2.1 Geography and geology

The newly discovered Paleolithic site of Piovesello (44°38'08.8“N 9°35'19.8“E) is located on the northern watershed of the Northern Apennines (Fig. 1), in the River Nure catchment area and lies at 870 m a.s.l. near the town of Cassimoreno (Ferriere, Piacenza) on the northern side of Mount Camolara-Ragola (Fig. 2).

Figure 1 Map of the Alpine region and of the Italian peninsula, showing the Late Pleistocene maximum extent of glaciers (after Ehlers and Gibbard, Reference Ehlers and Gibbard2004) and the continental belts during MIS2 (GIS elaboration based on bathymetric, elevation and WGS84 grid data sourced from: http://www.naturalearthdata.com; Sbrocco, Reference Sbrocco2014; Zickel et al., Reference Zickel, Becker, Verheul, Yener and Willmes2016). The hydrography of the Po plain – Adriatic basin during MIS 2 is from Monegato et al. (Reference Monegato, Ravazzi, Culiberg, Pini, Bavec, Calderoni, Jež and Perego2015). The location of the Piovesello and of the other Gravettian sites cited in the text is shown. 1) Grotte de L’Observatoire cave; 2) Riparo Mochi rockshelter; 3) Grotta delle Arene Candide cave; 4) Casaletto Ceredano pollen record (Ravazzi et al., Reference Ravazzi, Garozzo, Deaddis, De Amicis, Marchetti, Pini, Vezzoli and Zanchi2018); 5) Berceto paleolake pollen record; 6) Grotta Rio Secco cave; 7) Grotta Broion cave; 8) Lake Fimon pollen record; 9) Bilancino; 10) Fonte delle Mattinate; 11) Grotta Paglicci cave; 12) Grotta della Cala and Grotta La Calanca caves.

Figure 2 Digital Terrain Model and simplified geological setting (according to Elter et al., Reference Elter, Ghiselli, Marroni and Ottria1997, with updatings) of the northern watershed of the Mount Ragola-Camolara Massifs. The Gravettian site of Piovesello is shown. The view is from the North.

Piovesello is a hilly area comprised of a varied lithostratigraphic bedrock mosaic. The northern slope of Costa dei Cornini in the Massif of Mount Camolara-Ragola is shaped by ultramafites, massive, hard rocks of Jurassic age (Elter et al., Reference Elter, Ghiselli, Marroni and Ottria1997; Costa et al., Reference Costa, Vescovi and Zanzucchi1990, Fig. 2), but downstream the occurrence of carbonate and pelitic breccias, affected by karst, create a hilly plateau with differential erodibility and topographic complexity. Here, several depressions partly dissected by small creeks, host either accumulations of fine sediments affected by hydromorphic pedofeatures or palustrine basins (Fig. 2 in green). One of them (locally called “Lagasso”) is dammed by a ridge interpreted as a terminal moraine in the most recent State Geological Map 1:50.000 (Elter et al., Reference Elter, Ghiselli, Marroni and Ottria1997). We observed that several other depressions in the immediate surroundings contain only slope and palustrine deposits, thus excluding a glacial cover, even if in the area glacial deposits of late Pleistocene age have been mapped (Elter et al., Reference Elter, Ghiselli, Marroni and Ottria1997). We therefore reconsidered the Quaternary history of this area in the present work (yellow area in Fig. 2). A 2 m-thick sequence of hydromorphic fine sediments filling a small depression, locally called “Piovesello” (Fig. 2) yielded Palaeolithic artifacts. Over 750 (all sizes considered) knapped stones were recovered in 2007 by one of us (F.F.) in primary or subprimary position, embedded in loamy sediments exposed by a seasonally ephemeral creek.

2.2 Modern ecological setting

The modern vegetation of Mount Camolara-Ragola Massif includes mixed broad-leaved, partly cultivated forests (Hornbeam, Hop-Hornbeam, Chestnut, Mountain Maple, Black Alder), scrub growing on ophiolithic rocks (Amelanchier ovalis Medic.) and extensive beech forests up to a climatic timberline of 1700 m asl. A subalpine belt of conifers is missing, despite small, isolated areas of dwarf mountain pine (Pinus mugo/uncinata) that occur in and above the timberline ecotone up to the watershed maximum summits (Alessandrini, Reference Alessandrini1984). The natural timberline is depressed by pastoralism. Petrophytic vegetation on ophiolithic rocks includes several endemics (Saccani and Salvoni, Reference Saccani and Salvoni2015), among which Armeria arenaria subsp. apennina Arrigoni (Arrigoni, Reference Arrigoni2015) deserves a specific mention here, as it occurs in the fossil record of the studied site (see section 7.2).

6.1 Stone tool petrography

Most of the lithic industry was made on local red radiolarite beds belonging to the Monte Alpe Chert formation (Upper Jurassic), representing the sedimentary covers of the external Ligurid ophiolites (Abbate et al., Reference Abbate, Bortolotti and Principi1980; Cortesogno et al., Reference Cortesogno, Galbiati and Principi1987) and widely outcropping along the Apennine watershed at Monte Lama, Monte Menegosa, and Groppo di Gora (Negrino et al., Reference Negrino, Colombo, Cremaschi, Serradimigni, Tozzi and Ghiretti2017). At Piovesello up to 5 different subtypes of radiolarites were flaked (dark reddish, reddish, greenish-reddish, reddish-brown) attesting the exploitation of different varieties of this lithic resource. The cortexes attest that they gathered slabs and blocks from secondary deposits (soils, detritus, local streams). The petrographic features indicate a complete convergence with the local raw material represented by radiolarites interbedded with shales. The selected and flaked blocks are always highly crystalline, in some cases showing complex systems of fractures filled with quarz veins and hydrothermal mineralizations and recrystallizations. The exceptionally highly silified Monte Lama outcrops, situated 9 km east of Piovesello, represent one of the probable exploited area (Fig. 8). Similar outcrops are quite common in the Apennines between eastern Liguria and Emilia (Cortesogno and Gaggero Reference Cortesogno and Gaggero2002; Campana et al., Reference Campana, Martino, Negrino and Del Soldato2013), while different varieties have been surveyed in western Liguria in outcrops of limited extent (Arnasco radiolarites: Negrino et al., Reference Negrino, Colombo, Cremaschi, Serradimigni, Tozzi and Ghiretti2017). Few artefacts were made on cherts from the Calpionella Limestone (Tithonian- Early Cretaceous) lying above the Monte Alpe Cherts radiolarites, which contains few centimetres thick chert layers with medium to low rheological qualities.

Figure 8 A large scale Digital Terrain Model and simplified geological setting of the upper Val Nure on the northern Apennine watershed, showing topographic relationships between the new Gravettian site, the location of the Upper Palaeolithic workshop on Monte Lama ophiolitic sedimentary body, and the Mont Ragola ophiolitic Massif. Notice LGM glacier development on Mont Ragola and widespread landslides (L, light brown).

An allochthonous assemblage includes some chert tools and a bladelet core with end-products and refittable by-products, the biggest elements of which (one side-scraper and one end-scraper, Fig. 9) were obtained from the same chert block (thickness 6-7 cm), with micritic and calcareous whitish cortical zones 1 cm-thick. One artifact shows a rather clean bedding joint, thus suggesting an acquisition in subprimary contexts, very close to the outcrops. These two artifacts were probably introduced on the site as finished tools, while the core shows traces of bladelet reduction sequence carried out on site, since some unmodified end-products and by-products refit on it.

Figure 9 1) Position of the Oligocene Apt-Forcalquier basin (Provence, France). Gravettian sites with cherts from this area are also shown: O) Grotte de l’Observatoire cave; M) Riparo Mochi; AC) Arene Candide cave; P) Piovesello. 2) sample of Apt-Forcalquier lacustrine chert (loc. Pont Julien), collection of the Musée d’Anthropologie Préhistorique de Monaco. 2a, 2b) Well preserved algae structures (Charophyta) in a completely silicified fine texture sediment. Among the bioclasts, ostracods and gastropods are also somewhere represented, all belonging to freshwater environment. 3-6) Piovesello lithic tools made of Apt-Forcalquier Oligocene chert: backed point and fragment (3), lateral scraper (4) and micrograph showing a fragment of algal branch comparable to 2a and 2b (4a), bladelet core with refitted bladelets (5), end-scraper with refitted retouch microflakes (6) and fragments of algal branches comparable to 2a and 2b, in a microlaminated fine texture sediment with microdetrital bioclastic and iron oxides particles (6a).

The raw material is a fine crystalline brown brackish or lacustrine chert with a somewhat thick, whitish and planar micritic calcareous cortex. The paleontological content is represented by several carophyte algae (Charales) thallus and branches often still connected, ostracods and gastropods, all indicating a freshwater environment. The sedimentation took place in a quiet shallow water basin as attested by the finely textured microlaminated sediment enclosing both the microfossils and a few terrigenous grains (Table S2). The provenance of these artifacts is very likely the Oligo-Miocene basin of Apt-Forcalquier (Haute Provence, at least 300 km to the west, Fig. 9), which belongs to the complex of the north prealpine foreland molasse. A close comparison with a large collection of geological samples from the area was carried on in the laboratories of Monaco thanks to the cooperation of one of the authors (P. Simon). In the Apt-Forcalquier basin, chert exposures are localized in several places across a large area showing a certain variability (Slimak et al., Reference Slimak, Bressy, Guendon, Montoya, Ollivier and Renault2005; Tomasso et al., Reference Tomasso, Binder, Martino, Porraz and Simon2016). Close similitudes were detected with samples from Pont Julien (Apt-Bonnieux, Luberon Parc).

6.2 Lithic industry

The largest artifact concentration (Structure I – hereafter SI) was exposed at the base of SU 7. It is an assemblage of radiolarite flakes, articulated in two main concentrations with larger artifacts belonging to the operative chain of a single core, placed below a level indicative of in situ knapping. The northern concentration was characterized by artifacts fanned out across a limited area, while in the southern one the findings appeared more heterogeneously dispersed, and for the most part they were arranged with the dorsal face exposed. Not far from SI, a sediment core was recovered. Except for this concentration, the rest of SU 7 records a variable density of finds ranging between 1/sq and 15/sq, with completely devoid zones towards the periphery.

The composition of the lithic assemblage is reported in Table S3 and the key typological classes of retouched blanks are listed in Table 5. The presence of a rather wide diversity in technological products alongside with many complex refittings (more than 20 pieces each) hints to an on-place exploitation of radiolarite blocks and slabs. Reduction phases, albeit incomplete, are represented by refitted products ranging from core shaping and initialization to full production up to the discard of the residual core. In particular, the stages of core shaping and initialization, as well as the maintenance of the flaking face in more advanced stages, are by far the most represented, comprising 40% of the entire lithic assemblage both from surface collection and from archeological excavation. A good number of products referring to the progressive accomplishment of the laminar knapping (crested, side-crested and core-edge removal blades and core-tablets) shows that the entire lithic waste derived from knapping management exceeds half of the whole lithic assemblage or approximately two-thirds only when considering only the identifiable artifacts. Consequently, complete central blade products are almost completely absent in the lithic assemblage, being present almost only in the form of blade and bladelets fragments, thus suggesting an exportation of the desired products to be used in other zones of the settled area or elsewhere.

Radiolarite slabs were exploited along their maximum length for blade production (e.g. Fig. 10a). Preliminary shaping and longitudinal convexity were maintained through removing orthogonal flakes by hard stone percussion, while the transverse convexities through removal of blades using soft stone percussion. Production of long and rather symmetric blades was carried on with frontal and unipolar sequences using organic percussion. Transverse convexities and recovery of mistakes are maintained by neo-crest blades and seldom using an opposite striking platform, as seen in other Early Gravettian (Digan, Reference Digan2006; Taller and Floss, Reference Taller and Floss2011) and Middle Gravettian (Klaric, Reference Klaric2008) technocomplexes. Another refitted core, reconstructed from the artifacts found in structure SI (Fig. 10b), shows the development of varied technological behaviors, aimed at the production, in first instance, of short blades and later on of bladelets. These products were obtained through the exploitation of adjacent surfaces using the same striking platform thus avoiding the detachment of core tablets. Short blades were also produced from core N.153 (Fig. 10c), probably due to the constrained dimensions of the original block. The core was exploited in an almost entirely unipolar mode and with alternating phases of production and maintenance of the core face and the striking platform which was rejuvenated through the extraction of invasive and thick tablets. The core flaking face was exploited using unipolar removals and only during the final phase of production an opposite striking platform was opened, while lateral convexities were maintained by the removal of thick and partially cortical core-edge flakes.

Figure 10 Refitted blade cores from Piovesello made of local red radiolarite (pictures: A. Cocilova, G. Baldelli).

The retouched tools and backed implements are rather scarce (24 elements), both coming from surface collection and SU7. Tools are represented by burins (Fig. 11, n. 1), one end-scraper (Fig. 11, n. 2), scrapers, one splintered piece, and retouched blades. Simple burins and burins on fractures are made on various supports (thin flakes, laminar pieces, maintenance and cortical flakes) and show one or more long burin spalls. Burin-cores are made on thicker blanks and shares with maintenance and core-edge removal flakes the same width range (13-17mm).

Figure 11 The retouched lithic assemblage from SU7; 1) fracture burin; 2) end-scraper; 3) fragmented backed point; 4-5) marginally retouched backed points; 6) fragmented backed bladelet (pictures: A. Cocilova, G. Baldelli).

Backed points and backed bladelets complete and fragmented have been made from full production bladelets (Fig. 11 and Table S4). Backed points are intact and have invasive (Fig. 11, n. 3) or marginal back (Fig. 11, ns. 4,5) opposed to unretouched margins and additional retouches on the base, while backed bladelets (or points) are all fragmentary (Fig. 11, n. 6). A projectile point has been made on a burin spall. The size of the armatures is in agreement with the bladelet production: length 32-39mm, width 4-5mm, and thickness 2mm.

7.1 The terrestrial ecosystems of the northern Apennine at 30 ka and evidence of Gravettian human interaction at Piovesello

The availability of both microscopic plant remains and wood charcoal from a mountain site with traces of anthropic presence dating to the last glacial provides a chance to obtain unprecedented insights about terrestrial ecosystems and human activities on the northern Apennine watershed at around 30 ka ago. This is due both to sealing by water-saturated clayey deposits, allowing for pollen preservation, and to the occurrence of charcoal, allowing us to obtain ¹⁴C ages.

The identification of terrestrial habitats relies on the occurrence of high pollen producers from steppe and semidesert environments, like Poaceae, Artemisia, Ephedra fragilis and E. distachya types, but also on the occurrence of low pollen producers, which help in reconstructing the habitats near the archaeological site. This second group includes indicators of petrophytic desert - like Armeria (Fig. 7e) and Dipsacus fullonum type – and indicators of mountain steppe (Walter and Breckle, Reference Walter and Breckle1986), such as Plantago alpina type (including widespread species typical for Alpine and Apennine steppe - P. serpentina and P. alpina), Succisa pratensis type, Knautia arvensis type and Scabiosa columbaria type (these latter in Fig. 7f-g). These two indicator groups suggest that the dominant vegetation in the proximity of the site was an open petrophytic semi-desert characterized by both (i) orophilous, xeromorphic, basic-adapted vegetation occupying poorly developed soils on bedrock outcrops or regolith, and (ii) discontinuous mountain steppe on fine-grained substrates. The modern habitat of Plantago serpentina, P. alpina, as well as most of the European species of Armeria and Scabiosa encompasses garrigues, modern grassland steppes and semideserts in the Eurasian mountains and plateaux (Arrigoni et al., Reference Arrigoni, Ricceri and Mazzanti1983; Chiarucci et al., Reference Chiarucci, Foggi and Selvi1995; Horvat et al., Reference Horvat, Glavac and Ellenberg1974; Archibold, Reference Archibold1995; Walter and Breckle, Reference Walter and Breckle1986).

In the modern pollen-vegetation calibration dataset of the Alps (Davis et al., Reference Davis, Zanon, Collins, Mauri, Bakker, Barboni, Barthelmes, Beaudouin, Birks and Bjune2013; Furlanetto et al., Reference Furlanetto, Ravazzi, Badino, Brunetti, Champvillair and Maggisubmitted; Van der Knapp et al., Reference van der Knaap, van Leeuwen and Ammann2001), an arboreal pollen around 50% with very low concentration and pollen accumulation rate values is characteristic for sites at or above the timberline altitude, which also receive pollen from lower vegetation belts transported by convective air masses (Van der Knaap et al., Reference van der Knaap, van Leeuwen and Ammann2001; Pini et al., Reference Pini, Badino, Brunetti, Champvillair, Vallè, Ravazzi and De Amicis2017a; Badino et al., Reference Badino, Ravazzi, Vallè, Pini, Aceti, Brunetti, Champvillair, Maggi, Maspero, Perego and Orombelli2018). These pine pollen proportions are similar to those reported in a nearby pollen record from palaeolake Berceto (870 m a.s.l., Bertoldi et al., 2004, Reference Bertoldi, Chelli, Roma and Tellini2007, Fig. 1), with absence of woody plant macrofossils, and it was also likely located over the timberline during late MIS 3 and MIS 2. Furthermore, although runoff activity would have conveyed needles and other light organic particles from existing trees and shrub individuals to the Piovesello depression, the absence of stomata and of conifer periderms corroborates the idea that conifers did not grow in the vicinity of the site. The occurrence of pine charcoal in the single level SU7 at Piovesello appears to be clearly related to human activity instead of runoff and other processes related to slope movements (compare Figs. 6, upper panel and 12; see section 4.2). These data suggest that pines were locally absent in the study area which was above the timberline, while wood/charcoal was collected according to its abundance and carried to the site by humans together with other raw materials. Pine pollen is not well preserved at Piovesello, but SEM observation suggests occurrence of both Pinus mugo/uncinata (Dwarf Mountain pine) and P. sylvestris (Scots Pine) pollen (section 4.4). Dwarf Mountain pine is able to grow much higher than any other European pines above the timberline formed by the open forest, e.g. on warm southern aspects, as it occurs today in the Pinus mugo-timberline ecotone of Eastern Alps, in the Central Apennine and in other high Mediterranean and Balkanic and Carpathian Mountains (Ballian et al., Reference Ballian, Ravazzi, De Rigo and Caudullo2016). Thus, we cannot exclude the occurrence of sporadic, low density dwarf pines, likely on the southern slopes of the Monte Ragola-Camolara Massif, but most probably not in the immediate surroundings of the Piovesello small depression, or out of the runoff watershed. This observation accords with the modern biogeography of Pinus mugo/uncinata in the Northern Apennine, which maintains a few isolated living populations, including Monte Nero – M. Ragola (Alessandrini, Reference Alessandrini1984) close to the Piovesello site (see Fig. 8).

7.2 Implications for the historical biogeography of petrophytic plants in the northern Apennine

The find of Armeria fossil pollen in the sediment unit bearing human artifacts is of interest for the historical biogeography of the altitudinal open ecosystems in the N-Apennine. The actual living species in the region (A. arenaria subsp. apennina Arrigoni) is an endemic of the northern Apennines limited to petrophytic ecosystems on ultramafic rocks, either in the mountains and in the altitudinal belts (Chiarucci et al., Reference Chiarucci, Foggi and Selvi1995; Arrigoni, Reference Arrigoni2015), while our late Pleistocene fossil record is located in a gently sloping area which supported beech and fir forest until recent reclaiming (see modern vegetation in section 2.1.). The current distribution of Armeria in the northern Apennines may be seen as the result of a late Holocene expansion of the petrophytic ecosystems related to climate cooling and human-shaped pastures (Pini et al., Reference Pini, Ravazzi, Raiteri, Guerreschi, Castellano and Comolli2017b). However, these events occurred after an early Holocene forest rise causing a severe bottleneck for altitudinal plants in the low-altitudes of the northern Apennine chain. Our fossil data in turn provide evidence for a semidesert belt which likely persisted for a long time at 870 m a.s.l. during the last glaciation, including extensive habitats for petrophytic plants. Therefore, according to our data, the modern populations of Armeria arenaria subsp. apennina represent an orographic relict from larger occurrences in the last glaciation, regardless of glacier activity in the northern Apennine (see below). On the other hand, maximum refugial patterns are expected in the early Holocene phase of thermal maximum (Renssen et al., Reference Renssen, Seppä, Crosta, Goosse and Roche2012).

7.3 The paleoclimatic setting of the Piovesello site during Greenland Stadial 5

The occupation phase recognized at Piovesello in SU7 can be correlated on the basis of radiocarbon dates with the Greenland Stadial 5.1 of the recently revised GICC05 chronology of Northern Atlantic – European subcontinent (Rasmussen et al., Reference Rasmussen, Bigler, Blockley, Blunier, Buchardt, Clausen, Cvijanovic, Dahl-Jensen, Johnsen and Fischer2014) (Fig. 12), which is identified as one of the most pronounced cold spells with a high dust flux in Greenland ice sheet (Fig. 12). GS5 pre-dates the onset of the last glacial maximum in the Alps as it is currently understood (Monegato et al., Reference Monegato, Scardia, Hajdas, Rizzini and Piccin2017; see S. Margherita Stade in Fig. 12) and the culminations of the Boreal hemisphere major ice-sheets (Hughes et al., Reference Hughues, Carpenter, Rockström, Scheffer and Walker2013). Although major alpine valleys were glaciated as early as 32-33 ka cal BP (Spötl et al., Reference Spötl, Reimer, Starnberger and Reimer2013), glaciers draining the Italian side of the Alps reached their outermost piedemontane lobes (e.g. Garda and Tagliamento glaciers in Fig. 12) only after 28 ka BP. The valley glaciers developed on the northernmost watershed of the Northern Apennine may have reached a maximum extent between 29 and 27 cal ka BP, prior to the Alpine last glacial maximum, during the so-called Campo Imperatore Stade (see Giraudi, Reference Giraudi2015 and Fig. 12). Given these constraints, and the position of the Piovesello site which is located about 600 m from a ridge, possibly of morainic origin (site of Lagasso, see Fig. 2), we cannot exclude that human activities took place above the timberline in a semidesert vegetation, bounding the fronts of local glaciers at their maximum expansion. However, the Lagasso front moraine is currently undated and its deposition may not be in phase with the Alpine maximum.

Figure 12 The chrono-climatic setting of the Piovesello site compared with hemisphaeric, Alpine and Apennine climate proxies. Panel A – Lake pollen records as summarized by % of woody plants (Lake Fimon and Tenaghi Philippon), from Wick (Reference Wick2000), Pini et al. (Reference Pini, Ravazzi and Reimer2010), Muller et al. (Reference Müller, Pross, Tzedakis, Gamble, Kotthoff, Schmiedl, Wulf and Christanis2011); Panel B – Greenland ice core INTIMATE chronology from Rasmussen et al. (Reference Rasmussen, Bigler, Blockley, Blunier, Buchardt, Clausen, Cvijanovic, Dahl-Jensen, Johnsen and Fischer2014); Panel C – North Atlantic Heinrich Events; Panel D – Delta 18O record from speletheme 7H, Bernese Alps, from Luetscher et al. (Reference Lüetscher, Boch, Sodemann, Spötl, Cheng, Edwards, Frisia, Hof and Müller2015); Panel E – Glacier advances in the Apennines from Giraudi (Reference Giraudi2015); Panel F – Glacier advances of two major glaciers in their pedemontane moraine systems, eastern Italian Alps (Garda and Tagliamento glaciers); the distance from the valley outlet is shown in km, modified from Monegato et al. (Reference Monegato, Scardia, Hajdas, Rizzini and Piccin2017).

The forest cover of the Po Plain in this time span is recorded in the AP (arboreal pollen) from the Lake Fimon record in the south-eastern Italian foreland (Wick, Reference Wick2000; Monegato et al., Reference Monegato, Pini, Ravazzi, Reimer and Wick2011; Ravazzi et al., in prep., Fig. 1 and 12, panel A) and the Casaletto Ceredano site (Ravazzi et al., Reference Ravazzi, Deaddis, De Amicis, Marchetti, Vezzoli and Zanchi2012; Deaddis et al., Reference Deaddis, Garozzo, Cherubini, Croce, De Amicis, Gärtner, Lonelli, Maggi, Pini, Ravazzi and Schneider2017, Fig. 1). The Po Plain was extensively covered by mixed boreal forests during the whole span of MIS 3. The thermal timberline is estimated at around 700 m a.s.l. (Ravazzi et al., Reference Ravazzi, Orombelli and Tanzi2004). A forest withdrawal in northern Italy occurred at the onset of Heinrich Event 2, 27.5 cal ka BP (Fig. 12), in phase with the maximum of Alpine glaciers (Monegato et al., 2007, Reference Monegato, Scardia, Hajdas, Rizzini and Piccin2017) and with a marked lightening in the 7H cave speleothem isotopic record in the Alps (Lüetscher et al., Reference Lüetscher, Boch, Sodemann, Spötl, Cheng, Edwards, Frisia, Hof and Müller2015). The occupation phase at Piovesello not only pre-dates Heinrich Event 2, but also Greenland Interstadial 4 and 3 peaks, the latter bounding the end of Greenland Stadial 5.1 (Fig. 12). We conclude that the Po Plain was still forested at the time of the Piovesello occupation phase. Regional records estimate a timberline altitude not reaching the Piovesello location, corroborating the results of the palaeoecological study carried out on the archeological layer.

Cold to harsh climatic conditions have also been detected in cave and rock shelter deposits along the western Ligurian arch; respectively at Arene Candide and Mochi. At Arene Candide, micromorphological analyses consistently suggest that very cold and (semi)-arid periglacial environments bounded very reduced (unit B) up to absent (unit C) human traces between 32 and 28 cal ka BP. The data are consistent with a cooling trend marked by an abrupt increase of cold conditions inferred from both avifaunal (Cassoli, Reference Cassoli1980) and macromammal associations (Cassoli and Tagliacozzo, Reference Cassoli and Tagliacozzo1994). Unit C thus reflects a phase of arid and cold climate (ca. 31.2-32.0 ka cal B.P.) which has to be correlated with Heinrich Event 3 (Rellini et al., Reference Rellini, Firpo, Martino, Riel-Salvatore and Maggi2013). The chronology of the basal part of unit B is framed between 30.0-30.1 ka cal BP and partly matches the Piovesello time span, which is rather positioned in GS-5.1, the colder spell of GS5, which correlates to the more recent portion of HE3 (Fig. 12). The effects of Heinrich Event 3 on rock shelter deposits have also been suggested for the portion just below the middle part of unit D at Riparo Mochi, in the Balzi Rossi cliff. Here again, the faunal assemblage agrees with this assessment of cold and dry conditions and prevalence of Alpine grasslands in the middle parts of unit D (Tagliacozzo et al., Reference Tagliacozzo, Zeppieri, Fiore, Spinapolice and Del Lucchese2011), leading to consider that unit E may correspond to the return to cold stadial conditions during GS5. Unit F, on the other hand, does not contain sufficient elements to distinguish traces of D/O cycles in it (Douka et al., Reference Douka, Grimaldi, Boschian, del Lucchese and Higham2012; Tagliacozzo et al., Reference Tagliacozzo, Zeppieri, Fiore, Spinapolice and Del Lucchese2011).

7.4 The Piovesello cultural setting framed in the Early Gravettian phase

The Piovesello lithic assemblage is culturally homogeneous and its attribution to the Gravettian lies mainly on the broad technological and typological features recorded, despite the low number of tools and backed pieces. Burins and end-scrapers are a basic component of the common domestic Gravettian tool-sets, in addition to scrapers made on flakes and splintered pieces, as has been described in other contexts in central Southern Europe (Palma di Cesnola, Reference Palma di Cesnola1993). Although end-scrapers were almost systematically made on regular blades, the Gravettian is also known for the vast array of burin types (simple, on fracture and on retouch) which were realized on a range of blanks and are considered as diagnostic elements of specific cultural facies diffused in a short time-span at regional or supra-regional scale across Europe during the Middle Gravettian. This latter is the case of the Raysse–type and the Noailles–type burins; the first spread across a large area in western France where it follows the Noailles Gravettian facies in stratified sequences (Klaric, Reference Klaric2008). This specific artefact is made on blades with faceted longitudinal burin spalls and it looks different than the burins on fracture recovered at Piovesello. The Noailles burin type spreads at supra-regional scale from Spain (Bernaldo de Quiros et al., Reference Bernaldo de Quirós, Maíllo-Fernández, Castaños and Neira2015) to southwestern France and to the Tyrrhenian coast of Italy (Palma di Cesnola, Reference Palma di Cesnola1993) where its appearance is recorded at Riparo Mochi layer D (spits f3.5-1; f2; fl; f) at 29.4-28.5 ka cal BP (Douka et al., Reference Douka, Grimaldi, Boschian, del Lucchese and Higham2012) and at the open-air site of Bilancino on the western slope of the Central Apennine at 28.6-28.0 ka cal BP (Aranguren and Revedin, Eds, Reference Aranguren and Revedin2008). Raw material acquisition strategies at Bilancino (Aranguren et al., Reference Aranguren, Cavulli, D’Orazio, Grimaldi, Longo, Revedin and Santaniello2015) show a rather wide exploited territory which spans from the central Tyrrhenian coasts to the central portion of the Apennines, towards Umbria or Marche region, where the flint outcrops which provided at least 50% of the chipped stone assemblage at the site were located (Aranguren et al., Reference Aranguren, Cavulli, D’Orazio, Grimaldi, Longo, Revedin and Santaniello2015, p.445) and thus suggesting a well-established itinerary which crossed this mountainous chain. At the moment, this is the only indication of the presence of Noailles Gravettian east of the Apennine chain, but it is completely absent in Adriatic Italy (Gambassini, Reference Gambassini2007), in the whole Balkan region (Mihailovic and Mihalovic, Reference Mihailovic and Mihailovic2007) and Greece (Adam, Reference Adam2007). The available radiocarbon datings from Piovesello constrain the phase of human occupation between 30.7 and 29.4 ka cal BP (complete 2σ calibration interval), thus slightly preceding the appearance of the Noailles Gravettian (Fig. 13). Piovesello laminar reduction sequences share few common traits with Bilancino, where in-depth analyses of refitted cores allowed drawing a first picture of Evolved Gravettian lithic technology in the Italian peninsula (Grimaldi et al., Reference Grimaldi, Aranguren, Revedin, Gottardi and Cavulli2011), while there are some key differences which does not allow for a strict comparison of the two sites. At Bilancino laminar cores on allochthonous raw materials shows generally three striking platform and thus an advanced degree of reduction; the flaking angle is perpendicular or slightly tilted and the progression of the reduction sequences is carried on with bidirectional exploitation, while at Piovesello one striking platform is generally used, the flaking angle is rather steep (at least in bladelet cores with volumes comparable to the Bilancino ones) and the production is always unidirectional, while a second striking platform is only used during maintenance phases. Furthermore, in the Tuscan site some specific choices for lamellar production clearly mark a difference with Piovesello: narrow bladelets are produced using the débitage sur tranche d’éclat after a mesial fracture of the support and very often core-edge removal flake-blades invade the flaking surface, while at Piovesello only burin-cores are testified and lateral flakes or blades are used to maintain lateral convexities during a full frontal reduction sequence. Although the already stated different technological choices it is nonetheless to be remembered that Bilancino is a specialized site (Aranguren and Revedin, Eds, Reference Aranguren and Revedin2008) and its peculiar reduction sequences may bias the general frame.

Figure 13 Available early to middle Gravettian radiocarbon dates of the Italian Peninsula (this study; Aranguren and Revedin (Eds.), Reference Aranguren and Revedin2008; Boscato et al., Reference Boscato, Ronchitelli and Wierer1997; Douka et al., Reference Douka, Grimaldi, Boschian, del Lucchese and Higham2012; Palma di Cesnola, Reference Palma di Cesnola2004; Rellini et al., Reference Rellini, Firpo, Martino, Riel-Salvatore and Maggi2013) were calibrated using the IntCal13 curve (Reimer et al., Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards and Friedrich2013) on OxCal v. 4.3 (Bronk Ramsey, 2017).

Given the relative scarcity of lithic analysis of Middle Gravettian assemblages in the Italian peninsula, when widening the comparison to the Western Mediterranean area, a consideration of French Noaillian assemblages and their relative reduction sequences (e.g. Brassempouy: Klaric, Reference Klaric2008) shows some common points with Piovesello lithic technology. Shared common traits are unipolar exploitation for laminar production and the utilization of an opposite striking platform for maintenance purposes, a frontal progression of lithic production and the transversal convexities managed to obtain a rather narrow flaking surface. These characters which allow comparison of Piovesello to French Noaillan assemblages are nonetheless shared even by Early Gravettian assemblages in Italy and in other Western Europe areas (Wierer, Reference Wierer2013). All these considerations, together with the absence of Noailles burins, thus point to an attribution of this site to a late phase of Early Gravettian.

Backed implements are another diagnostic trait of the Gravettian and their typological features were used for its tripartite subdivision of Gravettian (Early, Middle and Late), being the first phase characterized both by invasive and by the occurrence of marginally backed points. Among these latter, a specific type, the fléchette point, a leaf-shaped point characterized by marginal semi-abrupt retouch, sometimes inverse, at one or both ends of the blade blank, is rather typical for the Early Gravettian (Pesesse, Reference Pesesse2008), but currently it is completely absent at Piovesello. The Gravettian is also typified from the invasive Gravette-type backed points and the Vachon points (Simonet, Reference Simonet2011). However, none of these pieces has been recovered at Piovesello except for very few fragmentary invasive backed bladelets which can be classified as micro-Gravette points (Fig. 11, n. 3). Most of the backed points at Piovesello have shallow abrupt retouches, are non invasive, made on regular straight bladelet in a way closely comparable to implements discovered in few earliest Gravettian assemblages. Similar products have been reported along the Southern Tyrrhenian coast at Grotta della Cala, where the earliest appearance of the Gravettian is testified by the marginally retouched points recovered in spit 3m which is overlaid by spit 3d dated to 31.9-31.2 cal ka BP and from spit 1m dated to 31.6-30.8 cal ka BP (Boscato et al., Reference Boscato, Ronchitelli and Wierer1997). A closer Gravettian complex which might offer further hints for the interpretation of the Piovesello lithic assemblage is the aforementioned Riparo Mochi in the Western Ligurian coast (Laplace, Reference Laplace1977). Here, layer F records an evolved Aurignacian phase dating to around 33 ka in the lower portion of the layer and a low-frequency human occupation phase in the upper part (spits 40-34), dated to 31.0-29.7 cal ka BP based on shells and charcoal, and possibly referred to an initial Early Gravettian. The dating from the overlying layer E (30.2-29.5 cal ka BP) is in good agreement with Piovesello, but this context, tentatively attributed to the early Gravettian, is unfortunately characterized only by rare artefacts and a small hearth, identified during the 20th century excavations (Douka et al., Reference Douka, Grimaldi, Boschian, del Lucchese and Higham2012). Similar observations may be proposed for the Ligurian site of Arene Candide, which yielded a thick Gravettian sequence with dating of Unit B (Rellini et al., Reference Rellini, Firpo, Martino, Riel-Salvatore and Maggi2013) bracketed between 27.8-31.0 cal ka BP, thus coherent with Piovesello, but again almost completely devoid of any anthropic trace, both in the older (Bietti and Molari, Reference Bietti and Molari1994) and in the recent excavations (Martino, unpublished data), where just over 20 elements were recovered.

In a broader perspective, the Piovesello assemblage could be thus ascribed to Early Gravettian (“Undifferentiated” or “Gravettian with backed points”; Laplace, 1966, Reference Laplace1977; Palma di Cesnola, Reference Palma di Cesnola1993, Reference Palma2006), a technocultural facies present in the whole Italian peninsula, prior to the differentiation between Adriatic and Tyrrhenian areas which characterize the middle Gravettian (Palma di Cesnola, Reference Palma di Cesnola1993; Gambassini, Reference Gambassini2007). This specific facies is testified mostly on chronometric than cultural base at Riparo Mochi layers F and E and Caverna delle Arene Candide in the lower portion of unit B (layer P10), at Grotta della Cala atrio-GB (e.g. layers 3m, 3d, 2, 1m and 1d), at Grotta della Calanca (lower level) and, on the other side of the Apennines, at Grotta Paglicci (levels 23C-A; level 23A dates to 33.0-32.2 cal ka BP, level 22F-B to 33.2-32.3 cal ka BP, level 22A to 31.8-31.1 cal ka BP), all characterized by an elevated incidence of backed points and by other typological and morphometrical indexes, despite marked differences have been reported among these sites (Boscato et al., Reference Boscato, Ronchitelli and Wierer1997).

7.5 The regional physical setting and the production and circulation of radiolarite and flint artifacts

The Piovesello site is just 9 km from the large lithic Mousterian and Aurignacian workshops recently described from Monte Lama (Fig. 8, Negrino et al., Reference Negrino, Colombo, Cremaschi, Serradimigni, Tozzi and Ghiretti2017), where thousands, if not tens of thousands, of radiolarite artefacts were collected. The Monte Lama - Prarbera Complex is the largest ophiolitic sedimentary body on the northern Apennine watershed (Bernini et al., Reference Bernini, Tellini, Vescovi and Zanzucchi1990) and includes at the top a thick portion of red radiolarite (with bluish varieties). The intense frequentation of Mt. Lama is linked to the presence of high-quality, strong vitreous raw material, suitable for the production of broad flakes as well as long blades. These raw materials were found up to 100 km west of the primary source, in the Aurignacian layers of Riparo Mochi and Riparo Bombrini in Western Liguria (Negrino and Starnini, Reference Negrino and Starnini2006; Negrino et al., Reference Negrino, Colombo, Cremaschi, Serradimigni, Tozzi and Ghiretti2017) and in coeval levels at the Grotte de l’Observatoire (Porraz et al., Reference Porraz, Simon and Pasquini2010) in Eastern Provence.

There are no radiolarite outcrops in the ultramafic masses of the Mount Ragola - Mount Nero, thus it is possible to conclude that the Piovesello radiolarite assemblage originated from Mount Lama. The ridge connecting Mount Lama to Piovesello has been extensively affected by slope activity during the last 30 ka (Fig. 8, areas in light brown), most probably triggered by intense fluvioglacial erosion connected to the last glacial lowstand and low fluvial baseline of the Nure and Ceno Rivers. This severe erosion dismantled ancient plateau and removed the Paleolithic deposits, remaining strips being preserved by massive ultramafic rocks at Mount Lama and in the Mount Ragola – Mount Nero ultramafic Massifs, but only outside of the glaciated area. From a geomorphic perspective, we therefore interpret our Paleolithic finds as the relicts of a late Pleistocene scattering over ancient highlands and subsequent dismantling. Last glacial maximum to Holocene fluvioglacial entrenchment and landsliding may have limited site visibility, and/or discouraged subsequent Paleolithic frequentation of the area.

An estimation of the degree of human frequentation around the wetlands of Lagasso-Piovesello and this part of the Apennine watershed seems problematic due to the aforementioned reasons. Moreover, it is also possible that extensive excavations might circumscribe the entire Piovesello settled area, thus allowing for the reconstruction of its inner spatial organization and for a deeper understanding of the site functions, since the current lithic scatters do not provide evidence for long-term use of this area. Low density of archaeological materials suggesting very ephemeral to short-term human presence during the early Gravettian is a shared feature with other sites in Italy. The actual archaeological evidence leaves at the moment some unanswered questions related to the effectiveness of such a scanty human presence which has to be tested against different variables such as subsistence modalities, mobility degrees, raw material procurement choices and settlement strategies to explain which concurrent factor may be involved. A context comparable to Piovesello has been reconstructed from geoarchaeological investigation on the Marche Apennine watershed, at Fonte delle Mattinate (770 m a.s.l.), where evidence of early Gravettian frequentation has been related to the interval between GI-5 and the HE3 (Giaccio et al., Reference Giaccio, Rolfo, Galadini, Messina, Silvestrini and Sposato2004; Silvestrini et al., Reference Silvestrini, Peresani and Muratori2005). In Liguria, at Riparo Mochi, the upper cuts of layer F and layer E, comprised between F42-41 and D28, testify to several episodes of very short and/or limited occupation episodes, bracketed by very short and/or periodical events of abandonment of the site (Douka et al., Reference Douka, Grimaldi, Boschian, del Lucchese and Higham2012). Comparably, Arene Candide cave has also produced very low density of artifacts in unit B, similarly to three caves in the northern area of the Veneto-Friulian plain. At Grotta del Broion, very few common tools, fragmented gravettes and ornamental beads from layer E were dated to 30.5-28.8 cal yr BP (Broglio and Improta, Reference Broglio and Improta1995). Grotta del Rio Secco yielded a modest assemblage from macro-unit BR1 composed of finished tools and backed bladelets associated to rare cut-marked caprid and marmot bones (Peresani et al., Reference Peresani, Romandini, Duches, Jéquier, Nannini, Pastoors, Picin, Schmidt, Vaquero and Weniger2014) and to charred wood dated to 33.4-30.0 cal ka BP (Talamo et al., Reference Talamo, Peresani, Romandini, Duches, Jéquier, Nannini, Pastoors, Picin, Vaquero, Weniger and Hublin2014). Higher densities in anthropogenic material have been recorded at southern latitudes at Grotta della Cala (Boscato et al., Reference Boscato, Ronchitelli and Wierer1997) and at Grotta Paglicci (Palma di Cesnola, Reference Palma di Cesnola2004).

Low density lithic scatters and other related indexes of human occupation can be referred to high mobility range patterns with a possible large-scale circulation of people and/or materials. The case of the Oligocene molasse flint used at Piovesello is here of considerable interest, since the outcrops are located in Provence (France), in an area roughly stretching south of the Lure Mountain, west of the Durance Valley and north of the Luberon Massif, where Lower and Middle Paleolithic sites and lithic workshops are widely known since the 20th century (Slimak et al., Reference Slimak, Bressy, Guendon, Montoya, Ollivier and Renault2005). The Upper Paleolithic evidences are more scattered and mostly connected to the provisioning of the Oligocene chert, among which the so-called “silex zoné du Largue”, a specific variant characterized by alternating light to dark brown parallel or concentric ribbon-like bands. In the Gravettian this raw material shows a wide diffusion and reaches the sites of Western Liguria, where it is preferably chosen for the production of the long blades, interpreted as funeral offerings and associated with the Barma Grande and the Young Prince from Arene Candide burials (Onoratini et al., Reference Onoratini, Simon, Negrino, Cauche, Moullé, Arellano, Borgia, Voytek and Arrighi2011). The reasons which may explain the presence of few retouched artifacts and a small lamellar core made on Oligocene molasse chert in an open site in the Apennines of Emilia, could be ascribed to its abundance, its high-quality rheological characteristics and to the lack of important tectonic fractures, since they originated in late Alpine molassic lacustrine basins not intensely involved in Alpine tectogenesis. Despite its rather low presence, the Provence chert seems to be chosen for laminar reduction sequences and for their future use. It is thus possible to hypothesize that these artifacts were part of the personal kit of one or more Gravettian hunters settled in this remote sector of the northern Apennine, who were nonetheless part of a long-distance exchange network (Tomasso and Porraz, Reference Tomasso and Porraz2016), ranging at least to the Western Ligurian sites, where Provence chert is abundant, together with rarer radiolarite artifacts (Porraz and Negrino, Reference Porraz and Negrino2008). At the same time the Oligocene chert artefact from Piovesello represents the easternmost area of diffusion of Provence chert, since at the moment such raw material is not testified in Central Italy. It is thus possible that these east-west displacements along the Ligurian and Provencal Arc may reflect precise economic choices and technical needs which were oriented toward the selection, in the two extreme poles, of high-quality cherts, as previously suggested (Tomasso and Porraz, Reference Tomasso and Porraz2016) and thus hint to the potential existence of a wide site network during Gravettian phases in the Liguro-Provencal arc, as already hypothesized for the Pyrenean area (Simonet, Reference Simonet2017).