Multi-proxy evidence for the impact of the Storegga Slide Tsunami on the early Holocene landscapes of the southern North Sea

Doggerland was a land mass occupying an area currently covered by the North Sea until marine inundation took place during the mid-Holocene, ultimately separating the British land mass from the rest of Europe. The Storegga Slide, which triggered a tsunami reflected in sediment deposits in the Northern North Sea, North East coastlines of the British Isles and across the North Atlantic, was a major event during this transgressive phase. The spatial extent of the Storegga tsunami however remains unconfirmed because to date no direct evidence for the event has been recovered from the southern North Sea. We present evidence that Storegga associated deposits occur in the southern North Sea. Palaeo-river systems have been identified using seismic survey in the southwestern North Sea and sedimentary cores extracted to track the Mid Holocene inundation. At the head of one palaeo-river system near the Outer Dowsing Deep, the Southern River, we observed an abrupt and catastrophic inundation stratum. Based on lithostratigraphic, macro and microfossils and sedimentary ancient DNA (sedaDNA) evidence, supported by optical stimulation luminescence (OSL) and radiocarbon dating, we conclude these deposits were a result of the Storegga event. Seismic identification of this stratum to adjacent cores indicated diminished traces of the tsunami, largely removed by subsequent erosional processes. Our results demonstrate the catastrophic impact of Storegga within this area of the Southern North Sea, but indicate that these effects were temporary and likely localized and mitigated by the dense woodland and topography of the area. We conclude clear physical remnants of the wave are likely to be restricted to inland basins and incised river valley systems.

The Holocene pre-inundation landscape of the southern North Sea, known as Doggerland, was an area associated with Mesolithic hunter-gatherer communities 1 . Sea level rise during the mid-Holocene period at the regional scale was episodic due to local variations in isostatic rebound, autocompaction and palaeotidal range, the precise timing and extent of which, and consequently impact on Mesolithic communities, remains unclear 2 . A key event however during this period was the Storegga Slide, which occurred off the Norwegian Atlantic coast 8.15 thousand years before present (Kya) [2][3][4][5][6][7][8] . This has been speculated to have triggered a catastrophic tsunami associated with the final submersion of Doggerland 2,9 . Despite the apparent magnitude of the Storegga Slide Tsunami evident in northern North Sea, reflected both in sediment deposits and model predictions [4][5][6][7]10 , there has been a surprising lack of physical evidence to suggest the tsunami reached the southern North Sea. 2,9,11 , Figure 1.

Reconstruction of the South Western Doggerland archipelago
In order to reconstruct the early Holocene palaeolandscape of the southern North Sea we used a seismic survey estimating the coastline at 8.2 Kya based on bathymetric contours 12 using existing sea level curve data 13 (SI Appendix Text S1). From these data we inferred a generalized landscape, Figure 2. The survey confirmed that the area that had been Doggerland was by this time represented by an archipelago and residual stretches of coastal plain off the eastern coast of England. Within the residual plain were a series of palaeochannels representing river systems within glacial valleys incised through Late Devensian terminal moraines 14 , including a channel associated with the Outer Dowsing Deep, for which we used the term the Southern River. This channel runs north to south terminating in north and south-facing headlands with a central basin. A restricted area around the central basin was associated with a distinct but discontinuous seismic signal suggestive of an anomalously distinct and partially eroded stratum (SI Appendix Text S1.1).

Identification of a Storegga Slide aged tsunami deposit
Using the palaeobathymetric reconstruction of the landscape as a guide, we investigated the distinct localized seismic signal further and collected 15 vibrocores across the central basin to track the Holocene inundation process. The core most strongly associated with the seismic signal (ELF001A) was characterized by a series of finely laminated silt and clay strata interrupted by a 40cm layer (horizon units 1A-6 and 1A-5) of clastic sediment consisting of stones and broken shells, overlaid by sands (Figure 3, Figure S1.3, Table S1.1). The clastic deposit rests on a sharp, eroded surface in the underlying sediments and exhibits fining upwards within the clay-silt fraction while the coarsest fraction (gravels) occurs in the middle of the unit. Unit 1A-6 is associated with a palaeomagnetic profile which contrasts with the preceding fine silt layer below indicating that it originates from a differing geology (SI Appendix Text S2.1), most likely from outside the local catchment ( Figure 3D). An elemental core scan (SI Appendix Text S2.2) confirmed the abrupt change associated with Unit 1A-6 ( Figure 3C) with Ca and Sr profiles indicating an influx of marine shells, while Ti and K indicate terrestrial environments towards the base of the core. Similarly, the alkane carbon preference index (CPI), which accounts for the ratio of odd to even carbon chained molecules, indicates a composition dominated by terrestrial forms in the lower core that abruptly switches to an aquatic and marine composition from the event stratum (SI Appendix Text S2.3, Figure 3A). Together, the geochemical and geomorphological evidence indicates that Unit 1A-6 is consistent with a powerful event, but no comparably large stratum was detected in adjacent cores.
We investigated the age and depositional rate of the sequence in ELF001A using optical stimulation luminescence (OSL), Figure 3B (SI Appendix Text S3.1), and directed AMS radiocarbon dating (SI Appendix Text S3.2). The luminescence profile shows a steady accumulation of signal over time either side of the clastic stratum reflecting a stable sedimentation regime (see Supplementary Information), but shows a disturbed inversion within the stratum consistent with an influx of extraneous matter, Figure 3B. OSL dates could be retrieved under steady sedimentation rate conditions indicating the tsunami deposit occurred at 8.14 ± 0.29 Kyrs. BP (SI Appendix Table S3.1), making this stratum closely contemporaneous to the Storegga Slide. Given the OSL profile, the extraneous matter in this stratum likely reflects material of various ages dredged up from older strata in the surrounding area. This was confirmed by radiocarbon dates from shell fragments within the clastic stratum dated to 8.34 ± 0.3 Kyrs BP (SI Appendix Text 3.2), two centuries prior to the Storegga Slide Tsunami. Together, this evidence convincingly establishes the clastic stratum in core ELF001A as a deposit resulting from the Storegga Slide Tsunami event.

Palaeoenvironmental impact of the Storegga Slide Tsunami
We further confirmed the catastrophic nature and source of this stratum using multiple palaeoenvironmental proxies,  3), and secondly applying a novel metagenomic assessment methodology in which all sedaDNA is assessed for deamination damage, which may be more suitable for this data type ( Figure S4.4). We then further tested sedaDNA for stratigraphic integrity to assess possible biomolecule vertical movement in the core column (SI Appendix Text S4.5). Figure 4 shows that the sedaDNA demonstrates highly significant differentiation between strata indicating a lack of movement post deposition. Together, these tests indicate that authentic sedaDNA was retrieved and most likely represent the original depositional environment. Interestingly, the same stratigraphic tests applied to pollen generally show a lack of differentiation between strata indicating both a consistent influx of pollen from the surrounding area from oak, hazel woodland, and that the sedaDNA derived from sources other than pollen as has been previously suggested in other sedimentary contexts 17 . This suggests a taphonomy in which the sedaDNA represents a local signal relative to a more regional palynological signal.
Unit 1A-6 is characterized by an abrupt change in both microfossil and sedaDNA evidence. There is an absence of diatoms and pollen; an increase in outer estuarine or marine taxa of ostracods and foraminifera; the appearance of fractured molluscan shells from different and incompatible habitats including sublittoral, intertidal and brackish species; and the sudden and significant influx of all woody taxa in the sedaDNA profile (figures S4.1, S4.5 and Table S4.1). A novel measure of relative biomass, biogenomic mass, based on sedaDNA and genome size (SI Appendix Text S4.5), suggests a higher biomass of trees than either Zostera or Potamogeton in this stratum, although these latter taxa dominate in other strata, Figure 4 ( Figure S4.5). Together, these proxies indicate a violent event that brought with it the debris of surrounding woodland, consistent with the impact, and backwash, of a tsunami that dates to that of the Storegga Slide.
The environment prior to this dramatic event and recorded in the underlying stratum Unit 1A-7 was an estuarine marsh typified by predominantly benthic epiphytic and epipelic diatom communities and brackish foraminifera and ostracods, with a sedaDNA floral profile of Zostera and Potamogeton as well as members of the Hydrocharitaceae and Araceae present. A small meadow influence is also apparent in the sedaDNA profile including buttercups, orchids, mallows and asterids suggesting proximal open terrestrial systems, Figure S4. 5. After the tsunami in units 1A-4 to 1A-1 the foraminifera and ostracod signal indicate a return to estuarine mudflats with a greater abundance of marine taxa such as Ammonium batavus indicating a more established marine signal than prior to the event, Table   S4. 1. The sedaDNA signal also indicates estuarine taxa such as Zostera, and a meadow influence, although the biogenomic mass appears greatly reduced suggesting more distant proximity of the flora. A faunal signal considerably weaker than the floral was present throughout the core, but shows a significant elevation in count towards the top units (p = 1.0014E-06), indicating the presence of rodents and larger animals such as bear, boar and cloven hoofed ruminants, as well as higher orders of fish (Acanthomorpha, Eupercaria, Osteoglossocephalai), Figure S4.6.
These data lead us to conclude the effects of the tsunami altered the immediate landscape, perhaps opening up the surrounding forest to larger animals, but given the return of the terrestrial signal in subsequent strata we conclude that the final marine inundation occurred at a later time in this area.

Geomorphological influence on the tsunami propagation
The occurrence of a Storegga tsunami deposit at ELF001A 42 km from its contemporary coastline ( Figure 2) is unexpected given previous models have estimated a magnitude of wave that would reach 21 km inland in this area 18 before becoming impeded by a glacial moraine belt 14 . Our reconstruction at 8.2 Kya suggests that the geomorphology of the landscape was likely instrumental in propagating the wave inland. The orientation of the coastline relative to the direction of travel is consistent with a funneling of the tsunami into glacial tunnel valleys that breach the moraine, of which the Outer Dowsing Deep is one, Figure 2. The passage itself is a steep sided U-shaped valley that becomes progressively deeper southwards towards the central basin, which being below the 8.2 Kya sea level supports the palaeoenvironmental proxy evidence for an estuarine system. Such a configuration is expected to lead to an intensification of the wave and consequently a significant impact 19 .
To explore this scenario further we tracked the probable height of the tsunami stratum from ELF001A across the landscape using Glacial Isostatic Adjustment models 13 , seabed mapping and Geographic Information System analysis (SI Appendix Text S1.1 and S1.3) and then used seismic data in order to predict where similar deposits would be expected to be preserved in other cores, Figure S1.1. The analysis indicated an absence of strata contemporaneous to the tsunami layer in many areas reflecting widespread erosion subsequent to the tsunami event, Figure S1.2. However, the seismic signal associated with the change in sediment density that was identified as the tsunami stratum in ELF001A was used to trace other core locations likely to contain signals originating from the same source, Figure   S1.2. Using this technique two other tsunami candidate cores were identified, ELF003 and ELF059A, situated within the central basin area and southern river channel, figures S1.1 and S1.3. It should be noted that this signal was not replicated in other seismic surveys, and the signal, which correlates in this instance cannot, as such, be regarded as diagnostic for the purposes of identifying the presence or absence of similar deposits elsewhere. The identification of the Southern River system by palaeobathymetry allowed the identification for two further cores ELF031A and ELF039, as containing these deposits in the likely outflowing channel to the south of the basin (SI Appendix Text S1.3). Significant surges in woody taxa similar to that seen in Unit 1A-6 at the predicted tsunami height were detected in ELF003 and ELF0039 based on sedaDNA. The corresponding height in ELF0059A occurred at the base of the core but was also associated with a significantly higher abundance of woody taxa than overlying strata, and ELF0031A generally yielded too little sedaDNA for interpretation, Figure S4.8. Radiocarbon dates from ELF003 confirmed the corresponding tsunami height to be contemporaneous with Unit 1A-6, supporting the notion that these cores carry a much-diminished signal of the tsunami event.

The Storegga Slide Tsunami in the southern North Sea and final inundation
Together, these data lead us to conclude that the current physical evidence for the Storegga tsunami in the study area is highly localized because of the channeling effect of the tunnel valley systems and barriers to wave propagation provided by the wooded glacial moraines.
This restricted distribution was further reduced by later erosive processes. It may be the case that physical evidence for the Storegga tsunami in the southern North Sea has not been previously observed because it only resides in restricted locales, such as incised river systems, where the geography and local conditions were favorable for preservation.
Our evidence shows that the Storegga Slide Tsunami impacted coastlines in the area of the southern North Sea covered by this study. In coastal areas where human populations may have resided for most of the year, settlement would have been badly affected. The multiproxy evidence suggests the landscape recovered temporarily and hence confirms that the final submergence of the remnant parts of Doggerland occurred some time after the Storegga Slide Tsunami. At the same time, the remaining local terrestrial landscape is suggested to have been more open. Occupation could therefore have continued after the tsunami retreated, but within a much modified coastal landscape before early-mid Holocene eustatic sea-level rise was responsible for finally submerging the remnant Doggerland lowlands and its associated Mesolithic communities.

Methods
All methods are described in the supplementary information texts. Geomorphological analysis including seismics, coring and palaeobathymetry are presented in SI Appendix S1.
Geochemistry analysis including palaeomagnetics, elemental core scans and organic chemistry profiling are presented in SI Appendix S2. Dating of sediments using OSL and organic materials using radiocarbon is presented in SI Appendix S3. Palaeoenvironmental analysis including foraminifera, ostracods, pollen, diatoms, molluscs and sedaDNA are presented in SI Appendix S4.  Diatoms

S.1. Seismic survey
The high resolution seismic geophysical dataset was acquired between October 2008 and March 2009 as two separate surveys by Gardline Surveys Limited. The data was obtained by the Gardline Vessel Vigilant, which was equipped with a surface-towed boomer system consisting of an Applied Acoustics 300 Plate powered by an Applied Acoustics CSP 1500 Pulse Generator. The receiver consisted of a 12-element single channel hydrophone eel recorded with a Gardline 2012. Digital data logging and initial processing was accomplished using an Octopus 760 geophysical acquisition package (CodaOctopus). During acquisition a swell filter was applied to the data when necessary to correct for the effects of sea swell. The system was operated at a power level of 300 joules with a 350-millisecond fire rate. This equipment setup was used on all profiles with useful data generally recovered to a depth in excess of 25 metres below seabed. The data was initially inspected, and processing were identified as candidates to also include traces of the tsunami deposit ( Figure S1.1).
These patches of similar seismic response were recorded during the re-examination of the data to assist interpretation and guide future survey in these areas. The initial lithological assessment of core ELF001A revealed seven different sediment types (units) separated by sharp, abrupt or diffuse contact, Table S1, Figure S1  Total sites cored in study. B. Core locations used in this study. 1A-4 ---Abrupt Contact---1.09 -1. 19 Dark grey silty fine sand with common shell fragments. Shells are <4mm commonly but with occasional fragments larger than 3cm. Shells are broken and sharp/fresh. Possibly some crude bedding. Moderately firm and compact. 1A-5 ---Diffuse Contact---1.19 -1. 51 Grey medium sand with very common shells fragments including whole shells and freshly broken shell fragments. Small stones throughout unit. Loose, unconsolidated and structureless. 1A-6 ---Sharp Contact---1. 51 -3.50 Mid to dark grey finely laminated silts and find sands. Sub-horizontal laminations from 2-3mm thick to 1cm thick. Occasional brown organic fragments. Moderately firm and compact. Occasional sand beds 2cm thick. Possibly becoming more silty with depth.

S1.3 Palaeobathymetry and estimation of 8.2Kyr shoreline and tsunami run-up
To understand the nature of this Tsunami deposit further, the seabed bathymetry of the Southern North Sea area was recovered from the European Marine Observation and Data Network (EMODNet) data portal 12 for use in a localized reconstruction. Palaeobathymetry was created by adding isostatic adjustment data 13 to the bathymetric data obtained from EMODNet using the method identified in Hill et al. (2014) 18 . Small, unresolved islands and features associated with the presence of modern sand banks were removed from all coastlines to aid clarity and consistency. Using the method described by Fruergaard et al. 6 , the local Tsunami height at ELF001A was then extrapolated from the topographic height of the top of the tsunami sequence within the core and applied to the palaeobathymetric data, Figure S1.4.
It should be emphasized that the purpose of this method is not to attempt to produce a full or detailed model of the tsunami, but rather to better to generate a visualization to improve our understanding of the spatial nature of this deposit and how it arrived in the basin within which it is situated. To avoid over-fitting, from these data we inferred a more generalized  1.90m -3.60m This phase represents little or no change in detrital input consistent with a low energy estuarine system. The S-ratio gradually increases down this part of the core suggesting a higher ratio of magnetite being deposited through this phase, which could be indicative of a change in salinity.

S2.2 Elemental core scan
The surface of each core section of ELF001A was scraped and cleaned to ensure a smooth, flat surface, and the top 3 metres was scanned using an Itrax® XRF core scanner at 500µm resolution with a dwell time of 15 seconds and x-ray tube settings at 30 kV and 50 mA. The modern material in the uppermost core section (Unit ELF001A-1) was not scanned. In individual core sections, the scanning line was adjusted to avoid sampling holes and some sub-sections were run individually to enable the core surface to be kept as flat as possible. Scanning data were compiled to produce a composite sequence.
Gaps in the data represent parts of the core which were not scanned. This may have been due to sampling gaps, those instances where the nature of the sediment did not provide a smooth surface for scanning, or the result of poor data quality (low kcps values). Selected elements are presented, normalised to the sum of the incoherent and coherent scattering which account for the effects of Compton scattering and Rayleigh scattering respectively [28][29][30] .
In general terms, analysis of the ELF001A core suggests that the unit interpreted as indicating a major storm event or tsunami is associated with a rapid switch to marine material. This is indicated by the sulphur signal above the red line in Figure 3. Calcium and strontium signals could derive from terrestrial conditions but these signals differ significantly to the pattern for titanium and potassium which must related to terrestrially derived material. The shaded bar in Figure 3 indicates where the data suggests an increased amount of terrestrial material above the proposed tsunami layer. Bromine may be a proxy for organic matter or sea spray, depending on context. Here it is more likely to represent organic content. Iron follows titanium and does not seem to represent changing redox conditions. Peaks in the calcium and strontium signals above the layers associated with an enhanced terrestrial signal are further indications for marine conditions and perhaps shells.

S2.3 Organic Chemistry profiling:
An Agilent 7890A gas chromatograph (GC) coupled with a 5975C Inert XL mass selective detector was used for the lipid analysis. The splitless injector and interface were maintained at 300°C and 340°C respectively. Helium was the carrier gas at constant flow. The temperature of the oven was programed from 50°C (2 min) to 350°C (10 min) at 10°C/min. The GC was fitted with a 30m x 0.25mm, 0.25µm film thickness 5% Phenyl Methyl Siloxane phase fused silica column. The column was directly inserted into the ion source where electron impact (EI) spectra were obtained at 70 eV.
Samples were analyzed using a full scan method from m/z 50 to 800. For the lipid extraction, fourteen sub-samples from core ELF001A were dried at room temperature for 48 hours, ̴ 3g of each was then solvent extracted using three portions of 12ml (dichloromethane: methanol 2:1 v/v) with ultrasonication and centrifugation. The solvent was transferred into a clean glass vial and removed under a stream of nitrogen at 40°C. The extracts were then silylated with ̴ 5 drops of BSTFA at 70°C for an hour. Excess BSTFA was removed under a stream of nitrogen and the samples diluted in 1ml of dichloromethane for analysis.
The lipids analysis of core ELF001A, yielded n-alkanes, fatty acids, n-alkanol and sterols, of these lipids the n-alkanes are the most informative in respect of the origin of the lipids. These show that; the area ELF001A -3 is dominated by marine organic inputs probably from submerged aquatic plants. In area ELF001A -4, aquatic plants are present, with the signals for bacteria and terrestrial plants in significant quantities. In addition, signals of sulfate reducing bacteria were also identified. This area has the chemical profile of an estuarine area, or it may be an area of water present, just before submergence. Area ELF001A -5 has a chemical organic profile similar to that obtained from area ELF001A -3, where submerged marine plants are dominant. In contrast, area ELF001A -6 is the most complex portion with evidence for terrestrial and marine plants and algae within just 15cm of the column, although marine inputs dominate this area. This suggests a major event associated with the deposition of these mixed deposits within well-defined strata. Area ELF001A -7 is the most homogenous of samples examined and is associated with terrestrial plants, bacteria and freshwater within the lower part of the core. The Carbon Preference Index CPI ratio for these samples distinguishes between terrestrial and marine sources. This indicates that terrestrial materials are increasingly present in the lower parts of the core 31,32 ( Figure 3). All NAR ratios are closer to one than zero, which indicates the origin of the lipids from sources other than petroleum 33,34 ( Figure 3).

OSL dating of sediments
In this supplementary text, we describe the protocols and procedures that were used to determine the quartz SAR OSL ages shown in The sediments revealed in core ELF001A were first appraised using portable OSL equipment  It was concluded that, with the dosimetry as presently constrained, the combined distributions were most appropriate for calculation of the stored dose. In justification of this, the stored doses thus obtained correlate well with the apparent dose-depth profiles obtained earlier (R 2 = 0.943).

D. Dose rate determinations (Figure S3.6)
The dose rates to these materials were assessed through a combination of X-ray Fluorescence core scanning, high-resolution gamma spectrometry (HRGS), and inductively-coupled plasma mass spectrometry (ICPMS) analysis.
Semi-quantitative element concentrations of K, U and TH, as obtained by X-ray Fluorescence core scanning at Aberystwyth University are shown in Figure S3. 6

E. Age determinations
Luminescence ages are calculated as the quotient of the stored dose (or burial dose, Gy section C above) and the environmental dose rate to these materials (mGy a -1 ; section D;    Obtained by X-ray Fluorescence using the Itrax® core scanner at Aberystwyth University.

S4.1 Foraminifera and ostracods A rapid assessment of the samples was undertaken on 15
samples. A range of materials were present in the samples including plant debris and seeds, molluscs, diatoms, and insect remains. Foraminifera and ostracods were present in all samples.
Three microfossil facies associations have been identified from the samples (Table S4.1).
The lowermost facies (associated with unit ELF001A-7) appears to be one indicative of estuarine mudflats. The microfaunas are very restricted (suggesting brackish conditions). The foraminifera are often very small and the ostracods are represented invariably by small juveniles. This is odd and may be, in part, a function of reduced salinities in the environment.
This lower part of the sequence also contains the remains of many spirorbid polychaete worms, which are normally attached either to a hard substrate or seaweed. In the absence of a hard substrate it is likely that seaweed was common and this suggests an abundance of algae on the mudflats. There are also a few juvenile molluscs in this part of the sequence.   0.96-0.97m 1.03-1.04m 1.10-1.11m 1.15-1.16m 1.20-1.21m 1.30-1.31m 1.35-1.37m 1.47-1.48m 1.66-1.67m 2.10 The results are presented as a pollen diagram produced using TILIA and TILIA*GRAPH 54 ( Figure 4, Figure S4.1A). Microscopic charcoal fragments were counted and are expressed as a percentage of total land pollen. The diagram has not been divided into biostratigraphic assemblage zones, but the position of the Tsunami deposit between 1.03-1.55m is indicated, dividing the sequence into pre-and post-Storegga; no pollen was preserved in this unit (Unit ELF001A-6). The diagram is dominated by relatively few, predominantly arboreal taxa: total tree and shrub percentages are generally above 90% total land pollen (TLP) with herbs accounting for a maximum of 20%. This implies the presence of dense woodland in the pollen source area. Corylus avellana-type (likely to be hazel, rather than Myrica gale in this situation), is dominant throughout (c. 60% TLP). Other trees which are consistently recorded but at lower percentages, are Quercus (oak) and Pinus sylvestris (Scots' pine) (both c. 10-20%), with lower values for Ulmus (elm; up to 5%), Alnus glutinosa (black alder; c. 5%) and Betula (birch, max 9%). Other trees/shrubs recorded sporadically at low percentages are Tilia (lime), Salix (willow), Fraxinus (ash) Hedera helix (ivy) and Ilex aquifolium (holly).
Herbaceous taxa account for a relatively low proportion throughout, but with Poaceae (wild grasses) consistently present (max 15%). Another herb recorded in almost every sample (max 5%) is Silene dioica-type (red campion), whilst Cyperaceae (sedges), Chenopodiaceae (Fat Hen family), Artemisia-type (mugwort). Ranunculaceae (buttercups), Filipendula (meadowsweet) and a few other herbs make occasional appearances including Sedum The most striking aspect of the sequence is the relative lack of fluctuation throughout, the curves of all the taxa are remarkably stable. Two comments are pertinent to how this relative homogeneity might be interpreted. Firstly, the data indicate that the vegetation within the pollen source area was broadly stable across the period of time represented by the diagram with no palynologically identifiable changes. Secondly, it is possible that this apparent stability of the environment through time, is related to the nature of the pollen source area for the silt dominated deposits that constitute the sequence. It is likely the pollen derived from a relatively large spatial area, including the dryland landscape adjacent to the lagoon but also terrestrial locations further upstream. In other words, the pollen record is resolving an area of landscape of potentially tens of square kilometres. Moreover, there is likely to have been a degree of mixing and reworking of the pollen within the water column, so interpretation must of necessity be tentative in terms of the extent or character of inferred vegetation dynamics throughout the sequence.
There are no pronounced changes in the spectra immediately above the hypothesised Tsunami layer, which might be taken to imply that this event had no identifiable impact on the local vegetation, with a predominantly wooded landscape both pre and post-Tsunami.
However, this interpretation must be tempered by the previous comments concerning the potential taphonomic complexity of the pollen record. However, there is evidence of potential changing woodland dynamics above this unit, towards the top of the diagram, between 1.03-0.84m. Total tree pollen percentages increase as a result of rising values for Pinus alongside reductions in Corylus. Again, this is difficult to interpret but may indicate an expansion in Scots pine, prior to the establishment of marine conditions at this location. Interestingly, there are also increased representation of microscopic charcoal across the same levels. It would be tempting to interpret these changes as potential evidence for increased dryness in the period before the final marine incursion at this location. Otherwise, there is no palynological evidence for woodland recession that would be expected to result from rising water tables in advance of rising relative sea levels. Brackish waters Intertidal to shallow sublittoral Sublittoral

S4.3 Diatom analysis:
A selection of 23 spot samples were prepared for initial diatom assessment from the sedimentary sequence of core ELF001A. Diatom preparation followed the methodology of Plater et al. (2000) 55 , with additional pretreatment using sodium hexametaphosphate, to assist in minerogenic deflocculation. Samples were sieved using a 10μm mesh to remove fine minerogenic sediments. The residue was transferred to a plastic vial, from which a slide was prepared, using Naphrax as the slide mountant, for subsequent assessment.
For samples in which diatoms were encountered in sufficient abundance during the initial assessment, a minimum of 300 diatoms were identified for each sample depth. If preservation was found to be poor, a complete slide was traversed in an attempt to extract the diatom data This is a common snail under stones and on weeds from mid tidal level down to 15m depth on rocky shores. Other molluscs present in this unit include Retusa obtusata, a predatory gastropod which was most likely preying on Rissoa; and the common mussel Mytilus edulis, which is usually found intertidally on rocky coasts. The mussel shells are all broken, which may suggest compression from overlying sediment or wave transport, however they are not especially rounded, which suggests they were not subjected to much wave rolling. Rather than rocky shores, common cockle, Cerastoderma edule, and European oyster, Ostrea edulis, are found in muddy and sandy environments. The intertidal to lower shore assemblage continues to dominate in Unit ELF001 A-5, however the brackish water fauna is now absent.
There is lower species diversity in this unit, however there is somewhat more equitability, Libraries were visualised on a 2% agarose gel. They were then cleaned using 45 µl SPRI beads and eluted in 20 µl TET buffer 75 . The cleaned libraries were quantified using a Qubit assay (Invitrogen) and a fragment size profile produced using a Bioanalyzer (Agilent).
Libraries were normalised to 4nM and pooled prior to sequencing on the Illumina NextSeq platform using the high-output, v2, 150-cycle kit (75x75 paired end), Table S4 An initial metagenomic BLASTn search (version 2.6.0) 79 was undertaken using the tab output (specified using -outfmt "6 std staxids"). This allows a large volume of data to be processed with a far smaller data footprint than the full BLAST output format. This was then converted to RMA format using the MEGAN5 command line (version 5.11.3) 80 , enabling the visualisation of the preliminary data. The patchiness of DNA sequence databases and the overrepresentation of model organisms leads to unreliable assignation of sequences. Reads were therefore stringently filtered using the Phylogenetic Intersection Analysis (PIA) 81 .
FASTA sequence reads with preliminary assignation to taxa of interest (in this case Viridiplantae, Chordata with primate reads excluded, Arthropoda, and a random subset of 10,000 bacterial reads) were extracted from the RMA files using MEGAN5 command line tools (version 5.11.3) 80 , These reads were subjected to a second round of BLASTn (version 2.6.0) 79 , this time to generate the full BLAST format as an output. These were then used as an input for Phylogenetic Intersection Analysis (Smith et al., 2015; default settings) 81 in order to retrieve stringent assignations. Taxa with >2% of the assigned reads also assigned to that taxon in the negative controls for that sequencing run were removed. Any taxa that remained after the stringent filtering that were not native to Europe were discarded, accounting for about 3% of the data.
Authentication: DNA damage analysis Two approaches were taken to establish whether damage patterns characteristic of ancient DNA were present to authenticate the sedaDNA. The constant 4˚C environment of the sea floor leads to an expectation of damage which may be as low as 2.5% of terminal overhang cytosines deaminated in the age ranges explored in this study 15 , which may be below levels of detectability as has been observed in previous studies 81,86 . Furthermore, the high ionic environment of marine conditions is expected to reduce deamination rates by reducing the rate of hydrolytic attack 87,88 , as has been observed for marine environments 89,90 . Here we observed deamination rates in the range of 7-15% which is in line with these expectations of low damage levels for the sediments of Unit ELF001A-6, dated to 8.14 ± 0.29 Kyrs. BP ( Figure S4.2).
The metadamage analysis across the sedaDNA set confirms this low level of damage signal, agreeing closely with the mapdamage assessment indicating damage levels are reflected across taxa ( Figure S4.4). We applied the metadamage analysis to sequence data both before and after the PIA filtration step. The mismatch base line in the post PIA analysis data is lower than the pre-filtered data, indicating a lower level of phylogenetic background noise as would be expected as less accurate phylogenetic assignations are rejected. In this way the metadamage analysis validates the PIA analysis.