Marine resource abundance drove pre-agricultural population increase in Stone Age Scandinavia

How climate and ecology affect key cultural transformations remains debated in the context of long-term socio-cultural development because of spatially and temporally disjunct climate and archaeological records. The introduction of agriculture triggered a major population increase across Europe. However, in Southern Scandinavia it was preceded by ~500 years of sustained population growth. Here we show that this growth was driven by long-term enhanced marine production conditioned by the Holocene Thermal Maximum, a time of elevated temperature, sea level and salinity across coastal waters. We identify two periods of increased marine production across trophic levels (P1 7600–7100 and P2 6400–5900 cal. yr BP) that coincide with markedly increased mollusc collection and accumulation of shell middens, indicating greater marine resource availability. Between ~7600–5900 BP, intense exploitation of a warmer, more productive marine environment by Mesolithic hunter-gatherers drove cultural development, including maritime technological innovation, and from ca. 6400–5900 BP, underpinned a ~four-fold human population growth.


Study sites, background information and methods for palaeoenvironmental reconstruction from Danish coastal sediments
Danish coastal palaeoenvironmental study sites sites. An additional off-shore site (in Aarhus Bay) is also included as this sediment core was collected almost adjacent to Norsminde Fjord and demonstrates efficient oceanic exchange of sediment (and by inference other materials) during the late Mesolithic period (ca. 8000-6200 cal. yrs BP; hereafter BP unless otherwise specified); i.e. sediment is being removed from the coastal waters and transported into the deeper basins offshore. The Aarhus Bay sediment accumulation data form part of an independent marine environmental change project 1 and is presented as supplementary information with permission of the authors.

Danish coastal analyses: chronology, additional methodology and proxy records
Accelerator mass spectrometry 14 C dating and age-depth models Prior to dating and age-depth modelling, core sequences were correlated via physical analyses (organic, carbonate and minerogenic matter; Supplementary Table 1), using losson-ignition techniques outlined in Dean et al. and Hieri et al. 2,3 . From the finalised core sequences, selected samples were sieved and picked for plant macrofossils for Accelerator Mass Spectrometry (AMS) 14 C dating. Appropriate material selected for AMS 14 C dating were analysed at the Aarhus AMS Centre (AARAMS) at Aarhus University (using an EN tandem accelerator) or at the 14CHRONO Centre, Queen's University Belfast (UBAnumbers in supplementary data). 14 C dating procedure follows methods outlined in Lewis, Philippsen et al. and Olsen et al. 4,5,6 . Results were reported as conventional radiocarbon dates ( 14 C yrs BP, before 1950 AD 7 ) and calibrated into calendar years using the IntCal13 calibration curve 8 . At Korup Sø no plant macrofossil analyses were performed 4, 9 and therefore the age-depth model for this site is based on 7 mollusc samples (5 dated at the Copenhagen 14 C laboratory and 3 at the AMS 14 C Dating Centre at Aarhus University 4,9 ; with a 400±50 year marine reservoir correction applied to each shell sample and calibrated with Marine13 8 . The Ulmus decline evident in the pollen record was used to verify the radiocarbon chronologies at Kilen, Norsminde Fjord and Korup Sø (with pollen samples being prepared using standard techniques outlined in Faegri and Iversen 10 ). No pollen data are available for Sebbersund, Horsens Fjord and Tempelkrog. Subsequent age-depth modelling was performed in either OxCal 11,12 or Bpeat 13 (see Supplementary Table 2 and below for further details).
A brief summary of the chronological records is provided in Supplementary Table 2. Agedepth models for all sites except Sebbersund are available in Lewis et al. 14 Figure 19).
Over the study period (8000-4000 BP) each site exhibits a continuous sequence with no evidence of hiati, though the Korup Sø record finishes at ca. 4200 BP, due to sediment infilling and isostatic uplift processes 4,9 . Lithology for all sites is largely homogenous (marine silty-clay gyttja, with low organic content) between ca. 8000-3700 BP. Plant macrofossil and mollusc remains ( Supplementary Figures 11-15) are present in all sequences though abundances vary markedly both within, and between sequences. Bulk sediment accumulation rates (AR) (Supplementary Fig. 1) were calculated for each site based on the AMS 14 C age-depth models for all sites. These age-depth models allowed for changes in the AR at times of marked change in physical sediment parameters (organic matter, CaCO3 and minerogenic -AR) to provide a more realistic picture of sedimentation and avoid overreliance on the 14 C dates as indicators of sediment rate change 6  The bone collagen extraction protocol from Brock et al. 23 , which is based on the Longin method 24 and revised with the inclusion of an ultrafiltration step 25,26 , was applied. Bone samples were cleaned and crushed into smaller pieces, followed by a demineralisation step (2% (0.6M) hydrochloric acid). The samples were dissolved at ~70˚C for 24 hours, followed by two filtering steps and lyophilisation. δ 13 C and δ 15 N values were measured using the Thermo Delta V IRMS with Flash 1112 Elemental Analyzer at the 14

CHRONO Centre in
Queen's University Belfast. δ 2 H measurements were executed using the TC/EA coupled to a Thermo Delta V IRMS at the Stable Isotope Facility in the School of Planning, Architecture and Civil Engineering of Queen's University Belfast.

Supplementary Notes 2 Archaeological and human-impact data
Shell midden abundance Human marine resource exploitation over the study period is estimated via relative shell midden abundance (i.e. greater marine resource exploitation = more abundant are larger shell middens present). Our estimate is based on the summed probability of 231 calibrated 14 C dates on Ostrea edulis shells present in Danish shell middens ( Fig. 2E; main text). There is substantial debate about the validity and methodology of using 14 C dates as a proxy for prehistoric activity 27,28,29,30,31,32 , but this method is being becoming more common 31,32,33,34,35,36 , and ultimately the reliability of this technique is determined by the quality of the dataset used. The dataset employed here is extremely comprehensive, incorporating 231 dates from 42 shell middens. This is a semi-quantitative measure, which we argue, in the absence of any other reliable method of quantifying shell midden abundance or volume, is representative of the general shell midden trend (in terms of presence and relative abundance) over the study period, agreeing with the widely accepted view among Danish archaeologists 37,38 . We acknowledge that these data may be biased by potential focussing of scientific interest in the Mesolithic, abundance/availability of oysters to 14 C date and the absence of dates submitted to other dating facilities, but we believe that this is a valid proxy of midden abundance over this period, agreeing well with archaeological observation of trends over the study period 37,38,39 . Furthermore, the great majority of oyster shells submitted for dating were sent to either the Aarhus University or Copenhagen National Museum dating facilities.
Following their first appearance around ca. 7600 BP 40 in Denmark, shell middens gradually increase in size/volume and abundance, reaching their maximum extent between ca. 6400-5700 BP (based on the data presented here; Fig. 2e main text). They decline in abundance in the Early Neolithic, though remain continuously present up until ca. 4200 BP and sporadic until ca. 3700 BP, before disappearing completely in the Bronze Age. Danish shell middens from the late Stone Age can contain diverse mollusc assemblages, but classically contain abundant Ostrea edulis (European flat oyster), Cerastoderma edule (edible cockle), Mytilus edulis (blue mussel) and Littorina littorea (common periwinkle). Across the Mesolithic-Neolithic transition (ca. 5900 BP) many of these middens exhibit a distinct faunal shift in mollusc composition from oysters dominance in the Ertebølle layers (ca. 7400-5900 BP) to predominately cockles and mussels in the early Neolithic (Funnel Beaker layers), which we have argued previously may be due to changing climate and sedimentary conditions 14 .
Another pulse of oyster dominated shell middens coincides with the Pitted Ware-Single Grave Cultural period ca. 4400-4800 BP.
Shellfish exploitation is indicative of a wider phenomenon of marine resource utilisation that is visible in archaeological remains (particularly shell middens but also in other types of coastal sites/settlements). From the many comprehensively excavated middens across coastal Denmark, it is clear that in the late Mesolithic, humans exploited a wide range of marine resources in addition to shellfish, including marine mammals, birds and fish species The variability in fish bone composition between sites is somewhat contrasting to the relatively consistent shellfish composition of Ertebølle sites. Whilst fishing was clearly an important activity, particularly in the summer months 43,45,46 , it appears that the catch (i.e. species exploited) was more strongly governed by localised factors, such as the topography, proximity to freshwater inlets or sites (e.g. rivers or lakes), water depth, protection, and availability of suitable breeding grounds (e.g. eel-grass beds, macrophytes).
For example, it has been suggested that eel might have been readily caught at Bjørnsholm Bay, at the mouths of small streams running into the former fjord inlet, whilst at Norsminde, the shallow waters of the former Kysing Fjord might have provided ideal breeding grounds for flounder 43,44,45 . Whilst fish bones are often abundant in Mesolithic coastal shell middens and deposits, they appear to be very scarce in Neolithic layers (e.g. Bjørnsholm Bay, Mesolithic: Neolithic fish bone ratio, 11,490: 252 bones 43,45 , Norsminde Fjord, all but one bone were found in the Mesolithic layers 44 ). The reason for this remains uncertain, particularly as species composition and relative proportions (e.g. Bjørnsholm Bay), do not appear to change markedly from the Mesolithic layers 38,45,47,48 , suggesting similar fishing patterns (i.e. methods and targeted species). Site abandonment is ruled out due the clear continuation of numerous middens (and other sites bearing coastal resources) in the Neolithic (including some sites, dated entirely to the Neolithic period). It remains possible that fish bones were deposited elsewhere, but this is hard to believe considering the fact that both molluscs and bones from other animals were deposited on these sites. Perhaps accelerated decomposition of fish bones also occurred in the Neolithic layers, due to a reduction in shell midden accumulation rates (as marine resources became less important) and/or changes in the physical environment (e.g. increased aerial exposure, weathering and erosion). The most plausible explanation on a regional scale, however, is that the fish catch was substantially reduced due to introduction of agriculture and reduced dependence on (declining) marine resources 49,50 .

Isotopes and bones
Following the pioneering work of Tauber et al. 49  6400-5900 BP) and the preceding 500 years for each population event (testing for unequal variances) shows that the SPD is significantly different (p<0.05) comparing Pop. 1 and Pop.
2 to the 500 years preceding each pulse for each region (see Supplementary Table 4). We further tested that the pre-agricultural population increase was a real phenomenon using the However, for the regions experiencing a later appearance of agriculture, several exhibit similar population increases (or pulses) in the late Mesolithic (i.e. northern Germany several pulses after ca. 7000 BP, but particularly after 6500 BP; Scotland, and western Sussex ca. 6500 BP). This suggests that similar population increases may have occurred further afield, potentially fuelled by abundant marine resources, though for these areas archaeological and/or marine environmental records are less comprehensive than those for south Scandinavia. For example, in Scotland there is some evidence for substantial marine resource exploitation from late Mesolithic shell middens 54, 55, 56 and a shift from a marine to terrestrial dominated diet 57, 58 across the Mesolithic-Neolithic transition. However, the archaeological shell midden records from Scotland are more ambiguous than those of south Scandinavia and, to date, no good quality marine palaeoenvironmental records detailing changes in marine productivity (e.g. long-term sedimentary pigment analyses) are available for Scottish coastal areas. Environmental hypotheses have been put forward to explain this potential dietary shift and the reasons behind the introduction of agriculture in Scotland 59 , but variable regional climate data (including evidence for drier, wetter conditions and no change) across the transition means that inferences of late Mesolithic/Early Neolithic climate remains contested and contradictory 55,59,60,61,62 .

Technological innovation
The number of (marine) fishing technologies is based on data presented in Fig. 26  here, demonstrated by both pollen and sedimentation records, is that during the Mesolithic period, there is little, or no, detectable impact on the land by humans, and it is only, after the introduction of agriculture that humans begin to manipulate the landscape more and more intensively to suit their needs.

Temperature
The Holocene thermal maximum (HTM) has been shown to be global in extent, though the precise dates are geographically variable 78,79,80 . For northern Europe, it is generally accepted that the HTM occurs between 8000-4000 BP, with peak temperatures between ca. 7500-6000 BP 81,82,83 , associated with intensification of the thermohaline circulation, northward migration of the polar vortex and persistent development of summer anticyclonic conditions over Scandinavia 79,80,83,84 . To show the HTM in the main text, pollen-inferred average air temperatures for Lake Trehörningen pollen record (from south Sweden) are plotted in Fig. 2A (main text). Lake Trehörningen exhibits rising temperatures in the period after 8000 BP, reaching maximum Holocene temperatures between ca. 7500-6000 BP, prior to a steady decline after ca. 6000 BP, falling more rapidly after ca. 5500 BP ( Fig. 2A main text), particularly in winter air temperature 81 . This curve has been chosen as we believe it is representative of the wider south Scandinavia area, the site is situated in close proximity to the study area and is relatively unaffected by human activity even in the Neolithic section of the record (i.e. post 6000 BP).
Whilst we feel this is the most appropriate temperature record to display in Fig. 2A (main text), it is important to clarify that similar conditions over the HTM are apparent in a number of other pollen records from south Scandinavia 84,85,86,87 . Of note here is the composite Holocene reconstruction curve for northern Europe from Seppä et al. 81 (see below) and a more recent reconstruction for Denmark by Brown et al. 87 , though we decided against plotting the Brown et al. 87 temperature reconstruction because Denmark has been shown to be heavily human-impacted following the introduction of agriculture 69,71,72,73 . The Seppä et al. 81 record shows consistently higher temperatures between 8000-4000 BP across northern Europe in quantitative reconstructions from 36 widespread stacked pollen records (supported by stacked chironomid temperature inferences), and therefore supports the Lake Trehörningen record, though is less specific to south Scandinavia than Lake Trehörningen 82 .
Higher temperatures during the HTM are further supported by characteristic floral and faunal changes throughout the study period 88,89 (Supplementary Figure 22), particularly by the widespread presence of a number of warmth-demanding plants and animals 46 Figure 22). We also highlight that this temperature data is from sediment core analyses from the remote Gotland Basin, and so less relevant to the shallow Danish and wider Southern Scandinavia coastal/marine waters. Warden et al. 92 infer that rising temperatures ca. 6000 BP drive agricultural innovation and subsequent population growth 92 .
In support of their SST reconstruction, Warden et al. 92 draw attention to two records in particular which show minor secondary increases in temperatures between ca. 6500-5700 BP and peaking around 5800-5700 BP 93,94 . However, in contrast to Warden et al. 92 , generally high temperatures precede this Early Neolithic peak in temperature in these two records, unlike the cold, unstable conditions apparent between 7000-6000 BP in the TEX86 SST reconstruction from the SE Gotland basin. Furthermore, it is important to highlight both problems/variability associated with SST reconstructions from biomarker techniques such as TEX86 95,96 and that these supporting records 93,94 are somewhat selectively chosen from a much larger body of evidence that shows the HTM peak temperature occurring between 7500-6000 BP 81 . Lastly, in these two records rising temperatures pre-date the introduction of agriculture and agree well with the early population increase hypothesized here and might even provide a supporting mechanism for increasing productivity beginning ca. 6400 BP. It is acknowledged here that in some regional temperature reconstructions, winter temperature do not peak until later, but in south Scandinavia, the presence of thermophilous species (e.g. ivy, holly and mistletoe) in the late Mesolithic period, suggest that winters must have been relatively mild 73,85 .

Sea-level change
The selected sea-level curve (in Fig. 2C of the main text) is from the Blekinge region of southern Sweden, chosen as it is broadly representative of the wider area and is one of the most comprehensive and well-dated records available for the mid-Holocene from southern Scandinavia. This curve is based on multiproxy evidence (magnetic parameters, sediment stratigraphy, organic carbon, pollen, plant macrofossils, diatoms) from two well dated cores, and from a region that has experienced similar rates of uplift (i.e. rebound) to northern Denmark and Zealand since the deglaciation of the Fennoscandian ice sheet 97,98,99,100 .
Comparison of the Blekinge sea-level curve with other regional sea-level data shows good overall agreement in terms of major trends, though subtle differences do occur due to regional differences in climate causing 'piling up' of seawater 101,102 and rates of isostatic rebound 103,104,105,106 . In general, sea levels were higher than present day (+4 to +8 m above present in the southern Baltic and Denmark 100, 107, 108 ) throughout the study period (8000-4000 BP). The broad pattern is characterised by rapid rise between 8000-7000 BP (particularly around 7600 BP 109 ) associated with the marine (Littorina) transgression and global sea-level rise 110 , reaching maximum levels between 6700-6000 BP, followed by a gradual decline after 6000 BP. Super-imposed upon this general trend are transgressive and regressive events (~1-3 m magnitude 106, 107 ), though these may vary spatially in number and/or amplitude (Fig. 2 main text) due to local rates of isostatic uplift 104,111 and climate conditions 101 .

Supplementary Figures
Supplementary Figure 1 ꞁ Sediment accumulation rates at all six Danish coastal study sites presented in this study. For Norsminde Fjord, the accumulation rate at an almost adjacent offshore site in Aarhus Bay 1 is also presented, along with regional sea level 107 , to demonstrate the anti-phase relationship over the Mesolithic late Kongemose and Ertebølle periods (i.e. lower accumulation in the inner coastal waters and higher sedimentation in offshore waters). This is interpreted as increased efficiency of water and sediment exchange between the inner coastal water and the more open offshore areas between 8000-6200 BP under times of higher sea level. Sediment (and likely other material) is being transported from the coastal areas and deposited in the deeper sea basins, hence slower accumulation rates are often apparent in coastal sites prior to the introduction of agriculture (e.g. Norsminde Fjord, Korup Sø, Horsens Fjord) and high accumulation rates are evident offshore. This is important for some mollusc taxa (e.g. Ostrea edulis) that favour coarse

Supplementary Tables
Site (coring summary) and references containing published methodology/data from site:

Description/details
Kilen: Two closely spaced boreholes were cored (in 2007) using Russian augers (10 cm and 5 cm diameter; 100 cm chamber length), with overlapping sections (of 50 cm). The overlapping sequences were correlated where possible by physical parameters* (subsampled at 1 or 2 cm resolution). Key References: Lewis 4 ; Philippsen et al. 6 ; Lewis et al. 14,19 Kilen is a west to east facing fjord situated in the western Limfjord (north Jutland) near Struer and today, almost entirely isolated from the Limfjord by a road and rail embankment (completed in AD 1856). A narrow connection with the Limfjord is retained via a small stream (termed the 'Kilerkanal') in the south-east corner. Modern average salinity within the fjord is ~8 g L -1 , but in the adjacent Struer Bay, salinities are 24-27 g L -1 . Maximum depth = 6.5 m, average depth = 2.5m with a surface area of ~3.3 km 2 and draining a catchment area of ~35.3 km 2 . Korup Sø: Two boreholes (Well no. 1 and 2) were drilled in 1982/83 (Petersen 9 ). All analyses presented here have been performed on material from Well No. 1, with a maximum depth of 880 cm. Core correlation was not attempted between the parallel drillings due to the significant distance between boreholes (~300 m) and the absence of physical parameters. Key References: Petersen 9 ; Lewis 4 ; Lewis et al. 14 Korup Sø is situated in the central part of the Djursland peninsula (today dry land, ~3 m above sea level) northeast of the city of Aarhus in east Jutland 4 . Due to sea level decline and sediment accumulation the fjord became isolated from the Kolindsund-Randers fjord system several thousand years ago 4, 9 . Norsminde Fjord: Cored (in 1996) using Russian auger (with diameters of 10, 7.5 and 5 cm; all with 100 cm chamber lengths) from 2 boreholes (1 m apart), with overlapping sections (of 20 cm), correlated by physical parameters (particularly minerogenic residue (following ignition) at 2 cm subsampling resolution. Made from wood, bone and antler for line fishing (demonstrating multiple methods of fishing in this period). Ertebølle fish hooks usually 2-3 cm in length (with no barb). Found from many sites across Denmark with some regional variations (and likely uses).

Middle Ertebølle
Leister A type of pronged spear made from wood to catch fish (likely used in eel fishing). Two different types; short stubby and long and slender varieties, the former used in 'hard' substrate and the latter in 'softer' substrates.

Harpoon
Made of antler or whale bone, harpoons were likely used in seal/whale hunting at sea.

Middle Ertebølle
Fish net Made from plant fibres. Associated floats made from wood and stones as sinkers for these nets found at multiple sites from the Late Ertebølle.

Terrestrial:
Polished flint axe Wooden (ash) axe helve with polished flint axe -a paramount tool behind expansion of first farmers. Used for a variety of tasks including felling trees and timber work.
Early Neolithic ~5900 BP (3900 BC) Ard A primitive plough enabling shallow ploughing of the land. Made of wood and had no mouldboard so only scraped furrows in the soil without turning it over.