Palaeoenvironmental interpretation

Descriptions of the preferred habitats for vegetation and NPP types and references are provided as Supplementary Tables 1 and 2.

The high abundance of Cyperaceae (sedges) pollen in the assemblages could have various palaeoenvironmental interpretations, given that Cyperaceae is a large family with a wide ecological range reaching from arctic tundra to tropical forests and seasonally wet grasslands (Simpson et al., Reference Simpson, Yesson, Culham, Couch, Muasya, Hodkinson, Jones, Waldren and Parnell2011). We assume that most Cyperaceae pollen reflects wetland environments, considering the overall MAR-1 pollen flora composition and palaeobotanical evidence retrieved from previous studies (Mädler, Reference Mädler1971; Nickel et al., Reference Nickel, Riegel, Schönherr and Velitzelos1996; Field et al., Reference Field, Ntinou, Tsartsidou, van Berge Henegouwen, Risberg, Tourloukis, Thompson, Karkanas, Panagopoulou and Harvati2018). Notably, plant macrofossil analysis on the Early Pleistocene lignite from Megalopolis (Mädler, Reference Mädler1971) identified various Cyperaceae wetland taxa such as Cyperus, Eriophorum, Scirpus, Eleocharis spp., and Rhynchospora. At MAR-1, Carex spp., Eleocharis palustris, Cladium mariscus, and Scirpus lacustris have been documented (Field et al., Reference Field, Ntinou, Tsartsidou, van Berge Henegouwen, Risberg, Tourloukis, Thompson, Karkanas, Panagopoulou and Harvati2018). Furthermore, high abundances of Cyperaceae—and Typha spp. (cattails) to a lesser extent—imply telmatic conditions (i.e., low water levels), given that typical wetland sedges and Typha latifolia prefer high soil moisture habitats, rich in nutrients (e.g., Uria-Diez et al., Reference Uria-Diez, Gazol and Ibáñez2014; Rasran et al., Reference Rasran, Hacker, Tumpold and Bernhardt2021). Accordingly, the predominance of sedges in the lower deposits of the sequence (UB10; MARPZ1; Fig. 3) indicates the development of fen (mire) peatlands during the formation of the lignite LII in MIS 13a. The peat originated primarily from the decomposition of Cyperaceae. Green algae might have also contributed to the formation of gyttja accumulations encountered in UB10 (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021), given that they can produce such accumulations when occurring in high concentrations (Shumilovskikh et al., Reference Shumilovskikh, O’Keefe, Marret, Marret, O’Keefe, Osterloff, Pound and Shumilovskikh2021).

Shifts in the abundances (high/low) of floating-leaved and submerged taxa indicate trends in the water column, reflecting the raising/lowering of water levels. Therefore, their fluctuating values between MARPZ1 and MARPZ2 suggest a transition from shallow wetlands where peat accumulations occurred in MAR-1 to temporarily open-water environments likely reaching up to 2 m depth (Dimopoulos et al., Reference Dimopoulos, Sýkora, Gilissen, Wiecherink and Georgiadis2005). NPP data reinforce this hypothesis; the decreasing green algae influx implies rising water levels at ∼478 ka (Fig. 6, Supplementary Table 2). This evidence suggests a variable environment at MAR-1, at the MIS 13/12 transition (interglacial–glacial).

In the early MIS 12, the MAR-1 site was characterised by marshy wetlands dominated by reed beds around the lakeshore area and shallow-water environments, as deduced from low abundances of wetland taxa and increased influx of dry environment fungal indicators (Figs. 3 and 6). However, in MIS 12a, the pollen assemblages—including those corresponding to the archaeological horizon—document a diverse, mosaic landscape. Thick stands of P. australis grew on periodically flooded soils, forming marshes on the littoral zone of the seasonal lake/pond near MAR-1. Sedges and cattails grew closer to the pond’s edges, as they require higher soil moisture. Furthermore, enhanced abundances of hydrophytes imply rising water levels (Fig. 3). The riparian zone was occupied by patches of riparian trees, which expanded between ∼424 and 422 ka, indicating the establishment of alluvial woodlands at MAR-1. At the same time, the increased influx of wood-decay and freshwater fungi further attests to the presence of shallow, freshwater habitats and woodlands at MAR-1 (Supplementary Fig. 2).

During MIS 11c, the previous mosaic structure of MAR-1’s wetland became simplified. The alluvial woodland at MAR-1 turned into an open dry marshland where thick stands of P. australis thrived around the pond/lake. Low abundances of Cyperaceae, Typha spp., and riparian trees suggest lower soil moisture. The rich content of green algae in MARPZ9 indicates a shallow, eutrophic water environment (Fig. 6). From 420 ka onwards, the surge of ferns implies the transition from a marsh to a fen/bog habitat where peat accumulations occurred during the UB1 formation. Identification of fern spores on a family/genus level was not feasible, albeit due to the correspondence of MARPZ9 with UB1 (Supplementary Fig. 1), fern presence is strongly associated with the wetland type. Therefore, the formation of the lignite LIII occurred in a shallow, stagnant waterbody where peat accumulations were derived from the decomposition of pteridophytes, helophytes, and other hydrophytes. The coeval presence of wood-decay and type 200 fungi (Fig. 6, Supplementary Fig. 2) suggests that temporal desiccation of the fen possibly activated these decomposers to decay organic material.

Vegetation and fire dynamics in response to palaeoenvironment and climate changes

The interval spanning from ∼485 to 478 ka (MIS 13a; MARPZ1–MARPZ2), the high abundances of mesophilous and montane tree taxa and the low amount of steppe elements in the catchment area of Megalopolis suggest a climate regime with relatively high temperatures and sufficient moisture availability for tree growth towards the termination of the MIS 13 interglacial period. Geochemical data from MAR-1’s sediments further reinforce this palynological evidence for wetter conditions. Particularly, the Rb/Sr ratios—which are considered a precipitation proxy (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021)—gradually increase through MARPZ2 (Fig. 4), attesting to a humid climate.

On a broader regional scale, our findings and climate inferences for this interval agree well with other pollen archives from the Mediterranean region. Pollen records from Tenaghi Philippon (northeast Greece; e.g., Tzedakis et al., Reference Tzedakis, Hooghiemstra and Pälike2006) and Vallo di Diano (southern Italy; Ermolli and Cheddadi, Reference Ermolli and Cheddadi1997) (Fig. 1) exhibit a similar vegetation character marked by mesophilous trees (mainly Quercus species), implying warm and wet conditions during MIS 13.

Regarding the local environmental conditions, low abundances of wetland taxa (i.e., submerged and floating-leaved taxa; Fig. 4) and the synchronous high influx of green algae until ∼479 ka point to eutrophic conditions in MAR-1’s waterbody at the lowlands of the basin during MIS 13a. Indeed, in shallow lakes, eutrophication is marked by the dominance of phytoplankton and the loss of submerged vegetation (Dong et al., Reference Dong, Yang, Li, Li and Song2014). Rising temperatures, shallow waters, light intensity. and increased nutrient concentrations promote eutrophication in aquatic environments (Nazari-Sharabian et al., Reference Nazari-Sharabian, Ahmad and Karakouzian2018). Therefore, this interval of excessive algal blooms at MAR-1 and the coincident increased temperatures at Megalopolis suggest that the local environment was influenced by climate change towards the MIS 13/12 transition.

Between ∼479 and 477 ka, a significant rise in wetland taxa (Fig. 4)—except a short-term drop in their abundances at ∼478 ka (see below)—indicates freshwater environments with still (to slow-flowing) waters at MAR-1. Specifically, the submerged vegetation in shallow lakes, like the one found at MAR-1, is usually associated with high water transparency (e.g., Moss et al., Reference Moss, Madgwick and Phillips1996; Zhang et al., Reference Zhang, Liu, Qin, Shi, Deng and Zhou2016), because its growth and survival is directly related to light availability in the water column (Dennison et al., Reference Dennison, Orth, Moore, Stevenson, Carter, Kollar, Bergstrom and Batiuk1993). The coeval decrease of green algae implies mesotrophic conditions and improved water oxygenation (Fig. 6). The presence of sponges and gastropods in the corresponding sediments further attests to favourable conditions in the waterbody at that period (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021). Possibly, the dense sedge stands near the palaeolake and the enhanced forest cover in the surroundings of the basin played a crucial role in improving MAR-1’s aquatic environment by eliminating potential sediment input from inflowing streams, thereby preventing sediment resuspension inside the lake and the subsequent deterioration of the underwater light regime. This hypothesis is confirmed by shoreline reconstructions at MAR-1, suggesting that the fine-clay composition of the corresponding deposits (UB10–UB9) and the absence of large sediment particles are associated with the presence of a thick vegetation belt at the lake’s margins (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021).

The transition from a floating-leaved to submerged-dominated vegetation after ∼478 ka is likely associated with a shift in ecological conditions inside the water environment. Specifically, higher oxygen concentration and light availability in the water column due to Nymphaea’s decline possibly led to the Myriophyllum spp. expansion. Floating-leaved plants in high abundances can substantially reduce oxygen and underwater light, affecting the growth of submerged plants (Caraco et al., Reference Caraco, Cole, Findlay and Wigand2006; Abarike et al., Reference Abarike, Amenyogbe, Ampofo-Yeboah, Abobi, Atindana, Alhassan and Akongyuure2015). Increasing Mn/Fe ratios (Fig. 4) corroborate the notion of higher oxygenation in the waterbody.

At the same time, high influx rates of macrocharcoal particles and pollen taxa (Fig. 7) suggest local fires in the Megalopolis Basin during MIS 13a. High values of terrestrial and waterside taxa and wood-decay fungi (Supplementary Fig. 2) point to sufficient living and dead plant material accumulations for burning in the catchment. The first peak of large-sized particles at ∼482 ka (MARPZ1) represents primary charcoal, given that the corresponding clayey sediments (UB10) were deposited in a stable waterbody protected by sediment resuspension. The low influx of micro- and mesoscopic charcoals further supports the hypothesis for a low-energy deposition environment and local fire signal. Likewise, the second macrocharcoal peak at ∼478 ka (MARPZ2) constitutes mainly primary charcoal, given the site’s high build-up of vegetation biomass. The notion of a local fire signal is strengthened by a short-lived event of local dry conditions documented at ∼478 ka; the temporal degradation of wetland taxa (Fig. 4) could have partially been a consequence of the disturbance caused by local fire activity in MAR-1. Nevertheless, secondary charcoal representing more than one fire episode cannot be excluded, considering the water-level fluctuations in MAR-1’s waterbody throughout MARPZ2 and the subsequent sediment mixing. The rise of mesoscopic charcoals might (partly) result from the secondary fragmentation of macrocharcoals due to sediment mixing.

A substantial change in the floristic composition of the Megalopolis Basin characterises the following period spanning the MIS 12 glaciation. More specifically, the palynological evidence indicates a significantly reduced forest cover in the catchment area from ∼478 ka upwards, with steppe elements dominating the basin and scattered patches of tree stands being restricted to the lowlands during MIS 12c (∼478.5–464.5 ka; Fig. 5). This apparent landscape opening reflects cold and dry conditions during MIS 12c. The sporadic presence of thermophilous Mediterranean taxa suggests that they reached close to their tolerance threshold for survival in this interval, thus further supporting the inferred low temperatures and moisture deficit. Geochemical data (decreasing Rb/Sr ratios; Fig. 4) from the corresponding deposits also corroborate the notion of increased aridity during MIS 12c.

Such climate conditions inferred by our pollen record for MIS 12c are also documented by other pollen archives from the south Balkans for the same period. Pollen data from Tenaghi Philippon (northeast Greece; e.g., van der Wiel and Wijmstra, Reference van der Wiel and T.A1987; Tzedakis et al., Reference Tzedakis, Hooghiemstra and Pälike2006; Pross et al., Reference Pross, Koutsodendris, Christanis, Fischer, Fletcher, Hardiman and Kalaitzidis2015) and Lake Ohrid (southwestern Balkans; Koutsodendris et al., Reference Koutsodendris, Kousis, Peyron, Wagner and Pross2019) (Fig. 1) show that the most severe tree-population contractions of the past 500 ka generally occurred during MIS 12 . Especially for the early part of MIS 12 (i.e., ∼478–457 ka; Railsback et al., Reference Railsback, Gibbard, Head, Voarintsoa and Toucanne2015), pollen data from Lake Ohrid show an average of 29.9% for arboreal pollen (AP) of which mesophilous taxa constitute 15.8% (Koutsodendris et al., Reference Koutsodendris, Kousis, Peyron, Wagner and Pross2019). For the Megalopolis catchment, the data exhibit more intense tree-pollen reductions; AP taxa account for 15.9% on average (mesophilous taxa 13% mean) for the interval between ∼478 and ∼464.5 ka (while the late part of MIS 12a is missing due to the lithological hiatus). Interestingly, pollen-based climate estimates for Lake Ohrid indicate mean temperatures of 17.6°C for the warmest month (Koutsodendris et al., Reference Koutsodendris, Kousis, Peyron, Wagner and Pross2019), which are similar to those inferred by the ostracod assemblages from MAR-1 for the same period (MARPZ4–MARPZ5/UB7), that is, summer water temperatures below 18°C (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021). Therefore, these lines of evidence imply that the moisture deficit was possibly the limiting factor for tree growth in Megalopolis rather than the temperature, reflecting more pronounced dry conditions at the southernmost latitudes of the south Balkans during MIS 12c.

While the terrestrial flora within the Megalopolis watershed was influenced by climate change, the wetland ecosystem was also affected by other environmental factors, such as changes in the hydrologic regime, limited underwater light, and eutrophication (e.g., Zhang et al., Reference Zhang, Liu, Qin, Shi, Deng and Zhou2016). In MAR-1, the degradation of wetland taxa (Fig. 4) reflects the dynamic character of the depositional environment between ∼476 and 464.5 ka. Micromorphological analysis of the sequence indicates that the deposition of its lower part (i.e., UB9–UB7) was occasionally affected by seasonal storm events, which in turn triggered hyperconcentrated flows reaching MAR-1’s aquatic environment (Karkanas et al., Reference Karkanas, Tourloukis, Thompson, Giusti, Panagopoulou and Harvati2018). Furthermore, a recent palaeoenvironmental study for the corresponding sediments implies a fluctuating water table owing to the unstable riverine-deltaic environment that prevailed in-between UB8–UB7 as deduced by low total inorganic carbon TIC values and ostracod taphonomy (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021). The reduced arboreal vegetation in the catchment could have also facilitated the input of external sediment load and dissolved nutrients in the water during MIS 12c. Interestingly, the waterside taxa (Fig. 4) remained unaffected by this unstable environment; P. australis thrives among emergent taxa, as it is a strong competitor in wetland environments and can persist in wave action and currents, protecting other emergent plants (Packer et al., Reference Packer, Meyerson, Skálová, Pyšek and Kueffer2017).

Short-lived increases of riparian trees and wet meadow herbs (Fig. 3) in MIS 12c suggest temporarily higher soil moisture in the riparian zone. At the same time, however, from a catchment-scale perspective, the tree-pollen taxa exhibit extremely low abundances, and geochemical proxies show low precipitation (Rb/Sr ratios) and increasing alkalinity (Mg/Ti ratios) (Fig. 4), reflecting overall dry conditions and low evaporation. A plausible scenario for temporarily moist soils in MAR-1 might have been the higher availability of near-surface groundwater. The presence of small streams (possibly developed from melting of mountain snow covers) in the western part of the basin and the proximity of MAR-1 to their inputs—as inferred by the sedimentological analysis of the clastic sequence (UB9 to UB2; Karkanas et al., Reference Karkanas, Tourloukis, Thompson, Giusti, Panagopoulou and Harvati2018)—could have allowed a more or less continuous recharge of the groundwater table in the low-lying areas of the catchment. Specifically, the sandy deposits of the sequence (UB8–UB6) would have facilitated the infiltration from water flows. Moreover, under the prevailing glacial conditions in Megalopolis, lower evapotranspiration rates might have also facilitated the rise of the groundwater table in the lowlands. Hence, temporarily damp soils can explain the vegetation development in the riparian zone under cold and particularly dry conditions. The same hypothesis was made for the formation of wetlands in the Po Plain (NE Italy) during the last glacial maximum by Miola et al. (Reference Miola, Bondesan, Corain, Favaretto, Mozzi, Piovan and Sostizzo2006), who inferred a higher groundwater table owing to the lower evapotranspiration.

While no environmental information for MAR-1 exists for the late MIS 12c and MIS 12b due to the hiatus, pollen data become available again in MIS 12a (∼438 ka). Particularly, the surge of the nitrophilous pioneer species P. aviculare—which shows substantial plasticity in response to disturbed areas (e.g., Meerts, Reference Meerts1995)—and of soil erosion fungi at ∼438 ka (Figs. 5 and 8) attest to an erosional event at the site. This evidence agrees with geochemical data and shoreline reconstructions for MAR-1; towards the UB7–UB6 (top MARPZ5) transition, all elemental and biological components show a sharp drop in their values, reflecting a change in the depositional regime, as argued by Bludau et al. (Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021). According to these latter authors, this event is accompanied by the surge of dust input in MAR-1’s sediments and is attributed to a period of warmer conditions that caused significant flooding; presumably, large volumes of meltwater from retreating glaciers—which had developed on the Peloponnese mountains during MIS 12 (Leontaritis et al., Reference Leontaritis, Kouli and Pavlopoulos2020)—eroded, transported, and deposited sediments into the basin. All these lines of evidence confirm that an erosional event caused these abrupt changes.

Regarding the Megalopolis’s fire regime, charcoal data suggest regional fire signal during MIS 12c (Fig. 7). The hypothesis for a distant fire source is further supported by low influx rates of pollen taxa, suggesting low biomass availability for burning in the catchment. Another plausible scenario could be that the increased microcharcoal influx reflects allochthonous input from the Megalopolis catchment, introduced in the sediments due to the unstable depositional environment. Patterson et al. (Reference Patterson, Edwards and Maguire1987) report increased microcharcoal values a few decades after a local fire event because of burned fallen down vegetation alongside the lake’s margin, while surface runoff following the fire event also transported secondary charcoal the following years. Considering the prevailing glacial conditions in the broader area of the Peloponnese during MIS 12 (Leontaritis et al., Reference Leontaritis, Kouli and Pavlopoulos2020), the latter scenario seems more realistic. Minor peaks of macrocharcoals are considered secondary and associated with MAR-1’s depositional history, given the deficit of plant fuels and local warm conditions—which favour fires—at Megalopolis at that time.

During the late part of MIS 12, the persistence of open vegetation (mainly steppe) in the catchment suggests the prevalence of a dry and cold climate at Megalopolis (Fig. 5). Such conditions are also documented by geochemical and biological proxies from the corresponding deposits (UB4–UB3) between ∼433 and 424 ka. Specifically, the ostracod assemblage indicates cold water temperatures (10–15°C), and decreasing Rb/Sr ratios imply low precipitation during this period (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021; Fig. 4). Nonetheless, superimposed on this long-term vegetation pattern during MIS 12, open mixed mesophilous woodlands temporarily developed between ∼433 and ∼430 ka, the interval encompassing the archaeological horizon (Fig. 5). The low, yet constant, presence of the thermophilous Mediterranean elements might be attributed to their local persistence in protected microhabitats at MAR-1’s wetland area, suggesting that the cooling was milder at the low-lying areas of the basin. Moreover, the coeval rise of the floating-leaved Nymphaea alba (Fig. 3) points to a temporal increase in moisture availability within the catchment in-between ∼434–433 ka, given that Nymphaea spp. habitats are regarded as precipitation dependent (Nzei et al., Reference Nzei, Mwanzia, Ngarega, Musili, Wang, Chen and Li2022). This hypothesis is reinforced by a synchronous short-lived moist episode (high Rb/Sr ratios; Fig. 4). Therefore, despite the overall glacial conditions, temporarily increased temperatures and sufficient moisture availability favoured the tree growth within the basin during MIS 12a.

The hypothesis of temperate conditions is further supported by palaeobotanical data from the corresponding sediments (i.e., UB4–UB3). Wood macroremains of mesophilous trees (i.e., Quercus spp. and Ulmus/Zelkova) and the presence of cf. Palmae in charcoal and phytolith assemblages attest to a relatively warm temperate climate between ∼433 and 424 ka (Field et al., Reference Field, Ntinou, Tsartsidou, van Berge Henegouwen, Risberg, Tourloukis, Thompson, Karkanas, Panagopoulou and Harvati2018). The present palynological analysis also documents the presence of identical mesophilous tree taxa in the same interval (MARPZ6–MARPZ8) (Kyrikou, S., unpublished data). According to Birks and Birks (Reference Birks and Birks2000), the presence of plant macrofossil taxa in an assemblage confirms that the respective pollen taxa identified in this assemblage originate from the catchment. From ∼428 ka onwards, the dry steppe biome gradually became replaced by grasslands, and the riparian trees increased, suggesting relatively humid conditions and damp soils at MAR-1.

Interestingly, pollen evidence from higher latitudes of the south Balkans indicates tree-population crashes during MIS 12a (∼442–424 ka), suggesting particularly cold and dry conditions. Especially, pollen data from Tenaghi Philippon document a total collapse of tree taxa (AP = 0% min.) in MIS 12a, the only one recorded at Tenaghi Philippon during the past 1.3 Ma (e.g., Tzedakis et al., Reference Tzedakis, Hooghiemstra and Pälike2006; Koutsodendris et al., Reference Koutsodendris, Dakos, Fletcher, Knipping, Kotthoff, Milner and Müller2023). Likewise, Koutsodendris et al. (Reference Koutsodendris, Kousis, Peyron, Wagner and Pross2019) report no forest expansion events (i.e., lowest mean AP abundances = ∼9% and min. 1.6%) in the catchment area of Lake Ohrid during MIS 12a, as well as the lowest temperature estimates for the entire MIS 12 (i.e., 0.2°C mean annual and 13.9°C mean warmest temperatures). These distinct differences in vegetation responses during MIS 12a, even within the south part of the Balkan Peninsula between its higher (i.e., Lake Ohrid and Tenaghi Philippon) and lower latitudes (i.e., the Megalopolis Basin; AP = ∼20% mean, 5% min.; mesophilous 15.4% mean in-between ∼438–424 ka), reflect regional-scale patterns concerning tree-population expansions and contractions. In addition, the mean summer temperatures at Megalopolis, as inferred by ostracods (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021), were higher by 1.1°C compared with those reported for Lake Ohrid. These lines of evidence suggest that climate conditions at the southernmost point of the Balkans were less severe during MIS 12a, thus allowing mesophilous and thermophilous trees to persist in Megalopolis; however, new climate-proxy records from MAR-1 and the Megalopolis Basin in general are necessary to draw robust conclusions about the climate, and particularly, the precipitation regime, as well as the responses of the Megalopolis ecosystem to climate oscillations during MIS 12a.

Despite prevailing climatic conditions that influenced the vegetation development at Megalopolis, local environmental parameters also controlled the wetland ecosystem at MAR-1 during MIS 12a. Specifically, the mosaic of habitats and mainly riparian trees (from ∼433 ka onwards; Fig. 3) was developed during a period marked by low precipitation (Fig. 8). A plausible scenario could be that the presence of a small spring/stream (surface water) close to MAR-1 during MIS 12a (Bludau et al., Reference Blackwell, Sakhrani, Singh, Gopalkrishna, Tourloukis, Panagopoulou and Karkanas2021) fed or recharged the groundwater table through seasonal seepage from its channel, thereby regulating periodically wet conditions in MAR-1 during the period of hominin occupation at the site. Furthermore, in the case of wetland taxa, their constant persistence could have also been facilitated by increasing oxygenation (Mn/Fe) and mesotrophic conditions (low algae content) in the water column (Figs. 4 and 6), given that a clear water environment with ample underwater light is an essential precondition for their growth. Such conditions are confirmed by the enhanced presence of diatoms and gastropods and the rich diversity of ostracods in the corresponding deposits (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021) in MIS 12a.

Towards the MIS 12/11 transition (termination V), the increasing forest cover and the establishment of alluvial woodlands (riparian trees; Fig. 3) at ∼424 ka point to a shift towards temperate conditions and sufficient edaphic humidity in the riparian zone. Peak values of Salix at ∼422.5 ka are synchronous with the rise of mesophilous trees and increased precipitation (Rb/Sr ratios; Figs. 4 and 5), also reflecting adequate moisture availability at Megalopolis. A similar pattern of Salix pollen has been recorded at Lake Ohrid between ∼427 and 424 ka (Kousis et al., Reference Kousis, Koutsodendris, Peyron, Leicher, Francke, Wagner, Giaccio, Knipping and Pross2018). According to those authors, this evidence signals the onset of the early interglacial forest phase that is characterised by increasingly warmer (wetter) climate and the presence of nutrient-poor soils. Considering that Salix is a pioneer tree with a competitive advantage for growth on poor soils, its dominance in MARPZ8 might be related to the onset of the MIS 11 interglacial period and increasing temperatures in the Megalopolis Basin.

Climate-proxy records from the broader area of the south Balkans evidence a similar climate regime to Megalopolis during the MIS 12/11 transition. Specifically, in Lake Ohrid (Fig. 1), this transition is marked by an abrupt increase in the mean annual precipitation (600 mm) and temperature (7°C) (Kousis et al., Reference Kousis, Koutsodendris, Peyron, Leicher, Francke, Wagner, Giaccio, Knipping and Pross2018). Likewise, the Tenaghi Philippon X-ray fluorescence and biomarker records indicate increasing precipitation from 425.9 ka onwards (Koutsodendris et al., Reference Koutsodendris, Dakos, Fletcher, Knipping, Kotthoff, Milner and Müller2023) and a sharp increase in the mean annual air temperatures (from 8°C to 16.5°C) from 431 ka (Ardenghi et al., Reference Ardenghi, Mulch, Koutsodendris, Pross, Kahmen and Niedermeyer2019).

The charcoal data suggest increased local fire occurrences in MIS 12a compared with the earlier periods. Local fires in MAR-1 were likely favoured by sufficient accumulations of local vegetation biomass for burning in the catchment area, as inferred by high abundances of both upland and wetland vegetation (Fig. 7) and wood-decay fungi; high Gelasinospora values also confirm local fire occurrence (Fig. 7). This notion is further strengthened by palaeobotanical data from MAR-1 that identify large wood charcoals from willows and alders (riparian), oaks, elms, and reeds (Field et al., Reference Field, Ntinou, Tsartsidou, van Berge Henegouwen, Risberg, Tourloukis, Thompson, Karkanas, Panagopoulou and Harvati2018), attesting to local biomass burning. Moreover, between ∼431 and 428 ka—the period including the archaeological layers—alkaline conditions (increasing Mg/Ti ratios; Fig. 4) documented in MAR-1’s sediments might be (partly) associated with local fires, given that fire alters soil chemistry through ash towards increasing alkalinity (see Shumilovskikh et al., Reference Shumilovskikh, O’Keefe, Marret, Marret, O’Keefe, Osterloff, Pound and Shumilovskikh2021). Local milder microclimate (temperatures) would have favoured fires during MIS 12a, and the precipitation regime might have also accounted for local fire events at MAR-1 towards the MIS 12/11 transition; mudflows identified in the corresponding sediments (UB3) are attributed to storm events (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021), which in turn could have triggered lightning strikes (the most common cause of fire; Pausas and Keeley, Reference Pausas and Keeley2009), promoting ignition. If so, the fuel would not have been saturated enough to prevent wildfires (Sadori and Giardini, Reference Sadori and Giardini2007).

The co-occurrence of micro- and mesoscopic particles points to regional wildfires; however, at least a portion of them could be a product of secondary fragmentation of larger particles, because the area was subjected to a minor postdepositional reworking process identified at the base of UB4 (Giusti et al., Reference Giusti, Tourloukis, Konidaris, Thompson, Karkanas, Panagopoulou and Harvati2018). Nonetheless, ascertaining microcharcoals’ origin in archaeological deposits is difficult, as natural and human-induced distortions should be considered (Lebreton et al., Reference Lebreton, Bertini, Ermolli, Stirparo, Orain, Vivarelli, Combourieu-Nebout, Peretto and Arzarello2019). According to these latter authors, in addition to potential taphonomic processes, increased charcoal fragmentation by hominin and animal trampling should be considered before, during, and after the main phase of occupation. In MAR-1, CFS suggest herbivore presence from MIS 12c, which peaks in MIS 12a (see Herbivore presence and environmental interactions at Marathousa 1). Thus, we assess, particularly in-between ∼431–428 ka (MARPZ7/archaeological horizon), that micro- and mesoscopic particles mirror regional and, to some extent, local origin of fires, even though the available data do not allow for determining the proportion. Moreover, significant microcharcoal input is anticipated in case of fire use by hominins (Lebreton et al., Reference Lebreton, Bertini, Ermolli, Stirparo, Orain, Vivarelli, Combourieu-Nebout, Peretto and Arzarello2019). Given the limitations of the pollen-slide method, it is difficult to evaluate the use of fire by hominins in MAR-1. So far, bones, artefacts, and macroscopic charcoal recovered from the site have not provided evidence of anthropogenic burning. Nevertheless, the potential role of hominins (or not) on the local fire regime in the Megalopolis Basin can also be evaluated by the nature of fire events, frequency, and intensity using our new sediment charcoal record (charcoal particles > 200 µm) from the MAR-1 sequence, to be assessed in our future work.

Within the early part of MIS 11, the increasing forest cover in the Megalopolis catchment points to a climatic amelioration with a shift towards warmer but still dry conditions, as deduced by the development of Mediterranean elements and the rise of pioneer shrubs (Fig. 5). The concurrent high concentrations of algal blooms (i.e., high productivity in the water column; Fig. 6) attest to higher temperatures in the Megalopolis catchment, similarly to the MIS 13a palynological evidence. In addition, at the lowlands of the basin in MAR-1, a continuous drying of the habitat is evidenced by shifts in the composition of wetland vegetation and increased accumulations of fungal indicators of dry environment (Figs. 3 and 6). Moisture deficit at Megalopolis is further indicated by elemental data (low Rb/Sr ratios; Fig. 4) from the corresponding sediments, which suggest low precipitation during MIS 11c.

Further evidence of increasingly warm and dry conditions during MIS 11c is provided from the broader region. Pollen data from Lake Kopais (southeast Greece; Fig. 1) show the expansion of Mediterranean woodlands (mainly comprised of Olea and Q. ilex), suggesting particularly warm and dry conditions at that time (Okuda et al., Reference Okuda, Yasuda and Setoguchi2001). In contrast, the Lake Ohrid (Kousis et al., Reference Kousis, Koutsodendris, Peyron, Leicher, Francke, Wagner, Giaccio, Knipping and Pross2018) and Tenaghi Philippon (Koutsodendris et al., Reference Koutsodendris, Dakos, Fletcher, Knipping, Kotthoff, Milner and Müller2023) pollen records show warm but wet conditions at that time. This evidence implies a N-to-S precipitation gradient in the south Balkans during MIS 11c.

Charcoal data also corroborate such climate conditions. Specifically, local fire occurrence, as deduced by the macrocharcoal peak at ∼421 ka, was likely favoured by the prevailing warm and dry climate. In addition to biomass availability, aridity is another necessary ecological parameter for natural fires, as dry conditions convert potential fuels to available ones (Pausas and Keeley, Reference Pausas and Keeley2009). The presence of Neurospora (Fig. 7) attests to local fire occurrence at that time, considering that it grows a few minutes after being subjected to moist heat (van Geel and Aptroot, Reference van Geel and Aptroot2006). At ∼416 ka, the last documented fire phase suggests an extralocal to regional fire signal, given the lack of macrocharcoals in MARPZ9.

However, some micro- and mesoscopic charcoals might reflect local sources, considering that microscopic particles can be produced by burning herbs and shrubs (Scott, Reference Scott, Cerda and Robichaud2009). Indeed, from ∼420 ka onwards, the transition from a marsh to peatland (characterized by high proportions of ferns and reeds) could have yielded high amounts of microcharcoals at MAR-1; fires occur naturally in peatlands (Link et al., Reference Link, Mclaughlin, Stewart, Strahm, Varner, Word and Wurster2023). Thus, adequate vegetation biomass at MAR-1, in addition to favourable climate conditions for fires, corroborates the hypothesis for local sources at ∼416 ka. However, due to limitations of the pollen-slide method, taxonomic identification of charcoal particles was not feasible, and any linkage to pollen taxa or (and) vegetation types that were burnt should be considered with caution.

Herbivore presence and environmental interactions at Marathousa 1

The first herbivore presence in the CFS record is documented between ∼473 and 464.5 ka within a steppe landscape where shallow, mesotrophic aquatic environments were present in the lowlands. Such a local setting possibly favoured the fauna, providing food sources and potable water under particularly dry conditions during MIS 12c.

Interestingly, the influx rates of Sporormiella-type peak at ∼437.5 ka. However, this peak is associated with a flood event that caused the hiatus in MAR-1’s sequence in-between ∼464.5–438 ka (see Vegetation and fire dynamics in response to palaeoenvironment and climate changes) rather than with a strong herbivore signal. Given that CFS can be transported from a non-local area during flooding (Lee et al., Reference Lee, van Geel and Gosling2022), this explains the surge of Sporormiella-type at that time.

Herbivore presence appears enhanced between ∼432 and 423 ka (high CFS influx; Fig. 8)—the period encompassing the archaeological horizon—suggesting that herbivore populations were concentrated near the MAR-1 wetland. This is corroborated by the rich faunal assemblage containing remains from micro- and macro-fauna (fishes, amphibians, birds, reptiles, mammals) and at least two individuals of the extinct straight-tusked elephant Palaeoloxodon antiquus (Doukas et al., Reference Doukas, van Kolfschoten, Papayianni, Panagopoulou and Harvati2018; Michailidis et al., Reference Michailidis, Konidaris, Athanassiou, Panagopoulou and Harvati2018; Konidaris et al., Reference Konidaris, Athanassiou, Tourloukis, Thompson, Giusti, Panagopoulou and Harvati2018, Reference Konidaris, Athanassiou, Panagopoulou and Harvati2022) excavated at the archaeological horizon.

Within this interval, the mosaic of habitats developed in the Megalopolis’s catchment (as mentioned above) possibly played a critical role in the survival of terrestrial and semi-aquatic fauna during MIS 12a. Wetland vegetation creates habitats that promote faunal diversity, providing shelter, nesting locations, and food sources; for instance, the floating-leaved blades of N. alba and Nuphar lutea and several parts of wetland sedges and grasses (e.g., rhizomes, nutlets, seeds, tubers, leaves, and culms) serve as animal food (Mishra et al., Reference Mishra, Tripathi, Tripathi, Chauhan, Azooz and Ahmad2016; Klok and van der Velde, Reference Klok and van der Velde2017). Moreover, the two short-lived forest expansion events—having ∼1 ka duration each—point to temporarily denser vegetation cover that likely attracted herbivores to the site during MIS 12a. A plausible scenario could be that the higher availability of food sources favoured browsing or grazing activities by herbivores at MAR-1, thereby reflecting their preference for areas with higher vegetal biomass during glacial periods. The persistence of Mediterranean taxa corroborates the refugial character of MAR-1 in this interval.

Concerning water availability, shoreline reconstructions at MAR-1 indicate that the area was a floodplain during the deposition of MARPZ6 (∼436.5–432 ka), where seasonal (MARPZ7) to permanent (MARPZ8) ponds developed between ∼432 and 421.5 ka (Table 1). These aquatic environments that persisted for ∼15 ka in the region and provided habitat to several aquatic organisms (Bludau et al., Reference Bludau, Papadopoulou, Iliopoulos, Weiss, Schnabel, Thompson and Tourloukis2021) would also have furnished food sources and access to potable water for the terrestrial fauna. The rich marshy vegetation around the pond and the enhanced upland vegetation would have protected the waterbody by eliminating sediments and other pollutants from water runoff. This notion is confirmed by the corresponding NPP assemblages, which mirror a shallow freshwater environment at MAR-1.

In contrast to MIS 12a, our data suggest herbivore biomass decline at MAR-1 during the beginning of the MIS 11 interglacial period. The postulated faunal decline (MARPZ9) is contemporaneous with climate and local hydrologic shifts. Climate and hydrologic regimes can affect the composition and abundance of fungal flora and might prevent the preservation of Sporormiella fungus on the site (Davis and Shafer, Reference Davis and Shafer2006; Wood and Wilmshurst, Reference Wood and Wilmshurst2013). Humidity is the most crucial factor in the growth of coprophilous fungi (Lee et al., Reference Lee, van Geel and Gosling2022). In MAR-1, local dry conditions are evidenced by the decline of riparian trees and wetland taxa and the rise of fungi, indicators of dry environments (Figs. 4 and 6) from ∼422 ka onwards. It could be that the local moisture deficit (i.e., soil moisture) affected the growth and preservation of CFS in MAR-1’s sediments during that time. If so, the CFS decline does not reflect a reduced faunal presence in the greater area of the basin but particular taphonomic conditions at MAR-1.

Alternatively, the dry area might have been no longer attractive to the fauna, which thus visited MAR-1 less often. From the ecological perspective, the degraded aquatic environment characterised by stagnant, anoxic waters and eutrophic conditions by ∼422 ka (Fig. 6) rendered MAR-1’s habitat inhospitable to certain taxa, supporting this interpretation. Furthermore, charcoal data seem to support the scenarios for local environmental stressors. Specifically, the CFS decline at ∼422 ka coincides with the termination of the most intense fire phase documented in the MAR-1 sequence (see Vegetation and fire dynamics in response to palaeoenvironment and climate changes). The more extended burning in the vicinity of the site—starting at ∼432 ka (MARPZ7) and culminating within MARPZ8—in conjunction with the overall climate and environmental oscillations, probably negatively affected the ecosystem, thereby forcing faunal populations to abandon the site.

On the other hand, the precipitation regime points to overall arid conditions in the Megalopolis catchment during MIS 11c. Geochemical proxies from the corresponding sediments have shown contrasting patterns in precipitation (fluctuating Rb/Sr ratios; Fig. 4) since ∼432 ka (MARPZ7). Particularly, moderate Rb/Sr ratios indicate a relatively stable state of moisture availability at the site between ∼432 and ∼425 ka when CFS present their maximal influx values. On the contrary, an abrupt increase in precipitation at ∼423 ka (high Rb/Sr ratios, upper MARPZ8) was followed by a sharp decrease from ∼422 ka onwards (low Rb/Sr ratios, MARPZ9), the time when CFS influx decreases. These rapid shifts in precipitation regime during MIS 11c—which diverged from the previous relatively steady state during MIS 12a—likely promoted changes in Megalopolis’s terrestrial and aquatic ecosystem, leading to a drastic herbivore population decrease in the region. Cooper et al. (Reference Cooper, Turney, Hughen, Brook, McDonald and Bradshaw2015) examined megafauna species along with major megafaunal transitional events (widely distributed geographically across Eurasia and North America) during the past 56,000 yr against the Greenland ice core record. Those authors suggested that the regional turnover of megafauna species was promoted during the onset of interstadial periods when the most rapid climate shifts in the late Pleistocene occurred. Our hypothesis for herbivore decline associated with abrupt shifts in the Megalopolis climate regime appears to agree with the proposed hypothesis of these authors; however, we underline that our hypothesis needs further analysis to draw a solid conclusion.

Despite the abovementioned scenarios, one can argue that the low temporal resolution of MARPZ9 is possibly masking the CFS signal in the record or, alternatively, that the otherwise increased influx of semi-CFS (Cercophora-type and Coniochaeta-type), along with the presence of few CFS (e.g., Sordaria-type) that grow primarily on dung, suggests persistence of faunal populations in the site. While such a hypothesis cannot be rejected, it is worth mentioning that this interval corresponds to the deposition of LIIIa (UB1) (Karkanas et al., Reference Karkanas, Tourloukis, Thompson, Giusti, Panagopoulou and Harvati2018) and the peatland development at MAR-1. Consequently, given that the former semi-CFS primarily grow on peat deposits, the synchronous presence of the latter is associated with the prevailing vegetation rather than with herbivore presence (see Supplementary Table 2 for preferential substrates of CFS).

Notwithstanding the proposed hypotheses for the CFS decline at the onset of MIS 11, the interpretation is still equivocal; the putative faunal decline might have been specified locally to MAR-1 or might reflect absence at the broader catchment of Megalopolis or even could mirror taphonomic bias in the signal of the CFS record. Therefore, only by applying a multi-core approach (Lee et al., 2022), that is, studying material from different coring sites within the basin covering the same interval, we can have better insights into whether the CFS data show the very local or regional signal of herbivore presence/absence in the Megalopolis Basin..

Could the Megalopolis Basin have served as a “safe haven” for the biota during the MIS 12 glaciation?

The pollen analysis of the MAR-1 sequence shows the local presence of tree populations in the Megalopolis Basin during MIS 12c and MIS 12a. More specifically, the persistence of mesophilous (and thermophilous in MIS 12a) trees, even in low abundances, indicates the persistence of moist environmental conditions in the lower altitudes of the catchment, reflecting the refugium character of MAR-1 and the Megalopolis Basin in general. The tree populations were possibly restricted to areas with well-drained soils near the riparian zone, which served as a protected microhabitat (small refugium) where the local microclimate and topography provided constant moisture availability, mediating their persistence. In addition, the continuous presence of aquatic vegetation and wet meadow herbs at the site, even during periods of a very shallow water environment, witnesses the capability of MAR-1 to retain water and support life within or in the nearby damp soils. Médail and Diadema (Reference Médail and Diadema2009) stressed the need to also consider herbaceous species when examining the potential of a site as a refugium, because their local survival and distribution patterns might vary from those of tree species, as phylogeographic studies have shown (e.g., Abbott and Comes, Reference Abbott and Comes2003; Cozzolino et al., Reference Cozzolino, Cafasso, Pelegrino, Musacchio and Widmer2003). Wetlands have also served as refugia for boreal herbs, such as marshes in northern Italy, southern France, and northeastern Algeria (see Médail and Diadema, Reference Médail and Diadema2009). Therefore, the MAR-1 wetland habitat, which was located near the palaeolake that existed in the Megalopolis Basin during the Pleistocene, allowed a diverse flora to survive during the glaciation of MIS 12 and to expand after its termination, as deduced by the tree-pollen abundance documented in MARPZ9 during MIS 11c (Fig. 5).

In addition to this long-term vegetation development during MIS 12, two distinct periods of changes in the vegetation composition can be identified in pollen data based on fluctuations in the arboreal pollen/ non-arboreal pollen (AP/NAP) ratios. In particular, the first period, spanning almost the entire MIS 12c, indicates drastic tree-population reductions, reflecting particularly dry conditions at that time. In contrast, during the second period (∼438–424 ka; MARPZ6–MARPZ8), which also includes the archaeological horizon, two short-term enhancements of tree populations—in addition to the increased average AP values—imply sufficient edaphic humidity and milder conditions at MAR-1 during the otherwise particularly cold and dry MIS 12a. Moreover, changes in floristic composition are also observed between these two periods ( see archaeological horizon in the Vegetation and fire dynamics in response to palaeoenvironment and climate changes subsection ). This is in agreement with the sedimentological evidence (Karkanas et al., Reference Karkanas, Tourloukis, Thompson, Giusti, Panagopoulou and Harvati2018) that indicates completely different depositional environments between the upper (UB5–UB1) and the lower part (UB10–UB6) of the sequence.

A first concise comparison of the MAR-1 record with other pollen archives from the south Balkans (see Vegetation and fire dynamics in response to palaeoenvironment and climate changes subsection)) suggests, particularly for the MIS 12a, that the Megalopolis ecosystem responded differently to prevailing climate compared with Tenaghi Philippon and Lake Ohrid (Fig. 1), reflecting less severe dry conditions at the southernmost part of the Balkans. This evidence appears to confirm our assumption about the role of Megalopolis as a glacial refugium during MIS 12, with an emphasis on MIS 12a, where all the archaeological evidence is found, representing a refugium within a refugium at that time. Nevertheless, in the presence of a significant lithological hiatus in the MAR-1 sequence spanning the entire middle part of MIS 12 (i.e., MIS 12b), we underline the fact that further research in the site examining high-resolution and continuous pollen records is necessary to explore in detail the vegetation dynamics of the Megalopolis ecosystem throughout MIS 12, and comparisons with other climate-proxy records are necessary to comprehend how climate influenced the ecosystem dynamics in the basin.

Concerning fauna, the fact that MAR-1’s wetland ecosystem attracted and sustained herbivores—even in low abundances—during MIS 12c, the period when Megalopolis was marked by the most intense forest contractions and a particularly dry climate regime, further supports the refugial character of the site. During the hominin occupation, the herbivore populations became abundant under wetter (milder) conditions and increased vegetation cover in the basin during MIS 12a. Considering the pollen evidence for severe tree-population contractions (and particularly cold conditions) at the higher latitudes of the south Balkans during MIS 12a, our CFS data, in addition to pollen data, also confirm the refugium hypothesis for the Megalopolis Basin at that time.

Consequently, the natural features of the site, such as ecosystem diversity (aquatic, semi-aquatic, and terrestrial), would also have provided essential food resources to hominins, such as plants and animals, as well as water and shelter from cold and arid conditions in the interval between ∼436.5 and 428 ka. Notably, the wide variety of plants (MARPZ6–8) might have created a potential for a broad diet. Phytolith analysis from the Theopetra Cave (central Greece) suggests that sedge rhizomes were consumed by MP hominins (Tsartsidou et al., Reference Tsartsidou, Karkanas, Marshall and Kyparissi-Apostolika2015). Despite the lack of evidence of plant exploitation either in MAR-1’s pollen record or in palaeobotanical remains retrieved from the corresponding sediments (Field et al., Reference Field, Ntinou, Tsartsidou, van Berge Henegouwen, Risberg, Tourloukis, Thompson, Karkanas, Panagopoulou and Harvati2018), this middle Pleistocene wetland was rich in plant resources, including edible taxa for hominins like seeds and rhizomes of Nymphaea and Nuphar, S. lacustris rhizomes, or Bolboschoenus (synonym = Scirpus) tubers (see Hillman et al., Reference Hillman, Madeyska, Hather, Wendorf, Schild and Close1989; Wollstonecroft et al., Reference Wollstonecroft, Hroudová, Hillman and Fuller2011). It could have offered the potential to hominins to utilise the vegetation directly, for example, by consuming fruits, seeds, and tubers/rhizomes, or indirectly, using the hard or fibrous parts of plants (e.g., reed stems) to produce tools for various purposes (see also Guibert-Cardin et al. [Reference Guibert-Cardin, Tourloukis, Thompson, Panagopoulou, Harvati, Nicoud and Beyries2022] for the use of MAR-1 artefacts on both animal and plant materials). Furthermore, hominins would have been attracted by the herbivore presence on the site. Indeed, abundant herbivore populations congregated in MAR-1 would have provided hominins a wide range of subsistence opportunities. Cut-marked elephant and other mammal bones from MAR-1 were found in association with lithic remains documenting human exploitation through butchering activity at the site (Konidaris et al., Reference Konidaris, Athanassiou, Tourloukis, Thompson, Giusti, Panagopoulou and Harvati2018; Guibert-Cardin et al., Reference Guibert-Cardin, Tourloukis, Thompson, Panagopoulou, Harvati, Nicoud and Beyries2022). In addition to the ecosystem diversity, the Megalopolis Basin had the potential to serve as a glacial refugium for the survival of all the biotic components of MAR-1’s ecosystem (i.e., to sustain wetland habitats and furnish potable water for faunal and hominin populations) due to its capability to retain water at its lowlands (see also Cuthbert et al. [Reference Cuthbert, Gleeson, Reynolds, Bennett, Newton, McCormack and Ashley2017] for the groundwater hydro-refugia) under the otherwise cold and dry conditions during MIS 12. We therefore hypothesise that the Megalopolis Basin could have acted as a long-term refugium offering survival over the course of the Pleistocene. However, further investigation from other sites with different temporal scales within the basin (see Konidaris et al., Reference Konidaris, Athanassiou, Panagopoulou and Harvati2022) is necessary to test this hypothesis.