Pena d’Água is located in central Estremadura, at 125 m a.s.l., ca. 100 km northeast of Lisbon and 47 km from the present Atlantic coastline. It is ca. 70 m wide, with a >10 m sedimentary deposit formed along an escarpment resulting from a tectonic fault, commonly known as the Arrife (from the Arabic “arriff”, scarp, coast) of the Aire Mountain. This tectonic fault separates the densely drained plains of the Tagus basin (to the east) from the dry, hilly limestone massifs (to the west), thus marking the boundary between two quite contrasting ecosystems. The site faces the former landscape with thick, extensive fluvial terraces; on its back, stands out the Aire Mountain (679 m a.s.l.) dominating the entire region (Fig. 2).

In the vicinity are several chert outcrops and the adjacent Cenozoic deposits are rich in chert nodules in secondary position along with good quality quartzite and quartz. The Arrife also has abundant permanent and seasonal freshwater springs, one of them just to the side of the archaeological site (“Pena d’Água” means “scarp” and “spring”, respectively), which may have made the site attractive for settlement. In sum, Pena d’Água is in a privileged location, with natural shelter, in an ecotone strip connecting a limestone mountain and wide river basin landscapes, with short-range access to high-quality lithic raw materials.

Eight field seasons between 1992 and 2000 in a 4 × 2 m test pit in the northeastern limit of the deposit revealed a ca. 5 m thick stratigraphic sequence encompassing eight main layers preserving six periods of almost uninterrupted human presence: Roman and Iron Age (layer B), Early to Late Neolithic (layers Eb-bottom to B) and Epipaleolithic (layer F). The primary observations regarding the site's stratigraphy, formation processes, cultural periods and their broader integration were published by Carvalho (1998a). In addition, special focus has already been given to Iron Age (Carvalho, 2008b) and Early Neolithic (Carvalho, 2008a: 58–62) occupations.

Layer F is the bottom unit of the stratigraphic sequence, lying on the local Miocene bedrock, and separated from the upper Eb-bottom layer by large boulders collapsed from the roof. Sediments are significantly different from those above. The matrix has very dense and compact coarse sand in a clayish matrix, and lacks the pottery, charcoal and bone that characterize all of the upper layers. It presents evidence for low-energy hydric action: calcium carbonate tuffs, rolled artefacts (sometimes accumulated in shallow pockets), faunal remains with taphonomic alterations (Valente, 1998) and conditions allowing the poor preservation of micromammal teeth (Póvoas, 1998). This hydric action might be related to a partially blocked spring located slightly above the excavation area, which may have been particularly active when layer F was formed (Simões, 2012).

Charcoal of wild olive tree predominates throughout the sequence, ranging from 74.4% in layer F to 92.7% in layer Eb, testifying to the Middle Holocene expansion of sclerophyllous vegetation and the decline of pine (Fiueiral, 1998). Layer F, however, shows the highest percentage of oak tree varieties (12.5%); cork oak (Quercus suber), in particular, reaches 7.5%. This is a slight, but relevant difference between layer F and the overlying layers at the botanic level. Faunal remains, on the other hand, are composed of a small number of rabbit (n = 1) and cervids (n = 2). In addition, some sheep/goat (n = 3) were interpreted as due to “[…] infiltrations from Layer Eb-bottom or to deficient dating of the Eb-base/F layers interface […]” during excavation (Valente, 1998: 93; Portuguese original).

The absence of hearths providing charcoal of unquestionably anthropic origin as well as deficient preservation of bones meant having to get a radiocarbon date from a bulk sample of cork oak charcoal. The result (Carvalho, 2008b: Table 18), was 7370 ± 110 BP (Wk-9213), corresponding to 8385–7983 cal BP (95.4% prob.), with a median value of 8.19 ky cal BP, after calibration with the INTCAL13 curve (OxCal program, version 4.2) (Fig. 3). Although caution is advisable due to the fact that it is a long-lived species, it is nonetheless a useful broad chronological estimate of when layer F was formed.

The plethora of data available from across Europe and Northern Africa seem to concur with an increase of demographic density in the end of the Pleistocene, probably as a result of adaptive strategies to survive through climatic crises such as the LGM or the Younger Dryas. Because the transition from the Pleistocene to the Holocene was one of melioration in climatic and environmental conditions, it might not demand an abrupt technological shift to overpass major difficulties. This may be the reason why we do not see an abrupt change in lithic technological traditions, at least in the western coast of Eurasia and probably the adjacent northern coast of Africa. However, by opposition, the 8.2 cal BP event was, in fact, an abrupt climatic crisis that affected all the Northern Hemisphere (Alley and Ágústsdóttir, 2005, Kleiven et al., 2008, Klitgaard-Kristensen et al., 1998, Rasmussen et al., 2007 and Renssen et al., 2007) and happened when demographic density was very likely higher than before. Such phenomenon demanded a rapid adaptation in the behavioural patterns of the hunter-gatherers populations, seen in regions such as the westernmost coast of Eurasia in multiple traits, such as settlement patterns, the appearance of shell-middens, more or less large burial grounds, and indirect data suggesting social inequality. All these aspects had consequences in the corresponding lithic assemblages.

Thus, the post-Younger Dryas amelioration does not seem to have had major impact in the settlement, technological and economic strategies of the hunter-gatherers of Estremadura (for a different view, see Araújo, 2009). Indeed, in this region (at least where research has been intensive), the archaeological record seems to be consistent in showing only very discrete changes on those cultural traits; as mentioned above, several authors even recognize strong similarities between the Late Magdalenian (Bicho, 2000, Gameiro, 2012 and Zilhão, 1997) and the Epipaleolithic (e.g., Araújo, 2003, Araújo, 2009, Araújo, 2011 and Bicho, 1994) with respect to lithic production. The question that is still to be answered is how fast the shift to the typically Mesolithic “blade and trapeze techno-complex” was, which can only be possible using assemblages dated to around the 8.2 ka cal BP event – as is the case of the Pena d’Água layer F – which seems to mark the transition to the Mesolithic.

The Pena d’Água assemblage shows that the techno-typological features prevailing at the time of the 8.2 ky event have much more resemblance to the Magdalenian industries (Araújo, 2003, Araújo, 2009, Bicho, 1994 and Zilhão, 1997) rather than with the subsequent Mesolithic ones (Carvalho, 2009, Marchand, 2001 and Roche, 1972). Those features are a close range acquisition of raw materials, even when these are not the best available within a wider region, a balance between chert, quartz and quartzite that typically depends on locally available sources, a strong coarse raw material component (e.g. quartzite and/or greywacke), and small and irregular bladelets knapped through Upper Palaeolithic-like technology (Araújo, 2012,Bicho, 1998, Bicho, 2002 and Marchand, 2005). According to the Pena d’Água evidence and the references cited above, technological features predating the 8.2 ky event show the predominance of flakes produced from cores with right angle (80°–90°), platforms with unidirectional parallel detachments, flat profiles, diverse cross-sections, irregular or parallel edges, and cortical or flat butts. With the exception of some backed and/or truncated points (such as Microgravette, Istres, Sauveterre, or La Malaurie points) and a smaller number of geometrics (usually trapeziums and rarely crescents or triangles), the large majority of the assemblages are comprised of marginally retouched flakes (notches, denticulates) or bladelets (such as Dufour bladelets), awls and splintered pieces. This is more obvious along the eastern borders of the limestone massifs (Santa Cita, Costa do Pereiro) and particularly in its southernmost tip (the Rio Maior area: Areeiro III and CPM V), where good quality chert occurs abundantly (Araújo, 2003, Araújo, 2009, Araújo, 2011 and Bicho, 1994), in contrast to the Tagus terraces (Araújo, 2012, Bicho, 1998 and Marchand, 2005).

This scenario strongly contrasts with that seen after the beginning of the Atlantic phase, the Mesolithic, where good quality raw material, especially chert, is acquired at longer distances in the areas where it has better quality, such as the Rio Maior valley (Pereira et al., 2015), and was used to produce longer, regular and standardized bladelets and blades probably through the soft hammer and/or indirect percussion techniques (Marchand, 2005). The retouched tools became dominated by regular geometrics while backed and marginal retouched tools almost disappear (Marchand, 2005 and Paixão, 2014).

Despite this is beyond the scope of this paper, it should be mentioned that considerable information already exists on local raw material sources (e.g., Aubry et al., 2014 and Matias, 2012), which will be crucial for far-reaching conclusions in the very near future. Indeed, ongoing surveys for chert sources in the Arrife area revealed new ones, still unpublished, some of which congruent with the assemblage in cause. These results will be published soon elsewhere.

A rather different pattern of raw material use emerges in areas where chert is rare. The nature and chronology of numerous macrolithic quartzite assemblages from the Tagus River have been long debated. The abundance of high-quality quartzite pebbles and cobbles made the production of chopper-like cores and semi-cortical flakes (with recurrent edge modification by retouch or use) easy. From the 19th century until the 1980s, research was based on surface collections, with sites being dated according to the terraces’ altimetry and the patina on the artefacts. This approach led to the proposal of a myriad of entities where Acheulean, Mousterian, Mirian, Languedocean and Clactonian were recurrently found associated within each surface collection (e.g., Breuil and Zbyszewski, 1942, Breuil and Zbyszewski, 1945, Raposo, 1986, Zbyszewski, 1943 and Zbyszewski, 1958).

Since the 1990s, long-term projects (e.g., Oosterbeek et al., 2002 and Oosterbeek et al., 2010), including geomorphological approaches (Corral, 1998, Grimaldi et al., 1998 and Mozzi, 1998), absolute dating (Burbidge et al., 2014, Cunha et al., 2008 and Martins et al., 2010) and the study of quartzite assemblages dated from the Acheulean to the Neolithic (e.g., Carvalho, 1998b, Carvalho, 2008b, Cura, 2002, Cura et al., 2004, Pereira, 2010, Pereira et al., 2012 and Rosina et al., 2004), permitted the confirmation of chronological markers, such as handaxes in the Lower Palaeolithic, or the Levallois in the Middle Palaeolithic. During the Upper Palaeolithic, the chronological differences in quartzite are more discreet, such as rarity of elongated blanks in the Gravettian, the dominance of prismatic and stepped reduction in the Solutrean and Magdalenian, and, in this case, the frequency of massive tools with edge damage. Consequently, the most used and – at present – reliable criterion to classify any post-Palaeolithic assemblage is the presence of diagnostic artefacts (pottery, leptolithic reduction sequences, retouched tools mostly in chert, or polished stone tools), even when the sites count with absolute dates. Such is the case of Amoreira (Fig. 1).

The interpretation by the same team on Amoreira resulted in two opposing views (for general overviews, see Cruz et al., 2000 and Rosina et al., 2010). Salvage works during 1992–1994 and 2005 recorded a stratigraphic sequence where layer C stands out for its quartzite macrolithic industry (with some leptolithic cherts), apparently associated to postholes, polished stone tools and potsherds. A radiocarbon date on a bulk sample of unidentified charcoal (I-17332: 7460 ± 120 BP) indicated a 8459–8014 ky cal BP (at 94.5% prob.) time interval, with the median at 8.27 ky cal BP, an Epipaleolithic result slightly older than Pena d’Água's. However, while the excavators consider the site “en place” (Cruz, 1997: 301), where “[…] pottery and polished stone axes first occur” (Oosterbeek, 2004: 85), the geoarchaeologists argued that layer C was formed over Pleistocene terraces where macrolithic industries were already present, thus presenting a more complex scenario based on the site's formation process (e.g., Grimaldi et al., 1998).

The detailed lithic analysis of Amoreira's layer C (Cura, 2002, Cura et al., 2004 and Cura et al., 2005) shows that quartzite dominates (55.8%) followed by quartz (22.7%) and chert (17.7%). These proportions are argued as most likely related to the high abundance of the first two raw materials and the low frequency of the third in the Tagus deposits. Chert was exploited via prismatic reduction to produce bladelets. Whereas core preparation elements are absent in the assemblage, core maintenance is observed, suggesting its long-distance acquisition with on-site knapping. Quartz reduction was in part similar to that of chert but this raw material was also used to produce a high frequency of flakes. The absence of refitting does not allow one to know if these flakes are related to an independent production or the configuration of prismatic cores. As in the case of Pena d’Água, quartzite is the most abundant raw material. It was exploited by knapping pebbles directly in the cortical platform, without preparation or configuration to produce flakes. The reduction sequence gave preference to detachments that kept some lateral cortex. Some cores seem to have been used also as choppers. Finally, there are some amphibolite fragments associated with polished stone axes and net weights.

When comparing Pena d’Água with Amoreira it seems that the chert assemblages are considerably different: the former has more retouched tools, while the latter shows a prismatic, more regular reduction strategy (closer to Mesolithic and Neolithic patterns). Relevant differences also exist with the quartzite assemblage. Although overall dominant in both cases, at Pena d’Água we do not have any net weight, but only a small granite groundstone that might have been used as anvil or pestle. Comparisons between the quartzite reductions in both sites show differences despite the use of the same river gravels: Pena d’Água has a wider variety of cores while Amoreira only has choppers; Amoreira flakes have almost only cortical platforms while in Pena d’Água the frequency is of 58%; and retouched tools are more diversified at Pena d’Água. However, some similarities also occur, such as cores reuse as choppers and predominance of notches and marginally retouched flakes.

In sum, the acknowledged differences between the two sites may be explained by an admixture of Epipaleolithic and Neolithic components at Amoreira, the first represented by the radiocarbon result, the second by the chert artefacts, potsherds, and polished stone tools. Similar traits in the quartzite technology and typology of both periods may have prevented a clear distinction between the two. This is the case of the Neolithic massive tools/cores component – which were made on large, relatively thin, elongated pebbles with unifacial detachments on one end (or “parallelepiped cores”, as Carvalho, 1998b named them), often showing macroscopic edge damage related with heavy duties – and the high percentages of retouched or notched flakes among the tools. This precise pattern of quartzite exploitation was observed in the Early Neolithic assemblages from the overlying layers at the Pena d’Água itself (Carvalho, 1998b). Thus, in addition to the site initial formation processes identified by Grimaldi et al. (1998), this admixture may be the result of two events: one during the Neolithic occupation, another during heavy disturbance due to modern construction.

Despite recent major developments on the Muge shell-middens, little is still known regarding the lithic management, thus making of Roche's work (1972; see also Marchand, 2001) the main frame of comparison. According to this author, quartzite in the form of cobbles and other non-chert rocks were randomly used throughout the midden sequences, while chert was used extensively in the production of small blades/bladelets that were systematically retouched into multifunctional tools (side-retouched, notched, truncated) or especially transformed into geometric microliths through the use of the microburin technique. Indeed, unpublished results (Paixão, 2014) from the upper levels of Cabeço da Amoreira (7685–7600 cal BP) show that the knapped assemblage is overwhelmingly dominated by chert (98.7%). Similar trends in other Mesolithic assemblages have also been recorded closer to the regional limestone massifs (Fig. 1 top), such as Forno da Telha (Araújo, 1993) and Pessegueiros (Carvalho, 2008b).

However, the striking difference between these and the production methods pre-dating the 8.2 ky event, including that from Layer F of Pena d’Água, is the much higher regularity of blank production (small blades and bladelets), probably due to the introduction of indirect percussion techniques (Marchand, 2001; see also Carvalho, 1998b, Carvalho, 2008b and Carvalho, 2009), and the much higher blank/core ratio. Geometric triangles in particular, a category completely absent in all Epipaleolithic sites from the region, become very abundant in Mesolithic shell-middens such as Cabeço da Amoreira (Paixão, 2014 and Roche, 1972) or Forno da Telha (Araújo, 1993), clearly reinforcing the idea of a consistent and abrupt shift in the technological patterns after the 8.2 ky climatic event.

In sum, the interpretation for the layer F of Pena d’Água here presented result from the congruency between the triangulation of data acquired by different and independent methods which are: •

a detailed and systematic geo-archaeological study that included micromorphology;

Each of these methods has roots in different fields of science, has its own consolidated strategies, is consistently accepted in archaeology and often stand-alone as diagnostic. Therefore, it is free of circular reasoning.

According to our current state of knowledge, the environmental transition from the Pleistocene to the Holocene did not include any abrupt impact on the social, economic, settlement and technological patterns of the hunter-gatherer-fisher communities in Portugal, who retained similar ways of life for millennia. Rather, these patterns changed abruptly – not progressively – at the 8.2 ky cal BP climatic event.

Layer F of Pena d’Água Rock-shelter, dated to ca. 8.1 ky cal BP, thus provides an excellent context for the analysis of the lithic industries predating the mentioned climatic event. In this paper we presented a detailed description of this assemblage and put it in context with evidence from both sides of this boundary event. Our results show that in layer F of Pena d’Água each raw material was reduced according to very distinct strategies, especially regarding the chert and quartz vs. quartzite, and that each raw material was reduced congruently within the available data published from contexts with accepted absolute dates that fit the Epipaleolithic from central Portugal. Moreover, our results show that at this time and place the assemblage is balanced between chert and quartz with a strong coarse raw material component (quartzite), small and irregular bladelets knapped through Upper Paleolithic-like technology, one of them transformed through back retouch and is absent of geometrics. Together, these features make it clearly closer to the Epipaleolithic rather than to the Mesolithic, reinforcing our argument of the abrupt technological change in the transition from the Borial to the Atlantic and not a progressive shift through the Epipaleolithic.

Comparisons highlighted important shifts in raw material acquisition (from opportunistic to strategically obtained) and main technological features (from expedient to standardized). Both aspects can be best explained by a generalized change in human mobility patterns (from residential to logistic) as an adaptive response to changing environments. The identified technological shifts show correlation with other cultural realms of these communities – such as funerary practices, palaeodiets, social organization, etc. – all of which deserve future inquiry.

Because of its coastal position, the western coast of Eurasia is one of the best regions to study such transition. In this context, the layer F of Pena d’Água is an exceptional study case because its chronology falls immediately before the 8,2 cal BP event. Our study clearly shows that its lithic assemblage has technological and typological characteristics that fits it within those known from the Pre-Borial and Borial climatic periods instead of those of the Atlantic, dated from immediately after.

The combination of the available record complemented by the present study shows that the major dramatic shift in the lithic assemblages occurred, not with the passage from the Pleistocene to the Holocene, but during the Holocene, from the Borial to the Atlantic. This dramatic shift was, thus, directly associated with the impact of the 8.2 cal BP event. Presently, the most probable explanation for such fact is that the Pleistocene–Holocene transition was of amelioration in environmental conditions, thus not demanding major changes in the tool-kits. By opposition, the 8.2 cal BP event was an abrupt climatic degradation with strong impact at the environmental and resource availability throughout the Northern Hemisphere, in an epoch when demographic density was high. Such crisis demanded a rapid adaptation of the dispersed hunter-gatherers populations, precipitating their confluence into the areas where resources were more available and stable, forcing an aquatic adaptation, sedentism and social complexity. At this level, the striking changes in the lithic assemblages seen with the 8.2 cal BP event and in the ways to produce them become the formula to build better tools that would suit them to face both the complex environmental and social setting in hand.

More detailed analysis and comparison between contexts dating immediately before and after this climatic event are needed across Eurasia, eastern coast of North America and the northern coast of Africa in order to understand the impact of the phenomenon in the daily live, social and economical patterns of the Early Holocene societies in the Northern Hemisphere.