Back to Uluzzo – archaeological, palaeoenvironmental and chronological context of the Mid–Upper Palaeolithic sequence at Uluzzo C Rock Shelter (Apulia, southern Italy)

The tempo and mode of Homo sapiens dispersal in Eurasia and the demise of Neanderthals has sparked debate about the dynamics of Neanderthal extinction and its relationship to the arrival of H. sapiens. In Italy, the so‐called ‘Transition’ from Neanderthals to H. sapiens is related to the Uluzzian technocomplex, i.e. the first archaeological evidence for modern human dispersal on the European continent. This paper illustrates the new chronology and stratigraphy of Uluzzo C, a rock shelter and Uluzzian key site located in the Uluzzo Bay in southern Italy, where excavations are ongoing, refining the cultural sequence known from previous excavations. Microstratigraphic investigation suggests that most of the deposit formed after dismantling of the vault of the rock shelter and due to wind input of loess deflated by the continental shelf. The occasional reactivation of the hydrology of the local karst system under more humid conditions further contributed to the formation of specific layers accumulating former Terra Rossa‐type soil fragments. Superposed on sedimentary processes, strong bioturbation and the mobilization and recrystallization of calcite have been detected. Optically stimulated luminescence (OSL) ages from Uluzzo C Rock Shelter are congruent with previously published radiocarbon ages obtained on shell beads and tephrachronology from adjacent sites preserving the Uluzzian technocomplex such as Grotta del Cavallo, confirming the onset for the Uluzzian in the area to ca. 39.2–42.0 ka. The OSL chronology from Uluzzo C also provides a terminus post quem for the end of the Mousterian in the region, constraining the disappearance of the Neanderthals in that part of Italy to ≥46 ± 4 ka.


Introduction
The Middle to Upper Palaeolithic (MP/UP) transition corresponds to the period between 50 and 40 ka BP, broadly identifying the time span when modern humans migrated into Europe and Neanderthals disappeared (Benazzi et al., 2011(Benazzi et al., , 2015(Benazzi et al., , 2020Douka et al., 2014;Higham et al., 2014;Fewlass et al., 2020;Hublin et al., 2020). Tracking the time and mode of the MP/UP transition is pivotal to understanding the biocultural processes that gave rise to the first major global replacement of populations and the establishment of humankind today, as indeed after 39 ka only modern humans inhabited Europe (Higham et al., 2009(Higham et al., , 2010Benazzi et al., 2011Benazzi et al., , 2015Benazzi et al., , 2020Hublin, 2015;Fewlass et al., 2020;Hublin et al., 2020). Hence, there has been a resurgence of interest in the technological transitions observed in southern Europe in recent years (i.e. Italy and Greece; Higham et al., 2009Higham et al., , 2014, and particularly for the Uluzzian, a lithic techno-complex associated with modern humans in the Italian peninsula, stratigraphically superposing the Mousterian, associated with Neanderthals (e.g. Benazzi et al., 2011;d'Errico et al., 2012;Moroni et al., 2013;Marciani et al., 2020). The taxonomic reassessment of modern human teeth retrieved from the Uluzzian deposit of Grotta del Cavallo (Benazzi et al., 2011), coupled with a clarification of the stratigraphic sequence (Moroni et al., 2018) and its geochronology Zanchetta et al., 2018), makes the Uluzzian technocomplex a crucial element to unravel the biocultural processes that occurred during the MP/UP transition in Europe.
There are very few well-preserved and well-documented (in terms of stratigraphic coherence, chronological control and cultural evidence) archaeological sites preserving a stratigraphic sequence with Late Mousterian and Uluzzian assemblages and thus encompassing the MP/UP transition. In addition to Grotta del Cavallo, the most notable examples include Grotta Castelcivita (Moroni et al., 2013), Grotta La Cala (Martini et al., 2018), Grotta di Fumane and Riparo Broion (Peresani et al., 2014(Peresani et al., , 2016(Peresani et al., , 2017(Peresani et al., , 2019, and Grotta Roccia San Sebastiano (Collina et al., 2020). A further important sequence preserving the Late Mousterian to Uluzzian transition is the Carlo Cosma or Uluzzo C Rock Shelter, one of the sites where the Uluzzian technocomplex was defined. This site, located a few metres from Grotta del Cavallo ( Fig. 1), was excavated in the1960s (Borzatti von Lowerstern, 1965, 1966Borzatti von Löwerstern and Magaldi, 1966), and almost forgotten by the scientific community for about 50 years. This paper discusses preliminary results of the reexcavation at Uluzzo C Rock Shelter ongoing since 2016 (Fiorini et al., 2018).

The archaeological context
Uluzzo C Rock Shelter is located on the western side of the Apulian Coast (southern Italy), on the Ionian Sea (40°9′27.84″ N, 17°57′35.34″E). It lies in the middle of the Uluzzo Bay, south to Grotta di Uluzzo, located downstream of the Uluzzo Tower, in the Parco Naturale di Portoselvaggio (Nardò).
Uluzzo C Rock Shelter now appears as a central hall with a smaller cavity on its right side, characterized by the absence of archaeological deposits, probably washed-out by a recent reactivation of the karst hydrological network.
During its first exploration (1961)(1962)(1963)(1964)(1965)(1966), the stratified deposit of the cave yielded rich lithic assemblages dating from the Mousterian to the Bronze Age, including the Uluzzian. Unfortunately, the upper levels, containing the signature of human presence in the Bronze Age, were destroyed by plunderers, and little archaeological sediments from that period have survived.
During the first excavations (1964)(1965)(1966), a 2.6 × 1.5-m test trench was opened in the middle of the deposit, reaching a depth of 2.5 m. In 1966, the original trench was extended by 1 m 2 , and at the end of the campaign, the bottom of the trench was about 8 m asl and had removed the central part of the deposit, identifying 22 spits within nine archaeologically distinct layers, labelled A to I by Borzatti von Löwerstern (Table 1).
According to Borzatti von Löwerstern and Magaldi (1966), layers H and I at the bottom of the deposit are analogous to the marine conglomerate identified a few metres from the cave made by large boulders perforated by lithodomes and covered by a conglomerate with small clasts. Moreover, Borzatti von Löwerstern and Magaldi (1966) suggested that the original  floor of the cave corresponds to the last Marine Isotope Stage (MIS) 5e transgression event, based on the presence of several specimens of marine gastropods (e.g. Strombus sp.), at around 8-10 m asl. This constrains the age of the deposit having MIS 5e as a terminus post quem, as it happens for most of the Mousterian deposits of the region (Mastronuzzi and Sansò, 2002;Mastronuzzi et al., 2007;Spinapolice, 2012Spinapolice, , 2018a.

Materials and methods
The current investigation at Uluzzo C started in 2015 with a preliminary inspection, and the excavation began in 2016 (Fig. 2). This paper includes the geoarchaeological and geochronological data generated in the course of four excavation campaigns conducted between 2015 and 2018. Related archaeological materials are described in Silvestrini et al. (this volume). The whole archaeological sequence has been sampled for optically stimulated luminescence (OSL) dating and sedimentological and microscopic analyses. Five OSL samples were obtained from the archaeological section, and sand-sized quartz grains (180-212 μm grain size) extracted using standard laboratory procedures (Wintle, 1997). Equivalent dose (De) values were obtained on the single-grain level using the singlealiquot regenerative-dose (SAR) protocol (Murray and Wintle, 2000) and appropriate quality assurance criteria applied to obtain single-grain De distributions (Murray and Wintle, 2000;Duller, 2003;Jacobs et al., 2011). Dose rates were obtained via a combination of in situ gamma counting (Mercier and Falguères, 2007;Guérin and Mercier, 2011) and beta counting via a GM-25-5 low-level beta counter (Bøtter-Jensen and Mejdahl, 1988).
The whole sequence was sampled for sedimentological (on bulk samples) and micromorphological (on undisturbed and orientated blocks) analyses. A detailed description of the methods used for excavation, geoarchaeological analyses, and dating is available in the Supporting Information (SI).

Field evidence and investigated units
One of the major goals of the reappraisal at Uluzzo C was an updated understanding of the archaeological and sedimentological sequence (Figs. 2 and 3).
Layers A-C correspond to the Upper Palaeolithic and were almost completely removed by Borzatti von Löwernstern, except for the sediment leaning against the NW-facing wall (Fig. 4). Archaeologically, we do not have additional data from layers A-C, because we did not excavate them systematically. We decided not to remove these layers but leave them as evidence of the original cave filling, and for eventual future analyses, we exclusively sampled these layers for micromorphology and OSL dating.
Cleaning and analysis of the exposed stratigraphic section ( Fig. 4) allowed us to identify the original archaeological layers left after Borzatti von Löwerstern's archaeological operations.
Thus, we have new data on the archaeological occupation in about half of the original surface of the rock shelter (for us Sector A) (Figs. 3 and 4), corresponding to Layer C of Borzatti von Löwerstern's stratigraphy at the beginning of our excavation.
Therein, we removed the uppermost (SU 6) and disturbed stratigraphic units (SU 16) contaminated with reworked sediment from the original excavation and bioturbated by recent vegetation (Fig. 2).
We stratigraphically excavated SU 3-19, which form the original Layer C, corresponding to the Final Uluzzian, while SUs 20-24 represent the beginning of Layer D and are under investigation. Nowadays, we consider layer D to be Uluzzian because of the characteristics of the lithics both from Borzatti von Löwerstern and our excavations (for details, see Silvestrini et al., this volume). Most of the sequence is composed of CaCO 3 -cemented layers; the exposed layers (C and D) also display the highest grade of cementation. The SUs investigated have a silty matrix and a reddish colour, are characterized by the abundance of faunal fragments (in some cases determinable; see also Silvestrini et al., this volume) and Uluzzian lithic industry. In particular, several flint bladelets were found, confirming the bladelet production as characteristic of this 'modern' techno-complex (for details, see Silvestrini et al., this volume). A variety of mammal species were identified, such as Cervus elaphus and carnivores such as Vulpes vulpes (Silvestrini et al., this volume).
Sector B corresponds today to the bottom of the sequence excavated by Borzatti von Löwernstern. The stratigraphic section of Sector A is accessible via Sector B, and the geological and OSL samples of layers E to G (including the η and σ layers) were taken from this stratigraphic section (Fig. 5). Archaeologically, the G and η layers are known from former studies (Spinapolice, 2012(Spinapolice, , 2018a as very dense occupation areas, with Mousterian technology and in situ fireplaces (Borzatti von Löwerstern and Magaldi, 1966).
Furthermore, the external area of the cave (Sector C, Fig. 4) was explored to test for the intact archaeological deposit. However, only reworked sediment from the old excavation was retrieved, lying directly on the bedrock.

Geoarchaeological evidence
The sedimentary sequence Exploration of the remnants of the stratigraphic sequence after the excavations of the 1960s almost confirmed the organization in SUs/Layers already described by Borzatti von Löwerstern and Magaldi (1966). The stratigraphic sequence here considered and described in Fig. 5 is what remains of previous excavations, and it is located in the innermost part of the rock shelter; layer labelling suggested by Borzatti von Lowerstern (1966) is adopted here for clarity.
The top of the stratigraphic sequence is sealed by a laminated, 10-20-cm-thick flowstone, probably formed during the Holocene. The uppermost part of the anthro-pogenic infilling of the rock shelter is Layer A, corresponding to a slightly reworked deposit. Layer A overlies a series of stratigraphic layers (B-F) bearing UP lithics, consisting of breccia-like layers with different amounts of fine matrix and in some cases displaying weak evidence of sedimentary structures. Layers B-F are cemented to moderately cemented by CaCO 3 . Layer B is a weakly cemented, sand and silt matrix-supported breccia. Layer C is a cemented clast-supported breccia, with very little silty matrix, alternating laminae with common silty matrix; rock fragments are centimetre-sized and platy and accumulated according to weakly inclined bedding planes. The following Layer D is similar to Layer C, but the matrix is less abundant. Conversely, Layer E is a matrix-supported breccia, cemented by CaCO 3 and abundant silt; at the bottom of Layer E a thin layer of silt and fine sand is present. Layer F is again a clast-supported breccia with a silty to sandy matrix and interlayered by at least two thin layers of silt and fine sand. The amount of fine sediments increases towards the bottom of the layer, where large blocks from the collapse of the roof of the rock shelter are present. The transition to the lower part of the sequenceand the MP occupation of the siteis marked by an abrupt (erosive) transition between the bottom of the blocks of F and σ. The latter is a massive, moderately CaCO 3cemented clay-rich matrix with some weakly rounded rock fragments. The following level η is a massive, silty to silty-clayey deposit interlayered by ash and charcoal-rich lenses; very few to common rock fragments are interspersed in the matrix along with charcoal fragments and centimetre-scale CaCO 3 nodules. The transition between η and G is an alignment of blocks collapsed from the roof of the rock shelter. Layer G is massive and moderately cemented, and silt and clay dominate the matrix, where very few rock fragments and CaCO 3 nodules are present. Occasionally, rock fragments form discontinuous alignments. According to Borzatti von Lowerstern (1966), these stratigraphic layers are in contact with an MIS 5e beach deposit, which actually is not evident along the sequence.

Sedimentological analyses
Chemical analyses on sediment samples showed mostly constant pH and total organic carbon (TOC) for the whole stratigraphic sequence (Table 2).
In particular, pH fluctuates around the mean value of around 8 by no more than 0.5 pH units. In absolute terms, the sediment is moderately alkaline, attributed to the high quantity of carbonates available in the environment and derived from the parent material and CaCO 3 recycling over time. TOC is always below 1% of the mass of the sediment.
Grain size analysis performed on the fine fraction of a selection of layers showed a general common trend in size distributions along the sequence. The main trend shows an abundance of fine sand and silt: silt notably represents between 40 and 60% of the total sediment. The finer portion of the sands are the second most represented group, with percentages ranging from 25 to 40%, while coarser fractions do not amount to more than 15% of the total. The least represented group is the finest: with the notable exception of layer σ, where it reaches 25%, clay never amounts to more than 10% on the total sediment. The graph of Fig. 6 shows two evident inflection points representing the main changes in relative abundance.
The first and clearest inflection point marks the steep increase at the passage from coarse to fine sand. The second, less pronounced, is located roughly around the shift from coarse to fine silt and shows a progressive decline of abundance towards clay. All investigated layers follow this model, with some minor exceptions. The most visible is in layer σ, which is shifted far more to finer material and shows a larger proportion of clay compared to silt, while sands remain similar to the general trend. On a smaller scale, layers ηb and Ga follow the opposite trend, where sand slightly increases while clay does not diverge. Layer B trends instead to a steeper curve, with a larger abundance of silts and a decrease in the sand and especially clay.

Micromorphology of thin sections
Essential micromorphological properties are summarized in the following parts and reported in Table 3, while the major observed features are illustrated in Fig. 7.
Layer A is composed of a heterogeneous yellowish-brown granular groundmass. It shows signs of transport from different sources based on the presence of subrounded quartz grains and rounded reddish pedorelicts originated from older and more developed soils. Bioturbation is evident around channel voids. Anthropogenic elements include bone, shell fragments and ash concentrations. The ash concentrations is often indistinguishable among the widespread calcite recrystallization pedofeatures, which in the upper part of the level form proper cementation, also associated with a general darker colour of the micromass. Unlike the previous Layer A, Layer C shows a higher degree of compaction, with a more coalescent groundmass and a blocky structure. However, the content is heterogeneous, with an increase in the frequency of quartz grains and bone fragments, while transported pedorelicts disappear. Signs of bioturbation are still evident. Microcharcoalssometimes grouped in horizontal lines/lensesand phosphatic nodules appear. Calcite recrystallization is less common than in the upper part of the sequence.
Within Layer D, two different microfacies types can be recognized (D1 and D2). The first is mostly granular, less compacted and yellowish, with more evidence of bioturbation; it is composed of a more homogeneous groundmass with less ash and recrystallized calcite. D2 is darker and almost massive, with few horizontal voids, rich in bones and calcite cementation pedofeatures. Quartz grains and phosphate nodules are similar to Layer C. D1 and D2 microfacies types are chaotically distributed in the slide. Layer E is characterized by a homogeneous and massive dark microstructure, with widespread calcite cementation in the groundmass. Quartz grains, phosphatic nodules and bone fragments are similar to the upper levels; pedorelicts are absent. The bottom of the level changes considerably. Here, the microstructure is composed entirely of quartz grains and rounded pedorelicts with a more developed fabric than the ones found above. Bioturbation is absent, and other elements such as bone fragments, charcoal and phosphate nodules are rare; around these components, calcite infillings cement all the porosity. Calcite infillings remain dominant in Layer F, where CaCO 3 fills voids around the dark granules that make up the partially compacted groundmass effectively cementing the layer. (2) sandy unit with rock fragments; (3) sandy unit with rock fragments and bioturbation; (4) matrix-to clastsupported breccia displaying weak oblique lamination (sandy-silty matrix); (5) clast-supported breccia (sandy-silty matrix); (6) large blocks due to roof collapse; (7) clay-rich deposit with scarce rock fragments; (8) slightly weathered silty deposit with scarce rock fragments; (9) silty deposit with scarce rock fragments; (10) CaCO 3 -cemented deposit; (11) weakly CaCO 3 -cemented deposit; (12) charcoal fragments and ash-rich lenses; (13) CaCO 3 nodules and/or concretions; (14)   Coarse components are less frequent: we noted a decrease in quartz grains, as well as pedorelicts (similar to those at the base of Layer E), bone fragments and ash accumulations. Below this level, all traces of bioturbation disappear. The underlying Layer σ is blocky and more compacted, with brighter colour and very few voids. Quartz grains become frequent and show more variability in their dimensions than in the upper part of the sequence. Pedorelicts such as those above (Layers E and F) are very few, and sometimes weathered into Fe and Mn nodules. Dark amorphous organic material is locally present in the groundmass; frequent ash concentrations can also be found. Calcite recrystallization is limited compared to other layers, except for the cemented portions of the level, which are almost completely filled by calcite crystals.
Layer η shows similar features to σ. Except for the less massive microstructure, the other components are found in similar frequencies, including quartz fragments and amorphous organic material, which is replaced at its base by clusters of microcharcoals. Phosphate concentrations reappear here, both as nodules and as coatings around bone fragments.
Calcite cementation progressively disappears downwards, where isolated concentrations of calcitic ash remain clearly visible. At the base of the level, several fragments of flint are present.
At the bottom of the section, Layer G is homogeneous and shows a dark granular microstructure inside dominant calcite crystals. The presence of other elements is limited to few fragments of quartz and rare bone and shell fragments, as well as pedorelicts similar to those above.

Geochronology
Optically stimulated luminescence dating OSL dating was conducted on the single-grain level for all OSL samples (samples ULOC 1-5), and the corresponding equivalent dose (De) values and dose rates, as well as optical ages are summarized in Table 4 (compare SI for methodological details). Typical OSL decay curves for a bright and a dim grain in response to the natural and a regenerated dose are shown in Fig. 8. For both grains the OSL signals decay rapidly Figure 6. Cumulative grain size curves for selected layers from the Uluzzo C stratigraphic sequence (the x-axis is on a logarithmic scale). The distribution of most of the matrix of stratigraphic units overlaps the mean cumulative grain size curve of Italian loess elaborated by Cremaschi (1987).
[Color figure can be viewed at wileyonlinelibrary.com]    to background level within the initial~0.5-1 s of stimulation time and the natural and regenerated OSL curves reveal similar shapes. These observations suggest that the quartz OSL signals are predominantly fast component-dominated, confirming that the SAR protocol is suitable for these samples (Wintle and Murray, 2006).
The overdispersion values (also σb values; Galbraith and Roberts, 2012) for samples ULOC 2-5 vary from 26 to 33% (Table 4) and De distributions are rather symmetrically distributed around the central weighted mean (Fig. 9 and SI), suggesting the presence of a single dose component. Indeed, σb values of~20% have been commonly reported for 'ideal' (i.e. well-bleached and undisturbed) sedimentary quartz samples (e.g. Olley et al., 2004;Arnold and Roberts, 2009). Hence, we opted for the central age model (CAM) to obtain an environmental burial dose for these samples . The resulting optical ages for Uluzzian layers C, D and E (samples ULOC 3, 4 and 5) range from 38.1 ± 2.2 to 42.7 ± 2.6 ka (grand weighted mean 40.6 ± 1.4 ka). These three OSL ages have relative age uncertainties of 6% and overlap within error, thus constraining the timing of Uluzzian occupation of this rock shelter to MIS 3. The clustering of OSL ages for the Uluzzian layers also suggests that the duration of Uluzzian occupation of the site did not exceed a few millennia (eventually lasting from ca. 39 to 42 ka), and was potentially much shorter than that.
ULOC 1 (Mousterian layer G) has an optical age of 46 ± 4.0 ka and is in stratigraphic order. This OSL age must be interpreted as the minimum age for the Neanderthal presence and site occupation because we did not obtain additional OSL samples from deeper parts of the Mousterian complex.
Sample ULOC 2 (layer B) was a relatively loose sand deposit that was penetrated by rootlets at the time of sampling. This sample also showed a high overdispersion value (75.5%; Table 4), and a distinctive group of low De values can be seen in the corresponding radial plot (Fig. 9). This sedimentological observation in combination with the high overdispersion value of this sample suggests that this deposit and thus the singlegrain De distribution is affected by younger intrusive grains that were worked in from above. The FMM (finite mixture model) was used to isolate this low De component from the main (i.e. high) De component David et al., 2007), for which an optical age of 22.4 ± 2 ka has been obtained (Table 4), and which is regarded as the depositional age for this sand deposit.

Radiocarbon dating
A single bone from US16 was pretreated at the Department of Human Evolution at the Max Planck Institute for Evolutionary Anthropology (MPI-EVA), Leipzig, Germany, and graphitized and dated at the Mannheim AMS laboratory (Lab Code MAMS) (Kromer et al., 2013). The AMS 14 C result of the sample from US16 is 25 310 ± 77 14 C BP (Table 5). In calibrated ages this date ranges from 29 820 to 29 320 cal BP at 68.2% probability and from 29 900 to 29 250 cal BP at 95.4% probability. The 14 C result was calibrated using IntCal20 in the OxCal v4.4 program (Bronk Ramsey, 2009;Reimer et al., 2020). The result fits well with the structure of US16, characterized by soft silty-sandy brown sediment, associated with bioturbation, also testified by the presence of modern glass fragments noted during the excavation. The bone is also associated with shell and pottery fragments.

Discussion
One of the most notable results presented here is the revised version of the archaeological section that was originally been excavated and described in the 1960s. We were able to identify all the original archaeological layers and describe them in much finer detail (Fig. 5), and have provided the first numerical age control for this archaeological section, based on OSL dating (Table 4). The micromorphological analysis, together with high-resolution excavation methods, allowed us to understand both the origin of the deposit and the way human groups may have used this specific rock-shelter.
Field and analytical data suggest that the formation of the sequence is related to the accumulation of sediments via a range of depositional processes, mainly during MIS 3 to 2. As reported by Borzatti von Löwerstern and Magaldi (1966), the bottom of the sequencenot reached during our explorationprobably consists of MIS 5e beach deposits, which are quite common along the shores of the Italian Peninsula (Rovere et al., 2016). Yet, the accessible part of the sequence (Layers A-G) was formed mostly by (i) the accumulation of sediments from the dismantling of the vault of the rock shelter due to spallation, (ii) the wind input of sediments probably deflated by the continental shelf that was subaerially exposed during MIS 4-2, (iii) the occasional reactivation of the hydrology of the local karst system and (iv) to a minor extent by human agency (anthropogenic debris). Superposed on the sedimentary processes, several post-depositional processes modified the deposits; among them, the most remarkable were bioturbation and the mobilization and recrystallization of calcite. Overall, the aeolian origin (Borzatti von Löwerstern and Magaldi, 1966) stated for most of the sediment is confirmed, and most of the layers show deposits with characteristics compatible with loess accumulation: the occurrence of allochthonous sub-rounded quartz grains detected in thin sections and the typical sigmoidal shape of cumulative grain-size curves (Fig. 6), dominated by the fine sand and silt fractions (see Cremaschi, 1987;Zerboni et al., 2015). Moreover, windblown sediments originating from the shelf or the reworking of tephra are quite common in the region (Cremaschi and Ferraro, 2007;Hirniak et al., 2020). As common in cave archaeological sites and considering sedimentological and micromorphological evidence, we can also consider humans as agents actively contributing to the accumulation of mineral grains and organic debris in the rock shelter (see Cremaschi et al., 2014;MacPhail and Goldberg, 2018).
From a palaeoenvironmental point of view, the dismantling of the roof and walls of the rock shelter and the accumulation  (Cremaschi and Ferraro, 2007;Hirniak et al., 2020). Such conditions have been recently confirmed for southern Italy by rainfall reconstruction based on isotopic data from a speleothem that grew during the last glacial cycle in southern Apulia (Columbu et al., 2020). Moreover, the breccia of layer C also suggests decreased temperatures and cooler environmental conditions at the time of deposition. The OSL ages of layer C-E overlap within error, yielding a grand weighted mean age for these Uluzzian layers of 40.6 ± 1.4 ka. This age thus also constrains the age of the breccia of layer C: it is thus likely that breccia deposition occurred during one of the stadial events (i.e. Greenland Stadial 9, 10 or 11; ca. 38.2-2.2 ka; NorthGrip Members, 2004;Rasmussen et al., 2014). A more precise temporal correlation between this breccia deposit and cold climatic conditions recorded in the Greenland ice cores is not possible with current dating precision. The breccia deposit, in fact, has many analogies with the grèzes litées (sensu Ozouf et al., 1995) and other frost-related coarse-grained deposits that form in rock shelters under glacial conditions (Laville et al., 1980). We thus interpret Unit C as a grèzes litées-like deposit that probably accumulated via freezing-thawing cycles during a cool climatic phase. Typical of the Uluzzo C Rock Shelter is the identification of the occasional accumulation of former soil fragments (pedorelict probably from cave infilling or Terra Rossa-type soils) in Layer E. Such pedorelicts formed after the occasional reactivation of the karst system, which was probably promoted by more humid conditions compatible with short interstadials or increased precipitation over the region. The speleothem record from Grotta Cucù Cave (Columbu et al., 2020), for instance, suggests several short-term increases in rainfall during MIS 3; according to the chronology of the infilling of Uluzzo C, one of the wet phases preceding the H4 event (for instance at ca. 43 or 47 ka BP; Rasmussen et al., 2014;Columbu et al., 2020;Allard et al., 2021) may have promoted  the reactivation of the karst system. The Uluzzo C sequence is sealed by a flowstone covering highly bioturbated sediments. Bioturbation and precipitation of the flowstone occurred after the abandonment of the rock shelter and are compatible with Early Holocene wet environmental conditions (Columbu et al, 2020).
Furthermore, based on our high-resolution excavation methods, we obtained insights into the range of site activities.
The stratigraphic layers identified (C and D) correspond to occupation levels where an anthropogenic origin is predominant. Both knapping activities and activities linked to prey consumption and the exploitation of game and possibly marine resources are evidenced on site (see Silvestrini et al., this volume). This is confirmed by the evidence of trampling in Layer D (regular horizontal void patterns and alignment of coarse mineral constituents, see Fig. 8), which we therefore interpret as a palaeosurface. The presence of microscopic bone fragments and phosphates possibly linked to the weathering of animal bones is coherent with the results from ancient DNA showing the occurrence of Cervidae and Equidae and the possible occasional presence of Hyenidae (Silvestrini et al., 2021 this volume). These data together show an intense occupation in the rock shelter and a wide set of activities performed on site, at least for the time span corresponding to the Uluzzian occupation (Layers C and D). Unfortunately, the portion of the excavation removed in former test pits does not allow us to comment on the internal organization of the rock shelter during the Pleistocene. The shelter was possibly used a basecamp by one or more small bands during hunting or foraging who were undertaking subsistence activities as far as the sea (as attested by the presence of marine shells) and gathering lithic raw materials both locally and further away (Silvestrini et al., this volume). The predominant presence of a recent Uluzzian at Uluzzo C is confirmed by technological analysis of the lithics (Silvestrini et al., this volume); in fact, the schemes of the chaînes operatoires, having as a main goal the production of bladelets and flakelets, characteristic of the Uluzzian technocomplex (contra Mussi et al., 2006;Zilhão et al., 2015), strongly matches the European UP , and can be broadly compared with layer D of Grotta del Cavallo (Moroni et al., 2018). Data from Uluzzo C thus confirm the Uluzzian as having fully modern behaviour expressed through techno-typologically complex lithic toolkits and highly adaptive subsistence strategies, a so-called 'Uluzzian package' Sano et al., 2019;Arrighi et al., 2020a,b,0c;Romandini et al., 2020).
The OSL time constraints generated for the Uluzzian layers at Uluzzo C (i.e. a grand weighted mean OSL age of 40.6 ± 1.4 ka, calculated from the OSL ages of samples ULOC 3, 4 and 5) match the chronology of other Uluzzian sequences of the region, particularly the one from Grotta del Cavallo, which wasuntil nowthe solely securely dated site in the region containing the Uluzzian technocomplex and human remains, marking a time frame for the arrival of modern humans in southern Italy. It has been suggested that during the~55-48 ka phase of MIS 3 natural environments were favourable for Homo sapiens to enter Europe Hublin et al., 2020), which is also in line with the earliest occurrences of H. sapiens in Turkey (Üçağızlı cave: Güleç et al., 2002;Kuhn et al., 2009), Lebanon (Ksar Akil: Copeland, andYazbeck, 2002;Yazbeck, 2004;Douka et al., 2013) and Israel (Manot Cave: Hershkovitz et al., 2015). Zanchetta et al. (2018) proposed a scenario for the human occupation of Apulia, based on the succession of tephra layers, and illustrated their hypothesis based on the correlation between regional human peopling and main climatic events, using Grotta del Cavallo as a reference site. At Uluzzo C, the sequence lacks macroscopic tephra layers, and further analyses are required to examine whether cryptotephras, which were identified at other Mediterranean archaeological sites (Hirniak et al., 2020), and are helpful in improving chronological precision, are present at Uluzzo C. Zanchetta et al. (2018) note that the sedimentation rate at Grotta del Cavallo was highly variable, and comparably thick layers (e.g. Cavallo layer M) alternate with highly condensed ones (Cavallo D and E); this is probably the case at Uluzzo C. Despite the stratigraphic diversity, the chronological constraints for the end of the Mousterian in the area are overlapping for the two caves: 45.5 ± 1.0 ka at Grotta del Cavallo, and OSL-dated to 46 ± 4.0 ka at Uluzzo C (ULOC 1, Table 4). This fits with the accepted range for the disappearance of Neanderthals from the European continent, constrained to ca. 47-40 ka BP (e.g. Müller et al., 2011;Hublin et al., 2020).
Furthermore, Uluzzo C Rock Shelter, together with Grotta del Cavallo (Fabbri et al., 2016) and Oscurusciuto (Boscato et al., 2011), are now the only sites in southern Italy with secure time constraints for the presence of Neanderthals (see for a review Spinapolice, 2018a). Additional OSL sampling from the 2.5 m-thick Mousterian complex at Uluzzo C, from layer G, providing a minimum age of 46 ka gives further temporal insights into the MP and UP occupation history of the area.
At Uluzzo C the Uluzzian has been OSL dated to 39-42 ka (ULOC 3-5, Table 4), with 42 ka marking a terminus post quem for Uluzzian site occupation (i.e. 42 ka is the earliest possible date for occupation of the rock shelter by groups of modern humans). At Grotta del Cavallo, tephra stratigraphic investigation identified the Campanian Ignimbrite (CI) layer (40 Ar /39 Ar dated at 39.850 ± 0.14 ka; Zanchetta et al., 2018); this acts as a terminus ante quem for the Uluzzian occupation (i.e. modern humans were present no later than ca. 39.9 ka ago). Although Grotta del Cavallo and Uluzzo C are only a few tens of metres apart from each other, we have been unable to positively identify the CI at Uluzzo C. Zanchetta et al. (2018) suggest that the Uluzzian transitional complex in the region is constrained between ca. 46 and 39.9 ka. The grand weighted mean OSL age for the Uluzzian at the Rock Shelter Uluzzo C in combination with the numerically dated CI at Grotta del Cavallo allows us to refine the time window of the Uluzzian transitional complex and suggests that modern humans arrived in southern Italy some time between ca. 39.9 and 43 ka.
Yet, the current stratigraphic and geochronological resolution at both sites hampers us in constraining whether the Uluzzian persisted significantly beyond 39.9 ka, i.e. beyond the CI eruption.
The post-Uluzzian human occupation at Uluzzo C has been investigated by Borzatti von Löwenstern and is present at Grotta del Cavallo and Grotta di Uluzzo (Borzatti von Löwenstern, 1963Palma di Cesnola, 1963. At Uluzzo C, Layer B, in the upper part of the sequence, has been OSL dated to the Last Glacial Maximum (ULOC2, 22.4 ± 2 ka, Table 4), and the radiocarbon date on a bone fragment from SU16 to 25 310 ± 77 14 C BP to the possible extent of the human occupation until MIS 3-2 will need further investigation. Disturbed Bronze Age evidence was also reported (Borzatti von Löwenstern, 1965), so possibly the occupation was interrupted in the Early Holocene and resumed only in the Late Holocene.

Conclusions
The arrival of H. sapiens in Europe and the demise of the Neanderthals are two potentially inter-related events that are hotly debated by the archaeological and palaeoanthropological community (e.g. Benazzi et al., 2011;Higham et al., 2011;Fewlass et al., 2020;Hublin et al., 2020). This debate is characterized by a wide range of assumptions, ranging from the idea of the relative inability of Neanderthals to survive competition with H. sapiens (e.g. Sano et al., 2019), to an assumed cognitive comparability of Homo neanderthalensis with H. sapiens (e.g. d 'Errico et al., 1998;Banks, 2013, 2015).
Our new chronological, geoarchaeological and sedimentological data from Uluzzo C (i) constrain the Uluzzian at the Uluzzo C Rock Shelter to 39-42 ka, (ii) thus confirming the time span of the Uluzzian in southern Italy, and (iii) contribute the growing body of evidence suggesting an early arrival of modern humans in the southern Mediterranean. Collectively, our new data from Uluzzo C (see also Silvestrini et al., this volume for analysis of lithic artefacts, faunal remains and ancient sedimentary DNA) provide us with more detailed insight into H. sapiens behaviour and organization of their technology and economy for a period when a crucialyet little understoodpopulation turnover occurred.

Supporting information
Additional supporting information can be found in the online version of this article.