, Earth Planet. Sci. Lett., 238, 217-224. MEGHRAOUI, M., S. MAOUCHE, B. CHEMAA, Z. CAKIR, A. AOUDIA, A. HARBI, P.-J. ALASSET, A. AYADI, Y. BOUHADAD and F. BENHAMOUDA (2004): Coastal uplift and thrust faulting associated with the Mw= 6.8 Zem- mouri (Algeria) earthquake of 21 May, 2003, Geophys. Res. Lett., 31, L19605. MITROVICA, J.X. and G.A. MILNE (2002): On the origin of Late-Holocene highstands within equatorial ocean basins, Quat. Sci. Rev., 21, 2179-2190. MITROVICA, J.X. and W.R. PELTIER (1991): On postglacial geoid subsidence over the equatorial oceans, J. Geo- phys. Res., 96, 20,053-20,071. MORHANGE, C. and P.A. PIRAZZOLI (2005): Mid-Holocene emergence of Southern Tunisian coasts, Mar. Geol., 220, 1-4. MORHANGE, C., J. LABOREL and A. HESNARD (2001): Changes of relative sea-level during the past 5000 years 761 Glacio and hydro-isostasy in the Mediterranean Sea: Clark’s zones and role of remote ice sheets in the ancient harbor of Marseilles, Southern France, Pa- leogeogr. Paleoclimatol. Paleoecol., 166, 319-329. NAKADA, M. and K. LAMBECK (1988): The melting history of the Late-Pleistocene Antarctic ice sheet, Nature, 333, 36-40. NAKADA, M. and K. LAMBECK (1989): Late-Pleistocene and Holocene sea-level change in the Australian region and mantle rheology, Geohpys. J., 96, 497-517. NIR, Y. and I. ELDAR (1987): Ancient wells and their geoar- chaeological significance in detecting tectonics of the Is- rael Mediterranean coastline region, Geology, 15, 3-6. PASKOFF, R. and P. SANLAVILLE (1983): Les Côtes de la Tunisie. Variations du Niveau Marin depuis le Tyr- rhénien (Maison de l’Orient Mediterraneen, Lyon), pp. 192. PELTIER, W.R. (1994): Mantle viscosity and ice-age ice sheet topography, Science, 273, 1359-1364. PELTIER, W.R. (2004): Global Glacial Isostasy and the Sur- face of the Ice-Age Earth: The ICE- 5G(VM2) model and GRACE, Ann. Rev. Earth Planet Sci., 32, 111-149. PELTIER, W.R. and T.S. ANDREWS (1976): Glacial-isostatic adjustment, I, The forward problem, Geophys. J.R. As- tr. Soc., 46, 605-646. PIRAZZOLI, P.A. (1987): Sea-level changes in the Mediter- ranean, in Sea-Level Changes, edited by M.J. TOOLEY and I. SHENNAN (Blackwell), Inst. British Geogr. Spec. Publ. Ser. 20, 152-181. PIRAZZOLI, P.A. (1991): World Atlas of Holocene Sea-level Changes (Elsevier, Amsterdam), pp. 300. PIRAZZOLI, P.A. (1996): Sea-Level Changes, the Last 20 000 Years (JohnWiley and Sons, Chichester), pp. 211. PIRAZZOLI, P.A. (2003): Mid- and Late-Holocene relative sea-level changes in the Mediterranean area: a review of selected sites with discussion of the possible eustat- ic, isostatic, and tectonic contributions, in Proceedings of the IGCP 437 Final Conference, 22-28 September 2003, Otranto-Taranto, Italy (Abstract). PIRAZZOLI, P.A. (2005): A review of possible eustatic, iso- static, and tectonic contributions in eight Late- Holocene relative sea-level histories from the Mediter- ranean area, Quat. Sci. Rev., 24, 1989-2001. RABAN, A. and E. GALILI (1985): Recent maritime archaeo- logical research in Israel - A preliminary report, Int. J. Naut. Archaeol. Underw. Explor., 14 (4), 321-356. SAMMARI, C., V.G. KOUTITONSKY and M. MOUSSA (2006): Sea-level variability and tidal resonancein the Gulf of Gabes, Tunisia, Cont. Shelf. Res., 26, 338-350. SCHMIEDT, G. (1972): Il Livello Antico del Mar Tirreno, Testimonianze dei Resti Archeologici (Olschki, Flo- rence), pp. 323. SIVAN, D., S. WDOWINSKI, K. LAMBECK, E. GALILI and A. RABAN (2001): Holocene sea-level changes along the Mediterranean coast of Israel based on archaeological observations and numerical model, Paleogeogr. Paleo- climatol. Paleoecol., 167, 101-117. SIVAN, D., K. LAMBECK, R. TOUEG, A. RABAN, Y. PORATH and B. SHIRMAN (2004): Ancient coastal wells of Cae- sarea Maritima, Israel, an indicator for relative sea-lev- el changes during the last 2000 years, Earth Planet. Sci. Lett., 222, 315-330. SNEH, Y. and M. KLEIN (1984): Holocene sea-level changes at the Coast Dor, Southeast Mediterranean, Science, 226, 831-832. SPADA, G. and P. STOCCHI (2006): The Sea-Level Equation – Theory and Numerical Examples (Aracne, Rome), pp. 81. SPADA, G. and P. STOCCHI (2007): SELEN: a Fortran 90 pro- gram for solving the «Sea-Level Equation», Comput. Geosci., 33 (4), 538-562. SPADA, G., A. ANTONIOLI, L. BOSCHI, S. CIANETTI, G. GAL- VANI, C. GIUNCHI, B. PERNIOLA, N. PIANA AGOSTINETTI, A. PIERSANTI and P. STOCCHI (2004): Modeling Earth’s post-glacial rebound, Eos, Trans. Am. Geophys. Un., 85, 62-64. STOCCHI, P., G. SPADA and S. CIANETTI (2005a): Isostatic re- bound following the Alpine deglaciation: impact on the sealevel variations and vertical movements in the Mediterranean region, Geophys. J. Int., 162 (1), 137- 147. STOCCHI, P., S. CIANETTI, L. GIROMETTI and G. SPADA (2005b): Holocene sea-level changes in the Mediter- ranean: the role of remote and near-field ice sheets, in Proceedings of the 9th International Workshop on Nu- merical Modeling of Mantle Convection and Lithospher- ic Dynamics, 8-14 September 2005, Erice (TP), p. 46. STONE, J.O., G.A. BALCO, D.E. SUGDEN, M.W. CAFEE, L.C. SASS III, S.G. COWDERY and C. SIDDOWAY (2003): Holocene deglaciation of Marie Bird Land, West Antarctica, Science, 299, 99-102. TEGMARK, M. (2002): An icosahedron-based method for pixelizing the celestial sphere, Astrophys. J., 470, L81- L84. TUSHINGHAM, A.M. and W.R. PELTIER (1991): Ice-3G: a new global model of Late-Pleistocene deglaciation based upon predictions of postglacial sea-level change, J. Geophys. Res., 96, 4497-4523. TUSHINGHAM, A.M. and W.R. PELTIER (1992): Validation of the ICE3G model of Würm-Winsconsin deglaciation using a global database of relative sea-level histories, J. Geophys. Res., 97, 3285-3304. TUSHINGHAM, A.M. and W.R. PELTIER (1993): Relative sea- level database, IGPB PAGES/World Data Center-A for Paleoclimatology Data Contribution Series # 93-106, NOAA/NGDC Paleoclimatology Program (Boulder, CO, U.S.A.). VOUVE, J., P. MALAURENT and J. BRUNET (1996): Caractéri- sation physique et environnementale d’un sanctuaire préhistorique sous-marin, semi-noyé et orné préalable- ment à l’élaboration d’une démarche conservatoire. Cas de la grotte Cosquer, France, C.R. Acad. Sci., Sér. 2, 322 (11), 935-942. WALCOTT, R.I. (1972): Past sea-levels, eustasy and defor- mation of the Earth, Quat. Res., 2, 1-14. WESSEL, P. and W.H.F. SMITH (1998): New, improved ver- sion of generic mapping tools released, Eos, Trans. Am. Geophys. Un., 79, 579. (received August 1, 2007; accepted January 19, 2008)