The earliest convincing evidence for human control of fire dates back to 500 kyr BP in Europe ^[1] and 790 kyr BP in the Near East ^[2]. Campfires recently found at Qesem Cave, Israel, support the idea that hominins possessed a controlled use of fire by 400–200 kyr BP ^[3]. This finding confirms previous views that fire became a widespread technology at the beginning of the Middle Palaeolithic ^[4]. Older possible evidence exists ^[5], but remains controversial ^[4], ^[6]. In Palaeolithic times fire may have been used for heating, cooking or roasting plant and animal food, illumination, and for protection from predators ^[3], ^[7].

Traditionally considered an innovation introduced to Europe by agriculturalists, slash and burn is a well known practice (^[8] and references herein) used to clear and maintain open vegetation, fertilize soil, and facilitate cattle grazing ^[9]. Fire-use for ecosystem management is not restricted to agrarian and herding societies: regular burning of landscape is reported for Native Americans and aboriginal people of Australia (“fire-stick farming”) to create favourable habitats for the foraging and the hunting of small and big game ^[10], ^[11]. Williams ^[12] reports the selective use of fire by Native American tribes to drive big game (deer, elk, bison) into small unburned areas for easier hunting; to attract game which like to dine on young grass ^[10]; to drive rabbits into small areas; and to obtain salt from grasses. Plant management by fire is also reported for production of straight branches for basketry by Californian Indian Tribes ^[13]. Australian Aborigines skilfully use fire in different environments during the annual round of hunting and gathering. They have a well-developed knowledge of how to produce ignition and control fire-spread extent by burning the vegetation at the beginning and the end of the dry season when rain and storms are likely to extinguish fires ^[14], ^[15]. Landscape burning is undertaken for a variety of purposes including clearing thick vegetation to facilitate travel, signalling, controlling insects and vermin, hunting and waging war, ceremonial activities, driving game, regenerating senescent vegetation, smoking animals from burrows and asphyxiating bats in caves (see a synthesis in ^[11]). Fire-stick farming creates fine-grained landscape mosaics with a greater amount of food by increasing small-animal diversity and consequent hunting productivity ^[16]. Hill and Baird ^[17] report the use of fire by aboriginal Kuku–Yalanji groups of the Australian rainforest during the dry season to encourage the growth of Cycas media whose seeds constitute an important source of carbohydrates.

Mesolithic hunter-gatherers are suspected to have manipulated vegetation in northwest Europe ^[18] by deliberately setting fires to improve the sight-lines for missile-based hunting and/or change vegetation to attract game species ^[19]. Bos and Urz ^[20] also consider Early Mesolithic people responsible for woodland opening by deliberate burning of the vegetation. Although it is likely that these techniques were not invented by Mesolithic hunter-gatherers, but rather stemmed from earlier traditions, evidence supporting this hypothesis remains exceedingly scant.

It has been proposed ^[21], ^[22] that a greater control and more extensive use of fire is one of the modern behavioural innovations that emerged in Africa among Modern Humans, that favored their spread throughout the world, and that determined their eventual evolutionary success. We would expect, if extensive fire use for ecosystem management were a component of the modern human technical and cognitive package, as suggested by some authors ^[23]–^[25], to find major disturbances in the natural biomass burning variability associated with and following the colonisation of Eurasia by Modern Humans.

Fire regime during this time span is poorly known due to the limited number of well dated and continuous charcoal records. A large increase in fire activity around 50 kyr cal BP, recorded in marine cores collected in the Pacific Ocean, has been interpreted as the result of burning produced by Modern Human populations colonizing Southeast Asia ^[26], ^[27]. Similarly, high incidence of biomass burning indicated by charcoal records in Australia between 60–45 ka has been attributed to the widespread use of fire for vegetation clearance during the initial colonisation of this continent ^[28], ^[29]. Dryness recorded in Australia after 45 ka has been interpreted as a feedback associated with changes in land-surface properties caused by human modification of the vegetation cover through fire ^[30], ^[31]. These interpretations, however, have faced repeated challenges in the last few years ^[32]–^[36].

This paper addresses for the first time the question of whether Palaeolithic communities modified natural biomass burning variability during Marine Isotope Stage (MIS) 3 and 2 (59.4–27.8 kyr cal BP and 27.8–14.7 kyr cal BP) in Europe and in particular whether Modern Human populations colonising this large area and replacing Neanderthals at the so-called Middle to Upper Palaeolithic transition introduced fire as a tool for ecosystem management, possibly leading to an advantage over Neanderthals.

To explore this issue, analyses of microcharcoal preserved in two deep-sea cores located off Lisbon (MD95-2042) and Bordeaux (MD04-2845) (figure 1) were undertaken to assess changes in biomass burning in Southwestern Iberia and Western France, respectively (see Material and Methods section). Multiproxy studies of core MD95-2042 ^[37]–^[43] and MD04-2845 ^[44] have documented the impact of the Dansgaard-Oeschger millennial-scale climatic variability and Heinrich Stadials (HSs) in the mid-latitudes of the eastern North Atlantic Ocean and Western Europe. Greenland Stadials (GSs) are characterised by a decrease in sea surface temperature (SST), high δ¹⁸O values of the planktonic foraminifera Globigerina bulloides and the development of the polar foraminifera Neogloboquadrina pachyderma (s.) left coiling. Greenland Interstadials (GIs) were characterised by warmer SST and low planktonic δ¹⁸O values. Extreme cooling episodes corresponding to HSs are marked by peaks of ice rafted debris (IRD), the strongest increase in planktonic δ¹⁸O values and peaks in magnetic susceptibility.

The age models of these two sequences are well constrained by numerous ¹⁴C AMS dates. These records provide a unique opportunity to correlate climatic fluctuations in the North East Atlantic region, vegetation change, palaeofire history, and human population dynamics during the MIS 3 and 2.

Concentrations of microcharcoal surface area (CCsurf) in core MD95-2042 (figure 2.1) reveal that the evolution of fire regime in Southwestern Iberia was in phase with vegetation shifts between GIs and GSs. High fire regime (see materials and methods for a definition) was contemporaneous with relatively warm and wet climatic phases (GIs), characterised by the development of an open Mediterranean forest and heathland. Low fire regime is observed during cold and dry climatic phases (GS including HS) characterised by semi-desert vegetation (Artemisia, Chenopodiaceae, Ephedra) ^[68]. As total plant biomass in forest and heathland communities is generally higher than open ground formations ^[69], these changes indicate the association of increased biomass accumulation with higher fire regime. CCsurf and biomass index, as reflected in the sum of arboreal Mediterranean taxa and heathland percentages, are strongly correlated (figure 3) and best fits a logarithmic function (b = 0.79, P<0.0001). This pattern indicates that in this region fuel availability, determined by climatically driven variations in vegetation biomass, is the main factor behind fire regime variation.

CCsurf from core MD04-2845 located off Bordeaux (figure 2.2) broadly follows the same pattern observed in the southern core for GSs and GIs. However, the correlation between warm/wet climatic conditions and corresponding fuel/fire regimes are not, after 40 kyr cal BP, as strongly expressed as in core MD95-2042. All HSs also reveal a more complex evolution, with a peak phase of CCsurf bracketed by two episodes of low fire regimes. Vegetation in Western France is mainly characterised by the expansions of coniferous forest (Pinus, Abies, Picea) and Atlantic forest (deciduous Quercus, Betula) during GIs and steppic vegetation during GSs ^[44]. The correlation of CCsurf and Arboreal Pollen (b = 0.525, P<0.0174, excluding Pinus) suggests that the fire regime of the last glacial period was still determined by changes in biomass. Pinus woodland certainly contributed to the biomass burned but the overrepresentation of this pollen taxon in marine cores precludes effective correlation tests without pre-corrections. Note that some GIs are characterised by relatively weak biomass/low fire regime and some HSs by relatively strong biomass/high fire regime. This is the case with HS 6, during which the expansion of Picea and Pinus woodlands may reflect the establishment of a fire-prone ecosystem such as that of boreal forest ^[70]. The three-phase structure that characterises HS suggests a more complex relationship in Western France between factors conditioning fire regimes (biomass, vegetation composition, precipitation, temperature, and lightning occurrence) ^[70]. The first decrease in fire regimes within HS may be due to a concomitant reduction in fuel and an increase in humidity, whereas the subsequent peak could reflect an increase in drought conditions and/or lightning associated with increasing fuel. The final reduction in fire regime could be the result of a new relative increase in humidity.

The described pattern persists after 41–42 kyr cal BP (GIs 9–10) in core MD04-2845, and after 38 kyr cal BP (GI 8) in core MD95-2042, the time accepted by most authors for the arrival of Modern Humans in France and Southern Iberia, respectively (figure 2). The only possible exception is represented by CCsurf signal for the beginning of GS 7 in the French core, which appears anomalously high for a GS. This, however, happens 6,000 years after the arrival of Modern Humans in Western France and is not recorded in the Southwestern Iberian margin. Both Neanderthals and Modern Humans are associated in Europe with periods of relatively high biomass and fire regime and periods of weak biomass and low fire regime (Figure 3). This implies that the observed large scale variations in fire regime were not determined by the population replacement. Figure 4 reveals a long term decreasing trend of fire regime in Western France between 70 and 25 kyr cal BP and, in Southwestern Iberia, an increasing fire regime until GI 10 followed by a long term decrease until 14 kyr cal BP. These long term trends also do not match the timing for the Neanderthal/Modern Human replacement.

Understanding cause-and-effect relationships between human-landscape interaction and regional fire regimes in Southeast Asia and Australia remains elusive despite both the abundant charcoal records available and the relatively simple settlement histories of these regions compared to Europe.

The strong increase in microcharcoal concentration observed in marine cores from Southeast Asia has been used to argue in favour of an extensive use of fire at around 50 kyr cal BP by modern human colonisers to clear vegetation and facilitate hunting ^[26], ^[27]. This would have perturbed natural fire regimes and dramatically changed vegetation composition ^[71]. Pollen analysis of the archaeological sequence from Niah cave (Malaysia), dated to the MIS 3, reveals an increase in Justicia, a fire-favoured taxon, associated with forest rich phases ^[72]. This has been interpreted as reflecting management of forest by modern human settlers to stimulate the growth of tubers and edible plants, and hunt animals attracted to clearings ^[73]. A similar interpretation has been suggested for the increase in biomass burning indicated by charcoal records in Australia between 60–45 ka and the following dryness observed in the area ^[30], ^[31]. Several lines of evidence, however, have recently challenged the anthropogenic hypothesis: 1) major changes in fire regimes have been documented by charcoal records for periods prior to the arrival of modern humans ^[32], ^[34]; 2) changes in the ENSO variability and its impact on rainfall may explain the increase in natural fire ^[33]; 3) a broadly-synchronous transition to more xerophytic vegetation is observed in New Caledonia, a region not colonised until 3,000 years ago ^[35]; 4) climate-model experiments show that feedbacks associated with fires, even catastrophic fires, would have been insufficient to dry out the region ^[36].

Contrary to Australia, where colonisers penetrated an unpopulated territory, Modern Humans entering Europe met a human population, the Neanderthals, well adapted to mid-latitude environments. Neanderthal knowledge of pyrotechnology is fully compatible with the hypothesis of their use of fire for ecosystem management. Numerous hearths have been found at Mousterian sites and in a number of cases there is clear evidence that they were deliberately constructed, maintained, and re-used ^[74]–^[77]. Neanderthals also used fire to harden wooden spears ^[78] and to process birch pitch to haft stone tools ^[79], ^[80].

Fire use for ecosystem management could be 1) a subsistence strategy already in possession of Neanderthals when Moderns arrived in Europe, 2) an innovation introduced in this region by moderns, 3) developed by them after their arrival, or 4) absent from both technical packages either because not a part of their cultural heritage or because ineffective in the European latitudes during MIS 3-2.

Our results show that contrary to Southeast Asia, no major increase in fire regime is recorded in Southwestern Iberia or in Western France at the onset or after the colonisation of these regions by Modern Human populations. CCsurf values associated in Southeast Asia with Modern Human impact are twice as great as the highest figures recorded in the same sequences for the period before colonisation by Modern Humans. Such a dramatic increase is not observed in our records. Also, no shift is observed in the vegetation apart from that expected by the impact of the millennial scale climatic variability on plant communities, and no increase in taxa that might be related to an increase in fire. Although the Southeast Asian and the European trends are difficult to compare considering the different latitudinal, paleoclimatic and vegetation settings, the coincidence in the former area between the peopling event and the increase in biomass burning makes it conceivable that the two phenomena are related in some way.

Our results strongly argue against the view that Neanderthals and Modern Humans were the driving factor of the large scale variations in fire regime observed in our records, which were clearly governed by the D-O millennial-scale climatic variability and its impact on fuel load. However, we cannot rule out at this stage the possibility that either one or both populations used fire for ecosystem management in ways that did not significantly affect the natural fire trend.

A number of quantitative analyses of the effects of Aboriginal fire treatment in Australia (e.g. ^[15], ^[16], ^[81]) have shown that under an anthropogenic regime the frequency of fire increases, but the intensity and extent of burns declines dramatically, resulting in a decrease in biomass burning (^[15], ^[81], ^[82]. Management of fire by Aboriginal people leads to fragmentation of the landscape, reducing fuel and therefore reducing fire. In contrast, the cessation of traditional aboriginal burning practices enhances vegetation development and fuel accumulation, allowing the occurrence of large and intense fires.

If this were the case in Europe and our record could detect such a change we should find a decrease in the biomass/CCsurf ratio after the introduction of this ecosystem management practice. The question, however, is whether such a decrease may be detected by regional sequences. Marlon et al. ^[83] have shown that the decrease in biomass burning, produced by expansion of intensive grazing, agriculture and fire management over the past 150 years, is recorded at a regional and global scale, but in this case, of course, we are dealing with a massive impact of humans on a variety of ecosystems.

Considering the relatively low population density predicted for Palaeolithic Europe ^[84], anthropogenic ignition sources may have simply amplified or otherwise subtly altered natural rhythms of fire frequency and intensity. If this were the case, a shift to an anthropogenic fire regime could only be detected by comparing the regional fire signal given by marine cores with multiple well dated continental sequences recording local fire frequency, intensity and resulting landscape patchiness.

Competing Interests: The authors have declared that no competing interests exist.

Funding: The work of A.-L. Daniau was funded by a BDI-CNRS-Région Aquitaine fellowship. This study was supported by the RESOLuTION (ESF-Eurocores-Euroclimate), the OMLL (ESF-Eurocores) and ARTEMIS programmes, Région Aquitaine project 20040206003N and the Université Bordeaux1 BQR/RAMSEMAR (Reconnaissance Automatique des Micro-objets contenus dans les SEdiments MARins)/project 2003. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

We thank S. Rigaud for her help in the preparation of figure 3 and W. Fletcher, W. Banks and S. Harrison for the critical reading of previous versions of the manuscript. We are grateful to two anonymous referees, R. J. Nevle, and the academic editor for their constructive criticisms, which greatly improved this paper.