Introduction

Connected to the establishment of the first agricultural societies in Eurasia during the late Pleistocene and early Holocene (ca. 12,000 years ago), the clearance of woodlands for cropland and pasture as well as for fuel and construction material is arguably the most evident manifestation of early anthropogenic impact on natural environments1,2. A multitude of proxy records, including pollen3, charcoal4, and stable-isotope5 data, as well as modeling of population density and land use change6 support the notion that anthropogenic impact further intensified during the Bronze Age (starting ca. 5200 years ago in southern Europe)7. While the earliest impact was confined to the close vicinity of settlements6,8, a surge in mining activity led to a hemisphere-wide human effect on the environment as evidenced by mining-derived lead (Pb) in Greenland ice cores from ca. 2600 years ago onwards9,10,11. Despite the manifestation of such supraregional human effects, it has yet remained unclear when early societies started to restructure their environment beyond the close proximity to sociocultural centers. Even more so, the impact of early societies on the marine realm has remained largely unexplored, although there is archeological evidence for human exploitation of nearshore resources already during the Last Glacial (ca. 23,000 years ago)12.

Here we investigate the anthropogenic impact on terrestrial and marine environments connected to the socioeconomic evolution of early cultures based on the analysis of sedimentary archives from the Aegean region. Besides having been exposed to exceptionally long anthropogenic influence due to its continuous inhabitation since at least the Middle Paleolithic (ca. 250,000 years ago)13, this region has yielded the earliest evidence for Neolithic sites in Europe dating to the 7th millennium BCE, and for state-based societies dating to the 2nd millennium BCE14. These Neolithic and Bronze Age centers were characterized by highly dynamic demographics15,16 closely associated with vibrant socioeconomic developments15,17.

Following the development of smelting technologies on the Balkan Peninsula ca. 7000 years ago18 and cupellation of Pb ores for the production of silver in the Aegean and Near East ca. 6000–5000 years ago19, the Pb content in Holocene geological archives is closely connected to socioeconomic change11,20,21,22,23. The smelting- and cupellation-related release of Pb into the environment is predominantly via the fine-particle fraction and, as such subject to large-scale atmospheric transport, resulting in a supra-regional to hemisphere-wide distribution9,10,11,20,21,22,23. We have analyzed the Pb content of terrestrial and marine sediment archives from the Aegean Sea and its borderlands using X-ray fluorescence (XRF) core scanning at decadal-scale resolution. To ground-truth the fidelity of the XRF-derived Pb signals, we have additionally determined the Pb content of selected cores using inductively coupled mass spectrometry (ICP-MS) (see the “Methods” section).

Extending to the early Holocene, the investigated archives comprise a terrestrial core from the Tenaghi Philippon peatland in the northern borderlands of the Aegean Sea and marine cores from different parts of the Aegean Sea (Fig. 1; Supplementary Table 1). The Tenaghi Philippon peatland receives negligible riverine input and has extraordinary fidelity in recording atmospheric signals24,25. In contrast, the marine cores are from sites that, along with atmospheric sediment input, also receive considerable riverine input of mixed provenance mainly from the central and southern Hellenic Peninsula, the northern borderlands of the Aegean Sea, and western Asia Minor26,27,28.

Fig. 1: Locations of geochemical and palynological records discussed in this study.
Fig. 1: Locations of geochemical and palynological records discussed in this study.
Full size image

Left: Position of the Aegean region (black box) within the North Atlantic realm and locations of the Pb records from Crveni Potok21 and NGRIP11,22 ice cores in Greenland. Right: Locations of the studied cores (closed blue and green circles) in the Aegean region, and selected previously published palynological records (open green circles) that are referred to in the discussion. Locations of known Pb ore deposits adapted from the OXALID database85 are also shown (closed violet circles). The maps were generated with QGIS software (version 3.16.10) using topographic data from NASA Shuttle Radar Topography Mission (SRTM) datasets86.

The Pb-content datasets are augmented by decadal- to centennial-scale-resolution palynological (pollen and spore) data from selected cores that provide insight into anthropogenic land-use change in the borderlands of the Aegean Sea as it results from deforestation and agriculture. Because the vast majority of pollen and spores are deposited within a few tens of kilometers from their parent plants29 or even less for some cultivars (e.g., Cerealia and Vitis)30,31, they predominantly portray terrestrial ecosystem change at local to regional scales. We have generated palynological data from cores in the northern (i.e., Tenaghi Philippon and SL152) and southern (i.e., EF-AR2) Aegean Sea regions to obtain a spatially differentiated picture of vegetation change, thereby also considering the locations of some of the earliest cultural centers. Whereas Tenaghi Philippon records vegetation and land-use change in the vicinity of one of the most prominent Neolithic settlements in SE Europe (i.e., Dikili Tash)32, marine core EF-AR2 records such changes on the northeastern Peloponnese, which was home to some of the most important palatial sites of Mycenaean culture32. Marine core SL152 integrates signals from the northern Aegean borderlands where numerous socioeconomic centers thrived from the Archaic to the Classic and Hellenistic periods (i.e., from ca. 2800 to 2100 years BP)32. Because these diachronous cultural centers were developed in settings of highly diverse vegetation33, the pollen analysis of the three cores is pivotal for deciphering human impacts across heterogeneous ecosystems of the Aegean region.

Based on excellent chronostratigraphical control through radiocarbon dating and additional inter-archive correlation via XRF-based bromine (Br) contents as a measure of the organic content of the sediments34, our datasets resolve Pb contamination and vegetation dynamics across the Aegean Sea region at timescales that do justice to the tempo of sociocultural transformations in the Ancient World. Whereas they clarify the onset and characteristics of land-use and terrestrial ecosystem change on regional to local scales, they yield information on the beginning of Pb pollution for both the terrestrial and marine realms, thus also providing a larger-scale perspective on early human impact. Together, they allow us to identify the transition from local-scale environmental impact connected to agropastoralism-based economies to more pervasive human effects on natural ecosystems linked to advanced monetized societies.

Results and discussion

Variability in Pb pollution across the Aegean region during the Holocene

While all cores show low Pb content and variability during the early and middle Holocene, they exhibit a conspicuous two- to threefold increase in Pb content during the Late Holocene relative to the natural baseline level. To assess whether the increase in the Pb content derives from anthropogenic activities and not from natural input, we have evaluated the downcore variability of the XRF-based Pb/Zr ratio based on the fact that Zr is a conservative lithogenic element. The Pb/Zr ratio shows a similarly strong increase as the Pb content of about 2000 years ago (Fig. 2; Supplementary Figs. 1 and 2), which indicates that the Pb increase indeed results from pollution rather than lithogenic input. Over the past 2000 years, the Pb/Zr ratio remains high and exhibits considerable internal variability displaying two or, in some cores, even three distinct peaks. Between these peaks, the Pb/Zr ratio decreases to almost natural baseline levels, with the amplitudes from maximum to minimum values being slightly different among the studied cores (Fig. 2). As measured for marine core SL152 and also for the terrestrial core from Tenaghi Philippon, the variability in the XRF-based Pb signal is closely reflected by ICP-MS-derived Pb concentrations (Supplementary Fig. 2). The Pbanthropogenic and Pb enrichment factors (see the “Methods” section) provide further unequivocal evidence that the Pb increase in both cores over the past two millennia is caused by anthropogenic activities. They confirm that the XRF-based Pb/Zr ratio is indeed linearly representative of the Pbanthropogenic content in the sediments and thus can be used to reconstruct the onset and magnitude of Pb pollution.

Fig. 2: Holocene Pb pollution across diverse environmental settings of the Northern Hemisphere.
Fig. 2: Holocene Pb pollution across diverse environmental settings of the Northern Hemisphere.
Full size image

From bottom to top: Pb/Zr variability in the marine cores and the peat record from Tenaghi Philippon in the Aegean region (this study), and Pb/Zr content in the peat record from Crveni Potok21 (NW Balkan Peninsula) and Greenland ice cores11,22. The bold lines above the Pb/Zr data represent 10 times moving average values. The records from Crveni Potok and Greenland are plotted against their previously published chronologies; for the development of the individual chronologies for each Aegean record see Methods. The vertical dashed line marks the mean age of the onset of Pb pollution in both marine and terrestrial settings of the Aegean region based on change-point analysis. I: Iron Age; C: Archaic and Classical periods; H: Hellenistic period; R: Roman Empire; BΥ: Byzantine Empire; O: Ottoman Empire. See Fig. 1 for site locations.

Onset of anthropogenic Pb pollution in the Aegean region

The application of change-point analysis (see the “Methods” section) on all dated records from the northern, central, and southern Aegean regions (Supplementary Table 2; Supplementary Fig. 3) reveals that the increase in the Pb/Zr ratio beyond the natural baseline occurs between 2400 and 1900 cal. years BP. The slight discrepancies among the individual cores can be attributed to regionally variable radiocarbon reservoir ages35, differences in the time-integrated into each dated sample due to heterogenous, site-specific sedimentation rates, different distances from the pollution sources, and the inherent uncertainties of the radiocarbon dating technique36. To minimize such dating-related effects, we estimated the onset of anthropogenic Pb contamination by calculating the mean age of the change points in the Pb/Zr ratios of all six records (Supplementary Fig. 4). This yields an age of 2150 cal. years BP for the onset of atmospheric Pb pollution in the Aegean region (Fig. 3), a datum that lies within the age-uncertainty ranges of all individual records. Our results further show that high anthropogenic Pb contents persisted for about 1000 years, with three distinct peaks at mean ages of ca. 2000, 1700 and 1200 cal. years BP (Fig. 3).

Fig. 3: Radiocarbon dating of Pb/Zr maxima and minima in six cores from the Aegean region (see Fig. 1 for locations).
Fig. 3: Radiocarbon dating of Pb/Zr maxima and minima in six cores from the Aegean region (see Fig. 1 for locations).
Full size image

The Pb/Zr record from marine core EF-AR2 (left) is used as an example of the variability in anthropogenic Pb content as discussed in the text: The colored circles indicate the positions of a the onset of the rise of Pb/Zr values beyond natural background levels based on change-point analysis (dark blue), b three Pb/Zr maxima (purple, light purple, blue), and c two Pb/Zr minima (yellow, orange). The respective age ranges of the events in the different cores are marked by black bars, with the boxes (in colors corresponding to the circles in the left panel) indicating the 1st and 3rd quartiles of maximum dating-error ranges.

Pb pollution in the Aegean region within a historical context

Our Pb records demonstrate that the onset of Pb pollution in both the terrestrial and marine realms coincides with the intensification of ore exploitation by advanced monetized societies as they were first introduced in the Aegean region ca. 2600 years ago15,37. More specifically, the surge in Pb pollution at 2150 cal. years BP occurs during the transition from the Hellenistic to the Roman period, and high Pb contents persist throughout the times of the Roman Empire and the first half of the Byzantine Empire until ca. 1200 cal. years BP. Within this time interval, two transient phases with reduced Pb pollution occur at ca. 1770 cal. years BP and ca. 1340 cal. years BP. Also registered in the peat record from Crveni Potok (NW Balkan Peninsula) and Greenland ice cores (Fig. 1), these transient events have been previously attributed to the disruption of mining activities during the Antonine (165–180 CE) and Justinian (541–549 CE) plagues11,21. Importantly, our results portray a higher amplitude drop in Pb pollution during these events than the Crveni Potok and Greenland records (Fig. 2). This may attest to a particularly devastating impact of these disease outbreaks on the socioeconomic situation in the Aegean region that caused the disruption of mining in the region (such as Pangaion mountain38, Limenaria on Thasos island19 and potentially SE Bulgaria19). In addition, a strong decrease in Pb content occurred during the decline of the Byzantine Empire (ca. 600 cal. years BP), and low levels of Pb pollution prevailed throughout the Ottoman period (Fig. 2). Interestingly, the latter result for the Aegean region is in contrast to the record from Crveni Potok, which shows increasing Pb pollution during the Ottoman period21 (Fig. 2). We attribute this discrepancy to enhanced mining activities on the central and western Balkan Peninsula, such as within the Dinaric Alps and the Carpathian Mountains21. At the same time, it may be connected with a reduction of mining in the Aegean region as a result of the socioeconomic transformation on the Balkan Peninsula following the collapse of the Byzantine Empire.

No Pb pollution is documented in the marine cores from the Aegean Sea for the Bronze Age, although such a signal emerges clearly in the peat record from Tenaghi Philippon (Fig. 1). There, a first Pb increase above the natural Holocene background level is manifested by change-point analysis at ca. 5200 cal. years BP. Even when considering maximum age model uncertainties, the onset of Pb pollution at Tenaghi Philippon is the oldest yet reported worldwide, predating the earliest previously known onset of pollution at Crveni Potok21 by up to 1200 years and the onset of Pb pollution in Greenland ice cores by up to 2600 years11,20,22. The considerably earlier onset of Pb pollution is a reflection of the archive’s proximity to the earliest silver mining centers of the Eastern Mediterranean region in the Balkan Peninsula18,19. It provides strong support to the notion that the northern Aegean region was home to the most productive mining centers in the Eastern Mediterranean during that time39.

Besides providing confirmation that anthropogenic Pb pollution commenced during the Early Bronze Age, the new data from Tenaghi Philippon also indicate a pronounced decrease in atmospheric Pb pollution from 3100 to 2800 cal. years BP that spans the very end of the Late Bronze Age and the Early Iron Age. Such a decrease has been noted previously in Pb-pollution records, e.g., at Crveni Potok21. However, the close proximity of Tenaghi Philippon to mining centers of the north Aegean region (such as Pangaion mountain39, Limenaria on Thasos island19 and potentially SE Bulgaria19) together with the magnitude of change implies that the reason for the decline has to be sought in a disruption of mining activities after the collapse of Bronze Age societies in the Aegean and the wider Eastern Mediterranean regions15.

Based on the above, the onset of Pb pollution in marine sediments from the Aegean Sea lags that recorded in terrestrial archives from the Balkan Peninsula by up to 3000 years considering age model uncertainties. The Pb pollution associated with agropastoral societies during the Bronze Age is only recorded in organic-rich terrestrial archives (e.g., peat) that are particularly sensitive recorders of atmospheric signals. Instead, the Pb pollution connected to the increasing demand of monetized societies for metals from the Classical period onwards is also registered in less sensitive recorders (e.g., marine sediments) with lower organic content and higher concentrations of heavy elements. The onset of ubiquitous Pb pollution in both terrestrial and marine environments at 2150 cal. years BP is linked to a fundamental political change. We argue that this change was brought about by the subordination of Greece to the Roman Empire after 146 BCE. The incorporation of Greek regions into the Roman political sphere provided the new rulers with the opportunity to benefit from the natural resources of the recently acquired provinces40, which led to an unprecedented increase in the exploitation of Greek mining districts in order to extract gold, silver, and other metal resources.

Terrestrial ecosystem change in the Aegean region

The Pb content in paleoenvironmental records has been widely connected to socioeconomic dynamics21,22,41, which are intimately linked to variable demands for raw materials for mining and smelting, particularly timber42,43. To assess potential links between human-induced terrestrial ecosystem change and increased Pb pollution associated with the establishment of monetized societies in the Classical period, we have examined a suite of palynological records from the Aegean region.

For the northern borderlands of the Aegean Sea, the pollen record from Tenaghi Philippon shows a minimum in deciduous tree-pollen percentages (such as deciduous Quercus, Corylus, and Ostrya) and a peak of Pinus percentages coinciding with the increase in Pb content at 2150 cal. years BP (Fig. 4). The surge in Pb content is further associated with an increase in the pollen percentages of agricultural indicators44 such as Olea, Cerealia type, Vitis, and Juglans (Fig. 4; Supplementary Fig. 5). Specifically, the increase in Olea percentages at ca. 2300 cal. years BP may reflect the regional onset of olive-tree cultivation. This is in agreement with archeobotanical evidence indicating that the cultivation of Olea in the northern Aegean region was introduced with the colonization during historical times45. Moreover, our palynological results document peaks of Cerealia and Cichorioideae at ca. 1950 and 1850 cal. years BP during Roman times associated with the presence of Vitis. These pollen types are sporadically present in the palynological record over the past 7000 years (Supplementary Fig. 5), which is in agreement with archeobotanical evidence from the nearby archeological site of Dikili Tas that suggests the cultivation of cereals and wine in the region since the Neolithic46,47. A similar picture emerges for marine core SL152 from the northern Aegean Sea, which provides a broader view of vegetation change in the northern Aegean borderlands than Tenaghi Philippon48. It also exhibits low percentages of deciduous tree-pollen as well as high percentages of Pinus, montane taxa (including Abies and Picea), and Cichorioideae at 2150 cal. years BP. Importantly, the percentages of steppic taxa (i.e., Artemisia, Chenopodiaceae, and Ephedra) do not increase in any of the two records, which excludes climatic forcing as a reason for the decrease of deciduous tree-pollen percentages25,48. Instead, the presence of agricultural indicators (Olea, Cerealia, and Cichorioideae) testifies to human-induced vegetation change. Similarly, the percentage increase in montane trees at the expense of deciduous trees at Tenaghi Philippon and in core SL152 is explained by the fact that the human occupation of lowland areas primarily impacted deciduous forests growing in low and mid-elevations, whereas montane forests at higher elevations remained largely unaffected. This interpretation is supported by archeological evidence for the use of trees from lowland areas (particularly Quercus) for use in buildings and furniture43 as well as for charcoal production and smelting42. Further evidence for lowland vegetation change in the northern Aegean borderlands at 2150 cal. years BP comes from the pollen record of Lake Dojran (Fig. 1). There, a strong decrease in Pinus percentages associated with increasing percentages of Cerealia (Fig. 4) suggests deforestation in the catchment area due to anthropogenic land-use change49. Deforestation and expansion of Cerealia and/or herbaceous plants as a result of human activity during the same time is also documented in more inland settings such as Lakes Zazari50, Kastoria51, and Prespa52 (Fig. 1).

Fig. 4: Pb/Zr and palynological records (selected taxa) as indicators for anthropogenic impact in the Aegean region.
Fig. 4: Pb/Zr and palynological records (selected taxa) as indicators for anthropogenic impact in the Aegean region.
Full size image

Tenaghi Philippon, M51/3 SL152, EF-AR2 (this study), Dojran from ref. 49, Belevi from ref. 56, Voulkaria from ref. 59, and Kournas from ref. 61. Quercus refers to pollen from deciduous oaks. The bold lines above the Pb/Zr data represent 10 times moving average values. The vertical dashed line marks the mean age of the onset of Pb pollution based on change-point analysis. I: Iron Age; C: Archaic and Classical periods; H: Hellenistic period; R: Roman Empire; BY: Byzantine Empire; O: Ottoman Empire.

For the southern Aegean region, our pollen record from core EF-AR2 off the northeastern Peloponnese exhibits a decrease in deciduous Quercus percentages and peaks of Olea and Abies percentages coeval with the increase in Pb pollution at 2150 cal. years BP (Fig. 4). Synchronous decreases in deciduous Quercus percentages and increases in Olea and Abies percentages are also registered at nearby Lake Lerna53, thereby providing evidence for the regional character of these vegetation changes. Together, these findings suggest an intensification of agricultural practices on the Peloponnese leading to the clearance of natural lowland vegetation. They are in agreement with archeobotanical evidence for Olea cultivation in the southern Aegean region having started already during the Bronze Age and having peaked during historical times45,54. Moreover, they are in concert with the widespread use of Quercus for domestic purposes and smelting on the northeastern Peloponnese55. Synchronous abundance peaks of agriculture indicators (mainly Olea) and drops in the abundances of lowland deciduous forest taxa (mainly Quercus) are also documented in vegetation records from lowland settings along coastal Asia Minor such as Lakes Belevi56 (Ephesos) and Elaia57 (Fig. 1). Moreover, concomitant increases in Olea and, to a lesser extent, Cichorioideae occur also in coastal Asia Minor (Lake Bafa)58, in western Greece (Lake Voulkaria and Klisova lagoon)59,60, and on Crete (Lake Kournas)61.

In light of the above, specific patterns in vegetation change emerge for the northern and southern Aegean regions at ca. 2150 cal. years BP when anthropogenic Pb increases in all cores. In the northern borderlands of the Aegean Sea, the Pb pollution at that time was associated with a decrease in forest cover and a modest increase of agricultural indicators in lowland settings, whereas high-altitude forests remain rather unaffected. In the borderlands of the southern Aegean Sea, the rise in Pb pollution is accompanied by a strong increase in agricultural indicators and locally, also a decrease in deciduous forest cover in the lowlands. Although ubiquitous vegetation patterns prevailed within the Aegean region during the Classical period, a different picture emerges in vegetation dynamics for the northern and southern Aegean during the Bronze and Iron Ages (i.e., from 5200 to 2750 cal. years BP). The latter periods are characterized by asynchronous onsets of deforestation and expansion of agriculture, as well as the absence of spatially homogeneous vegetation patterns. For instance, deforestation, as reflected particularly in deciduous Quercus percentages starts earlier in the southern (e.g., at ca. 4500 cal. years BP in core EF-AR2 and Lake Belevi) than in the northern Aegean region (e.g., at ca. 3600 cal. years BP at Tenaghi Philippon and ca. 2800 cal. years BP in core SL152) (Fig. 4). In addition, a transient forest recovery during the Iron Age, as depicted by an increase in Quercus percentages in some records from the southern (e.g., Lake Belevi and core EF-AR2) and northern (e.g., core SL152) Aegean regions (Fig. 4), is of local importance only as there is no reforestation signal in the other examined pollen records during that time (e.g., Dojran, Kournas, Tenaghi Philippon, and Voulkaria). Also, the onset of agricultural practices shows no regional patterns among the studied records. Increasing percentages of agricultural indicators occur only locally in the southern Aegean region (e.g., Lakes Kournas and Belevi) during the Bronze Age (Fig. 4) despite abundant archeological evidence for agricultural practices in the Aegean region at that time45,47,62. Our results of deforestation and arboriculture in the Aegean region from the Bronze Age onwards bear strong similarities with human-induced environmental change in Asia Minor and the Levant known as the Beyşehir Occupation Phase63 (BOP). The time-transgressive onset of the BOP at different localities in the Eastern Mediterranean region rules out a climate trigger such as a regional-scale extension of dry conditions63. This interpretation is supported by paleoclimatic evidence for spatiotemporally complex long- and short-term dynamics in Aegean hydroclimates during the Bronze Age and historical periods64. Together, these results suggest that the deforestation in the Eastern Mediterranean region from the Bronze Age onwards was connected to anthropogenic land-use change. We therefore conclude that the anthropogenic impact of agropastoral societies on terrestrial ecosystems of the Aegean region during the Bronze and Iron Ages was predominantly confined to the vicinity of then-existing socioeconomic centers. Collectively, the pervasive vegetation change and the onset of Pb pollution in the terrestrial and marine realms at 2150 cal. years BP marks the onset of unprecedented, broad-scale environmental impact on the Aegean region by monetized societies.

Implications of Pb pollution in the Ancient World

The marked increase in Pb content at 2150 cal. years BP can be traced across the Aegean region in a multitude of environmental archives from a wide range of depositional settings. Although not representing an isochron sensu stricto, it can serve as a marker for the direct alignment of marine and terrestrial archives from the Aegean region that lack robust chronologies. Considering that Pb pollution can also be traced in tree rings65 and speleothems66, this marker has the potential for systematic correlation of a wide range of environmental archives in the Aegean region. As such, the Pb variability expands the chronological toolbox for a highly dynamic period of antiquity and can prompt a better understanding of human–ecosystem interactions during the rise and fall of ancient societies in the Eastern Mediterranean region.

Our combined analysis of Pb pollution and vegetation change in the Aegean region demonstrates that anthropogenic impact started to affect both marine and terrestrial environments during the end of Classical antiquity, at ca. 2150 cal. years BP. Specifically, the Pb increase caused by the intensification of extractive metallurgy and the synchronous human-induced vegetation change at that time testify to an extensive exploitation of natural resources during the transition from the Hellenistic to the Roman period. Our data also demonstrate that the extensive use of natural resources persisted for one millennium throughout the Roman and Byzantine Empires.

We conclude that the Pb pollution marker at ca. 2150 cal. years BP, when the Roman Empire began to occupy Ancient Greece, represents a turning point in the environmental history of the Aegean region: Until ca. 2150 cal. years BP, the environmental variability in that region was predominantly naturally driven, and anthropogenic signals occurred on a local scale only. From that time onwards, it was subject to persistent, extensive human impact. We propose that the human impact on both the terrestrial and marine realms as it emerges at ca. 2150 cal. years BP can serve as a marker for the environmental history of the Aegean region. Further research on historical Pb pollution signals, notably for Anatolia, the Levant, and the Italian Peninsula will show to what extent this marker can be applied to the greater Eastern Mediterranean region.

Methods

Regional setting

Being part of the Eastern Mediterranean Sea, the Aegean Sea is delimited by the Balkan Peninsula in the north and west, and Asia Minor in the east. In the southwest and southeast, it is connected to the Ionian and Levantine Seas, respectively (Fig. 1). Climatically, the Aegean Sea and its borderlands are characterized by a Mediterranean climate with relatively wet winters and dry summers. The wind field is generally dominated by northerly directions, with the occasional occurrence of south-southwesterly winds; the northerly winds reach maximum strength in December–February and in mid-June to mid-September67.

Material

One core from Tenaghi Philippon and 13 cores from the Aegean Sea have been examined (Supplementary Table 1). The marine cores SL151 and SL152 were retrieved during R/V METEOR cruise M51/3 in 200168, cores M144 3-2, 5-2, 5-6, 7-2, 10-5 were retrieved during R/V METEOR during cruise M144 in 2018/201969, and cores EF-MY3, EF-MY4, EF-AR1, EF-AR2, EF-AR3, EF-AR5 were retrieved by R/V AEGAEO during Eurofleets+ cruise MYRTOON in 202170. Lithologically, all cores consist of greenish-gray to dark-gray mud containing nannofossils, pteropods, and foraminifera.

Terrigenous input into the Aegean Sea is primarily riverine and predominantly originates from the Balkan Peninsula (ca. 25 × 106 tonnes suspended sediment load annually), with subordinate contributions from western Asia Minor (ca. 10 × 106 tonnes suspended sediment load annually)28. The North Aegean Trough, which is home to cores M51/3 SL151, SL152, and M144 3-2 from the North Aegean Sea, receives fluvial sediment input predominantly from Macedonia and Thrace. The western region of the central Aegean Sea, including the Skopelos/Kymi basin from which cores M144 5-2, 5-6, and 7-2 were retrieved, receives fluvial input primarily from Thessaly and Euboia Island. The Lesvos basin, where core M144 10-5 has been recovered from, is located in the eastern region of the central Aegean Sea and receives fluvial input from western Asia Minor and Lesvos Island. The southwestern region of the Aegean Sea includes the Myrtoon and Argolikos basins, from which cores EF-MY3, EF-MY4 and EF-AR1, EF-AR2, EF-AR3, EF-AR5 have been retrieved, respectively; this region receives fluvial sediment input from the eastern Peloponnese.

The TP-2005 core from Tenaghi Philippon was retrieved in 2005 and consists of fen peat24. While the core represents a high-fidelity archive for at least the past ca. 1.35 Ma, its topmost interval (corresponding to age from ca. 300 CE until today) is disturbed by human activity related to the drainage of the peatland and agriculture; hence, this interval has been excluded here.

Chronology

The age models of marine cores EF-AR2, EF-MY4, M144 5-6, M144 10-5, and M51/3 SL152 are based on a total of 60 new and 7 previously published48 14C accelerator mass spectrometry (AMS) dates. All 14C dates and their uncertainties (1σ) are reported in Supplementary Table 2 and Supplementary Data 1. The dating was carried out on mixed planktic foraminifera, predominantly Globigerinoides ruber and, in case not enough G. ruber specimens were available, Globigerina bulloides, Globigerinoides sacculifer, and Orbulina universa from the >250 μm fraction as well as pteropods (Supplementary Table 2; Supplementary Data 1). The chronology of the uppermost 5 m of the core from Tenaghi Philippon is based on two new AMS radiocarbon ages from bulk peat samples, and three previously published AMS ages from snails (i.e., Oxyloma elegans and Viviparus contectus)71.

Bayesian age-modeling was performed in R (Version 4.2.3) using the package rbacon (Version 2.5.0)72; all previously published radiocarbon dates from cores M51/3 SL15248 and TP-200571 were also recalibrated. The sapropel S1 interruption centered at 8.2 kyrs BP73 was used as an additional chronological tie point in core EF-AR2. For all marine cores, a reservoir-age correction of 395 ± 85 years35 was applied, and the ‘Marine20’ calibration curve74 was used. The ‘IntCal20’ calibration curve75 was applied for the Tenaghi Philippon core. Moreover, the chronologies of the pollen records from Lakes Belevi56, Dojran49, Kournas61, and Voulkaria59 were recalculated using the ‘IntCal20’ calibration curve75 (Supplementary Data 2); the pollen data from Lakes Dojran76 and Voulkaria77 were downloaded from the Pangaea database (www.pangaea.de), and these from Lake Kournas78 from the Neotoma database (www.neotomadb.org).

X-ray fluorescence (XRF) core scanning

All cores were scanned with an AVAATECH (GEN-4) XRF core scanner equipped with an OXFORD ‘Neptune 5200’ series 100 W X-ray source with a Rhodium anode at the Institute of Earth Sciences, Heidelberg University. Prior to scanning, the core surfaces were smoothed and covered with 4-μm-thick Ultralene® foil to avoid contamination of the detector window and desiccation of the cores. Scanning was carried out at 30 kV energy level, with a Pd-thick filter and slit sizes 5 mm downcore and 10 mm crosscore. A current of 1000 mA and a measuring time of 20 s was used for the Tenaghi Philippon peat core, and a current of 500 mA and a measuring time of 10 s was used for all marine cores. Data acquisition was via a RAYSPEC SiriusSD 65 mm2 Silicon Drift Detector (Model 878-0616B), and a BRIGHTSPEC Topaz-X Multichannel Analyzer. Processing of the X-ray spectra was performed using the BRIGHTSPEC bAxilBatch software (Version 1.4). The Br and Pb counts were normalized to the total counts of all processed elements for each measurement, excluding Ag and Rh because these elements are biased by the signal generation as they are included in the beam collimator of the detector and the X-ray source, respectively (Supplementary Data 3). Based on the individual age models, the mean temporal resolution of the XRF data is 8 years (range: 4–37 years) for the TP-2005 core from Tenaghi Philippon, 15 years (5–53 years) for Core SL152, 20 years (8–33 years) for Core M144 5-6, 9 years (4–55 years) for Core M144 10-5, 35 years (6–190 years) for Core EF-AR2, and 60 years (13–200 years) for Core EF-MY4.

Pb and Zr concentrations

Lead concentrations on 35 samples from core SL152 and 34 samples from Tenaghi Philippon (Supplementary Data 4) were determined by inductively coupled mass spectrometry (ICP-MS) in clean labs at the Institute of Geosciences, Goethe University Frankfurt, and the Institute of Earth Sciences, Heidelberg University, respectively. For core SL152, 200 mg of sample powder was first treated with 8 ml of concentrated HCl:HNO3 for 14 h at 120 °C. The acid mixture was evaporated, the digested sample re-dissolved in 6 M HCl, and centrifuged to separate the undissolved silicate fraction (<40% of total weight). All work was performed in PicoTrace laminar flow boxes using only double-distilled acids and 18 Ohm Milli-Q water.

For the Tenaghi Philippon core, the samples were treated following the protocol of ref. 79. Specifically, between 70 and 200 mg of weighted, freeze-dried, and powdered peat samples were dissolved in 6 ml HNO3 and 0.1 ml HBF4 in PTFE vessels using a microwave-heated autoclave. Heating to 240 °C for 75 min resulted in homogeneous digestion solutions. Instrumental accuracy was checked versus the SLRS-6 river water reference material for trace metals. The quality of the entire analytical process was verified by co-processing the MESS-4 reference material. Full procedure blanks reached, on average, 0.03% of the peat concentrations, not exceeding 4.8% for the lowest concentrated samples.

To differentiate the anthropogenic Pb from the natural lithogenic influx, we adapted the methodology of ref. 21. First, we calculated Pblithogenic using the composition of the conservative lithogenic element Zr in the upper continental crust (UCC)80 based on the equation of ref. 81

$${{{{\rm{Pb}}}}}_{{{{\rm{lithogenic}}}}}={{{{\rm{Zr}}}}}_{{{{\rm{sample}}}}}\times ({{{{\rm{Pb}}}}}_{{{{\rm{UCC}}}}}/{{{{\rm{Zr}}}}}_{{{{\rm{UCC}}}}})$$
(1)

The Pblithogenic value was then subtracted from the overall Pb concentration of each sample to estimate the likely anthropogenically derived Pb fraction81:

$${{{{\rm{Pb}}}}}_{{{{\rm{anthropogenic}}}}}={{{{\rm{Pb}}}}}_{{{{\rm{sample}}}}}-{{{{\rm{Pb}}}}}_{{{{\rm{lithogenic}}}}}$$
(2)

The Pb enrichment factor (EF) was calculated considering the concentration of the Zr in the UCC and on each sample following the equation of ref. 82

$${{{\rm{EF}}}}=({{{{\rm{Pb}}}}}_{{{{\rm{sample}}}}}/{{{{\rm{Zr}}}}}_{{{{\rm{sample}}}}})/({{{{\rm{Pb}}}}}_{{{{\rm{UCC}}}}}/{{{{\rm{Zr}}}}}_{{{{\rm{UCC}}}}})$$
(3)

Palynology

Palynological preparation comprised sediment freeze-drying, weighing, spiking with Lycopodium spores, treatment with HCl (10%), NaOH (10%), HF (40%), heavy-liquid separation with Na2WO4 × 2H2O (when necessary), acetolysis, sieving through a 7 μm mesh, and slide preparation using glycerin jelly. For the Tenaghi Philippon record, 138 new samples were counted in the uppermost 5 m of the sequence, and between 193 and 444 (mean: 294) pollen grains (excluding pollen from aquatic plants and fern spores) were counted per sample. Pollen percentages were calculated based on the number of pollen grains from terrestrial plants, excluding Cyperaceae and Poaceae because of their natural over-representation at the site25. For core SL152, 73 samples from the uppermost 1.5 m of the core were counted; this dataset was further augmented by the previously published dataset of ref. 48. For core EF-AR2, 32 samples were counted. Between 69 and 533 (mean: 208) and 65 and 275 (mean: 209) pollen grains were counted per sample for SL152 and EF-AR2, respectively, excluding Pinus pollen grains because of their natural over-representation in marine sediments48. The percentages of the excluded taxa in each core were calculated on the basis of the main sum plus the counts for the excluded taxa (Supplementary Data 5). Based on the individual age models, the mean temporal resolution of the pollen data is 44 years (range: 8–147 years) for the Tenaghi Philippon core, 72 years (8–265 years) for core SL152, and 253 years (110–701 years) for core EF-AR2.

Change-point analysis

A linear regression model was applied to the XRF-based Pb/Zr ratios in order to identify change points using the strucchange package in R83. The algorithm detects deviations from stability in a classical linear regression model, allowing for a maximum of 5 breakpoints. The integrated optimization algorithm determines the optimal locations of these breakpoints and provides their confidence intervals at the 2.5% and 97.5% levels (Supplementary Data 6). The code of the R script has been adapted from ref. 84.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.