Radiocarbon dating distal tephra from the Early Bronze Age Avellino eruption (EU-5) in the coastal basins of southern Lazio (Italy)

Distal tephra from the major Somma-Vesuvius Avellino (AV) eruption is widespread in the coastal basins of Southern Lazio (Central Italy). Dated to 1995 ± 10 cal yr BC in 2011, later on doubts arose about the reliability of this frequently cited age. This led to a major effort to date AV tephra holding sections, based on a thorough methodological approach. Various aspects were studied to identify sections yielding reliable 14 C ages, including bioturbation, inbuilt age, and variable sediment accumulation rate. Lowered rates upon deposition of tephra, particularly in anoxic marshy environments and attributed to toxic F contents, showed up as sharp increases in pollen density. The ‘sampling error ’ was quantified for specific sedimentary environments and derived from coring data and published data on accumulation rates for similar Central Mediterranean sites. Next, two Bayesian analyses were performed, a traditional using the full set of samples and a novel, based on samples that were deemed as suitable (no bio-turbation, inbuilt age, etc.) and of which the age was corrected for the sampling error. The age obtained by the novel analysis had the smallest range (1909 – 1868 cal yr BC), differs about a century, and is virtually identical to the ages published by Passariello et al. (2009) and Alessandri (2019). The earlier found age (2011) is ascribed to a statistical coincidence. The results solve a long debate on the age of the AV eruption, which is the youngest of the three major eruptions in the Central Mediterranean Bronze Age. Ages of the other two, the Agnano Mt Spina (Phlegrean) and FL eruption (Etna), are still uncertain and disputed. This study illustrates the need for a thorough approach in 14 C dating tephra holding sediment archives in the Central Mediterranean, and employed a methodology that can be applied in such approach. Attention is called for potentially toxic fluorine concentrations in Campanian tephra, which may have had a serious impact on the contemporary environment and induced chronological hiatuses, but hitherto were not reported for the early tephra.


Introduction
The Avellino eruption of the Somma-Vesuvius counts among the major Holocene eruptions in the Central Mediterranean (see Zanchetta et al., 2011) and resulted in 'Pompeii type' sites close to the volcano, where the Avellino tephra covers an exceptionally well conserved Early Bronze Age landscape (see e.g.Albore Livadie, 1999;Vanzetti et al., 2019).These sites were discovered shortly before the end of the 20th century (Albore Livadie et al., 1998) and gave rise to numerous excavations (see e.g.Albore Livadie et al., 2019;Di Vito et al., 2019;Vanzetti et al., 2019.These allowed for a deep insight into the contemporary prehistoric cultures and their land use.The relevance of this eruption for paleoclimatic archives lies particularly in the wide distribution of its tephra and the associated possibilities for long distance correlation.It is enhanced by its specific chemical and mineralogical signature that allows for its easy recognition (see Zanchetta et al., 2011 and2019).This is exemplified by identification of the Avellino tephra in well known cores from locations as far apart as Lago d'Accesa, in Tuscany (Magny et al., 2007), Lago Grande di Monticchio, in Basilicate (Wulf et al., 2004), Lake Veliko jezero, in Croatia (Razum et al., 2020), Lake Ohrid and Lake Shkodra, both in Albania and Montenegro (Wagner et al., 2008;Sulpizio et al., 2010;respectively), and the Sea of Marmara, in Turkey (Çagatay et al., 2015).
Early radiocarbon dating attempts focused on pre-eruption materials found at proximal sites and dating from immediately before the eruption, which was mainly because of the uncertainty about the length of time over which these sites remained uninhabited after the eruption.It also explains why the most accepted age (3.86 ± 0.03 cal Ka BP) was based on radiocarbon analyses of a goat that was killed by the eruption (Passariello et al., 2009).Radiocarbon datings from truly distal areas, based on multiple samples from sedimentary sequences with an intercalated Avellino tephra layer and a Bayesian approach, did not exist till the discovery and dating of such layer in the Agro Pontino (see Fig. 1).It was identified as a tephra layer from the EU-5 eruption phase (Sulpizio et al., 2008) and dated to 1995 ± 10 cal yr BC (or 3935-3955 cal yr BP), based on radiocarbon dates for two sites: Migliara 44,5 and Campo inferiore (Sevink et al., 2011).
The age obtained was slightly older than ages published in previous studies (Albore Livadie et al., 1998: 1880-1680 cal yr BC;Passariello et al., 2009: 1935-1880 cal yr BC) and this led to quite some scientific debate (see for example Jung, 2017).There still is no agreement on its age as is exemplified by Zanchetta et al. (2019) and by the recent study of Razum et al. (2020), in each of which both dates (Passariello et al., 2009;Sevink et al., 2011) are mentioned.This uncertainty led Alessandri (2019) to review all available radiocarbon dates connected with the eruption, of which most stem from proximal archaeological sites, but also include distal sediment sequences.He found a calibrated age that is close to the age found by Passariello et al. (2009): 1929-1856 cal yr BC.Alessandri additionally stressed the importance of the Avellino tephra and its age for the stratigraphy of the Early and Middle Bronze Age in South and Central Tyrrhenian Italy, which has been heralded by many other archaeologists.Remarkably, it is not only the Avellino eruption of which the absolute age is still disputed.The same holds for the Agnano Mt Spina eruption (Phlegrean) and FL eruption (Etna), which are the other two major Bronze Age tephra layers in the Central Mediterranean mentioned by Zanchetta et al. (2019).They state that the 'Agnano Mt Spina chronology is supported by poor radiocarbon dating, which needs to be significantly improved' and that 'The chronological constraints for FL are even less robust'.
A palaeogeographical reconstruction of the Agro Pontino and Fondi basin, focussing on the Bronze Age and using the Avellino EU-5 tephra layer (further referred to as the AV layer) as a marker bed, is a central topic in the Dutch Avellino Impact Project, which started in 2015.Results from this project were published in a number of papers (Doorenbosch and Field, 2019;Van Gorp and Sevink, 2019;Van Gorp et al., 2020;Sevink, 2020), and include a recent paper on the identification and characteristics of Bronze Age tephra in these basins (Sevink et al., 2020b).Apart from the AV layer, these comprise far less common and thinner tephra layers, ascribed to the earlier Phlegrean Astroni6 and the later Somma-Vesuvius AP2 tephra.Radiocarbon ages that were obtained for samples from directly below the AV layer in the Agro Pontino and Fondi basins (Sevink et al., 2020b) varied little and converged to c. 3570 yr BP, conforming to results for proximal sites (e.g.Passariello et al., 2009;2010;Albore Livadie et al., 2019;Alessandri, 2019).However, samples from above this AV layer exhibited a much larger variation in radiocarbon age, which was rather surprising and could not be straightforwardly explained.These results raised serious questions about the reliability of the earlier established age of the AV layer in the Agro Pontino (Sevink et al., 2011), which were analogous to questions that had been raised by Jung (2017) and Alessandri (2019).It led us to a critical evaluation of the radiocarbon dating of the AV layer, paying attention to aspects that thus far had not been considered in such studies of distal sites.
In the first place, in most Bayesian analyses that served to date intercalated tephra layers in sediment archives it is implicitly assumed that the sedimentation rate was constant.An example is the dating of the AV layer by Sevink et al. (2011).However, evidence that such assumption is valid is often scant and our results strongly suggested that in several of the sections studied a more or less significant stratigraphic hiatus or decline in the sedimentation rate had occurred upon tephra deposition.This seriously hampers a straightforward application of a Bayesian analysis and led us to the following connected questions: 1) What evidence may exist for the occurrence of a post-tephra stratigraphic hiatus/lower sedimentation rate?And 2) What might be the cause of such hiatus/lower sedimentation rate?We tried to answer both questions; the first by studying pollen densities in representative sections, which is a technique that has often been used for the identification of changes in sedimentation rates (see for example Maher, 1972); the second by paying attention to the potentially toxic impact of tephra on ecosystems/vegetations (see for example Grattan and Pyatt, 1994).
Another aspect is that lagoonal/lacustrine sediments in the Mediterranean generally are low in materials that are suitable for reliable radiocarbon dating (such as from non-aquatic plants, see Sevink et al., 2013).In such situations, the 'sampling error' may become significant, being defined as the difference in age between the material that was sampled and dated, and the tephra layer.This difference is determined by the vertical distance between the two and the sedimentation rate but generally values for the latter are rather uncertain.A further complication is that seeds and other plant materials are selected from individual samples, which have a certain thickness, thus introducing further uncertainty in the sampling error.In particular in environments with low sedimentation rates and scarce suitable plant macro remains, for which large samples need to be taken for the retrieval of the material required for radiocarbon dating, this 'error' may easily be in the order of 100 years ±50 years.In most studies that aim to date tephra from this type of sedimentary environment little attention is paid to this sampling error, but it can be accounted for in a dedicated Bayesian analysis, that also considers variations in sedimentation rates.
Based on answers obtained for the two questions described above, from a larger set of sites we selected those sites with stratigraphies/ ecosystems that were least affected by a toxic impact of the tephra deposited and lacked indications for a significant change in sedimentation rate upon tephra deposition.On these selected sites we performed such dedicated Bayesian analysis, which included a check of the effect of the 'sampling error'.Results are reported in this paper and solve the still existing controversy about the age of the Avellino eruption.Additionally, we discuss the relevance of our seemingly novel methodological approach for solving uncertainties in tephra ages for similar sediment archives.

General information and backgrounds
During the last glacial period, when sea level was very low, in the Agro Pontino and Fondi basin rivers cut deep valleys into a thick complex of predominantly fine-textured Quaternary sediments (e.g.Sevink et al., 1984), of which the youngest were described as the Borgo Ermada marine complex.These valleys gradually filled in with the Holocene sea level rise.Deposits from this Holocene transgression were described as the Terracina marine complex.Towards c. 2 ka BC, when sea level rise slowed down (Lambeck et al., 2011;Vacchi et al., 2016), beach ridges could build up and lagoons came into existence (Sevink et al., 1982(Sevink et al., , 1984;;Van Gorp and Sevink, 2019;Van Gorp et al., 2020).Near the coast, these were mostly freshwater lagoons, which overall were shallow and underlain by sandy beach ridge deposits, while further inland the lagoons graded into marshy valleys.
In the central part of the Agro Pontino basin a different situation existed.Upon sea level rise the Amaseno river built up an alluvial fan, which shortly before the AV eruption started to block the single outlet of the fluvial system that drained the northern part of this basin (see AF in Fig. 2).This led to a gradual 'drowning' of the earlier inland landscape and created a large lake and associated marshes (Van Gorp and Sevink, 2019;Van Gorp et al., 2020).In the NE part of this lake, peat with intercalated lacustrine marls (calcareous gyttja) and some travertine accumulated, and in the SW pyritic black organic clays.In the NW, these lacustrine deposits graded into fluvio-deltaic sediments and, further upstream, genuine fluvial sediments.The waters that ran into the lake from the adjacent mountains were largely fed by springs with highly calcareous and often sulphuric waters (Boni et al., 1980;Tuccimei et al., 2005;Sappa et al., 2014).A similar situation existed in the Fondi basin, where an inland lake formed also with Holocene peats and calcareous gyttja.
Agro Pontino: CL = coastal lake; IL = inland lake.AF = Amaseno fan.1/2 = oxic lacustrine/lagoonal sediments and 3 = anoxic lacustrine sediment with shaded transitional zone; 4 = fluvio-deltaic sediments.conspicuous and easily identifiable tephra layer.It was preceded by tephra from a slightly earlier and smaller Astroni eruption (most probably the Astroni 6 eruption, see Sevink et al., 2020b), which seems to be restricted to the Fondi basin and followed by a minor tephra fall from the Monte Somma-Vesuvian AP2 eruption, recorded in the south of the Agro Pontino.Occurrences, characteristics and ages of these three tephra are extensively described in Sevink et al. (2020b).The age of the AP2 tephra is less well known than that of the Avellino eruption, its most recently published age being c. 1700 cal yr BC (Jung, 2017;Sevink et al., 2020a).The age of the Astroni 6 eruption is rather uncertain.According to Smith et al. (2011) it is slightly younger than 4098-4297 cal yr BP (Astroni 3) and older than 3978-4192 cal yr BP (Fosso Lupara), while  (2011).
In the Agro Pontino and Fondi basin, four types of sedimentary environments were distinguished in which the AV layer was encountered (Van Gorp and Sevink, 2019;Sevink et al., 2020b).Type 1: oxic aquatic to marshy, with peat to peaty clay; Type 2: oxic aquatic (lacustrine/lagoonal), with calcareous gyttjas to calcareous marls ('gyttja'); Type 3: anoxic marshy, with pyritic more or less peaty black clays ('pyritic clays'); Type 4: oxic, fluvio-deltaic, with calcareous clays to loams.Their overall distribution is depicted in Fig. 2. In this figure no distinction is made between areas with type 1 and with type 2 sediments, because the intricate pattern in which they occur cannot be depicted at the scale of this figure.
Following on the large scale development of agriculture in Southern Lazio, which started in the Late Bronze Age (Attema, 2017), over large areas these sediments were gradually buried under mostly fine textured, reddish-brown reworked soil material, described as fluvio-colluvial deposits (Sevink et al., 1984), whereas outside these areas peat continued to accumulate, if not stopped by land reclamation and associated drainage (Van Joolen, 2003;Feiken, 2014;Attema, 2017).The overall situation is depicted in Fig. 3.
The AV layer was identified in the field as a 2-3 cm thick intercalated sandy grey-creamy coloured tephra layer.This layer holds very conspicuous idiomorphic 'golden' mica and sanidine crystals, of which the mica reaches sizes up to c. 4 mm.Where the tephra layer is intercalated in gyttja, such as in the interior basin of the Agro Pontino near Mezzaluna and in the coastal lagoonal deposits near Borgo Hermada, it forms a virtually continuous horizontal layer with often sharp upper and lower boundaries, which could be followed over large distances.Though this type 2 sediment holding the AV layer is extremely well suitable for palaeoecological studies, plant macro remains are invariably from truly aquatic plants, potentially affected by reservoir effects, and therefore unsuitable for radiocarbon dating.The pyritic (peaty) clays (type 3) are marked by low sedimentation rates, demonstrated by the shallow occurrence of the AV layer in such sedimentsoften less than 50 cm below current ground level or below the fluvio-colluvial deposits -and are also low in suitable plant macro remains, posing problems for reliable radiocarbon dating.Aquatic to marshy peats and clays (type 1) hold a more discontinuous AV layer, but these sediments are basically well suitable for radiocarbon dating and palaeoecological studies.Fluviodeltaic sediments (type 4) were rather unsuitable for palaeoecological studies and radiocarbon dating for a variety of reasons: common hiatuses, discontinuous strata, low pollen content due to oxidative conditions, etc.
Most of the radiocarbon dates were on samples from type 1 sediments.For the palaeoecological studies cores and monoliths were sampled to cover the two basins concerned.These samples allowed for an analysis of the impact of the AV tephra fall on the relevant major types of vegetations encountered.For a study on the potential toxic impact of the tephra, cores were selected to cover the same range of environments (types 1, 2, and 3), but based on the systematic differences in geochemical conditions (oxic versus anoxic/sulphidic; super saturated calcareous watersnon saturated waters).The approach was to assess whether 'spikes' in element concentrations occur in the AV layer and immediately above that layer, which can be linked to the deposition of this tephra, the elements concerned having been immobilized in the specific environment.

Methods and materials
For radiocarbon ( 14 C) analysis of samples from immediately above and below tephra layers, plant macro remains were handpicked under the microscope from subsamples that were obtained by sieving over a 105 or 150 μm mesh sieve to remove fines.Remains from plants that might obtain their carbon dioxide from water were excluded.In some instances, suitable plant macro remains were absent and humic material (black organic clay with very finely divided organic matter) was used.For 14 C analysis, samples were subjected to an ABA pre-treatment.Samples were analysed by the AMS-method at the Centre for Isotope Research (CIO) of the University of Groningen, The Netherlands.For an extensive description of the methods and AMS systems (GrA and GrM) used at CIO, see Dee et al. (2020). 14C dates given (in yr BP) are based on the standardized calculations, including correction for isotopic fractionation (Mook and Van Der Plicht, 1999;van der Plicht and Hogg, 2006).Dates have been calibrated using the software OxCal 4.3 (Bronk Ramsey, 2017) and the IntCal13 calibration curve.
For the palaeoecological study, sediment subsamples of 100 cm 3 were wet sieved and plant macrofossils were picked from the resulting residues.For separation of pollen, samples were treated with 10% KOH, 37% HCl, bromoform/ethanol (specific gravity 2.0) and acetolysis.To every sample Lycopodium spores tablets were added for the determination of the pollen concentration.Results for plant macrofossils, pollen and spores have already been published elsewhere (e.g.Bakels et al., 2015;Doorenbosch and Field, 2019) or will be published in the near future.Here, focus is on changes in vegetation upon the deposition of the AV tephra inferred from the pollen assemblages encountered and in pollen concentration as indicator for changes in sediment accumulation rate.
Thin sections of undisturbed samples for microscopic study were produced at the RCE (Amersfoort, The Netherlands) by impregnation with resin, followed by cutting and polishing to a thickness of ca. 30 μm.
Sections were studied under a petrographic microscope.
For chemical analysis of potentially toxic elements, cores were taken with a gouge auger at sites with a distinct AV layer.Core sections of 10-12 cm length were cut into 5 to 6 samples that were each 2 cm thick, with the tephra-bearing layer being one of the central layers.From each sampled layer 250 mg was transferred to a 50 ml Teflon PFA microwave vessel and 6 ml HCl 37% and 2 ml HNO 3 65% was added.The samples were left to react for 60 min before digestion in a microwave (Multiwave 3000, Anton Paar GmbH, Ostfildern, Germany).The sample was then transferred to a 50 ml volumetric flask and after addition of 1 ml of lanthanumnitrate (3.1 g La(NO 3 ) 3 in 100 ml water) diluted to the mark using 18 MΩ-water.The whole procedure was performed in duplicate.Trace elements (Zn, Pb, Hg, Cr, V, Cu, Sr) were measured in triplicate using an ICP-OES (Optima-8000, PerkinElmer, Waltham, U.S.A.).Analyses were performed at the IBED lab (University of Amsterdam).
Fluorine contents were estimated at Actlabs (Canada) by FUS-ISE.Samples 0.2 g in size were fused with a combination of lithium metaborate and lithium tetraborate in an induction furnace.The fuseate was dissolved in dilute nitric acid and prior to analysis by ISE the ion strength of the solution was adjusted.The chloride ion electrode is immersed in this solution to measure the fluoride-ion activity directly.The detection limit for F was 0.01%.Actlabs also performed a full chemical analysis of sample 192 using procedures and techniques which are similar to the one described above.
Bulk densities of undisturbed core segments were established by drying and subsequent weighing of samples of known volume.After treatment with H 2 O 2 and HCl to remove organic material and calcium carbonate, respectively, the weight percentage of the fraction >20 μm was established.Samples were sieved over a 63 μm sieve.Silt and clay fractions in the remaining suspension were estimated with a sedigraph (Sedigraph III Plus, Micromeritics, Norcross, USA).Tephra contents were defined as the fraction >63 μm, since that fraction was found to be dominantly composed of tephra particles (see also Sevink et al., 2020b).
Post-Avellino mean sediment accumulation rates for the various sites and cores studied were estimated using all available information (archaeological, chronometric) on the age and phasing of the younger, anthropogenic fluvio-colluvial sediments that were encountered in the sections concerned (see e.g.Van Joolen, 2003;Feiken, 2014).
Additionally, for a small number of sites holding two tephra layers (AV layer and either tephra from the Phlegraean Astroni eruption or the Vesuvian AP2 eruption), mean accumulation rates could be estimated based on the ages of these tephra layers and thickness of the intercalated sediment.Lastly, for comparison accumulation rates for other Central Italian sites were taken from published cores.

Results and discussion
Sections that were obtained with a gouge corer, by digging a pit, or from existing pits are indicated in Fig. 1.An overview of the various sections and their major characteristics is presented in Table 1, while Table 2 shows the results from the 14 C dating for relevant samples.A full overview of samples for which 14 C dates are available is given in appendix A. First, attention will be paid to the dates and the potential factors that play a role.These factors will be dealt with individually, discussing their potential role in the sections studied and the relevant analytical results.The objective is to assess the value of each of the dates and how they might best be incorporated in the Bayesian chronological analysis (Bronk Ramsey, 2009).

14 C dates: first evaluation
In Table 2 an overview is given of the dates for samples that may provide reliable indications of the age of the AV layer.To facilitate the discussion, samples from below and from above the AV layer are separately presented.Additional information presented in this table concerns aspects that are dealt with later on and pertain to factors that limit their suitability for the Bayesian analysis, and the sampling error.For details on the Bayesian analyses, reference is made to Appendix B, while OxCal codes used can be found in Appendix C.
The dates for materials from below the AV layer at first sight provide a precise and reliable terminus post quem for this tephra, given the fact that 10 of the 14 samples analysed were close to the previously published age (Sevink et al., 2011).The exceptions are from the sites of Frasso 500 (500), Femmina Morta (197),and Mezzaluna (405).For the Femmina Morta site, the set of 14 C ages for the various samples (see appendix A) strongly suggests that the latter two dates concern younger materials that for some reason were present in the sediment underneath the AV layer (for a full discussion see 4.2).The date for Mezzaluna concerns wood from an older tree trunk (at 10 cm below the AV layer) and its higher age is therefore not surprising.
A factor not accounted for is the aforementioned sampling error, which combines the accumulation rate and depth of sampling relative to the AV layer.It represents the offset in age between the sedimentary layer from which the 14 C date was obtained and the tephra layer.At Mesa for example, the sample was from 0 to 3 cm below the AV layer, which assuming an accumulation rate of 3 cm/century would imply that the AV layer is between 0 and 100 years younger.Whether this rate is realistic will be discussed in section 4.4, as well as the potential impact of such an error.
For the samples from above the AV layer, already at first sight it is clear that the 14 C ages obtained for the samples from Migliara 44.5 and Campo inferiore are unexpectedly high and the result for Mesa too young to be consistent with the rest of the dates in Table 2.A number of causes can be distinguished for the wayward dates, several of which have already been mentioned in section 2. These include bioturbation, the old wood effect and the sampling error, being dependent on the accumulation rate and thickness of layers sampled for extraction of material to be dated.

Bioturbation
For most sections, the 14 C dating was performed on a small number of plant macro remains of limited size, with emphasis on seeds and small plant remains (twigs, small branches and leaves).Downward transport of such plant macro remains and charcoal by soil fauna is well known (Eisenhauer et al., 2009;Forey et al., 2011;Domene, 2016).If by earthworms, it is known to be accompanied by transport of pollen in their excreta (see for instance Van Mourik, 1999).This would likely lead to absence of temporal variability in the pollen record, betraying the homogenizing earthworm activity.The situation is different if such transport was preferentialonly plant macro remains -or the transport of pollen was of too limited magnitude to induce a recognizable homogenization of the pollen archive.Preferential downward transport of seeds is for example known from ants (Majer et al., 2008;Robins and Robins, 2011;Kovář et al., 2013).Evidently, such faunal activity will not extend below the mean lowest groundwater level, but short dry spells may have sufficed for some bioturbation to have occurred.
Only site 197 at Femmina Morta shows some evidence of preferential downward transport of plant macro remains.Its palaeoecological record exhibits well defined temporal variations in the composition of the pollen assemblages and macro remains that were identified (see Fig. 5 and Doorenbosch and Field, 2019).However, the 14 C ages obtained for plant macro remains are virtually identical, independent of their depth of sampling.Even where taken from below the Astroni layer (indicated as Astr layer in appendix A) the material is slightly younger than the AV layer.At the Tumulillo site this Astr layer was found to date from between 3720 and 3765 B P (c.2150 cal yr BC, see section 2), The pollen record demonstrates a drier phase with slightly better drainage and lower accumulation rate, following on the deposition of the AV layer (Doorenbosch and Field, 2019).It is this period with better drainage in which the bioturbation may well have taken place, which evidently also would explain the younger age of the plant materials dated.Moreover, in the thin section of the AV layer and adjacent peat layers clear indications exist for such bioturbation in the form of some large biopores (pedotubules, Brewer, 1964), postdating the AV layer (see Fig. 4).In none of the other sections studied one or more of the above discussed indications for bioturbation have been observed, nor are indications for homogenization found in the pollen records.

Old wood effect ('inbuilt age')
This effect can only be expected for samples that indeed consist of wood, but it is more complex than simply 'old wood' and is now commonly referred to as 'inbuilt age' (see Dee and Bronk Ramsey, 2014).In cases where bark is analysed, outer annual rings of larger wood fragments, or wood in the form of small twigs, a significant 'traditional' old wood effect can be ruled out, such as for the sample from Mezzaluna (see appendix A: GrM-17418) and some wood samples from Campo inferiore (appendix A: Tree trunk, outer 2-5 rings; see also Sevink et al., 2011).Wood samples of uncertain origin (i.e.unknown position within a trunk or larger branch) were excluded during our sampling.In contrast, samples from above the AV layer at Campo inferiore dated by Sevink et al. (2011) were indeed such undefined wood samples (appendix A: GrA-46210 and -32454).They were obtained from a peat layer above the AV layer, which overall was strongly decomposed.Sevink et al. (2011) described that it was only in a small depression that this peat held recognizable plant macro remains (wood) and they assumed that this wood was from a tree postdating the AV layer.
As described by Sevink et al. (2011) and more extensively by Feiken (2014), the many large beams and poles encountered below the AV layer at that site must have been cut down by contemporary Early Bronze Age peoples, who worked trunks and large branches into poles and beams by removing bark, wood and side branches.This happened prior to the deposition of the AV tephra.It cannot be excluded that some trees survived the tephra fall and earlier downcutting of the alder forest, to die and fall on top of the AV layer at a later stage.This would also be an example of the inbuilt age effect and would readily explain the unexpected age of the wood samples from above the AV layer at Campo inferiore.
A rather similar effect may occur in cases where older sediment is eroded and subsequently deposited on top of the AV layer.Such phenomenon might occur when tephra deposition causes a major reduction in biomass and biomass production (see 4.4.2),leaving nearby older sediment exposed to erosion while insufficient 'fresh' biomass is produced to mask the effect of the reworked organic matter on the overall

Sites
Nr.In Fig. 1

Sampling error
As described previously, to obviate hard water effects, macro remains of plants known to take their carbon dioxide from the atmosphere were selected for 14 C dating wherever possible, but concentrations of such remains were low, implying that often sediment slices several centimetres thick were needed to retrieve sufficient material.In cases where the accumulation rate was low this may induce a significant 'sampling error'.Various strands of evidence are relevant here, including a) potential sampling errors based on sample size and estimated accumulation rate; b) indications for post-tephra accumulation rates from the palaeoecological data; and c) potential toxic impacts of tephra.These various aspects are discussed separately.

Accumulation rates
Mean accumulation rates, calculated for the sediment sequences studied, are presented in Table 3A and testify to the very low mean rates for the deposits in both coastal areas.However, these mean values may be biased by very low accumulation rates from the Early Iron Age onwards (ca.1000 BCE), post-AV accumulation rates having been much higher initially and strongly slowing down later on.This scenario seems very realistic, given the outcomes of numerous studies on early land use in the Agro Pontino (Van Joolen, 2003;Feiken, 2014;Attema, 2017;De Haas, 2011, 2017).These studies showed that though fluvio-colluvial sediments (indicated as 'young colluvial deposits' in Fig. 3) are rather widespread, which suggests a massive supply of eroded soil material that was mainly transported by canal-like streams, the accumulation rate was very much lower outside the locations where this sediment was deposited.In these areas, rates have been very low since at least the Early Roman times.The major argument for such lack of significant accumulation is that remains of Early Roman and later land use are found in topsoils of the Holocene Terracina deposits all over the interior Agro Pontino basin (Feiken, 2014;Attema, 2017;De Haas, 2017).Moreover, the Early Roman artefacts are frequently found at the current land surface, testifying to very low accumulation rates outside the fluvio-colluvial areas for this later period.
Higher accumulation rates and thus lower sampling errors are found for the cores holding two tephra layers (Table 3B).The age of the AP2 and Astroni 6 tephra layers is not truly precisely known (see section 2 above) but can be set at c. Two centuries later and earlier, respectively.Rates calculated for such intervals are much closer to the values from the literature for cores from central Mediterranean sedimentary complexes (Table 4) and thus seem much more reliable and for that reason have been used to estimate the 'sampling error'.The five cores for which we could calculate these rates represent specific sedimentary environments types 1 and 2 -for which toxic impacts of the tephra are expected to be limited or even absent (see 4.4.2).

Potential toxic impact of tephra
As discussed before, four types of sedimentary environments were distinguished each of which has its characteristic geochemical conditions.In a sulphidic anoxic environment (type 3), for example, many heavy metals are immobilized in the form of highly insoluble metal sulphides (e.g.Krauskopf, 1967;McBride, 1994;Sundelin and Eriksson, 2001).Thus, if present in the tephra, these heavy metals may show up as spikes, whereas in the highly calcareous lacustrine environment in which type 2 sediments were formed, fluorine (F) will be immobilized in the form of CaF 2 (e.g.Garand and Mucci, 2004).Spikes of elements that are known to occur in the Vesuvian ejecta (e.g.Signorelli et al., 1999;Balcone-Boissard et al., 2012) thus can be explained as resulting from tephra deposition.Evidently, if immobilized, these elements are very unlikely to have had a toxic impact, but in contemporary environments in which such immobilization did not occur, the opposite may be true.Pb for example will be immobilized in an anoxic sulphidic environment, but because of its mobility in well aerated aquatic environments may have reached toxic levels in the latter.On the other hand, if fallen into well aerated waters which were rapidly refreshed, this dilution is very likely to have prevented toxic levels to be reached.
Table 5 shows that overall concentrations of specific elements in the sediments sampled are quite variable.For example, Zn reaches high levels in the Mezzaluna section but is low at other sites such as Mesa.We did not find distinct spikes of heavy metals that can be linked to the AV layer and are also commensurate with specific sediment types.Zn seemingly exhibited such spike in the Mezzaluna core, but in none of the other cores.The only clear spike is the F-spike found at both Mezzaluna 405 and Borgo Hermada 601.These are both from gyttja deposits with intercalated tephra layer.Since F will remain mobile in aquatic environments with low Ca concentrations, we refrained from extensive Fanalyses of such sediment types (types 1, 3, 4) and assumed that F-spikes would not be encountered.This is corroborated by the analyses for Borgo Hermada 362 (type 1 sediment), in which F-values are slightly higher, but no distinct spike occurs.
The data for the Mezzaluna 405 and Borgo Hermada 601 sections thus strongly suggests that the tephra contained F, which was most probably immobilized in the highly calcareous lacustrine environment in the form of CaF 2 .F-concentrations in this tephra were in the order of 2 g/kg, given the concentration of tephra in the layers analysed, which is defined as the fraction >63 μm (see Table 5).These F-concentrations are in the same order as reported critical health values for fluorine (e.g.Cronin et al., 2003;Weinstein and Davison 2004;Hansell et al., 2006;Petrone et al., 2011) and in line with its reported values in Vesuvian ejecta (Signorelli et al., 1999;Balcone-Boissard et al., 2012).In fact, Cubellis et al. (2016) reported even higher concentrations for the 1944 Vesuvius eruption (up to 0.5 wt percentage -5 g/kg) and described the serious environmental health problems resulting from these ashes.That the F-levels from earlier Vesuvian eruptions also led to serious human health problems was demonstrated by Petrone et al. (2011).
To what extent this F may have affected the vegetation is less clear, since available studies on the impact of F (e.g.Grattan and Pyatt, 1994;Koblar et al., 2011;Kumar et al., 2017;Banerjee and Roychoudhury, 2019) point to diverse and species dependent impacts and results published cannot readily be applied to wetland ecosystems.Moreover, studies on the impact of tephra on wetland ecosystems rarely pay attention to the impact of F (e.g.Hotes et al., 2004;Ayris and Delmelle, 2012).Nevertheless, it is clear that fluorine is indeed a biohazardous  agent.That the tephra deposition may indeed have led to a major reduction in biomass production is strongly suggested by the results for Mesa, where material from immediately above the AV layer returned an age of 3085 years BP.A standstill in the accumulation of material evidently may lead to significantly younger 14 C ages and cause a typical 'sampling error'.
In summary, fluorine concentrations in the AV tephra will have been at least in the order of 2 g/kg.If deposited in a semi-aquatic environment in which dilution was restricted, this may well have led to a toxic concentration of fluorine negatively affecting the vegetation, such as in the very tranquil anoxic and sulphidic environments described as type 3.If directly diluted or immobilized in an environment that was oversaturated relative to calcium carbonate, a toxic impact is far less likely to have occurred (types 1, 2 and 4).

Palaeoecological data
The first question is whether regrowth of vegetation after tephra deposition was instantaneous and the accumulation of organic material, calcareous gyttja or clastic sediment was a rather continuous process or was more or less seriously interrupted by the tephra deposition.
Indicators for such impact may be a) a sudden increase of the pollen concentration of plant species growing outside the wetlands where the deposits were formed (upland plants) and b) a sudden decline in the amount of pollen released by peat-forming plants (wetland plants).
In Fig. 5 data are presented on the pollen concentrations in the various sections studied.Trees and shrubs are depicted separately from herbs in order to allow the consideration of a possible influence of deforestation around the basins and within their margins which might affect conclusions.In some diagrams the original authors set the Poaceae (grasses) apart because they might represent peat-forming reeds on the one hand, but upland grasses at the other.The clearest example of absence of any effect of tephra deposition is the calcareous gyttja of Mezzaluna (type 2), but also the gyttja of Borgo Hermada 192/601 (type 2) and the peat of Ricci (type 2-4) show no immediate reaction after the AV event.They markedly differ from the pattern in the section from Migliara 44,5, with type 3 sediment, in which a massive increase is observed.Trends in type 1 sediment are rather variable.In the Femmina Morta and Fondi 122 sections the effect is considerable.In Tumolillo 2 change apparently set in before the deposition of the volcanic ash, whilst the Frasso 500 section provides no clues at all.It is striking that in the Table 3 A: Mean accumulation rates for the post-AV sediment in the sections studied, based on depth in cm of the AV-layer below the land surface (in the absence of a colluviaalluvial cover) or below the base of the colluvio-alluvial cover.Estimated age of burial is based on 14 C dating of the base of the colluvio-alluvial cover (see Van Joolen, 2003;Feiken, 2014).B: Accumulation rate for the inter-tephra sediment (between Astroni and AV-layer, and between AP2 and AV-Layer).two sites, which display a possible adverse effect on peat accumulation, the kinds of vegetation responsible for peat formation, wetland herbs and wetland trees, do not behave different from the others (Fondi 122) or even lag behind (Femmina Morta).
Overall, the data suggest that the vegetation and accumulation of calcareous gyttja and peat in a type 2 environment were not affected by the toxic tephra, whereas the accumulation of sediments in a type 3 environment was lowered to a considerable degree.The situation in type 1 environments proved to be ambiguous.Eventually, slight differences in water depth and associated water flow may have existed, which explain these differences within the latter type of environment.

Estimating sampling errors
Estimations of the accumulation rates for the various types of sedimentary environments can be based on the data presented in Tables 3B and 4 and formed the basis for our distinction of three ranges for the various sedimentary environments: 3-6 cm/century for type 3, 6-12 cm/century for type 1, and 15-20 cm/century for type 2. These values have been used to calculate the sampling error for specific samples, based on accumulation rate and mean sampling depth.Values thus obtained are presented in Table 2 for samples that are not deemed to be unsuitable for other reasons (inbuilt age, bioturbation, hiatus or uncertain stratigraphic position) and are calculated for minimum and maximum accumulation rates (listed as 'Rate', with minimum and maximum values).The uncertainty linked to the thickness of the sampled layer, which is not taken into account in the calculated 'absolute range', is of minor importance, since it does not affect the error (μ) itself, but rather the significance (±σ) of this error.For Frasso 500 GrM 17,223, for example, the 'absolute range' given in Table 2 is 8-17 years with an error of 13 years, whereas if including the uncertainty linked to the thickness of the sample (2 cm) this range would become 0-33 years.For samples that are from layers that are 1 cm thick the inclusion of the uncertainty linked to thickness of the layer sampled leads to even smaller differences, such as for Fondi 122 GrN 17,227 (range 4-8 years versus range 0-16 years).Evidently, in case that accumulation rates are low and thicker layers are sampled, larger and more relevant differences will occur between these calculated sample errors and in that case have to be accounted for.For a full discussion of this aspect, reference is made to appendix B. Characteristics that in our set of sites restrict the use of 14 C dates for a Bayesian analysis include inbuilt age, bioturbation, hiatus, and uncertain stratigraphic position.Hiatuses are typical for post-AV sections in type 3 sedimentary environments, where clear indications exist for a toxic impact of tephra deposition and associated major reduction of accumulation rates, based on the pollen density profiles.These ecosystems are evidently fragile (Bagarinao, 1992) and the vegetation may need considerable time for recovery, if impacted by toxic tephra.The pollen concentration graphs indicate that accumulation rates in type 1 and 2 environments were not or only marginally affected by tephra deposition.Unfortunately, our type 2 environments are fundamentally unsuitable for 14 C dating for other reasons (no suitable plant macro remains).This is exemplified by the Mezzaluna site, where such plant macro remains are indeed absent in the calcareous gyttja and a relatively large error is found for the wood sample from below this gyttja, which hampers the use of this date.Sediment accumulation rates for the fluvio-deltaic sites (type 4 environment) are highly uncertain and sequences are likely to have hiatuses.The overall conclusion is that for samples from type 1 environments, taken as thin layers from directly above and below the AV layer the sampling error is of minor relevance only and that such samples are ideally suitable for 14 C dating and constraining the age of the AV layer.
Table 2 thus shows the offset linked to the sampling error for those samples for which such offset can be reliably calculated.The mean offset in years (calculated from the mean distance and mean sedimentation rate) is a tentative measure for the age of the AV layer relative to the calibrated age of the sample concerned.Offsets for the sites Frasso (500), Tumolillo (1005), and Fondi (122) are small, while those for Campo inferiore (Campo), Mezzaluna and Borgo Ermada 601 are quite large.

Configuration of the bayesian models
Two different modelling approaches were applied to the 14 C data traversing the AV layer in an attempt to bring together all the evidence and refine the eruption date.All analyses were conducted in the program OxCal (version 4.3, Bronk Ramsey, 1995;2017).For details reference is made to Appendices B and C.

Model 1
The first approach consisted of a simple two-Phase Sequence with one Phase for the dates obtained above the tephra and one for the dates below the tephra.This is a legitimate configuration for such an analysis because, although the samples come from disparate sites, in chronological terms they are clearly distinguishable by this one criterion.Further, by placing the dates in Phases no relative ordering information between the samples is assumed and the underlying premise is that within each Phase the results are uniformly distributed over a period of time determined by the program itself.This is advantageous for our study because it is clear some results are significantly more offset than others from the true date of the eruption.Furthermore, in order to ensure the final outputs were not biased by a small number of particularly outlying data, use was made of OxCal's Outlier functionality.This feature automatically filters out individual results that are inconsistent with the data set as a whole.In Model 1, every 14 C date was given an equal (5%) prior probability of being an outlier, and during the iterative process the algorithm downweighed the contribution of the most wayward dates to the final outputs.All 25 14 C dates listed in Table 2 were included in the model.A comparative model (S1) is given in appendix C which, instead of outlier analysis, employed the removal of the dates with the lowest Agreement Indices until the overall model reached 60% Agreement.The OxCal codes for both Model 1 and S1 are given in appendix C.

Model 2
A second approach was taken where closer attention was paid to the sampling error discussed in detail in this paper.In practice, this involved using the estimated sedimentation rates to correct for the sampling error inherent in each result.As shown, not all of the dates in Table 2 were eligible for this analysis.It could only be applied to those from sites where reliable and uninterrupted sedimentary rate information was available.In all, thirteen dates met these criteria.Using OxCal, the ten eligible dates before the eruption were shifted to younger ages and three dates after the eruption shifted to older ages.The shift was applied using a simple Normal distribution centred on the mean sedimentary rate, with a 2-sigma range that extended from the minimum to the maximum absolute sampling error.To clarify, a sample centred 2 cm below the tephra at a site with a sedimentation rate of ranging from 5 (min.) to 10 (max.) cm/century required a shift of 20 (min.) to 40 (max.)years.In such a case, the calibrated date was shifted by a factor of 30 ± 5 years (1sigma).Finally, all the adjusted dates were averaged by reusing them as prior probabilities in a bounded Phase containing a Sum function (see Dee et al., 2014).For this last step, each adjusted date was also given a 5% prior probability of being an outlier, to ensure no extreme results biased the final average.The configuration adopted for Model 2 is a simplification, given the notorious irregularity of calibrated 14 C dates.However, as the shifts were only of the order of decades and very minor in comparison to the breadth of the unmodelled calibrations themselves, it was considered acceptable for exploratory purposes.A comparative model (S2) is given in the appendix which, instead of outlier analysis for the final step, employed the removal of the dates with the lowest Agreement Indices until the overall model reached 60% Agreement.The OxCal codes for both Model 2 and S2 are given in appendix C.

Outputs of bayesian models
Fig. 6 shows the position of the Avellino tephra in relation to all the individual dates included in Model 1.The model identified three dates as being extreme outliers: one in the pre-eruption group (GrM-17418, Mezzaluna) and two in the post-eruption group (GrA-46203, 46,205, Migliara 44.5;GrM-17888, Mesa 700).These were essentially eliminated from the analysis.The model also downweighed the contribution of four other dates (GrA-46210, Campo;GrN-32454, Campo;GrM-18970, Femmina Morta, GrM-16626, Femmina Morta) as they were only deemed partially in agreement with the sequence as a whole.The date for the Avellino eruption from Model 1 is somewhat bimodal in nature and extends from 1934 to 1841 cal BC at 95% probability.The alternative model S1 generated a congruent date of 1910-1810 cal BC at 95% probability, producing a broader range than using outlier analysis.Fig. 7 shows the dates used for Model 2 before (7a) and after (7 b) their shifts to account for sampling error.As is apparent, the adjustments had a very limited effect on the absolute position of the dates.Nonetheless, the adjusted dates should in theory overlie the eruption date, and thus can be averaged to estimate the date of the eruption (Fig. 8).This smaller data set was internally very consistent, and only sample (GrM-17418, Mezzaluna) was regarded as an extreme outlier.The modelled date for the Avellino eruption obtained by Model 2 was 1909-1868 cal BC at 95% probability.The alternative model S2 generated an almost identical date of 1907-1869 cal BC at 95% probability.
For comparative purposes, the modelled dates for the eruption from Model 1 and from Model 2 are shown alongside each other in Table 6 and in Fig. 9.The overwhelming picture is that the results obtained from the two different methods are highly compatible with each other and, notwithstanding all of the preceding discussions about the challenges involved in dating this tephra layer, they represent a credible sciencebased date for the eruption.

General discussion and conclusions
Our study set out to identify factors that might complicate the 14 C dating of materials from above the AV layer, paying particular attention to the causes for the wide range in dates obtained.Apart from wellknown factors such as the old wood/inbuilt age effect and bioturbation, we looked closer into the sampling error connected with dating plant macro remains from layers that formed in environments with low accumulation rates.The resulting sampling error can be assessed by combining the accumulation rate and thickness of the layer sampled.Particularly in environments with low sedimentation rates (6-8 cm/century or lower) and with sediment holding a low percentage of suitable plant macro remains (>2 cm thick layers sampled), the uncertainty of 14 C dates for such materials rapidly increases (see also Appendix B).This rendered these dates less suitable for reliable and precise assessment of the absolute age of the intercalated tephra layer.
The chemical analyses and pollen concentration curves together strongly suggest that the AV tephra held toxic concentrations of F and that its deposition in stagnant anoxic shallow aquatic to marshy environments induced a serious reduction of the sediment accumulation rate, most probably as a result of a reduced biomass/necromass production by the vegetation.Such a toxic impact was rather surprising and not reported earlier for distal areas, even though comparable toxic impacts were well known for proximal Vesuvian areas.Remarkably, it has also not been mentioned as a relevant phenomenon in studies of sites with the AV tephra from such proximal areas and may well have been overlooked.Far more optimal for constraining the age of the AV layer are samples from oxic aquatic to marshy environments in which peat to peaty clays accumulated, where this presumed toxic impact played an at most subordinate role.In these settings, accumulation rates were higher, as were contents of suitable plant macro remains, and thus sampling errors are lower, increasing the statistical reliability of the ages Interestingly, samples that in the Bayesian model for the whole dataset (model 1) were discriminated as being 'extreme outliers' or 'partially in agreement with the sequence as a whole' are identical to the samples that based on the factors discussed above (bioturbation, inbuilt age, and sampling error) were considered as suboptimal (see Table 2).This lends further weight to the outcome of model 2, which is based on the Bayesian analysis of the more limited set of samples and includes corrections for the sampling error.This set dominantly concerns samples from type 1 sediment with thicknesses of layers sampled of 2 cm or less, and thus with a small sampling error.
The current results provide a clear explanation for the discrepancy between the earlier published date for the AV layer -2010-1958 cal BC, 1 σ (Sevink et al., 2011) -and the date established by model 2-1909-1868 cal BC, 1 σ.This earlier dating was based on results that now can be deemed biased by a large sampling error and/or inbuilt age, in combination with a small set of dates, which by statistical coincidence resulted in an age of the AV layer that was interpreted as robust.That

Table 6
The 68% and 95% probability ranges for the Avellino eruption obtained from Models 1 and 2. this was indeed statistical coincidence is clear from the analysis of the much larger data set that is discussed in this paper.
Our results stress the need to base truly robust dating of distal tephra layers, intercalated in lowland sedimentary sequences, on a critical evaluation of large sets of dates, accounting for potential effects of bioturbation, inbuilt age, and sampling error.By doing so, we now can solve the controversy over the age of the AV layer that resulted from our earlier dating (2011).Our current estimated age is remarkably congruent with the age obtained by Alessandri (2019Alessandri ( : 1929Alessandri ( -1858 cal BC, 1 σ) and earlier by e.g.Passariello (2009Passariello ( : 1935Passariello ( -1880 cal BC, 1 σ) and Jung (2017Jung ( : 1908 ± 12 ± 12 cal BC, with a probability of p = 86%).
More general implications of our research are that ages of distal tephra layers from lowland sedimentary archives in the Mediterranean, if based on a limited set of 14 C dates and not supported by a thorough analysis of the potential sampling error, are deemed to remain unreliable.Such analysis should focus on changes in sediment accumulation rates following upon the deposition of tephra, which can be assessed by scrutiny of pollen density profiles across sections containing tephra layers.Our results also suggest that particularly in archives that are more proximal and contain more massive tephra layers, toxic impacts may play an important role and may have led to significant postdepositional hiatuses hampering reliable and precise 14 C based assessments of the age of these tephra.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.J. Sevink et al.

Fig. 1 .
Fig. 1.Location of the Agro Pontino and Fondi basin, and of the sites studied.Numbers refer to sites described in Table 1.Numbers of locations with palaeoecological data are in bold.A = higher complex of Pleistocene marine terraces; B = Agro Pontino graben; C = Monti Lepini; D = Monti Ausoni; E = Monte Circeo.

Fig. 3 .
Fig. 3. Simplified geological map of the area studied, showing the distribution of the Holocene Terracina deposits and beach ridge, and the Holocene Young colluvial deposits.> 1 m = thicker than 1 m.

Fig. 6 .
Fig. 6.The outputs from Model 1.The modelled posterior distributions (dark grey) overlie the prior probabilities (light grey).The transitional Boundary marking the Avellino Eruption is shown in red.In each case, the 68% and 95% probability ranges are indicated by the square brackets beneath the modelled estimates.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 7 .
Fig. 7.The calibrated dates utilized for Model 2 before (a) and after (b) the adjustments for sampling error.

Fig. 8 .
Fig. 8.The averaged date for the Avellino Eruption from Model 2 based on all of the dates adjusted for sampling error.

Fig. 9 .
Fig. 9. Comparison of the modelled dates for the Avellino Eruption obtained by Model 1 and Model 2.
Sevink et al. (2020b)report an age of 4.23 cal kyr BP for the Astroni 6 eruption.The latter age is earlier than the age reported bySevink et al. (2020b), which is in between 2168 ± 118-2117 ± 84 cal yr BC (4118 ± 118-4067 ± 84 cal yr BP), and is more in line with the age reported bySmith et al.

Table 2
Data on selected samples from below (upper series) and above the AV-layer (lower series).EBA/MBA = Early/Middle Bronze Age.Inbuilt age: x = unsuitable for further analysis because of inbuilt age; ?= possible error because of inbuilt age.Biot.=Bioturbation;x = unsuitable for further analysis because of bioturbation.Hiatus = unsuitable because of hiatus.Strat.Pos.: ?= uncertain origin of material.For facies and sediment type: see Table1.
14 C age of the accumulated sediment.Such an effect may have occurred in the anoxic pyritic environment at Migliara 44,5 and explain the aberrant high age of the black organic clay samples from above the AV layer at that site.

Table 4
Overview of data on accumulation rates and sampling for published cores from the Tyrrhenian coastal area of central Italy.For Facies, see Table1.

Table 5
Chemical composition of tephra and adjacent strata.For sediment type symbols, see Table1.AV-tephra layer is indicated in grey.