Stratigraphy and chronology of the Cala Mosca site, SW Sardinia (Italy)

Abstract The relict beach deposit of the Cala Mosca marine terrace is considered an important section of Marine Isotopic Stage (MIS) 5e (ca. 125 ka) sea-level highstands. Analysis of the stratigraphy and sediments of the deposit indicates the presence of a composite marine terrace comprising two superimposed marine units, luminescence dated to the MIS 5e (137 ± 7, 134 ± 7 ka) and MIS 5c (92 ± 6 ka) substages. The stratigraphic superimposition of the two highstands, both placed ~5 m above present sea level, agrees with other areas along the Sardinia coasts. The evident superimposition of two sea-level highstands and development of the composite terrace cannot be accounted solely by high-frequency sea-level oscillation that occurred within MIS 5 for the Mediterranean Sea. This suggests controversial, but significant, regional versus local tectonic activity occurred during the Late Pleistocene.


INTRODUCTION
Mediterranean Quaternary marine terraces have been the subject of research since the early years of the last century. Following a simplified scheme of a sea-level change cycle, each of these marine terraces is a gently seaward-dipping, wave-cut erosional surface developed on the substratum during transgression, and often covered by a veneer of shallow marine deposits during sea-level highstands. Typically, the terrace forms during a single sea-level highstand and provides insights to better understand the interaction between sea-level fluctuations and regional tectonic activity (Anderson et al., 1999;Zecchin et al., 2004Zecchin et al., , 2011Muhs et al., 2011;Gurrola et al., 2014;Pedoja et al., 2014;Ferranti et al., 2021). Under a combination of sea-level changes and low uplift rate, successive transgression-regression cycles could result in shallow-marine sediments reoccupying the same stratigraphic position over the same marine erosional surface (Cantalamessa and Di Celma, 2004). Consequently, successive shallow marine deposits could be superimposed and separated by a composite surface (Leonard and Wehmiller, 1992;Carobene and Ferrini, 1993;Ortlieb et al., 1996;Cantalamessa and Di Celma, 2004;Lucchi, 2009), resulting from multiple phases of erosion and deposition associated with different sea-level fluctuations. The younger transgressive wave ravinement surfaces (WRS) (sensu Zecchin et al., 2019, and reference therein) may obliterate or rework older subaerial and/or transgressive deposits of the previous cycle hosted by the terrace (Cantalamessa and Di Celma, 2004).
Because of the composite nature of the surface, the marine terrace is the result of multiple transgressive-regressive cycles (T-R) (Nalin et al., 2007(Nalin et al., , 2020Lucchi, 2009;Zecchin et al., 2011). Recognizing the composite nature of the marine terraces along the coast of areas considered "almost tectonically stable" may reveal: (1) different sea-level fluctuations, and (2) the interaction between sea-level changes and local tectonic activity.
The Mediterranean coasts preserve many sedimentary and geomorphological features correlated to the last interglacial period (LIG, MIS 5e), (127-117 ka) (Lisiecki and Raymo, 2005;Railsback et al., 2015). The MIS 5e sea-level markers (e.g., marine terrace and marine sedimentary covers for areas considered tectonically stable or affected by mild tectonic activity), have received particular attention in order to reconstruct past sea-level fluctuations and evaluate long-term tectonic rates along the coasts of the Mediterranean (Pedoja et al., 2014;Rovere et al., 2016;Benjamin et al., 2017;Stocchi et al., 2018). MIS 5e represents one of the substages within the MIS 5 stage (ca. 128-73 ka) during which the mean sea-level was 5-9 m higher than present (Kopp et al., 2009;Dutton and Lambeck, 2012;Dutton et al., 2015). Within MIS 5, there are two other interstadials when sea-level reached a relative highstand-MIS 5c (ca. 100 ka) and MIS 5a (ca. 80 ka). These, however, had sea-level peaks lower than MIS 5e, and ∼20 m below the present. A recent global compilation of field data with different glacio-isostatic adjustment (GIA) calculations placed the sea level during MIS 5c and MIS 5a in a range between −24 ± 2 m and −22 ± 1 m, respectively (Creveling et al., 2017).
The Island of Sardinia is located in the center of the western Mediterranean. It hosts well-exposed MIS 5e coastal deposits Sechi et al., 2020), and it is generally considered tectonically stable Ferranti et al., 2006;Rovere et al., 2016;Vacchi et al., 2016), evinced by lowintensity seismicity (Rovida et al., 2020;Gorshkov et al., 2021). The marine sequence of Cala Mosca (southwest Sardinia, Italy) is one of the well-known MIS 5e marker sites and has been considered the Tyrrhenian locus typicus since the beginning of the last century (Cerrone et al., 2021). Since then, it has been used as a proxy to calibrate the Pleistocene sea-level curve and to support the tectonic stability of Sardinia (Lambeck et al., 2004;Antonioli et al., 2006Antonioli et al., , 2007Antonioli et al., , 2017Rovere et al., 2016). However, generalized aseismic vertical motions along the island have been reported and validated by GPS measurements, relief inversion, and deep river incision (Buttau et al., 2011;Serpelloni et al., 2013;Antonioli et al., 2017;Quye-Sawyer et al., 2020). In addition, deformation of MIS 5e tidal notches (Mariani et al., 2009) and offsetting of Middle to Upper Pleistocene shallow marine deposits by normal faulting have been reported recently (Cocco et al., 2019;Casini et al., 2020). Furthermore, the pioneering stratigraphic study of Ulzega and Hearty (1986) described the presence of a thin carbonate crust and patchy thin soil layer separating two possible superimposed marine sedimentary units in the Cala Mosca marine succession. These are both observed to lie above the present sea level and interpreted to be formed due to a double sea level fluctuation within MIS 5e (Hearty, 1986;Hearty et al., 2007). This interpretation is similar to that observed in other semi-stable areas of the Mediterranean region (Cerrone et al., 2021, and references therein).
The aims of this paper are: (1) to review the stratigraphic meaning of the section cropping out at Cala Mosca, and (2) to perform new chronometric dating of this section using luminescence methods, and thereby to tentatively determine the interaction between sea-level changes and local tectonic activity.

GEOLOGICAL SETTING AND STUDY AREA
The Corsica-Sardinia block is located in the western Mediterranean basin. It represents a segment of the south-European plate that was detached with an anticlockwise rotation to the present position after opening of the Liguro-Provençal Basin (LPB) during the Oligocene-Early Miocene (Fig. 1a) (Carminati et al., 2012). The Sardinia block consists of a Variscan crystalline basement and Permian, Mesozoic, and Cenozoic volcano-sedimentary covers (Carmignani et al., 2016;Andreucci et al., 2017a) (Fig. 1b). During the Cenozoic, tectonic evolution related to Northern Apennine dynamics led to the formation of several basins bounded by a series of normal and strike-slip faults with several orientations, usually inherited from Variscan anisotropies (Oggiano et al., 2009). The southern sector of the island is dominated by the NW-SE oriented Campidano graben (Fig. 1b). The graben developed during multiple tectonic phases as a consequence of (1) transtensional tectonics related to the Northern Apennines collision in the Oligo-Aquitanian, (2) extensional tectonics due to the opening of the Balearic basin in the Burdigalian-Langhian, and (3) further extensional tectonics due to the opening of the southern Tyrrhenian back-arc basin during Plio-Pleistocene ( Fig. 1a) (Casula et al., 2001;Faccenna et al., 2002;Cocco et al., 2012Cocco et al., , 2013. Despite the graben being considered almost tectonically stable and subjected to a post-rift stage since the Middle Pleistocene (post 700 ka); Casula et al. (2001) and Cocco et al. (2019) reported that the northwestern side was offset by vertical tectonic movements during the last 100 ka.
The southern part of the graben is characterized by normal faults offsetting the pre-Quaternary basement forming a complex horst and graben system bounded to the west by the Campidano fault and to the east by the Sarrabus fault (Fig. 1b). The study area is on a NW-SE oriented horst between the Campidano and Molentargius grabens to the west and on the east, respectively (Fig. 1b). Two minor normal faults bound the horst-the Fangario fault on the west and the Molentargius fault on the east (Fig. 1b, c). Even if the area of Cagliari is considered stable under slow subsidence (0.2-0.5 mm/yr), Ferranti et al. (2008) suggested the possible existence of an active tectonic structure associated with the Fangario fault due to lateral-shear deformation. Recently, low-magnitude earthquakes (M <4.5) and slow vertical movement of 1-2 mm/yr have been recorded for the surrounding area of Cagliari and the north-western part of the island (Serpelloni et al., 2013;Rovida et al., 2020).
Presently, the Cala Mosca area consists of a small southward-opened embayment bounded by two promontories (Mt S. Ignazio and Mt S. Elia) (Figs. 1c, 2a) (Barca et al., 2005(Barca et al., , 2019. The Late Pleistocene marine succession crops out along the southern part of the bay, resting on a wave-cut platform carved into Miocene rocks. The succession is characterized by highly fossiliferous sandy to gravelly beach deposits (Fig. 2a). It is possible to observe the double gravel marine deposits reported in Ulzega and Hearty (1986) divided by a surface with a carbonate crust and patchy thin soil described by Hearty (1986) (Fig. 2b).

Facies analysis and stratigraphy
A facies analysis was performed on the Upper Pleistocene sediments cropping out along Cala Mosca bay. The deposits were subdivided into facies based on their sedimentary (e.g., lithology, grain size, sedimentary structures) and macrofossil characteristics (see Fig. 3 for more details on facies nomenclature). Facies interpretations are based on Pascucci et al. (2009) for shallow-marine deposits and Andreucci et al. (2014) for alluvial-colluvial deposits. Based on vertical and lateral facies relationships and lithology, facies were grouped into five facies associations (Fig. 3). Facies relationships are shown schematically in Figure 4. Twelve vertical logs were measured in the field (Fig. 5) and correlated according to sedimentary body lithologies and geometries. The unconformitybounded stratigraphic units (UBSU) were defined and divided by major bounding surfaces visible on the outcrop. Three UBSUs were recognized and named U3a, U3b, and U4, based on the nomenclature of Pascucci et al. (2014) (Fig. 5). Elevation above the present sea level was measured in the field with a metric tower rule. The chronological framework of the studied section was established by luminescence dating on shallow marine deposits.

Luminescence dating procedure
Three samples (CLM1, CLM2, CLM3) (Table 1; Fig. 5) from the beach deposits were collected for luminescence dating purposes as blocks (∼50 × 50 × 50 cm) and treated at the University of Sassari Luminescence Dating laboratory under subdued red light. The outer light-exposed part of the block (∼10 cm) was removed and used to determine the field and saturation water content and total dose rate. Water content was measured following Casini et al. (2020) and Sechi et al. (2020). The inner part was gently crushed, and the material was placed in 40% hydrochloric acid (HCl) and 30% H 2 O 2 baths to dissolve carbonate cement and organic matter. The material was dry-sieved, and a sediment fraction of 90-180 μm was chosen for further laboratory analysis. Quartz and K-feldspar grains were isolated via density separation (see supplemental material). All measurements were performed with an automated Risø DA-20 TL/OSL reader, with quartz luminescence detected through a 7.5 mm UV filter (Hoya U-340), and the K-feldspar IRSL with combined Schott BG-39 and Corning 7-59 filters.  (Gmelin, 1791) ex Strombus bubonius, and other warm-water fauna are found above a marine abrasion platform covered by >4 m of debris. For aeolian and colluvial deposits, Coltorti et al. (2007) reported a U/Th age of 122 ± 5 ka performed on Cladocora caespitosa. Erosional surface overlies the pre-Quaternary limestone (CGI) bedrock. (c) Field view of double gravel marine deposits (as reported in Ulzega and Hearty, 1986) divided by erosional surface with carbonate crust and patchy thin soil described by Hearty (1986). Dashed red line and arrow indicate the erosional surface. Figure 3. Simplified description of the observed facies in the studied area. Facies are labeled using the following nomenclature: G = conglomerate; S = sandstone; W = reddish sandy-siltstone; b = boulders to cobbles; c = cobbles to pebbles; d = pebbles to granules; e = medium-to very coarse-grained sand; w = reddish silty-clay matrix; l = laminated; t = trough cross-bedded; u = low-angle cross-bedded; t = tabular; m = massive; r = root traces; i = good imbrication; o = openwork; and f = high fossiliferous. Thus, facies labeled Seu corresponds to sandstone (S) medium to very coarse grained (e) and low-angle cross-beds (u). Facies interpretation based on (Massari and Parea, 1988;Pascucci et al., 2009;Andreucci et al., 2010aAndreucci et al., , 2014. Wd = inferred water depth range (in meters). Small-sized aliquots, 2 mm in diameter (i.e., containing some hundreds of grains), of quartz and K-feldspar grains were placed on stainless-steel discs using silicon oil as adhesive and stainlesssteel cups. The quartz OSL signal was measured with the standard SAR protocol Wintle, 2000, 2003), while K-feldspar grains were measured with the protocols, pIRIR 225 (Thiel et al., 2010(Thiel et al., , 2015Ferranti et al., 2021) and pIRIR 290 Thiel et al., 2012;Zucca et al., 2014a;Andreucci et al., 2017b;Casini et al., 2020;Sechi et al., 2020). To detect the portion of interest of OSL-quartz signal to calculate the equivalent dose (De), the first second (s) of signal decay was taken to which the following 2.6 s were subtracted as background (early background, i.e., Cunningham and Wallinga, 2010), whereas the first 5 s were subtracted by the last 20 s (late background) for pIRIR signal. At least 21-22 reliable aliquots were measured for the De estimation.
The OSL quartz signals have good luminescence characteristics, dominated by fast component and satisfactory laboratory performance (recycling close to unity and low recuperation). However, OSL signals were beyond the saturation limit for all samples, and thus not used for dating. The pIRIR 290 and pIRIR 225 signals have suitable luminescence characteristics, thermal stability, and good laboratory performance (recycling ratios within 10% of unity and recuperation <5%). A representative dose response curve for both protocols is shown in Figure SM1. Both protocols gave low fading values of 1.21 ± 0.31%/decade for pIRIR 290 and 1.51 ± 0.10%/decade for pIRIR 225. However, pIRIR 290 gave De ∼26% (average of three samples) larger than pIRIR 225 . In order to investigate the source of this difference, dose recovery tests for both protocols were carried out by adding different laboratory doses (close to natural) onto sun-bleached  Figure 2b. Due to modern coastal morphology, the studied sections and their correlation (12 stratigraphic logs) were arranged along an idealized SW-NE transect. Assuming this direction is roughly perpendicular to the paleoshoreline, the transect crosses the paleo-marine system. Labels such as Gem indicate facies described in the text and Figure 3. Major unconformity-bounded units (U3a, U3b, and U4) are delimited by their respective unconformity surfaces (WRS 1 , SU 1 +WRS 2 , and SU 2 ). Table 1. Summary of samples, stratigraphic unit interpreted, (n) the number of reliable aliquots used for dating, (OD; %) overdispersion of De expressed in percentage; K-feldspar pIRIR 225 doses, (RD) residual dose; (g 2days ) fading rate; and (Dr) total dose rate (Gy/ka). Estimated ages in ka and related MIS (marine isotope stage). For more details of determination of ages and dose rates, see the Supplemental Material (Supplementary Table SM1 aliquots of each sample (see supplemental material for details). The average recovery ratio after residual subtraction for pIRIR 290 is 1.26 ± 0.03 (average of all samples), suggesting that the laboratory-measured dose overestimates the given dose by ∼30% as observed. PIRIR 225 shows a good dose recovery ratio (0.95 ± 0.01), with low unbleachable residual (∼5 Gy) and fading rates, and thus it was the chosen protocol for dating. Given the low fading value, we decided not to correct the final pIRIR 225 ages for fading. A potential problem in dating is the incomplete resetting of the luminescence signal at deposition. However, by comparing signals with different bleaching ratios, it is possible to check the completeness of bleaching. When corrected for the dose overestimation pIRIR 290 De for each sample, pIRIR 290 De agreed with pIRIR 225 De, and therefore final ages, suggesting that both signals were potentially well bleached at the same degree prior to deposition (for more details, see the Supplemental Material and Supplementary Fig. SM1).
The total dose rate was calculated using a Risø low-level gas flow beta multi-counter GM-25-5 (GMBC; Bøtter-Jensen and Mejdahl, 1988) based on the proposed method by Cunningham et al. (2018). Measured mass and activity concentrations of K, U, and Th (and/or progeny) were obtained with inductively coupled plasma mass spectrometry (ICP-MS) at SGS Minerals (Canada) and high-resolution gamma spectrometry (HRGS, Denmark DTU) analysis. The activities obtained with HRGS were converted to mass concentration with conversion factors (Guérin et al., 2011(Guérin et al., , 2012 (Table SM1).
The advantage of the Cunningham et al. (2018) method is that the beta counter can provide a precise proportion of β dry dose rate for the sample when the concentrations of U, Th, and K for bulk sediment are known. The analysis also highlights that the K-contents (expressed in K%) in the studied samples are quite variable (Table SM1). To account for this potential variability, K% concentrations obtained for each sample by the three methods were averaged and used as a natural concentration of K% for bulk sediments to estimate the proportional contribution of this element on beta counts. (Table SM1).
However, the K-content of individual grains, and therefore the magnitude of the internal beta dose to the overall dose rate, is variable between grains, and it is difficult to estimate this for multiple-grain samples. Thus, as reported widely in literature, the internal contribution of β-dose is calculated on average K content of 12.5 ± 0.5% (Huntley and Baril, 1997) and internal abundance for 87 Rb mean value of 400 ± 100 ppm, and corrected for grain-size attenuation (average of 90-180 μm) (Huntley and Hancock, 2001).
The dry dose rates were corrected for moisture content, sample grain size, β-grain size attenuation (Guérin et al., 2012), and cosmic contribution. For moisture content, a 4% error on sample water content was used, and the cosmic contribution was based on the geographical position, altitude, and the measured sample burial depth following Prescott and Hutton (1994). Table SM1 reports the U, Th, and K concentrations, beta and gamma and measured dry and final total dose rates (for more details, see the Supplemental Material); final ages are listed in Table 1.

FACIES ANALYSIS AND DEPOSITIONAL PROFILES
Eleven facies were recognized and grouped into three shallowmarine (Tl, Rgb, Mpb) and one continental (Cs) facies associations (Fig. 3). A detailed description of facies associations is provided below, along with the depositional profiles for the studied deposits and a schematic of the lateral distributions of facies (Fig. 4).

Facies association: Tl transgressive lag
Facies association Tl is well developed along all the study area where it unconformably rests on the substrate or erosionally lies on facies associations Rgb and Mpb (rocky gravel beach, mixed pocket beach). The Tl association is dominated by clast-supported conglomerates with a sandy matrix and characterized from land to sea by facies Gcf and Gem (coastal pothole deposits; conglomerate lag) ( Fig. 3; Fig. 4).
Facies Gcf: coastal pothole deposit Facies Gcf crops out discontinuouslyonlyat the centerof the bay, where it fills narrow depressions and grades upward into facies Gem or Gcoi (Figs. 6, 7). The base is carved into the substrate and occasionally is draped by a thin (maximum 5 cm thick) carbonate crust. Gcf ranges in thickness from 20-50 cm and is composed of clast-supported conglomerate with highly fossiliferous, chaotic-to-poorly sorted, granule-to-coarse sandy matrix ( Fig. 6; Fig. 3). Clasts are sub-rounded to well-rounded (spherical or flat disc) pebbles and cobbles. Marine shells are mainly broken ( Fig. 6a-d).
Facies Gem: conglomerate lag Facies Gem crops out continuously along the study area and unconformably rests on the substrate or overlies facies Gcf (Fig. 6e, f). This facies consists of cm-thick (up to 1 m thick) layers characterized by a poorly organized openwork conglomerate of pebble-sized clasts ( Fig. 6e; Fig. 3). Clasts are sub-rounded, spherical, and/or disc-shaped pebbles to small boulders (Figs. 6e, f; 7e, f). Gravel composition reflects the local bedrock lithology. Carbonate clasts become dominant close to the limestone promontory and are pervasively bored by mollusks (lithofaga), poriferans, and algae. Clasts show faint seaward dipping imbrication, but often follow the morphology and inclination of the basal surface. In places, granules to very coarse sandy matrix with abundant fragmented marine shells fill the spaces among the clasts.

Interpretation of facies association Tl and depositional profile
The relationship of facies association Tl with the underlying erosional surface and overlying beach deposits (facies associations Rgb and Mpb, see below) suggests that Tl represents transgressive deposits draping a wave-cut platform formed by wave erosion and coastal-cliff retreat (Zecchin et al., 2019). In particular, facies Gcf represents the trapped sediment infilling erosional potholes or pools (Figs. 4, 6c)  . The Mediterranean Sea is microtidal, with tidal ranges ∼35-50 cm. Potholes form above the intertidal zone in the supratidal zones from 0.5-2 m above the present sea-level. Facies Gem, in contrast, is interpreted as the transgressive lag (sensu Parea, 1988, andZecchin et al., 2019, and reference therein) draping the shore platform in a fully submerged condition (subtidal zone; 0.5-2 m below present sea-level). Overall, facies association Tl shows a deepening upward trend indicated by the evolution from supratidal (facies Gcf) to subtidal conditions resulting from the marine transgression and shore retreat. Moreover, the depositional profile of Tl highlights the typical three-step evolution of a marine terrace that occurs during the early phase of transgression: (1) coastal cliff retreat and wave-cut platform development; (2) filling of the potholes in the intertidal zone, (3) mantling of the shore platform in subtidal conditions (Fig. 4) (Postma and Nemec, 1990).

Facies association Rgb: rocky coast gravelly beach deposit
Facies association Rgb crops out discontinuously as patches along the area and either rests unconformably on the substrate or overlies facies association Tl. Although Rgb does not show important lateral facies variations at the outcrop scale, different features reflect the local variability of substrate lithology (limestone or sandstones) and paleotopography (Fig. 4). Facies association Rgb is dominated by pebbly to bouldery clast-supported conglomerates and characterized from land-to-sea by facies Gcoi, Gbo, and Gtf (Fig. 3).

Facies Gcoi: Extraordinary gravelly berm
Facies Gcoi is observable in the northern inland part of the bay with E to W orientation. It consists of 20-70 cm thick tabular, slightly seaward dipping conglomerates of moderately to well-sorted, wellrounded, disc-shaped rounded cobbles and spherical oblate pebbles. Clasts have seaward dipping imbrication and an openwork clastsupported framework (Figs. 6-8). In places, it contains disseminated large (outsized) sub-rounded clasts, disarticulated chaotic organized broken marine shells, and rare patches of granules.

Facies Gbo: gravel beach face
Facies Gbo consists of a 1.5 m maximum thickness clastsupported conglomerate. Clasts range in size from pebbles to boulders showing various degrees of reworking, with a general openwork arrangement and slight traces of landward imbrication (Fig. 7). This facies is well preserved at the center of the bay, and  at outcrop scale a slight seaward dip is observable. Generally, clasts range in size from pebbles to cobbles with variable shapes from subangular to well rounded, and scattered small subrounded boulders can be observed (Fig. 7a-c). Where the facies is well expressed, it may grade upward into 50 cm thick structureless coarse-grained conglomeratic sandstone with rounded granules and pebbles, and dispersed sub-rounded cobbles and boulders. Whole and broken marine shells are occasionally organized in layers with faint sandy cross lamination (Fig. 7d, e).

Interpretation of Facies association Rgb
Although there is a distinctive lateral and vertical trend, the general gravel characteristics and the relationships with the underlying facies association Tl indicated that Rgb represents a gravelly beach system developed along the rocky shore ( Fig. 4) (Sanders, 2000;Felton, 2002;Felton et al., 2006;Bizzarri, 2010;Sechi et al., 2013Sechi et al., , 2020. Facies Gcoi, dominated by moderate-to well-sorted openwork gravel (rounded and disc-shaped clasts) with a clear seaward dipping imbrication, represents the gravelly berms that most likely accumulated at the inland part of the marine terrace (supratidal gravel berm; Pascucci et al., 2009). Moreover, Gbo and Gtf are characterized by spherical-/rolled-shaped and/or outsized clasts in seaward dipping strata, which allow interpretation of this facies as a gravelly beach face (Figs. 4, 8d, e). Facies Gbo and Gcoi show moderate to good stratal organization, indicating that this beach system developed over a gently inclined and wide shore platform. By contrast, Gtf is interpreted as the beach face formed at the base of the cliff proximal to the cliff-talus system and fed both by the colluvial debris flows and marine sediments deposited during storms (Fig. 8a, c) (Sanders, 2000;Sechi et al., 2020). This facies does not show a good stratal organization suggesting that the beach face accumulated at the cliff base on a narrow marine terrace on hard limestone substrate.

Facies association Mpb: mixed pocket beach association
Facies association Mpb widely crops out on the eastern side of the study area, where it displays a maximum thickness of ∼4.5 m and lies above facies association Tl (Fig. 9). Towards the SW, closer to the high cliff of S. Ignazio promontory, Mpb laterally grades into gravel-dominated deposits of facies association Rgb (Figs. 6, 7). Facies association Mpb is dominated by mixed sandy to gravelly well-stratified strata and characterized from land-to-sea by facies Semr, Gci, Seu, and Sedt ( Fig. 3; Fig. 4).
Facies Semr: backshore Facies Semr consists of a 1-2 m thick, poorly structured, wellsorted coarse-to medium-grained sandstone with rare and scattered rounded limestone pebbles. Pervasive plant root phytoturbation almost destroys the primary sedimentary structures. However, traces of flat laminations can still be discerned (Fig. 9b, d). Sedimentary structures become more evident moving seaward, and Semr grades into facies Gci and/or Seu.
Facies Seu: foreshore Facies Seu consists of 1-1.5 m thick seaward dipping low-angle/ parallel cross-laminated, well-sorted sandstone with scattered and/or aligned rounded pebbles, cm-thick layers of rounded granules, and broken or complete marine shells lying upside down. Heavy minerals are often aligned, forming a few millimeter thick dark laminae (Fig. 9c, d). This facies is occasionally interlayered with strata of facies Sedt.

Facies Sedt: upper shoreface
Facies Sedt is a 30-50 cm thick wavy sinuous, often trough-cross bedded coarse-to medium-grained sandstone interlayered with 10-20 cm thick layers of rounded granules and pebbles rich in fragments of seashells and sparse small, disc-shaped cobbles (Fig. 9d).

Interpretation of Facies association Mpb and depositional profile
The relationship of facies association Mpb with the underlying deposits of facies association Tl and its internal strata architecture suggest that it can be interpreted as a mixed sandy/gravelly pocket beach system developed on a rocky shore (Fig. 4). Facies Semr represents the backshore part of the mixed beach system grading landward from the extraordinary storm berms zone toward the incipient dune field. The low-angle cross-stratification of facies Seu indicates that these deposits formed in the foreshore zone of the mixed beach system dominated by swash and backwash (Postma and Nemec, 1990;Pascucci et al., 2009). The alternation of bioclastic-rich and black laminae is related to winnowing by fairweather waves, while clast-supported conglomerates (Gci) represent berm accumulations during storms (Sechi et al., 2013 (Fig. 8e). The pervasive lithofaga boring testifies that clasts spent a significant period underwater before being carried onshore by major storms. Finally, facies Sedt is characterized by trough crossstratification and sandy/gravelly bars most likely developed in the upper shoreface dominated by longshore currents. The occasional planar layers of granules and/or cross-stratified beds are interpreted as related to major storm events. The rock fragments were reworked from the nearby limestone cliff and/or carved from the sandstone shore platform. Facies association Mpb shows a clear shallowing-upward trend with trough-and-bar strata (upper shoreface) passing gradually into low-angle, cross-stratified sandy/gravelly bodies (foreshore and berms), and finally into structureless sandstone deposits (backshore).
Overall, Mpb is a prograding (normal regressive phase) mixed sandy/gravelly beach system (Fig. 4). The lateral variation from facies association Mpb to a gravelly beach system of Rgb is most likely due to the interplay between the paleotopography of the bay and wave currents. In particular, along the bay flanks and at the toe of the cliff, a gravel-dominated beach system developed, whereas the mixed sandy/gravelly deposits tend to be preferentially accumulated at the center of the bay, as occurs on several modern pocket beaches .

Facies association Cs: Cliff slope deposits
Facies association Cs crops out along the study area with a maximum thickness of ∼8 m on the southwestern side and rests unconformably on the bedrock or facies association Rgb (Fig. 8a). Cs is mainly formed by alternation of facies Gdw and Wed (Fig. 3).

Facies Gwd: Cliff slope debris flow and rockfall deposits
Gwd is composed of meter-thick tabular strata or meter-wide lenses of normally graded matrix-supported breccia (Fig. 8a). It occurs as strata inclined at >30°, characterized by disorganized fining-upward, clast-supported framework with reddish to brownish sandy-silt matrix. Clasts reflect the cliff bedrock lithology ranging in size from pebbles to boulders, which are angular to subangular shaped.
Facies Wed: Cliff slope colluvial deposits Facies Wed is characterized by a 0.5-1 m thick tabular or lenticular layer of medium-to fine-grained sand to siltstone with aligned very angular pebbles to cobbles at the base. Angular dispersed boulders may be present (Fig. 8a).

Interpretation of facies association Cs and depositional profile
Facies association Cs is interpreted to represent rock falls along the cliff flank, alternating with some generation of colluvial/incipient soil accumulated in concave-up depressions .

LUMINESCENCE CHRONOLOGY
The pIRIR 225 ages derived for the specific samples are summarized in Table 1 and Figure 5. Samples CLM1 and CLM2 were collected on the northeastern side of the study area from the base (∼3.5 m above present sea level = asl) and the top (∼5 m asl) of mixed sandy/gravelly beach system (facies association Mpb; Logs 3 and 4 of Fig. 5). Derived pIRIR 225 ages are 137 ± 7 ka and 134 ± 7 ka, respectively, showing an excellent consistency with each other (within errors) and with the previously published U/Th and amino acid racemization (AAR) ages of 139 ± 10 ka, 138 ± 8 ka, and 122 ± 5 ka (Ulzega and Hearty, 1986;Belluomini and Delitala, 1988;Coltorti et al., 2007Coltorti et al., , 2015. Sample CLM3 was collected on the southwestern side at the base of the cliff beach deposit (∼1.8 m asl) of the gravelly beach system (facies association Rgb; Log 10 of Fig. 5) and gave a pIRIR 225 age of 92 ± 6 ka.

STRATIGRAPHY AND BOUNDING SURFACES
Lateral correlations along the study area with the facies association stacking patterns, key discontinuity surfaces, and luminescence ages indicate the presence of three major unconformity-bounded units: U3a, U3b, and U4 (Fig. 5).
The lowermost unit, U3a, rests unconformably on the Miocene sandstone substrate (ADP; Figs. 5, 9a) and is bounded at the base by a smooth erosional surface cropping out from the NE to the center of the study area (Logs 1-6, Fig. 5). Unit U3a is characterized at the base by deposits of facies association Tl mantling the erosional surface on ADP. On the northeastern side of the bay (Logs 1-5, Fig. 5), unit U3a consists of a 3-4 m thick, welldeveloped mixed beach system (facies association Mpb), showing a clear shallowing-upward trend (progradation) from the upper shoreface to the backshore deposits (Fig 4, 5, 10). The erosional surface is interpreted as a wave ravinement surface (WRS 1 ) developed during a transgression (sensu Zecchin et al., 2019) (Fig. 10).
From the NE to the central part of the bay towards the SW, a few cm-thick calcrete beds associated with a reddish palaeosol mantle the top of the U3a and mark a period of subaerial exposure (SU 1 , Logs 1-7, Fig. 5). At the center of the study area (Logs 6, 7, Fig. 5), U3a is characterized from the base upward by facies association Tl (transgressive lag) grading into a rocky gravelly beach system (facies association Rgb) (Figs. 4, 5, 8a-c). Unit U3a from the center toward the NE side is overlain by unit U3b, which does not crop out on the SW side of the study area (Figs. 5,7,8) and has luminescence ages of 137 ± 7 ka and 134 ± 7 ka (CLM1, CLM2, Table 1), and thus is referred to substage MIS 5e. U3b rests on, and is bounded by a second erosional surface at the base. U3b is observable along the entire study area, slightly increasing in elevation from SW (1 m asl) to NE (5 m asl; Fig. 5). U3b is mainly composed of transgressive deposits (facies association Tl;Figs. 4,5,7,8), and along the inland part of the bay by extraordinary gravel-berm relict deposits (facies Gcoi) (Fig. 9e, f). This deposit lies unconformably on the subaerial surface SU 1 at the top of U3a. The erosional surface on the NE side of the study area cuts the top of lithified U3a and forms erosional potholes (facies Gcf), which are draped by a few-cm thick carbonate crust and filled by facies Gcf (Fig. 9e, f). The second erosional surface represents a wave ravinement surface (WRS 2 ) related to a second sea-level transgression. The WRS 2 almost obliterates and reworks the first subaerial surface SU 1 . However, traces of SU 1 are still observable as calcrete crust. The interplay of the two surfaces is interpreted as a composite surface (SU 1 + WRS 2 ) (sensu Nalin et al., 2007) (Fig. 9f). On the southwestern part of the bay, U3b is bounded at the base solely by WRS 2, and it rests unconformably on the Miocene limestone substrate (Figs. 5, 8). In places, traces of pothole structures mantled by transgressive deposits (facies association Tl) can be observed at the base. Here, U3b mainly consists of ∼1 m of rocky cliff talus to incipient gravel beach deposits (facies Gtf, facies association Rgb; Figs. 4, 8). A luminescence age of 92 ±6 ka (CLM3) indicates that U3b was formed during the MIS 5c substage. The subaerial surface SU 2 separates the top of U3b from U4 (Figs. 5, 8a). The U4 is characterized by cliff slope deposits (Cs facies association) comprised of an alternation of debris flow and rockfall (Gwd) with meter-scale colluvial deposits (facies Wed) (Fig. 10a, c). Subaerial surface SU 2 most likely developed during a sea-level fall, and the overlying deposits (U4) could be associated with cold/glacial phases during MIS 5b or the following MIS 4-MIS 2.

ELEVATION OF PALEO SEA-LEVEL MARKERS
Luminescence chronology indicates that the shallow marine sediments of U3 and U3b formed as a result of sea level fluctuations during MIS 5e (135-115 ka, U3a) and MIS 5c (ca. 100 ka, U3b). The stratigraphic superimposition of MIS 5c on MIS 5e has been observed in other areas along the NW coast of Sardinia. In particular, the relict sandy beach deposits of MIS 5c are superimposed on MIS 5e deposits, which crop out as sandy beach deposits or algal ridge bioconstruction (deposit build-up by intertidal red coralline algae) (Andreucci et al., 2010a;Sechi et al., 2013Sechi et al., , 2018bSechi et al., , 2020Pascucci et al., 2014;Zucca et al., 2014b). Moreover, uplifted MIS 7 marine terraces and related beach cover placed above the present sea level between 0.6 ± 1 m and 2.9 ± 0.2 m has been reported by Andreucci et al. (2009) andCasini et al. (2020).
Cemented beach deposits have been used as relative sea level (RSL) indicators in many MIS 5e studies (Pedoja et al., 2014, and references therein). The paleo relative sea level (pRSL), Figure 10. (a) Global sea-level curve from Waelbroeck et al. (2002) and Shackleton (2000); for the MIS 5e plateau, we use data from Sechi et al. (2020); for Sardinia and for Holocene sea transgression, we use data from Pascucci et al. (2018) uncorrected for glacial isostatic adjustment (GIA) for the double relict beach deposits, is here estimated for Sardinia based on Rovere et al. (2016, equations [1]-[4]).
To calculate pRSL, equations consider different parameters of processes acting near modern mean sea level environments, which shape both rocky and sedimentary coasts (e.g., the tide amplitude, the elevation of sea level indicator [measured in the field] and the mean high water height [calculated average of the elevation of the lower and upper limit of water height]). The choice of upper and lower limits depends on the morphology of the deposit studied .
For our calculation, field elevations (E) of sea level indicators are based on the elevations measured using a meter tape and the facies and stratigraphic interpretation. In the absence of a modern analogue, to calculate some of the parameters needed for pRSL calculation, we considered the elevations of the wave ravinement surfaces as the wave-breaking depth (the point of interaction between approaching waves and sea floor) as the lower limit of beach deposits, and the relict beach berms (facies Gci) as the upper limits. An error of ±10% of elevation to the measurement method and the average significant wave height (Hs 1.2 m) along with the tidal range (0.35m Cagliari tidal gauge) was used to solve the equations (Stocchi et al., 2018). The average calculated pRSLs (GIA-uncorrected) for sea level markers for the NW coast of Sardinia are 2.15 ± 0.9 m (MIS 5c) and 3.3 ± 0.6 m (MIS 5e), respectively, while pRSLs calculated for Cala Mosca differ slightly from these and are 5.8 ± 1 m (MIS 5c) and 5 ± 1 m (MIS 5e), respectively.

DISCUSSION
The composite terrace, regional tectonics, and sea-level marker at Cala Mosca The revised stratigraphic analysis performed on the marine sedimentary deposits of the Cala Mosca marine terrace indicates the presence of two unconformably superimposed shallow marine depositional units (U3a and U3b) bounded at their base by two ravinement surfaces, WRS 1 and WRS 2 . WRS 1 cuts the top of the pre-Quaternary substrate, while WRS 2 cuts the top of U3a and is the base of the second unit (U3b). Unit U3a is partially reworked, and U3b lies at the same elevation as U3b (i.e., at ∼5 m asl). The interaction between the WRS 2 and the older SU 1 surface forms a composite unconformable surface (WRS 2 + SU 1 ). Recognition of the composite bounding (unconformities) surface led to the term for the marine terrace of Cala Mosca as a "composite marine terrace," characterized by two superimposed shallow marine units.
This stratigraphic framework agrees with that previously reported by Hearty (1986), who observed the presence of double beach deposits and interpreted them as the consequence of millennial-scale sea-level fluctuation within MIS 5e (Hearty et al., 2007). However, this hypothesis remains controversial owing to uncertainties of the corrected estimation of paleosea markers, post-depositional tectonic vertical motions, precise age estimation reliability, and the contribution of glacial isostatic adjustment (Barlow et al., 2018;Mauz et al., 2018;Stocchi et al., 2018).
(1)Is Sardinia as tectonically stable as previously considered? (2) Is the sea level curve of the Mediterranean region well constrained? (3) Does Sardinia record local sea-level fluctuations that differ from those at the global scale? If the island has been tectonically stable at least since the Pleistocene, one of the major causes of the differences between RSL indicators and the global sea level (GSL) curve might be related to GIA effect.
A compilation of field data related to the GIA rebound has placed the peak of MIS 5e sea level between 2-2.5 m at Cala Mosca (Stocchi et al., 2018). The modeling, however, was based on the assumption of elevation of the MIS 5e double beach interpretation of Hearty (1986). By contrast, the MIS 5c-and MIS 5a-dated sea-level indicators are less constrained than MIS 5e, and the lack of available GIA correction for this time period makes it challenging to compare the regional to global data for the Mediterranean Sea. Nevertheless, Creveling et al. (2017) placed these two between −24 m and +2 m, and between −22 and +1 m, respectively, using a large dataset of globally distributed sites. This uncertainty makes it difficult to place the Cala Mosca MIS 5c deposits into the global pattern of sea level.
Stratigraphic superimposition of MIS 5c on the MIS 5e sea highstands conflicts with sea-level fluctuations that occurred within the MIS 5 stage (Fig. 10a) and with what might be expected for a stable region such as Sardinia Ferranti et al., 2006;Rovere et al., 2016). Shallow marine deposits related to MIS 5c and stratigraphically superimposed on the MIS 5e deposits were observed in other areas along the Sardinian coast (Andreucci et al., 2010a;Sechi et al., 2013Sechi et al., , 2020. Moreover, a tilted MIS 5e notch and different faultbounded domains recording vertical movements offsetting shallow marine deposits of MIS 5e and MIS 7 have been reported for the northwest and east coasts of Sardinia (Mariani et al., 2009;Buttau et al., 2011;Cocco et al., 2019;Casini et al., 2020). Recent high-resolution DGPS measurements also showed a significant variability of vertical velocities along the western Sardinia coast, with a few sites recording localized uplift at moderate rates (1.6 ± 0.6 mm/yr to 1.0 ± 0.3 mm/yr; Antonioli et al., 2017) and occurrence of low-magnitude earthquakes (M <4.5) along the Campidano graben (Rovida et al., 2020).
The evidence suggests different possible scenarios.
(1) If Sardinia, as a whole block, was tectonically stable and affected only by slow subsidence (negative displacement) according to the worldwide standard sea-level curve and GIA model for the central-western Mediterranean , the MIS 5c marine terraces should be placed at −22 m or lower under continuous subsidence. The MIS 5e marine terraces should be placed at +5-9 m under stable conditions, or at lower positions under active subsidence (Fig. 10a). In this case, it is necessary to admit that highstand peaks (MIS 5c and MIS 5a) that occurred after MIS 5e are still not well constrained in amplitude and elevation (Creveling et al., 2017) and/or that the Quaternary RSL curve in the Mediterranean is characterized by unrecognized highfrequency sea-level oscillations below the sensitivity of presently available dating methods (Pasquetti et al., 2021). (2) If, by contrast, we assume a continuous regional uplift (positive displacement) as the driving factor in rising the MIS 5c to the present, the MIS 5e beach deposits would have risen to a higher elevation (∼27-31 m) with a 0.27 mm/yr rate with a down-stepping terrace staircase pattern (Ferranti et al., 2021).
Because the two highstands, MIS 5e and MIS 5c, differ in time, amplitude, and position, we have to infer that Sardina is not an entirely stable block, but is characterized by fault-bounded domains that have been affected by local vertical movements (negative and/or positive), with respect to the regional setting. This scenario might include syn-depositional interaction between positive and negative vertical movements triggered by the Quaternary tectonic evolution of the Campinano graben domain (Ferranti et al., 2008;Cocco et al., 2013Cocco et al., , 2019. The combination of local subsidence due to lateral-shear deformation of the southern sector of the basin with slow regional uplift led to differential vertical movements of the structural horst and graben forming the Cagliari area (Valensise and Pantosti, 1992;Doglioni and Prosser, 1997;Thompson and Parsons, 2016).
At first, syn-depositional MIS 5e negative vertical movement ([subsidence + high fault slip rate] > regional uplift) offset the Cala Mosca horst downward and brought the MIS 5e deposits to the MIS 5c sea-level position (i.e., 22 m below the present). Later, positive (low regional uplift > [subsidence + slip rate]) uplifted the horst and the composite terrace to the present sealevel position (Fig. 10b).
In conclusion, the chronostratigraphic framework of the Cala Mosca marine section, together with other areas along the Sardinia coast, mismatches with the model of Quaternary regional stability of the whole Sardinia block and suggests that the island is characterized by several fault-bounded domains that were affected by local vertical movements during the Pleistocene (Mariani et al., 2009;Cocco et al., 2013Cocco et al., , 2019Casini et al., 2020). Although this model seems to be the simplest explanation and our preferred alternative, different questions remain unaddressed, and further investigations need to be carried out at the regional scale.

CONCLUSIONS
Detailed facies analysis and stratigraphic interpretation of Pleistocene terraced marine deposits of Cala Mosca have allowed us to recognize the presence of a composite marine terrace. This is characterized by two unconformably superimposed depositional marine units (U3a and U3b) bounded at the base by two different ravinement surfaces (WRS 1 and WRS 2 ). Surface WRS 1 cuts the top of the pre-Quaternary substrate, while WRS 2 is formed at the top of U3a and bounds the bottom of the second marine deposit (U3b). During deposition of unit U3b, it partially reworked U3a, and reoccupied almost the same stratigraphic position now at ∼5 m above present sea level. Luminescence chronology (pIRIR 225 protocol) dated U3a to MIS 5e substage (137 ± 7, 134 ± 7 ka), which is in good agreement with independent ages obtained with the U/Th and AAR methods, and pIRIR 225 protocol dated unit U3b to MIS 5c (92 ± 6 ka).
The superimposition of the two units and their stratigraphic position above present sea level suggest different scenarios in which high-frequency sea-level oscillation and tectonic activity that occurred during MIS 5 have to be considered. This leads to alternative hypotheses that might explain the present chronostratigraphic framework. (1) The two MIS 5c and MIS 5a highstands are still not well constrained in amplitude and elevation and the well-established and widely accepted/documented position of MIS 5e substage. (2) The Quaternary relative sea level curve of the Mediterranean region is characterized by unrecognized highfrequency oscillations, mostly occurring below the sensitivity of the presently available dating methods. (3) Late Quaternary deposition in Sardinia was controlled by localized vertical uplift and subsidence.
Whichever is the preferential hypothesis, recognition of composite unconformity surfaces and the measured chronological data for the Cala Mosca Pleistocene marine deposits indicate that the site cannot be linked to a single sea-level highstand, and thus used to reconstruct sea-level fluctuations for MIS 5e. In this scenario, recognizing erosional structure and bounding surfaces such as composite surfaces is needed to discern a single marine terrace from one concealing more than one sea level event. Composite marine terraces may highlight undetected local tectonic activity that must be considered when using relict beach deposits as paleo sea-level indicators.
Supplementary Material. The supplementary material for this article can be found at https://doi.org/10.1017/qua.2022.45 Financial Support. This research was supported by funds and financial support to SA, VP, and DS provided by a grant from Reagione Autonoma Sardegna: L.R. 7/2007, Bando 2017-Cambiamenti climatici e neottettonica -La Sardegna un continente semi stabile (P.I. V. Pascucci). Partial funding to VP has been provided by Fondo di Ateneo per la Ricerca 2020 and to SA by Project FdS 2020.