Page 20 - Vacchi_Marriner_2016

M. Vacchi et al. / Earth-Science Reviews 155 (2016) 172–197 191
based transfer functions to reconstruct former sea-levels are seldom re- extends from the early Holocene, the database documents rapid rising
-1
ported in the Mediterranean (e.g., Anadón et al., 2002; Silvestri et al., rates (≤8mm a ) between ~12 and ~7.5 ka BP corresponding to the
2005; Triantaphyllou et al., 2005). This is mainly due to the microtidal major deglaciation phase (e.g., Peltier, 2004; Lambeck et al., 2014). Ris-
setting of most Mediterranean coasts, which is not conducive to estab- ing rates significantly slowed in the remaining part of the mid-Holocene
lishing a well-defined intertidal zoning. However, our results indicate and reached values b1mm a -1 during the late Holocene where ice
the potential of inner or semi-enclosed lagoon samples as RSL index equivalent meltwater input is negligible (Milne et al., 2005; Church
points in the Mediterranean Sea. In this database, we assigned an indic- et al., 2008). Comparison between predicted and observed RSL changes
ative range of 0 to -1 m MSL to these samples, on the basis of the avail- are only robust at those sites where the RSL reconstruction is mainly
able literature. Future development of local or regional transfer based on virtually incompressible and basal data points. Conversely,
functions could significantly reduce this vertical uncertainly, possibly caution is required for comparisons done with RSL reconstructions
making them a high-resolution RSL indicator. based only on intercalated samples.
Our compilation underlines the importance of L. byssoides fossil rims The database documents a general overestimation of the nominal
as sea-level index points in the Mediterranean. The majority of our sam- ICE-5G (VM2) model prediction in most of the regions. This is not sur-
ples, collected at the beginning of the 1990s, show a large chronological prising since the ICE-5G (VM2) model was not constrained by RSL ob-
error bar due to 14 C dating uncertainty. Recent developments in AMS servations from far-field sites with respect to the formerly glaciated
techniques, pretreatment and calibration (e.g., Bronk Ramsey, 2008) regions (Peltier, 2004), such as the Mediterranean Sea. Data from
could drastically reduce these uncertainties (e.g., Faivre et al., 2013). most of the regions located in the northwestern sectors of our database
Mediterranean beachrocks can be considered precise indicators (#2, #3, #4, #5, #6, Figs. 6 and 7) and from northwestern Sicily (#12,
when they are corroborated by chemical and fabric information of the Fig. 9) are in good agreement with the lower boundary of the predicted
cement and, when possible, by the description of sedimentary struc- curve (i.e. between nominal and minimal, see Section 4, Table 2). Dis-
tures (Mauz et al., 2015b). Without such information, the resolution of parities increase eastwards and southwards where the minimal curve
beachrocks index points decreases considerably, making them a weak slightly overestimates the RSL position, notably in the Ionian and mid
proxy for Mediterranean sea-level reconstruction. As an example, in to southern Adriatic Seas (Figs. 9, #14 and 11).
southwestern Sardinia (#8), the large error bars of the youngest Such underestimation is also visible in regions considered tectoni-
beachrock index point did not significantly improve the late Holocene cally stable. However, sediment compaction cannot account for this
RSL history of the region (Fig. 12C). Microstratigraphic analysis of the misfit. For instance, the underestimation is also present in the mid-
cement of these samples may lead to a considerable reduction in the eastern Adriatic region where the database is only composed of virtually
error bar and, therefore, result in a better understanding of the RSL incompressible samples (Fig. 11, #19). It is possible that the chronology
changes in this area. of the GIA model adopted in this study (and its rheological profile) could
Our database further confirmed the importance of archaeological be modified in order to provide a better fit with the new RSL index
RSL data-points. We decided to only produce index points from coastal points. However, we did not pursue this goal, since modifying these pa-
structures closely related to former sea-level (i.e. fishtanks, fish ponds rameters would probably disrupt the agreement between the RSL ob-
and ancient harbour interface structures such as quays and jetties). servations available from polar regions and modelling predictions.
When coupled with fixed biological indicators, these structures may This would require the implementation and the adoption of a fully 3D
produce index points with uncertainty only represented by the leveling GIA model accounting for the lateral variations in viscosity, between
and the tidal errors (Morhange and Marriner, 2015). Conversely, we the shield area and the remote sites, which is not the purpose of this
produced limiting points with such coastal structures whose relation study.
with former sea-level is less clear (i.e. coastal quarries, roads, tombs, In our database, we avoided RSL reconstructions in areas affected by
wrecks, sewage). major tectonic deformation such as the Calabrian arc (see Section 2.1,
Our multi-proxy approach provides new insights into the ongoing Ferranti et al., 2006; Antonioli et al., 2009). However, comparison be-
debate about sea-level position during Roman times in the Tyrrhenian tween predicted and observed RSL gave new insights into the influence
Sea. Fish tank interpretations by Evelpidou et al. (2012) place RSL be- of mild but significant long-term uplift on the sea-level history of a
tween −0.58 and −0.32 m MSL while Lambeck et al.’s(2004b) inter- number regions. In mid-eastern Siciliy (region #13) our record agreed,
pretation places RSL between −1.37 and −1.2 m. On the Tiber Delta, using a larger dataset, with the results of previous studies (Dutton
Goiran et al. (2009, 2010) placed the RSL at −0.8±0.2 m MSL at et al., 2009; Spampinato et al., 2011) that are consistent with long-
-1
~2.0 ka BP on the basis of the fixed biological indicators found in the term uplift rates of between ~0.4 and ~0.7 mm a .Inthe western Ligu-
Roman habour of Claudius’ basin and the Darsena (Fig. 7,#9).One rian Sea (region #4), the Holocene data-points collected near Nice de-
salt-marsh index point from the Tiber Delta indicates RSL was at scribe a RSL history higher than those of the remaining areas of the
-0.6±0.6 m at ~1.9 ka BP (Di Rita et al., 2010). Long term uplift rates region. The elevation of these data suggests that the low long-term up-
-1
of ~0.11 mm a -1 are reported in the Tiber Delta area (Ferranti et al., lift (average rate ≤0.06 mm a since the last interglacial) reported by
2010; Marra et al., 2013). If we apply the uplift correction to the two lat- Dubar et al. (2008) has been active for more than half of the Holocene
ter data-points we obtain a RSL of between −0.8 and −1mMSLduring because these data are above the nominal curve from ~12.5 to ~5.5 ka
Roman times. This value is similar (with errors) to the Roman RSL ob- BP. Our compilation confirms that such an uplift trend is restricted to
tained in the Gulf of Gaeta (#10). Here, one lagoonal index point places the Nice area and rapidly decreases in the surrounding sectors, as al-
the RSL at −1±0.5 m at 2.1 ka BP. In summary, our database seems to ready hypothesized on the basis of the MIS 5e elevations (Dubar et al.,
suggest that RSL rose to within 1 m of modern MSL from ~2.0 ka BP to 2008; Rovere et al., 2011). Similarly, in the Tiber Delta (region #9)
present. Nonetheless, the associated error bars do not allow us to recon- and on the Sele coastal plain (region #11) the long-term uplift rates re-
cile the different functional and architectural interpretations of Roman ported by Ferranti et al. (2006, 2010) most likely control the position of
fish tanks. Further high-resolution studies are required in this area, if RSL data points, especially for the early-to mid-Holocene. The offset be-
possible that couple fixed biological indicators and archaeological coast- tween predictions and observations in central Spain (Fig. 6,#1) remains
al structures. an open issue. Here, the early to mid-Holocene RSL data-points placed
above the nominal curve suggest an underestimation of the RSL position
6.2. Predicted vs observed RSL changes in the western Mediterranean by the ICE- 5G (VM2) isostatic model. At ~8.3 ka the RSL was ~6 m
higher than in northern Spain (Fig. 6, #1). On-going vertical movements
-1
There is a generally good qualitative agreement between the pre- (Serpelloni et al., 2013) do not exceed ±0.5 mm a with slightly higher
-1
dicted and the observed RSL changes. In regions where the record subsidence rates (b2mm a ) near Valencia (Fig. 2); these data meant