Human occupation continuity in southern Italy towards the end of the Middle Palaeolithic: a palaeoenvironmental perspective from Apulia.

After the last interglacial [Marine Isotope Stage (MIS) 5e] Europe was affected by several harsh climatic oscillations. In this context southern Italy acted, like the rest of peninsular Mediterranean Europe, as a 'glacial refugium', allowing the survival of various species, and was involved in the spread of 'cold taxa' (e.g. woolly mammoth and woolly rhino) only during the coldest phases (MIS 4 and MIS 2). Both late Mousterian and early Upper Palaeolithic sites testify to a human occupation continuity in southern Italy and especially in Apulia in this time span. Here we present a focus on three key Apulian Palaeolithic sequences (Grotta di Santa Croce, Riparo L'Oscurusciuto and Grotta del Cavallo - layers F-E) jointly spanning from the late MIS 4 to the demise of Neanderthals around 43 ka. Novel chronological, sedimentological and zooarchaeological data are discussed for the first time in the light of the palaeoenvironmental information provided by recent analyses carried out on a speleothem from Pozzo Cucù cave (Bari) and the results of the magnetic susceptibility analysis from Riparo L'Oscurusciuto. This integrated reading allows a better understanding of the role played by the Apulian region as both a refugium for late Neaderthals and a suitable habitat for the early settling of modern humans.

high frequency

low frequency

Modern Humans

Marine Isotopic Stage

magnetic susceptibility

number of identified specimens

Pozzo Cucù

Introduction

Recent genetic studies indicate that European Neanderthals faced drastic demographic crises during the last phases of their history. Genetics points to possible local extinctions of some Neanderthal lineages and to a later spread of other groups into the newly unpopulated areas. Since available data are still scarce, the dynamics of population turnovers after 60 ka bp are still unclear (Hajdinjak et al., 2018), but the proposed scenarios imply a demographic reduction during Marine Isotope Stage (MIS) 3, when Europe experienced rapid climatic oscillations on millennial to centennial scale (Hemming, 2004; Long and Stoy, 2013). Regardless of what the more reliable re‐colonization scenario is, the existence of one or more refugia – which fostered resilience of Neanderthal populations – must be postulated. Refugia guaranteed survival of taxa and preservation of biodiversity when Pleistocene glacial–interglacial cycles shaped the European ecological setting (Ordonez and Svenning, 2017; Morales‐Barbero et al., 2018). Several areas, like Iberia, the Italian peninsula, southern Balkans, the Carpatian basin and the Atlantic French margin, are considered to have acted as important refugia for more termophilous/temperate species, whose geographical distribution contracted southwards during the coldest phases (e.g. Sommer and Zachos, 2009). Noteworthy, refugia should have also guaranteed the persistence of the species during warmer phases (e.g. Bennett et al., 1991).

The Italian peninsula played this role during MIS 2 both for flora (Petit et al., 2002) and for fauna (Sommer and Benecke, 2005), and available data (although scanty) suggest similar circumstances also in earlier phases (MIS 5–3). For instance, the persistence and survival of several arboreal taxa occurred during the whole Late Pleistocene during both stadial and interstadial phases (e.g. Allen et al., 2000; Ricci Lucchi, 2008; Allen and Huntley, 2009). Considering continental large mammals, on which data are more abundant, Late Pleistocene faunal associations from Italy fall within the Late Aurelian age, an assemblage difficult to define due to the regional fragmentation and the lack of representative faunal units (Masini and Sala, 2011). In a first phase, corresponding to the Last Interglacial and to the first MIS 5 cooling stages, species more adapted to warm conditions spread throughout the Italian territory. Hippo (Hippopotamus amphibius) and straight‐tusked elephant (Palaeoloxodon antiquus) were recognized in northern Italy at Grotta della Madonna dell'Arma (Liguria, archaeological cave site, Valensi and Psathi, 2004), in central Italy at Campo Verde (Latium, archaeological open‐air site, Marra et al., 2018) and in southern Italy at Grotta Grande (Campania, archaeological cave site, Ronchitelli et al., 2011) and Contrada Iannì (Calabria, archaeological open‐air site, Marra, 2009). Hippo was also found in layers 1 and 6 at Avetrana (Apulia, stratified karstic infill deposit, Petronio et al., 2008). Fallow deer (Dama dama) was attested to in all the above‐mentioned sites, as well as in northern Italy, at Caverna degli Orsi (Venezia Giulia, archaeological cave site, Berto and Rubinato, 2013) and possibly at Ciota Ciara, an archaeological cave site in Piedmont (identified as ‘cf. Dama’ in Berto et al., 2016) (Fig. 1). Data related to MIS 4 are scanty and less clear, also due to the lack of a good framework of chronometric studies. Commonly, this phase is said to be characterized by the decline and disappearance of hippo and straight‐tusked elephant, and by the appearance of cold stenothermal taxa: woolly rhino (Coelodonta antiquitatis) and woolly mammoth (Mammuthus primigenius). Regretfully, most remains of these large mammals are from undated open‐air sites, or from contexts whose chronological attribution is uncertain (e.g. for the woolly mammoth, MIS 4–3 at Riparo Tagliente, an archaeological cave site in Veneto – northern Italy, Reggiani and Sala, 1992; age >41 ka bp at Buca della Iena, archaeological cave site in Tuscany – central Italy, Pitti et al.,  1976). Evidence tentatively associated with MIS 4 seems to be present in the southernmost part of Apulia (southern Italy, woolly rhino at Grotta Romanelli, Pandolfi and Tagliacozzo, 2013). Warm climate taxa are also tentatively reported for this period, as for instance Paleoloxodon antiquus at Grotta Guattari (Latium, central Italy) associated with an age of 57 ± 6 ka (Schwarcz et al., 1991a; Stuart, 2005) or fallow deer at Arma delle Manie (Liguria, northern Italy, Valensi, 2009). MIS 3 well‐dated woolly rhino and mammoth specimens have been detected so far only in north‐eastern Italy at Grotta di Fumane (Veneto, archaeological cave site, layer A3, Tagliacozzo et al., 2013) and Settepolesini di Bondeno (alluvial deposits in Emilia, eastern Po valley, Beta‐128160: 40 884–35 912 cal bp; Gallini and Sala, 2001), respectively. Thus, during MIS 3, the area north‐east of the Apennine chain was still influenced by the arrival of cold taxa, while in the area stretching from the north‐western Tyrrhenian belt to southern Italy the presence of more thermophilous fallow deer was quite constant (Fig. 1). This taxon was found in archaeological cave sites from Liguria (Riparo Bombrini, Holt et al., 2019), Tuscany (Buca della Iena, Grotta La Fabbrica, Pitti et al., 1976; Grotta dei Santi, Spagnolo et al., 2020a), Latium (Grotta Breuil and Grotta del Fossellone, Alhaique and Tagliacozzo, 2000), Campania (Grotta di Castelcivita, Masini and Abbazzi, 1997; Grotta della Cala, Benini et al., 1997; Riparo del Poggio, Boscato et al., 2009) and Apulia (Grotta del Cavallo, Boscato and Crezzini, 2012a; Riparo L'Oscurusciuto, Boscato and Crezzini, 2012b). Fallow deer made its last appearance at Grotta della Cala (Campania, southern Italy) in the Aurignacian layers (Benini et al., 1997). Its extinction is in line with the reduction of the signal of temperate forests from the pollen record of central–southern Italy, which is counterbalanced by the increase of steppe‐related herbaceous taxa (Allen et al., 2000; Badino et al., 2020). The presence of cold taxa restricted to north‐eastern Italy makes the Italian Peninsula quite different from Iberia, where woolly mammoth was able to reach the southernmost regions during some of the coldest MIS 3 oscillations (Álvarez‐Lao et al., 2009).

Figure 1

‘Warm’ (Dama, Palaeoloxodon and Hippopotamus) and ‘Cold’ (Coelodonta and Mammuthus) large mammal taxa in Italy during MIS 5 and MIS 4–3. Circled area: geographical position of the archaeological sites analysed in this work (Nos. 10–12) and Grotta di Pozzo Cucù, (speleotheme, No. 8, white triangle). The digital elevation model is the European digital elevation model from the GMES RDA project (https://www.eea.europa.eu/data-and-maps/data/eu-dem#tab-originaldata/eudem_hlsd_3035_europe). The bathymetric model is from the European Marine Observation and Data Network. The map was generated using ArcGIS version 10.5. In A are represented both the minimum extension of the coastline during MIS 5e and a mean of the maximum estimated extension of the coastline (grey line) during the cooler oscillations of MIS 5. Sites: A (MIS 5): (1) Contrada Iannì, (2) Grotta Grande, (3) Avetrana, (4) Campoverde, (5) Grotta della Madonna dell'Arma, (6) Ciota Ciara, (7) Caverna degli Orsi, (8) Grotta di Pozzo Cucù. B (MIS 4–3): (8) Grotta di Pozzo Cucù, (9) Grotta Romanelli, (10) Grotta del Cavallo, (11) Riparo L'Oscurusciuto, (12) Grotta Santa Croce, (13) Grotta della Cala, (14) Riparo del Poggio, (15) Grotta di Castelcivita, (16) Grotta di Sant'Agostino, (17) Grotta Reali, (18) Grotta Guattari, (19) Grotta del Fossellone, (20) Grotta Breuil, (21) Grotta dei Santi, (22) Grotta la Fabbrica, (23) Buca della Iena, (24) Settepolesini di Bondeno, (25) Riparo Tagliente, (26) Grotta Fumane, (27) Riparo Bombrini. See text for references and explanations. [Color figure can be viewed at wileyonlinelibrary.com]

In this scenario, a number of dated contexts testify to Neanderthals’ persistence in peninsular Italy during the Late Pleistocene. Among the main sites attributed to MIS 5 are Riparo del Poggio (Boscato et al., 2009), Grotta Grande di Scario (Ronchitelli et al., 2011) and Riparo di Santa Caterina (Gambassini and Ronchitelli, 1998) in Campania. An MIS 5–4 site is Grotta Guattari (Schwarcz et al., 1991a) in Latium. Sites dated to MIS 3 are present in Tuscany (Grotta dei Santi, Moroni et al., 2019; Spagnolo et al., 2020a), Latium (Grotta dei Moscerini, Kuhn, 1995; Sant'Agostino, Tozzi, 1970; Schwarcz et al., 1991b), Campania (Riparo del Poggio, Boscato et al., 2009), Molise (Grotta Reali, Peretto et al., 2020) and Apulia (Grotta Romanelli, Sardella et al., 2014; Grotta di Santa Croce, Arrighi et al., 2009; Riparo L'Oscurusciuto, Boscato et al., 2011; Grotta del Cavallo, Zanchetta et al., 2018). From about 45k to 39k cal a bp (Higham et al., 2014), Neanderthals were replaced in Europe by Modern Humans (MHs). According to the current state of knowledge, the early dispersal of MHs into Europe followed two penecontemporaneous routes: the first, perhaps slightly older, through the inner part of the Balkan peninsula, is attested to by the recent identification of MH bone remains in association with an Initial Upper Palaeolithic assemblage at Bacho‐Kiro Cave in Bulgaria (Hublin et al., 2020). The other, towards Apulia (possibly along the Ionian fringe of the southern Balkan peninsula – Greece and across the Adriatic Sea), is related to the spreading of the Uluzzian technocomplex, whose occurrence is testified to in several localities. The main sites are located in Greece (Higham et al., 2014), Apulia (southern Italy, Ionian area, Benazzi et al., 2011; Moroni et al., 2013, 2018), Campania (southern Italy, Tyrrhenian area Martini et al., 2018; Collina et al., 2020; Marciani et al., 2020a), Latium (central Italy, Villa et al., 2018), Tuscany (central Italy, Villa et al., 2018) and Veneto (northern Italy, Peresani et al., 2019). When MHs reached southern Apulia (Grotta del Cavallo), the last Neanderthals were still present in the central part of the region (Riparo L'Oscurusciuto, Fig. 1) and the co‐existence of the two groups possibly lasted for several hundred years (Higham et al., 2014). Thus, Southern Italy, and more specifically Apulia, is a key region for studying the role of the environment in relation to both late Neanderthals and the newcomers. The aim of this work is to refine our knowledge about the ecological setting of Apulia during the last part of MIS 4 and the first half of MIS 3, by integrating faunal, chronological and palaeoclimatic data from four sites: Grotta di Santa Croce, Grotta del Cavallo, Riparo L'Oscurusciuto and Grotta di Pozzo Cucù (Fig. 1). We test the possible role that the Apulian region played, initially as a refugium for late Neanderthals, and later as the scenario of their replacement by MHs.

The sites

Grotta di Santa Croce (Bisceglie – Barletta‐Andria‐Trani)

Grotta di Santa Croce (Fig. 2) is a horizontal cave containing a rich archaeological deposit mainly located immediately in front of its entrance. Early research at this site was carried out between 1939 and 1958 (Cardini, 1954; Segre and Cassoli, 1987). In 1997, fieldwork was resumed by the University of Siena in collaboration with the local Heritage Office. Except for a test trench carried out inside the cave, the main excavation activity was focused on the talus of the external deposit, with an 18‐m‐wide and 1.5‐m‐deep trench (Fig. 3). Excavations did not reach the bottom of the archaeological deposit. Even if the peripheral part of the talus is about 40 m far from the present‐day cave entrance, in Palaeolithic times, the rocky vault most probably covered a large part of the extant outer deposit. Moving from the bottom of the already excavated part of the stratigraphy, the external sequence is subdivided into three main lithostratigraphic units: (i) reddish sandy levels yielding a rich assemblage of Mousterian lithic implements and animal remains, which is characterized by large limestone blocks testifying to the episodic collapse of the vault – the lithic assemblage is characterized by a discoidal reduction sequence; (ii) pinkish sandy layers affected by erosive events, with a combination of Mousterian and Protoaurignacian evidence; and (iii) greyish sandy/silty layers with few Gravettian implements (AMS Beta 147996: 29 526–28 446 cal a bp) (Boscato et al., 2006a, 2010; Arrighi et al., 2009; Ranaldo et al., 2017) (Fig. 3).

Figure 2

The studied sites. A: Riparo L'Oscurusciuto (photo by Paolo Boscato); B: Grotta di Santa Croce (photo by Paolo Boscato); C: Grotta del Cavallo (photo by Vincenzo Spagnolo); D: Grotta di Pozzo cucù (photo by Orlando Lacarbonara). [Color figure can be viewed at wileyonlinelibrary.com]

Figure 3

Stratigraphic sequences of the studied archaeological sites (C modified after Palma di Cesnola, 2001). [Color figure can be viewed at wileyonlinelibrary.com]

The presence of the Protoaurignacian in the pinkish sandy layers is also confirmed by a 14C date (AMS Beta 181166: 36 755–34 277 cal a bp). No structures, such as fireplaces, were found. Small mammals, birds and molluscs are very scarce (or even absent) as well as the vegetal remains.

Grotta del Cavallo (Nardò – Lecce)

Grotta del Cavallo (Fig. 2) opens in the Uluzzo bay. The cave is formed by a main chamber with a wide entrance facing NE. Excavations were directed by A. Palma di Cesnola (University of Siena) from 1961 to 1979, and later by other colleagues of the same University until 2019. A thick stratigraphic continental deposit lies on a marine beach and includes the Middle Palaeolithic, the Uluzzian, the Late Upper Palaeolithic (Romanellian) and the Neolithic (Palma di Cesnola, 1967). The upper part of the Mousterian sequence (layer FI‐III) lies between two tephra layers dated, respectively to 108.7 ± 0.9 ka (X‐6, layer G) and ~45.5 ka (Y‐6, layer Fa) (Zanchetta et al., 2018). Layer FII, which has been 14C‐dated to 47 900–42 100 cal a bp, yielded a Neanderthal deciduous tooth (Fabbri et al., 2016; Sarti et al., 2017). The Uluzzian sequence (layers E–D) is sandwiched between the tephra layer Fa and another tephra corresponding to the Campanian Ignimbrite (Y‐5, 39.85 ± 0.14 ka). The chronology of the Uluzzian package at Cavallo is also confirmed by 14C dates obtained in the Oxford laboratory (Benazzi et al., 2011; Higham et al., 2014). In layer EIII (Fig. 3) two deciduous teeth were found (Benazzi et al., 2011; Moroni et al., 2018). These represent the early evidence for MHs in Italy and one of the oldest in Europe. No data are available so far from bird, mollusc and vegetal remains, while the only data available for small mammals are from Dalla Valle (2008).

Riparo L'Oscurusciuto (Ginosa – Taranto)

Riparo L'Oscurusciuto (Fig. 2) is a key site for the study of late Neanderthal groups in southern Italy. It is located at the bottom of the ravine of Ginosa. Excavations (which are still underway) started in 1998 under the direction of the University of Siena in collaboration with the local Heritage Office. This site has a long‐term stratigraphic sequence in which the archaeological evidence is exclusively related to the Mousterian occupation (Fig. 3). The part of the deposit already excavated is dated between 55 ka (volcanic tephra, layer 14) and 45 335–40 711 cal a bp (14C date, bottom of layer 1) (Marciani et al., 2020b and references therein). Some extensively excavated layers have yielded important information about settlement strategies and space organization patterns of late Neanderthals. Layer 11 is a long‐term palimpsest characterized by the presence of several hearths, whose diameters range from about 25 to 50 cm. They have an SW–NE alignment and are associated with areas devoted to knapping and butchering. A possible sleeping area, connected to this phase, was detected in the NW corner of the shelter. Layer 9 displays a similar alignment of hearths allocated in sub‐circular shallow pits. By contrast, layer 8 does not show clear structures and was probably affected by slight erosional phenomena. The phase represented by layers 5–7 is characterized by a change in the space management: a single large hearth (about 2 m in diameter) was set in the NW corner of the shelter. Layer 4 is a thick palimpsest very rich in faunal and lithic remains. Several stones, some of them intentionally imported into the site, were found in the upper part of the layer (Boscato and Ronchitelli, 2007; Spagnolo et al., 2016, 2019, 2020b; Martini et al., 2020). Small mammal, bird, mollusc and vegetal remains are sporadic.

Grotta di Pozzo Cucù (Castellana Grotte – Bari)

This cave (not an archaeological site) opens into the Cretaceous limestone plateau belonging to the Calcare di Altamura Formation, and is located around 10 km from the Adriatic coastline (Fig. 2). The modern entrance, a narrow ~8‐m‐deep pit at ~270 m a.s.l., leads to a sub‐horizontal environment that stretches for ~300 m. A ~50‐cm‐long stalagmite from this cave has recently produced the first continuous Western European speleothem palaeoclimate record for the last glacial period (Columbu et al., 2020). By applying high‐resolution U–Th dating (n = 27) and δ13C–δ18O (n = ~2700), the Pozzo Cucù (PC) record allowed the determination of rainfall amount, moisture circulation and soil bio‐productivity from 106.0 ± 2.8 to 26.6 ± 0.9 ka, with average ±2σ age model uncertainty of around 1.3 ka. Average ±2σ age model uncertainty drops below ~1 ka over the period between ~50 and ~40 ka.

Materials and methods

Chronology at Santa Croce

Three herbivorous teeth recovered in the external talus of Santa Croce cave were analysed by the ESR/U‐series method. The following analytical protocol was used (Bahain et al., 2010): after its extraction, enamel was cleaned of any track of sediment or dentine using a dental drill, crushed and sieved to retrieve the 100–200‐µm grain‐size fraction, which was then split into 10 aliquots. One aliquot was kept as a natural reference and the nine others were irradiated using a panoramic source of 60Co (LNHB, CEA, CEN, Saclay, France) at doses ranging between 25 and 11 700 Gy. These aliquots were then measured by ESR spectrometry with a Bruker EMX spectrometer (band X, frequency modulation 100 kHz) at room temperature. The following parameters were used: microwave power of 1 mW, modulation amplitude of 0.1 mT, scan range of 10 mT, scan time of 4 min. At least four measurements for each aliquot were performed on different days. The ESR intensities of the g = 2.0018 enamel ESR signal were measured according to Grün (2000) and the equivalent doses D e were then extrapolated from the obtained dose–response data sets using an exponential + linear function (according to Shao et al., 2015).

U‐series analyses were performed on each dental tissue using the chemical protocol treatment described by Shao et al. (2011). Measurements were performed on a Neptune Multi‐Collector Inductively Coupled Plasma Mass Spectrometer at Nanjing Normal University, China (Shao et al., 2015). The radioelement contents of sediments associated with the teeth were measured by high‐purity low‐background gamma spectrometry, and conversion contents‐doses factors of Guérin et al. (2011) were used for the age calculation. For each dental tissue, Rn loss was estimated from both gamma and alpha/inductively coupled plasma measurements (Bahain et al., 1992). Additional in situ gamma measurements were performed with an Inspector 1000 Canberra gamma spectrometer using an NaI detector. The cosmic dose rate was estimated for each sample by the formula of Prescott and Hutton (1994) considering the depth of the samples in the sequence.

ESR/U‐series ages (Table 1) were estimated from the whole data set as well as dose‐rate contributions and U‐uptake parameters considering the US model (Grün et al., 1988) and using the ‘USESR’ software (Grün, 2009).

Table 1

ESR/U‐series data and ages of teeth from Santa Croce.

Samples	Tissue	U content (p.p.m.) (2σ)	234U/238U (2σ)	230Th/234U (2σ)	222Rn/230Th	Initial thickness (µm)	Removed thickness internal side (µm)	Removed thickness external side (µm)
SC535	Dentine	20.01 ± 0.69	1.025 ± 0.049	0.066 ± 0.009	0.191	1339 ± 167	35 ± 4	29 ± 4
	Enamel	0.43 ± 0.03	1.033 ± 0.092	0.130 ± 0.029	1.000
SC546	Dentine	94.88 ± 2.71	1.075 ± 0.059	0.109 ± 0.011	0.110	1467 ± 183	38 ± 5	26 ± 3
	Enamel	1.76 ± 0.08	0.952 ± 0.059	0.133 ± 0.013	0.261
SC549	Dentine	6.42 ± 0.26	1.068 ± 0.061	0.403 ± 0.080	1.000	1246 ± 156	29 ± 4	27 ± 3
	Enamel	0.28 ± 0.03	1.005 ± 0.128	0.403 ± 0.080	1.000

A k‐value (α efficiency) of 0.13 ± 0.02 (Grün and Katzenberger‐Apel, 1994).

The water content of the sediments was estimated after 1 week of drying in a 40 °C oven but, as the sampled section has been open for a long time, the obtained values were considered as minima and a representative value of 15 ± 5% was assumed for the age calculations. Water contents of 0 wt% in the enamel and 7 wt% in the dentine were used.

The beta dose attenuation was evaluated from the enamel part destroyed on each side of the enamel layer during the preparation process (according to Brennan et al., 1997).

The age uncertainty was calculated with a Monte Carlo approach (Shao et al., 2014).

Magnetic susceptibility analysis

Magnetic susceptibility (MS) analysis was performed at Riparo L'Oscurusciuto to obtain palaeoclimatic information, especially for the lower part of the succession currently not excavated. A huge part of siliciclastic sediments in caves and rock shelters derive from infiltration processes, through which fine‐grained sediments are carried from overlying soils (Iacoviello and Martini, 2012, 2013; Martini et al., 2018, 2020). In turn, pedogenesis during cool and arid climates produces sediments with low MS magnitudes, while warmer and humid climates yield higher MS magnitudes. As a consequence, MS data in cave and shelter sites can be an excellent palaeoclimatic proxy (Ellwood et al., 1998).

Samples (n = 22) were collected in the 5.7‐m thick siliciclastic succession (i.e. low‐resolution sampling). Two samples were not considered in this palaeoclimatic‐orientated study because they were exclusively made of volcanic ash (i.e. not providing palaeoclimatic information). Samples were sieved in the laboratory at 1 mm, following the procedure proposed by Ellwood et al. (2001): 10 g of the <1‐mm fraction of each sample is placed in a cylindrical bottle and MS is measured with a Bartington Instrument MS‐2 susceptibility meter combined with an MS2B dual frequency sensor. The instrument operates at 465 and 4650 Hz, thus providing the opportunity to measure low‐frequency (LF) and high‐frequency (HF) MS values (χmass). The difference between these LF and HF χmass values corresponds to the χfd (χfd = χmassLF − χmassHF), a value that has a close positive relationship with pedogenesis and highlights the MS signal of minerals produced during pedogenesis (cf. Maher, 1986; Liu et al., 2003).

Data elaboration and presentation follow the suggestion of Ellwood (2007), even if raw data are not smoothed using splines due to the low‐resolution sampling. Palaeoclimatic trends are expressed with the bar log format; that is, changes are assumed to be significant if the MS cyclical trends increase/decrease by a factor of two or more and if the change is represented by two or more data points, and consequently the highs and lows associated with these cycles are differentiated by filled (high MS values) or open (low MS values) bar logs (Ellwood, 2007).

Faunal remains

The study focuses on the already published faunal remains from Grotta di Santa Croce (excavation fieldwork 1998–2005), Grotta del Cavallo (excavation fieldwork 1980–1986) and Riparo L'Oscurusciuto (excavation fieldwork 1998–2011). All these sites were stratigraphically investigated with modern methodologies and sediments were water‐sieved with 1‐mm size mesh. In addition to the already published data (Boscato and Crezzini, 2006, 2012a,b; Boscato, 2017; Spagnolo et al., 2020b) new samples from Grotta di Santa Croce (layer 549, Ciullo, 2016) and Riparo L'Oscurusciuto (layers 4–7, Batini, 2018) are presented. The sample from Grotta di Santa Croce is from the 2011 fieldwork and is composed of 789 bone fragments. The sample from Riparo L'Oscurusciuto was recovered during the 2017 fieldwork, carried out in a stratigraphic baulk located in the northernmost part of the shelter. This sample is composed of 8659 bone fragments. Specimens were identified using the osteological reference collection at the University of Siena. The number of identified specimens (NISP) is used to describe the abundance of taxa. The calculation of the minum number of individuals has not been carried out due to the ongoing debate about its validity (Lyman, 2008, 2018). Ecological considerations on ungulate taxa are from Heptner et al. (1988).

Results

The obtained ages range between 57 ± 4 and 46 ± 3 ka. A Bayesian treatment of these results including stratigraphical and radiocarbon constraints was also realized using Chronomodel software (Lanos et al., 2016) (Fig. 4). It suggests that the Mousterian occupation levels are clearly younger than MIS 5 and were coeval with MIS 4 to early MIS 3.

Figure 4

Bayesian age model of the Santa Croce external deposits (performed with Chronomodel). The very young age of the modelled end of the Mousterian sequence is probably biased by erosional events.

Magnetic susceptibility

Raw MS data from Oscurusciuto are presented in Supporting Information Table S1 and the results are synthesized in Fig. 5. HF and LF values display a similar trend over the entire succession, and absolute values for each sample show only slight differences. The same trend is also expressed when plotting χfd values; indeed, the bar log format obtained for χmassLF, χmassHF and χfd are the same. This result could be partially influenced by the low resolution of sampling, but surely highlights the high influence of climate‐driven pedogenetic processes in the observed MS values.

Figure 5

Sedimentary log of the Oscurusciuto succession with the locations of samples collected for magnetic susceptibility (MS) analysis. Results of MS measurement are reported and accompanied by bar logs that highlight the main palaeoclimate fluctuations as deduced by MS data. Samples OSC 229 and 230 have not been analysed due to their composition, almost exclusively represented by volcanic ash. χLF and χHF: low‐ and high‐frequency MS values. [Color figure can be viewed at wileyonlinelibrary.com]

The data suggest that the deposition of the succession started in relatively cool/dry settings (from the base of the succession up to 83 cm), followed by a period of warm/humid climate (interval 83–184 cm), in turn overlaid by sediments deposited during cool/arid settings (interval 184–281 cm). This part of the stratigraphic sequence has still not been investigated; therefore, these data are – so far – the only available proxies for palaeoclimatic reconstructions.

The upper part of the succession, which has already been excavated, shows a more monotonous situation with MS values, suggesting a warm and humid setting in the interval between 281 and 467 cm, corresponding to the archaeological layers 14–3. Only the samples collected in the upper layers (SU1 and possibly a part of SU2, i.e. the interval from 467 to 570 cm) show MS values documenting a rapid shift towards more cool/arid settings (Fig. 5).

Faunal remains are listed in Table 2. The oldest sample, referred to the final phase of MIS 4 to early MIS 3, is from Grotta di Santa Croce. The mammal assemblage is poor in terms of biodiversity and points to a quasi‐exclusive predation upon horse and aurochs in all the layers. Zooarchaeology on the new sample from layer 549 shows the same pattern already observed in the other layers (Boscato et al., 2006b; Boscato and Crezzini, 2006; Boscato, 2017): isolated teeth count for the 75% of the identified remains (NISP = 72), long bones are represented only by diaphysis fragments and no phalanges have been identified. Considering NISP and the unidentified fragments together (total specimens = 789), diaphysis and compact bone fragments account for 74.6% of the whole sample. Only 0.5% of the remains are spongy bones/epiphyses. Anthropic modifications (impact marks, cone flakes and burned bones) account for 2.3% of the faunal remains, while carnivore‐induced modifications were not observed. About 92% of teeth belong to adult/elderly individuals.

Table 2

Number of identified specimens (NISP) according to sites and layers.

Site	Layers	Equus ferus	Equus hydruntinus	Equus sp.	Bos primigenius	Capra ibex	Rupicapra sp.	Caprinae	Cervus elaphus	Capreolus capreolus	Dama dama	Cervidae	Sus scrofa	Stephanorhinus sp.	Panthera leo	Crocuta crocuta	Canis lupus	Vulpes vulpes	Lepus sp.	Total
GC	EIII5	53	1	1	86	/	/	/	52	/	/	/	3	/	/	1	2	4	1	204
GC	FI	14	/	/	25	/	/	/	22	/	3	2	/	/	/	/	/	6	4	76
GC	FII	21	/	/	46	/	/	/	26	1	7	2	1	/	/	/	/	7	2	113
RO	1	2	/	/	20	/	/	/	11	1	6	/	/	/	/	/	/	/	/	40
RO	2/29/30/31	48	/	/	82	5	/	/	27	8	12	3	/	/	1	/	/	/	/	186
RO	3	16	/	/	15	1	/	/	15	2	1	/	/	7	/	/	/	/	/	57
RO	4	7	/	/	165	1	4	/	26	4	10	2	/	/	/	/	/	/	/	219
RO	5‐6‐7	1	/	/	160	/	/	/	9	2	20	5	/	1	/	/	1	2	/	201
RO	8	1	/	/	79	/	/	/	4	/	7	/	/	/	/	/	/	/	/	91
RO	9	4	/	/	69	/	/	/	12	4	6	1	/	/	/	/	/	/	/	96
RO	11	3	/	/	61	/	/	/	5	1	3	/	1	/	/	/	/	/	/	74
GSC	525	142	/	/	111	/	/	/	/	/	/	/	/	/	/	/	/	/	/	253
GSC	535	73	/	/	96	/	/	/	/	/	/	/	/	/	/	/	/	/	/	169
GSC	546	68	/	/	131	/	/	/	1	/	/	/	/	/	/	/	/	/	/	200
GSC	549	15	/	/	53	/	/	/	4	/	/	/	/	/	/	/	/	/	/	72

Abbreviations: GC, Grotta del Cavallo; GSC, Grotta di Santa Croce; RO, Riparo L'Oscurusciuto.

New data from Oscurusciuto were collected in the northernmost sector of the shelter, layers 4–7 (excavation ongoing). Faunal remains from this area are less affected by concretions than those from the rest of the site (only 2% of them are partially covered by thin calcite films). Bones are usually very fragmented (91% are within the size class 1–3 cm) and only 0.05% are unbroken. Among the 71 specimens taxonomically identified, 56 are from layers 5–7 and 15 from layer 4. They mainly include isolated teeth and diaphysis fragments (93% in layer 4 and 76% in layers 5–7). The only identified epiphyses belong to the proximal metacarpal and to first and second phalanges. Considering also the unidentified sample, spongy elements are represented by 16% of specimens in layer 4 and 17% in layers 5–7. Several specimens are burned (33% in layer 4 and 65.7% in layers 5–7), while other anthropic marks (cut marks, percussion notches and cone flakes) were identified on 1.7% of remains in layer 4 and on 1.6% in layers 5–7. No carnivore‐induced bone modifications were identified. These data confirm the picture that has emerged from previously studies (Boscato and Crezzini, 2006, 2012a,b; Boscato, 2017; Spagnolo et al., 2020b).

As far as Oscurusciuto is concerned, this faunal assemblage is distributed across a period between 55k and ~43k cal a bp. Aurochs is almost always the most abundant species. The only exception is layer 3, where horse counts for the 28% of the faunal remains. Again, taxa well adapted to open environments (steppe and forest steppe) are those showing the highest frequency. Moreover, caprines (ibex and chamois), wild boar and cervids were also identified. In layer 1 cervids account for the 45% of NISP. Fallow deer was identified in all the phases, while rhino and carnivore specimens are very rare.

At Grotta del Cavallo it was possible to consider both Mousterian (Layer F, Sarti et al., 2000) and Uluzzian (Layer E, Boscato and Crezzini, 2012a) data. Generally, cervids are more abundant than at the other sites, but again aurochs is the most targeted prey. Such a faunal association seems to resemble the quasi‐contemporaneous upper part of the sequence from Oscurusciuto. A slight increase of taxa related to open environments is visible during the Uluzzian, but the difference with the preceding Mousterian is not remarkable. At Grotta del Cavallo, carnivore and lagomorph remains are consistently very rare. The analysis carried out by Boscato and Crezzini (2012b) points to an anthropogenic accumulation of animal remains without any trace of carnivore activity.

Discussion and conclusions

From a chronological perspective, our data anchor – for the first time – the external deposit of Santa Croce to a period between the end of MIS 4 and early MIS 3.

This result enables us to evaluate human presence over a long time frame that covers some of the harshest phases of the early Late Pleistocene. At Santa Croce, Oscurusciuto and Cavallo, the layers where zooarchaeological and MS analyses were performed are palimpsests. As a consequence, data from these layers represent ‘mean values’ that describe timespans of different duration. This must be considered when they are compared with palaeoenvironmental results from Pozzo Cucù, as the datasets have different scales (Fig. 6).

Figure 6

Comparison between palaeoclimate proxies (rainfall amount and soil bioproductivity) from Pozzo Cucù Cave (PC stalagmite, Columbu et al., 2020), magnetic susceptibility data from Riparo L'Oscurusciuto and zooarchaeology from the studied archaeological sites. [Color figure can be viewed at wileyonlinelibrary.com]

The new zooarchaeological data from layer 549 of Grotta di Santa Croce and layers 7–4 of Riparo L'Oscurusciuto confirm the pattern outlined by previously published samples: significantly, the Middle Palaeolithic faunal assemblages share similar characteristics in all the considered sites. First, we note the scarcity of remains related to small game and birds. These resources had an extremely marginal – if not absent – role in the subsistence of Apulian Neanderthals, who mainly hunted aurochs and/or horse. Second, zooarchaeology also points to repetitive behaviours in hunting strategies and carcass reduction over time. Carcasses were intensely exploited, including within‐bone nutrients (bone and marrow). Epiphyses and spongy bones were probably smashed to extract these substances. However, in spite of the general scarcity (in all sites and layers) of spongy bone fragments, there is no evidence of either the use of bone as fuel (e.g. Spagnolo et al., 2019) or the alteration of osteological samples by carnivores, such as hyenas (Boscato and Crezzini, 2006, 2012a). This could suggest the hypothesis that Neanderthals systematically exploited fat contained in epiphyses and spongy bones. Hunting strategies, as far as for aurochs and horses, were mostly focused on the killing of mature/elderly individuals (Boscato and Crezzini, 2006, 2012b).

The highest representation of these two species together – which are both adapted to open landscapes (forest steppe and steppe, respectively) – is reached in the cold/arid phases at Santa Croce, where cervids are very rare, and fallow deer is completely absent. At Oscurusciuto and Cavallo, although horse and aurochs count for most of the faunal remains, the other species are more abundant compared to Santa Croce. Fallow deer, which was observed at Avetrana during MIS 5, appears again in Apulia, thus placing this region within the ‘warm’ belt described in the Introduction and represented in Fig. 1B. The presence of cervid taxa indicates also a more humid habitat in that area, which remained quite constant in the period, thus indicating that the extreme climatic fluctuations of MIS 3 did not affect Apulia. Macromammal data from layer F of Grotta del Cavallo are in line with those from small mammals from the same layer, where the dominant taxon Microtus savii, related to open environments in a Mediterranean climate, is accompanied by Apodemus sylvaticus, Apodemus flavicollis and Elyomis quercinus, which attest to more vegetated areas (Dalla Valle, 2008). Even if the dates from Santa Croce show a high degree of uncertainty, the pattern displayed by the faunal remains from this site seems to be in agreement with the evidence from Pozzo Cucù, which attests to a long arid phase within MIS 4 and – after some rapid oscillations that probably are not registered by zooarchaeological analysis – a very harsh peak at ~55 ka followed by another relatively arid phase until about 50 ka. A subsequent more humid phase characterized by quite stable conditions in soil productivity (Columbu et al., 2020) is in line with the data from L'Oscurusciuto and Cavallo (Fig. 6). The harshest peak detected in the PC speleothem at about 55 ka, together with all the other rapid climate oscillations provided by this record, is probably too short to be perceived through the faunal record. Even if the faunal composition at Oscurusciuto reflects a continuous presence of open environments (with the spotted presence of vegetated areas) between 55 ka and ~43k cal a bp, the increase in horse remains in layers 3 and 2 could testify to local slightly more arid conditions, which were not recorded at Pozzo Cucù. This evidence is supported by MS data, which register a cool/dry phase in layer 1. Such apparent inconsistency can be explained by the fact that the response of animal communities to climate changes is faster than that of pedogenetic processes. Indeed, a certain amount of time is needed for pedogenesis, for the infiltration of fine‐grained sediments through the rock massif and finally for their deposition in the environment of the rock‐shelter. Similar shifts have already been observed in soils and soil‐related loess in China (Liu et al., 2003). Even if shifts in MS detected at Oscurusciuto do not provide the exact magnitude of the corresponding climatic oscillations, the two cool/dry phases identified in the unexcavated part of the sequence can be tentatively linked to the peaks in δ18O registered in Pozzo Cucù before 55 ka. If so, the oldest part of the sequence of Oscurusciuto could be attributed to MIS 4 or to a very early phase of MIS 3. The ecotonal environment of the ravines surrounding Oscurusciuto was composed of a mixture of very different habitats (the wet, possibly vegetated, bottom of the ravines, the rock walls and the flat arid uplands) that guaranteed a high biodiversity of ungulates. This feature, along with the availability of lithic raw materials and the presence of (at least seasonal) water streams, was probably the reason why Neanderthals were able to persist in the territory (e.g. Spagnolo et al., 2020b). Even if some more arid phases are indicated by the spread of horse and by MS data, the fairly constant presence of fallow deer indicates the absence of shifts towards very cold conditions.

When MHs reached Apulia about 45k cal a bp, climate and habitats were still the same as those experienced by late Neanderthals. MHs hunted the same taxa as Neanderthals, but showing a different strategy in the exploitation of herds (younger individuals; Sano et al., 2019) and carcasses (less intensive exploitation of spongy bone tissues). This different exploitation model could originate from a diversity in hunting technology and behaviours between the two populations. In particular, Sano et al. (2019) demonstrated that MHs were fully equipped with a new projectile technology (bow and arrow or spear‐thrower) that allowed them to hunt while maintaining a greater distance from prey. However, further research is needed to understand in detail nutritional patterns displayed by Neanderthals and MHs, as well as any possible diversity between their carcass exploitation strategies. MHs arrived in the Uluzzo bay when Neanderthals had most probably already abandoned this area. The survival of Neanderthals at Oscurusciuto until ~43k cal a bp points to a possible coexistence between the two populations in Apulia for some hundred years, but no evidence of an interaction has been found so far. In conclusion, Apulia offered resources and suitable stable climatic conditions for both late Neanderthal persistence and the early settling of MHs. Data from Pozzo Cucù and zooarchaeology do not support a population replacement due to climatic changes.

Therefore, even if a comprehensive explanation for the abandonment of the Uluzzo bay area by Neanderthals before 45k cal a bp and for their disappearance ~43k cal a bp at Oscurusciuto is still lacking, the possibility that more advanced hunting strategies implemented by MHs could have played a significant role in the replacement between the two groups cannot be excluded (Sano et al., 2019).

Author contributions

F.B.: conceptualization of the paper and analysis of part of the faunal remains; F.B., P.B., A.R. and A.M.: coordination of research and excavation at the three archaeological sites; P.B. and J.C.: identification of faunal remains; I.M.: sedimentological analysis; J.J.B. and C.F.: dating at Grotta di Santa Croce; V.S.: sedimentological sampling and realization of Figures 1 and 2; A.C.: palaeoenvironmental analyses at Pozzo Cucù; all the authors collaborated in writing, reviewing and editing the paper.

Additional supporting information may be found in the online version of this article at the publisher's web‐site.

Supplementary Table S1. Raw MS data from Oscurusciuto.

Click here for additional data file.