journal of archaeological science - cornell university · 2014. 1. 28. · 1516 ( masters, 2009, p....

lable at ScienceDirect

Journal of Archaeological Science 41 (2014) 772e783

Contents lists avai

Journal of Archaeological Science

journal homepage: http: / /www.elsevier .com/locate/ jas

High-precision dating the Akko 1 shipwreck, Israel: wiggle-matchingthe life and death of a ship into the historical record

Brita Lorentzen a,b,*, Sturt W. Manning b, Deborah Cvikel c, Yaacov Kahanov c

aDepartment of Earth and Atmospheric Sciences, 2120 Snee Hall, Cornell University, Ithaca, NY 14853, USAbMalcolm and Carolyn Wiener Laboratory for Aegean and Near Eastern Dendrochronology, B48 Goldwin Smith Hall, Cornell University, Ithaca, NY 14853-3201, USAc Leon Recanati Institute for Maritime Studies, University of Haifa, Haifa 31905, Israel

a r t i c l e i n f o

Article history:Received 25 January 2013Received in revised form19 September 2013Accepted 13 October 2013

Keywords:Akko 1 shipwreckWiggle-matchingRadiocarbonAkkoMuhammad Ali PashaEgyptianeOttoman WarsHistorical archeology

* Corresponding author. Malcolm and Carolyn WienNear Eastern Dendrochronology, B48 Goldwin Smith HNY 14853-3201, USA.

E-mail address: [email protected] (B. Lorentzen).

0305-4403/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.jas.2013.10.013

a b s t r a c t

The Akko 1 shipwreck is the remains of an eastern Mediterranean naval or auxiliary brig, which wasfound inside the ancient harbor of Akko, Israel. The shipwreck and finds were recorded underwater, andsome of the ship components, along with the majority of the finds, were retrieved and analyzed. ABayesian dating model, incorporating 14C wiggle-matching of the ship timbers, tree-ring analysis, and 14Cdates from short-lived finds, is used to model the ship’s construction and wrecking dates. These newdata, combined with the results of archaeological research and available historical records, suggest thatthe ship was built during the first third of the 19th century as part of Muhammad Ali’s fleet. Akko 1 thenpossibly plied the eastern Mediterranean under the Egyptian flag during the First EgyptianeOttomanWar in 1831e1833. The wrecking event apparently occurred during the 1840 naval bombardment ofAkko. This is the first time that 14C wiggle-matching and Bayesian analyses have been used to date theconstruction and wrecking of a shipwreck in the southeastern Mediterranean. The results show thatBayesian analysis and 14C wiggle-matching techniques are valuable tools for analyzing the region’sshipwrecks, including those from recent historical periods.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

1.1. Historical setting

The historic walled port city of Akko (Acre, St. Jean d’Acre, Akka)is located at the northern extremity of Haifa Bay in northern Israel.It has a continuous settlement history from the Early Bronze Age tothe modern era, serving as an important naval and trading port(Dothan, 1993; Makhouly and Johns, 1941; Negev and Gibson, 2005,pp. 27e28).

The town and its harbor were conquered by the Ottomans in1516 (Masters, 2009, p. 9), and were the scene of several navalcampaigns during the Late Ottoman Period. In 1799, British controlof the sea and the Akko harbor prevented Napoleon Bonaparte fromtaking the town, and stopped his advance northwards (Alderson,1843, p. 28; Anderson, 1952, pp. 372e373; La Jonquière, 1900, IV).Muhammad Ali’s Egyptian flotilla was severely damaged during a

er Laboratory for Aegean andall, Cornell University, Ithaca,

All rights reserved.

heavy bombardment of Akko in December 1831, and the Egyptianstook the town by land on 27 May 1832, after a 6-month siege. Theperiod of Egyptian rule over Akko lasted until 3 November 1840,when a British-Austrian-Ottoman fleet bombarded the town. Dur-ing this event, the main powder magazine of Akko exploded,causing enormous damage to the town, which was taken thefollowing day (Alderson, 1843, pp. 39e48; Anderson, 1952, pp.561e564; Kutluo�glu, 1998, pp. 62e73, 173; Rustum, 1926). Analysisof the hull timbers and finds from the Akko 1 shipwreck indicatedthat the ship was built, sailed, and sank during this same dynamictime period (Cvikel and Kahanov, 2009, 2013).

1.2. The Akko 1 shipwreck

The Akko 1 shipwreck was found in 4 m of water and under athin layer of sand inside the ancient Akko harbor (see Fig. 1). Itwas excavated over three seasons in 2006e2008 by the LeonRecanati Institute for Maritime Studies at the University of Haifa.The shipwreck remains, lying northwest to southeast, were 23 mlong from the bow to aft extremity, and a maximum 4.66 m widefrom the line of the false keel to the uppermost remains of the portside. It is suggested that the original ship was a small Egyptianarmed vessel, or auxiliary brig, about 26 m long, built in the eastern

Delta:1_given nameDelta:1_surnameDelta:1_given nameDelta:1_surnamemailto:[email protected]://crossmark.crossref.org/dialog/?doi=10.1016/j.jas.2013.10.013&domain=pdfwww.sciencedirect.com/science/journal/03054403http://www.elsevier.com/locate/jashttp://dx.doi.org/10.1016/j.jas.2013.10.013http://dx.doi.org/10.1016/j.jas.2013.10.013http://dx.doi.org/10.1016/j.jas.2013.10.013

Fig. 1. Location of the Akko 1 shipwreck and Akko in Haifa Bay (drawing: S. Haad and N. Yoselevich).

B. Lorentzen et al. / Journal of Archaeological Science 41 (2014) 772e783 773

Mediterranean. A complete description of the ship components andfinds is given by Cvikel and Kahanov elsewhere (2009, 2013). Ourwork here is part of an ongoing series of studies about the Akko 1shipwreck (e.g., Ashkenazi et al., 2011; Cvikel and Kahanov, 2009,2013; Kahanov et al., 2012; Mentovich et al., 2010).

Given the large amount of political upheaval and frequent navalcampaigns that occurred in Akko during this relatively short time,high-precision absolute dates are required to place Akko 1’s con-struction and wrecking within their appropriate historical context.Dendrochronology (tree-ring dating) can provide extremely precisedates or a terminus post quem for when ship timbers and woodenelements were cut. However, it is not always possible to obtain adendrochronological date from sampled wood. Even whendendrochronological dates are available, it may be necessary torelate tree-ring data to dates from other materials from the ship-wreck site, in order to date events post-dating the ship’s con-struction. Typological dates derived from the ship’s hull and smallfinds do not provide the level of dating precision required, nor do

single AMS 14C dates, because the history of past radiocarbon levelsin the atmosphere is such that for radiocarbon ages in the last 300years, there will always be several wide ranges of possible cali-brated calendar ages if the 14C data are considered in isolation. Thisproblem has been encountered in dating several other historicalshipwrecks (Cvikel et al., 2008; Cvikel and Kahanov, 2006) andbuildings (Biger and Liphschitz, 1991) in the region.

1.3. Bayesian modeling and 14C wiggle-matching

In order to obtain useful, high-precision 14C dates for historicalsites, buildings, or shipwrecks, 14C data can be analyzed togetherwithin a Bayesian chronological model (Bayliss, 2007; BronkRamsey, 2009a). A more detailed description of Bayesian princi-ples and its applications in dating archaeological sites for non-specialists is given by Bayliss (2007), so only a short summary ofthese methods is given here. Within the Bayesian framework, asite’s scientific dates, such as 14C and dendrochronological dates

Fig.

2.Plan

oftheAkk

o1hu

llremains

(drawing:

C.Bran

don);thelocation

sof

thetimbe

rssampled

forwiggle-match

ing(A

KO-1,A

KO-11,

andAKO-12)

arelabe

led.

B. Lorentzen et al. / Journal of Archaeological Science 41 (2014) 772e783774

(known as the ‘standardized likelihoods’), are analyzed in thecontext of one’s additional knowledge about the dates and site(‘priors’) (Bayliss, 2007). For a shipwreck, this additional data mayinclude information relating ship materials and finds to relativechronological sequences (e.g., whether materials belong to theship’s construction or its later period of use), typical use periods ofmaterials onboard, or documentary evidence about the ship orshipbuilding practices during a particular time period. The Bayesianmodel combines the raw scientific dates with the other prior in-formation in a quantitative manner to produce formal statisticaldate estimates (‘posterior density estimates’) for a given event, suchas a ship’s construction or wrecking date (Bayliss, 2007; BronkRamsey, 2009a).

Radiocarbon wiggle-matching is a specific application ofBayesian analysis, which combines tree-ring analysis and radio-carbon data (Bayliss, 2007; Bayliss and Tyers, 2004; Bronk Ramseyet al., 2001; Galimberti et al., 2004). 14C dates are obtained frommultiple tree-ring samples, whose relative order is known by anexact number of years. This sequence of 14C dates can then be fittedto the fluctuations of past atmospheric radiocarbon levels recordedin the radiocarbon calibration curve, in order to provide high-precision dating estimates for a given year in the tree-ringsequence. Wiggle-matched data can in turn be incorporated withother dates and prior information into a larger Bayesian model,which can increase the precision of a model’s dating estimates,including for relatively recent historical events (e.g., Turetsky et al.,2004).

In this study, we present a comprehensive Bayesian analyticalmodel for Akko 1 to determine high-precision absolute dates forthe ship’s construction and last voyage/wrecking event. We use 14Cwiggle-match dating of tree-ring sequences from the ship’s timbersand additional 14C dates from the ship’s wooden elements to pro-vide a terminus post quem for the ship’s construction, and a series of14C dates on short/shorter-lived samples found on the shipwreck, aswell as a terminus post quem dating estimate from an earlier anal-ysis of the ship’s cannonballs (Mentovich et al., 2010), in order todetermine themost likely date of the ship’s last voyage or wrecking.

Our study is, to our knowledge, the first time that Bayesiananalysis and wiggle-matching have been used to date the con-struction and wrecking of a shipwreck in the southeastern Medi-terranean. Our results show that these techniques are extremelyvaluable tools for the analysis of shipwrecks, particularly from thehistorical period, when there can be rich, detailed historical docu-mentation available, but the traditional dating of undocumentedphysical remains is not always precise enough to link the archae-ological and textual records. Bayesian chronological models, likethe onewe employ for Akko 1, can narrow possible shipwreck datesto a range that allows the physical ship remains to tie into, andgreatly enhance, the period’s historical documents, and may e andshould e be used in the future to date and interpret other ship-wrecks that have been found in the region.

2. Material and methods

2.1. Tree-ring analysis and radiocarbon wiggle-matching of the shiptimbers

The Akko 1 ship timbers were first analyzed for potentialdendrochronological dating. Cross-sections were cut from twelveof the ship timbers, which appeared upon initial examination to besuitable for dendrochronological analysis. The surface of eachsample was prepared with a razor blade, and the tree-ring widthsmeasured along two to three radii for each sample using the Tell-ervo (Brewer et al., 2010) dendrochronological analysis package.Sample metadata, including presence of injuries and species, were


also recorded. Since many of the sampled timbers are oak (Quercussp.), samples were also examined to determine if any sapwood (theanatomically distinct area of the tree containing the outermosttree-rings) was present, since accurate estimation of the tree’sfelling date is often possible when sapwood is preserved (Hillamet al., 1987; Hughes et al., 1981; Wazny, 1990).

Since the samples do not cross-match against one another oragainst any of the available Aegean or European reference tree-ringchronologies, three samples e all oak (Quercus sp.) (labeled AKO-1,AKO-11, and AKO-12) ewere selected for 14C wiggle-matching (seeFig. 2). Twelve 10-year segments of tree-rings (three each fromAKO-1 and AKO-11, and six from AKO-12), whose relative positionto one another on each sample was known from exact ring counts,were dissected and sent to the Heidelberg and Oxford RadiocarbonLaboratories for analysis. Ten-year increments were selected forwiggle-matching, because most of the samples include periods ofextremely narrow ring growth with a limited amount of material tosample. Thus taking decadal sections allowed sufficient material tobe sampled for both routine high-precision 14C dating at Heidelbergon two larger timbers, and AMS 14C dating at Oxford on the smallersample, while gaining adequate dating precision against theradiocarbon calibration curve.

The sequences of AMS (for AKO-12) and conventional (for AKO-1and 11) 14C dates for each sample were then wiggle-matched usingOxCal 4.2 (Bronk Ramsey, 1995, 2009a; Bronk Ramsey et al., 2001)and compared against the IntCal09 radiocarbon calibration curve(Reimer et al., 2009) with curve resolution set at 1 year and noallowance made for possible regional 14C variation (DR). Eachradiocarbon date was treated as dating the center-point of thedated rings (e.g., the date for rings 1001e1010 was treated as ring1005.5). Since the Akko 1 tree-ring sequences do not dendrochro-nologically crossdate one another, they were treated as indepen-dent series in a single model, in which there was no priorassumption about which sample was older/younger.

None of the samples contain sapwood. The average number ofsapwood rings in oak species varies depending on the tree’sgrowing location (e.g., northern Europe versus Anatolia) and age(Griggs et al., 2009; Hillam et al., 1987; Hughes et al., 1981; Wazny,1990). Thus, a minimum number of additional rings was added tothe wiggle-matched date of each sample’s last extant ring, in orderto determine aminimum felling date for the trees. This numberwasbased on the average number of sapwood rings for oaks from theship timbers’ likely area(s) of origin. This simpler method was usedinstead of Millard’s (2002) proposed model for estimatingsapwood, which has been developed for British oaks in the researchof Miles (2005, 2006) and is implemented in OxCal 4 (BronkRamsey, 2009a). This is because all of the necessary metadata forcalculating such sapwoodmodels are not currently available for oakpopulations from most of the ship timber’s likely source areas.Additionally, the Akko 1 timbers sampled here may be one ormultiple oak species (see below). The appropriatewood anatomicaldata for calculating sapwood models for each of the possible oakspecies in the eastern Mediterranean are likewise unavailable, andthis is a topic requiring further research before more sophisticatedsapwood models can be developed.

The Akko 1 ship finds (Cvikel and Kahanov, 2009, 2013); themodern distributions of the tree species from which the ship tim-bers were cut (Akkemik and Yaman, 2012, pp. 166e170; Ducoussoand Bordacs, 2003; Schweingruber, 1993, pp. 188e201); and his-torical evidence (McNeill, 1992; Mikhail, 2011) indicate that theship timbers likely came from Anatolia/the Aegean, and possiblyItaly. We therefore first employed a sapwood estimate based ondata from a group of mixed north Aegean oak species (Griggs et al.,2007, 2009, pers. comm., 2012), and then a second alternative es-timate from a group of oaks in northwest Turkey with higher

sapwood counts that were reported as either Quercus robur L. orQuercus petraea (Mattuschka) Liebl., but which may in fact be asimilar Turkish endemic species (Quercus vulcanica Boiss. ex Kot-schy). We then made a third sapwood estimate based on sapwoodcounts for Italian oaks reported in Haneca et al. (2009, Table 1).

It is assumed that more or less all of the outer heartwood (theresin-filled inner part of the tree preceding the sapwood) is presentin the samples, and that it is largely the sapwood rings (which areless resistant to decay) that have been removed from the outer partof the timbers. We then added a minimum estimate of 2 � 1 yearsin which the oak was seasonedewhich we based on 19th centurystatements that two years of seasoning is a minimum for oak(although it should be noted that in some cases, seasoning couldtake up to about a decade) (Tredgold, 1875).

The estimated minimum felling dates for the wiggle-matchedsamples were then incorporated into a dating model with fouradditional AMS 14C dates from wooden elements analyzed by theRadiocarbon Dating Laboratory in the Institute of Particle Physics inZurich, in order to calculate a terminus post quem date for the ship’sconstruction (the minimum construction date, or ‘MCD’ in theOxCal model). The fifth dated timber fragment (ETH-32620) fromthe ship has a ‘modern’ (post-AD 1950) radiocarbon ageof�80� 40 BP (see Table 1). Whether this date reflects a laboratoryerror or contamination is not known. This sample is the onlyextreme outlier in the overall set (see Table 1) and was excludedfrom analysis.

The non-wiggle-matched timber fragments were modeled us-ing Bronk Ramsey’s (2009b) charcoal outlier model in OxCal, whichmodels the 14C dataset as an exponential distribution. This modelwas used, since many of the wooden fragments were taken fromthe outer edges of the timbers, which include younger (morerecent) tree-rings. Therefore the 14C dates on the wooden fragmentdataset will likely have an exponential distribution toward thetimber felling date. However, the charcoal outlier model also allowsfor the fact that some of the timber fragments may include older(less recent) tree-rings toward the center of the tree, whose datesare substantially older than the general felling period for the ship’stimbers.

The calculated MCD for Akko 1 was constrained to a very con-servative uniform date range (in which there are no prior datingpreferences within the specified interval) between AD 1600 and1950 (which marks the end of the “pre-bomb” 14C period, afterwhich nuclear bomb testing altered atmospheric 14C levels and theshape of the radiocarbon calibration curve) (Randerson et al.,2002). After an initial run of the models using the three differentsapwood estimates, a second iteration of the model using the northAegean sapwood estimate was run, which excluded two dates(ETH-32621 and ETH-32623) identified as outliers.

2.2. Dating the ship’s final voyage/wrecking

The date of Akko 1’s last voyagewas estimated using a set of fiveAMS 14C dates, measured by the Radiocarbon Dating Laboratory inthe Institute of Particle Physics in Zurich, on short or shorter-livedsamples found in the shipwreck (see Table 1). It is likely that thedates on the short or shorter-lived samples are distributed towardthe last few years or voyages of the ship (e.g., the two rope samples,and the leather flask), or e in the case of the dated food remains elikely the literal last voyage or season of sailing, and give a closeterminus post quem for the ship’s wrecking date. The 14C dates onthe short/shorter-lived samples are therefore grouped in the samePhase in OxCal (see Fig. 4) and modeled as an exponential distri-bution (‘Tau_Boundary’ in OxCal) (Bronk Ramsey, 2009a)toward the end of this Phase.

Table 114C data and calibrated ages of wood and organic samples from the Akko 1 shipwreck. Sample B2012 (marked in bold and italicized) was excluded from the dating model.

Serial number Lab number Samplenumber

Sampletype

d13C & 14C Age (BP) 68.2% Probability 95.4% Probability

1 ETH-31975 e Wood �21.6 � 1.1 275 � 40 1520e1590 (37.9%)1620e1670 (30.3%)

1480e1680 (89.7%)1770e1800 (5.7%)

2 ETH-32619 A2012 Wood �24.5 � 1.1 250 � 40 1520e1580 (18.8%)1620e1680 (29.1%)1760e1800 (15.3%)1930e1960 (5.1%)

1480e1690 (64.2%)1720e1810 (23.8%)1920e1960 (7.4%)

3 ETH-32620 B2012 Wood �27.2 � 1.1 �80 � 40 1950e1955 (68.2%) 1950e. (95.4%)4 ETH-32621 702 Olive pit �29.5 � 1.1 5 � 40 1700e1720 (6.6%)

1820e1840 (3.5%)1880e1920 (25.8%)1950e1960 (32.3%)

1690e1730 (13.3%)1810e1920 (47.1%)1950e1960 (35.0%)

5 ETH-32622 700 Rope �30.9 � 1.1 95 � 40 1690e1730 (17.9%)1810e1920 (50.3%)

1680e1780 (30.1%)1800e1940 (65.3%)

6 ETH-32623 777 Leather flask �26.7 � 1.1 15 � 40 1700e1720 (8.5%)1810e1840 (7.3%)1880e1920 (30.8%)1950e1960 (21.6%)

1690e1730 (16.3%)1810e1920 (56.0%)1950e1960 (23.1%)

7 ETH-34103 2034 Wooden nail �26.0 � 1.1 190 � 50 1650e1690 (16.7%)1720e1810 (38.8%)1920e1960 (12.7%)

1640e1960 (95.4%)

8 ETH-34105 2035 Wood �29.3 � 1.1 115 � 45 1680e1740 (20.7%)1800e1930 (47.5%)

1670e1780 (35.8%)1790e1950 (59.6%)

9 ETH-35813 2008/1 Rope �23.6 � 1.1 80 � 45 1690e1730 (18.4%)1810e1920 (49.8%)

1680e1770 (29.2%)1800e1940 (66.2%)

10 ETH-35814 2008/2 Nut shell �24.9 � 1.1 135 � 45 1680e1780 (28.1%)1800e1890 (29.6%)1910e1940 (10.5%)

1660e1950 (95.4%)


It is assumed that the estimated cutting dates for samples AKO-1, 11, and 12, and the dates for the non-wiggle-matched ship timberfragments (e.g., materials relating to the ship’s construction) pre-cede the dates for the shorter-lived samples onboard the ship.Therefore the initial boundary of the ‘Last Voyage Phase’ wasmodeled as the ship MCD, while this Phase’s end boundary isconsidered the best estimate for Akko 1’s last voyage or wreckingdate (‘LV’ in the OxCal model). We added an additional timeconstraint, allowing a distribution of up to 20 years for the dates inthe ‘Last Voyage Phase’ (i.e., a time constant of 0e20 years to theexponential distribution). This provides a generous allowance forthe lifetime of the short/shorter-lived sample objects, and indeed20 years is estimated as the typical lifespan reported for shipssimilar to Akko 1 (Cvikel and Kahanov, 2013; Dodds and Moore,2005, p. 17; Elkin et al., 2007, p. 37). A second iteration of thismodel was also run, which excluded the two outliers (i.e., ETH-32621 and ETH-32623).

An additional model was run, in which a terminus post quem of1839 for the date of the last voyage was added. This constraintincorporates dating information from Mentovich et al.’s (2010)analysis of the Akko 1 cannonballs, in which they found highconcentrations of manganese added to the cast iron. This is amanufacturing technique used in cannonballs produced only after1839, which suggests a post-1839 date for the ship’s last voyage.

2.3. Testing model robustness and possible regional 14C offsets

Additional tests were conducted to determine if any regionaloffsets between the calibration curve and sample radiocarboncontent might affect the model’s calculated dates. Available data(Manning and Kromer, 2012; Manning et al., 2010) indicate thatsuch an offset is extremely small to negligible for the ship timbers,if they are from the Aegean, Turkey, or Italy. Archaeological evi-dence (Cvikel and Kahanov, 2013) suggests that Akko 1 was a shipfriendly to Egypt and was possibly even built there. If, then, theshort/shorter-lived samples are from Egypt, there may be a small

but significant growing season-related offset of 19� 5 14C years likethat which Dee et al. (2010) found in samples from the Nile Valleyin periods prior to the Aswan dam construction. However, it shouldbe noted that sample material from Egypt’s Mediterranean coastwould have a smaller offset, while the offset in material from theLevant would be smaller still. To consider the relevance of such anoffset to our dating models, we compared the modeled minimumconstruction dates against the calibration curve in OxCal first with ageneral DR allowance of 0 � 10. Then, in a second test, we added a19 � 5 14C years adjustment only to the modeled date for the short/shorter-lived samples to determine the effect on the modeled datefor the ship’s final voyage/wrecking.

3. Results

3.1. Tree-ring analysis and wiggle-matching the ship timbers

Of the twelve ship timbers sampled, only six e all deciduousoaks e had sufficient rings (�50) for dendrochronological analysis.Reliably identifying European and Mediterranean oak wood at thespecies level is not possible; thus we give wood identifications onlyat section level, following the recommendations of Akkemik andYaman (2012, p. 178), Schweingruber (1990, pp. 401e403), andWazny (pers. comm., 2012). Five samples were cut from oaksbelonging to the section Quercus (white oaks), which in Europe andthe Mediterranean includes the common species Q. petraea (Mat-tuschka) Liebl., Q. pubescens Willd., Q. vulcanica Boiss. ex Kotschy,and Q. robur L. One sample was cut from an oak belonging to thesection Cerris (red oaks), which in Europe and the Mediterraneanincludes the common species Quercus cerris L. All of the sampleshave relatively short tree-ring sequences of less than 100 rings, andnine of the twelve include the innermost, juvenile tree-rings at ornear the tree’s pith. The bark, vascular cambium, and sapwood areabsent from all of the timbers sampled. Most of the samples havescarring and abrupt growth suppression and releases in their tree-rings (see Fig. 3), which altered tree-ring growth. None of the

Fig. 3. Photos of sample AKO-1, and graph showing the AKO-1 ring-width measurement sequence (relative years beginning at 1001 from the center of the tree outwards areshown), which features abrupt growth releases due to an injury (A), and abrupt growth declines in both the section’s middle (B) and center (C). The lighter color around the sample’souter edges gives the appearance of sapwood, but closer inspection revealed it to be only discoloration (photos: B. Lorentzen).

Table 214C data from the AKO-1, AKO-11, and AKO-12 oak samples employed in the 14Cwiggle-match.

Lab ID Relative rings(each sample)

d13C & 14C age (BP) SD

AKO-1Hd-27447 1008e1017 �24.52 117 22Hd-27465 1028e1037 �26.6 150 19Hd-27487 1048e1057 �25.92 142 20

AKO-11Hd-30362 1006e1015 �26.2 140 12Hd-30391 1036e1045 �26.3 185 26Hd-30392 1046e1055 �25.7 152 19

AKO-12OxA-19432 1001e1010 �25.0 119 25OxA-19466 1011e1020 �24.0 128 21OxA-19433 1031e1040 �24.5 202 25OxA-19467 1041e1050 �23.7 166 22OxA-19434 1061e1070 �23.9 191 26OxA-19468 1071e1080 �24.2 191 22


samples yielded significant dendrochronological crossdates withone another or with existing European and Mediterranean oakreference chronologies.

The three samples selected for 14C wiggle-matching, AKO-1 and12, and AKO-11, were sampled from hull timbers: two frame tim-bers and a hull plank, respectively (see Table 2). AKO-1 has 74 rings(plus one unmeasured, partial ring); AKO-11 has 65 rings (plus oneunmeasured, partial ring); and AKO-12 has 93 rings. The 10-yearsections taken from each sample and used in the wiggle-matchare given in Table 2, and their calculated placements from themodel (Fig. 4) are shown in terms of their fit on the IntCal09radiocarbon calibration curve (Reimer et al., 2009) in Fig. 5. All the14C data from the wiggle-matched wood samples lie in the 18thcentury AD on the upward 1715e1790 slope. When the OxCalcharcoal outlier model is used, the non-wiggle-matched woodsamples lie in the late 18th through early 19th century AD. The datafrom the subsequent short/shorter-lived samples fit best aroundthe wiggle in the calibration curve between approximately 1820and 1840.

All three samples were cut from Quercus section Quercus. Spe-cies in this section have a wide distribution stretching from theBlack Sea and southwest coasts of Anatolia westward throughoutmost of Europe (Akkemik and Yaman, 2012, pp. 166e170; Ducoussoand Bordacs, 2003; Schweingruber, 1993, pp. 188e201). For reasonsdiscussed above (also: see discussion below), the ship timbers weremost likely culled from forests in Anatolia/the Aegean, and also

possibly northern Italy, so average sapwood estimates from theseregions are employed. According to Griggs’ research (pers. comm.,2012; modified from data in Griggs et al., 2007, 2009) on a multi-species group of 167 oaks growing in northern Turkey andGreece, oaks ages < 100 years generally have an average of

Fig. 4. The Akko 1 dating model, comprising all 14C data on the ship’s wood elements/samples; three independent 14C wiggle-matches for timbers AKO-1, AKO-11, and AKO-12; theestimated minimum construction adding the estimated missing number of sapwood rings and minimum of 2 � 1 years for seasoning; a group of other 14C dates on wood samplesfrom the ship; and 14C dates on short/shorter-lived samples from the shipwreck. The A values are the OxCal agreement values for the individual samples; a satisfactory value is �60.The O (Outlier) values are shown only where the posterior values from the General Outlier Model (Bronk Ramsey, 2009b) are greater than the prior value (of 5); these outliers areindicated by an arrow and bold text in the figure. Note that the modern sample ETH-32620 has already been excluded.


19 � 4.33 sapwood rings, while oaks ages 101e150 years have anaverage of 21.19 � 4.61 sapwood rings. If we use Griggs’ (pers.comm., 2012) sapwood estimates from the smaller dataset of oaks(possibly Q. vulcanica Boiss. ex Kotschy) from northwest Turkey,then the average for oaks < 100 years old is 27.2 � 7.19 sapwoodrings, and 27.81 �7.78 for oaks 101e150 years old. According to thedata from Martinelli reported in Haneca et al. (2009, Table 1),Italian oaks have an average of 13.23 � 6.06 sapwood rings.

Sample AKO-1 was cut from a tree that was at least 75 years old,and the pith and innermost rings of the tree are present. SampleAKO-11 was cut from a tree at least 66 years old; the pith is notpresent, although based on the sample’s ring curvature, its

innermost rings are likely from very close to the tree’s center(within approximately 10 rings). Sample AKO-12 was cut from atree at least 93 years old, although the pith is not present, nor isthere clear indication that the sampled timber was cut close to thecenter of the tree (thereby placing AKO-12 in the older sapwoodestimate age class). Thus the sapwood estimate for trees in the

Fig. 5. The placement of the 14C data in the Akko 1 dating model in Fig. 4 against the IntCal09 (Reimer et al., 2009) radiocarbon calibration curve (�1s). The 14C data from thedifferent samples are shown as the 68.2% probability modeled calendar age ranges from Fig. 4 on the x axis (calendar date scale), and as the 14C age BP (�1s) on the y axis (14C years).


across the different sapwood models between samples AKO-1 and11 versus AKO-12, which may reflect very small differences be-tween the Heidelberg (which measured AKO-1 and 11) and Oxford(which measured AKO-12) radiocarbon laboratories, or a slightdifference in the actual timber ages. However, the date ranges for allthree samples comfortably overlap one another at both 68.2% and95.4% probability. This indicates that the timbers were cut atroughly the same time; our assumption that the timbers had onlytheir sapwood rings removed (plus possibly a few heartwood rings)is likely correct; and the calculated minimum timber felling dateand subsequent minimum construction date is robust.

3.2. Dating the ship’s construction

The calculated minimum construction dates for Akko 1 usingthe different sapwood estimates are given in Table 3, and theirprobability distributions are shown in Fig. 6. If we calculate aminimum construction date using the north Aegean oak sapwoodestimate (since it is the most likely origin for the timber), the mostlikely period for the ship’s construction is between AD 1817 and1834 at 68.2% probability (AD 1810e1856 at 94.3% probability and1889e1897 at 1.1% probability) (see Fig. 6A). If we use the other twosapwood estimates instead, the likely ship construction dates varyby less than 10 years at 68.2% probability (see Fig. 6B, C, Table 3). Ifwe rule out the very small probability of a minimum constructiondate in ca. 1880e1900 in two of the models, the likely minimumconstruction dates produced from using different sapwood esti-mates vary by less than 20 years at 95.4% probability.

The different models give a possible (but very unlikely) mini-mum construction date range from ca. AD 1880e1890. However,none of the finds from Akko 1 indicate that the ship was con-structed any later than the mid-19th century (Cvikel and Kahanov,2013). Our models likely produce these date ranges because ofchanges to the shape of the radiocarbon calibration curve aroundthe early 20th century, when burning of fossil fuels started altering

radiocarbon levels in the atmosphere (the ‘Suess Effect’) (Suess,1955). Thus, while these post-1880 construction (and also wreck-ing) date ranges are included in our results and shown in Figs. 6 and7, they may be excluded from our analysis of both the ship’s likelyconstruction and wrecking dates.

3.3. Dating the ship’s final voyage/wrecking

The overall model produces slightly different estimates for thecalendar date of the ship’s final voyage/wrecking depending on thesapwood estimate employed (see Fig. 7A, C, and D). The most likelyprobability ranges are all very similar at AD 1822e1842 (northAegean oaks); AD 1825e1848 (the northwest Turkey oak subset,‘DEV’); and AD 1821e1837 (Italian oaks) at 68.2% probability (seeTable 3). If an additional terminus post quem of 1839 for the ship’slast voyage, based on Mentovich et al.’s analysis (2010), is added tothe model using the north Aegean sapwood estimate, OxCal pro-duces a most likely date range of AD 1839e1852 at 68.2% proba-bility, with probability strongly skewed toward the earlier part ofthe date range (see Fig. 7B).

3.4. Testing model robustness and possible regional 14C offsets

The overall dataset employed in our model (see Fig. 4)ecomprising the 14C wiggle-matched timbers, additional woodsamples, and short/shorter-lived sample setefits nicely to the var-iations in the atmospheric radiocarbon record (see Fig. 5). Theoverall model using the north Aegean sapwood estimates producesan analysis with a good OxCal agreement index (Amodel 69.7>60),and the alternative northwest Turkish subset and Italian sapwoodestimate models also offer Amodel values >60 at 69.8 and 67.5,respectively. There are only two minor outliers produced usingBronk Ramsey’s General Outlier model (2009b) (see Figs. 4 and 5). Ifthemodel using north Aegean sapwood estimates is re-runwithout

Table 3Calendar date ranges (years AD) for the minimum felling dates (MFD) for the three wiggle-matched timbers, for the ship’s minimum construction date (MCD), and for the lastvoyage (LV) according to a range of dating models and criteria (see text for model descriptions). The OxCal agreement values (‘Amodel’) are also listed for each model; asatisfactory value is �60.

North Aegean(all data)

North Aegean(minus 2 outliers)

Northwest Turkey(all data)

Italy(all data)

North Aegean(all data and1839 TPQ)

North Aegean(all data andDR 0 � 10)

North Aegean(all data andDR 19 � 5 onshort-livedsamples)

Amodel value 69.7 > 60 94 > 60 69.8 > 60 67.5 > 60 68.8 > 60 81 > 60 60 � 60AKO-1 MFD (68.2%) 1800e1821 1799e1820 1806e1825 1795e1814 1800e1824 1800e1823 1801e1821AKO-1 MFD (95.4%) 1787e1829 1785e1830 1792e1833 1782e1822 1785e1834 1785e1841 1786e1829AKO-11 MFD (68.2%) 1801e1820 1801e1820 1807e1824 1795e1812 1801e1823 1800e1823 1801e1820AKO-11 MFD (95.4%) 1789e1831 1789e1835 1794e1838 1785e1825 1790e1841 1789e1843 1788e1832AKO-12 MFD (68.2%) 1812e1827 1811e1828 1813e1830 1805e1820 1816e1832 1812e1829 1812e1827AKO-12 MFD (95.4%) 1804e1836 1803e1838 1804e1843 1797e1829 1807e1841 1804e1842 1803e1837MCD (68.2%) 1817e1834 1815e1837 1818e1837 1813e1830 1824e1843 1816e1836 1817e1833MCD (95.4%) 1810e1856

(94.1%),1889e1897(1.1%)

1808e1856 1812e1860(91.1%),1880e1900(4.3%)

1806e1844 1818e1861(89.4%),1879e1902(6.0%)

1810e1864(93.8%),1887e1897(1.6%)

1810e1853(92.4%),1888e1902(3%)

LV (68.2%) 1822e1842 1823e1849 1825e1848 1821e1837 1839e1852 1823e1847 1822e1841LV (95.4%) 1815e1866

(94.1%),1897e1905(1.3%)

1814e1868 1817e1873(91.4%),1894e1909(4.0%)

1814e1854 1839e1874 (89.4%),1882e1885 (0.7%),1894e1911 (5.3%)

1815e1874(94%),1897e1905(1.4%)

1815e1864(92.5%),1897e1908(2.9%)


these two outliers, the resulting date ranges are almost the same(see Table 3).

The DR test of 0 � 10 years for calculations using the northAegean oak sapwood model returns little evidence of any signifi-cant offset. The mean offset for the minimum timber use date isjust �3.3 � 7.7 years (�1 standard deviation), which indicates thatthere is no substantial offset. This modifies the last voyage/wreck-ing date range by only a few years, so that it is AD 1823e1847 at68.2% probability (AD 1815e1874 at 94% probability and AD 1897e1905 at 1.4% probability). If the pre-Aswan ‘Egyptian’ Nile Valleygrowing season adjustment of 19 � 5 14C years is added to theshort/shorter-lived samples, the modeled last voyage/wreckingdate range is placed only slightly later at AD 1823e1847 at 68.2%probability (AD 1815e1874 at 94% probability and AD 1897e1905 at1.4% probability). The overall agreement of the model also goes up(Amodel ¼ 81), which may suggest that this consideration is rele-vant. Nevertheless, even if a small 14C offset applies to the short/shorter-lived materials, its overall effect on our dating results isquite small and does not affect our overall interpretation of thedates (see discussion below), but may create a slightly morecompatible model.

4. Discussion

According to the dating models calculated above, the minimumconstruction dates for the Akko 1 ship timbers lie somewhere be-tween AD 1813 and 1843 at the extremes of the 68.2% probabilityranges (see Table 3). The ship’s final voyage most likely occurredbetween AD 1821 and 1852 at 68.2% probability when no otheradditional information is included in the model. If the terminus postquem indicated by Mentovich et al.’s (2010) analysis of the can-nonballs is incorporated and the north Aegean sapwood estimateused, then the most likely final voyage/wrecking date can be nar-rowed to AD 1839e1852 at 68.2% probability. The model andcalculated dates fit well with the approximate date ranges given tothe Akko 1 finds, arms, and rigging elements (Cvikel and Kahanov,2009, 2013; Shalev, pers. comm., 2007, 2010).

Our model indicates that the ship was constructed after Napo-leon Bonaparte’s 1799 siege of Akko (see Table 3). Even if the Italiansapwood model (which provides the earliest possible construction

date) is used, the earliest likely construction date at 95.4% proba-bility still post-dates the siege of Akko by 9 years. It is instead mostlikely that the Akko 1’s timbers were cut, and building of the shipinitiated, during the political ascendency of the Egyptian valiMuhammad Ali Pasha.

Muhammad Ali invested considerable resources in obtainingtimber, which he deemed critical for building up the Egyptian navy,and even personally oversaw timber importation to Egypt (Mikhail,2011). The most easily accessible and largest quantities of availabletimber came from the Black Sea and south Mediterranean coasts ofAnatolia, which traded heavily with Egypt during the early 19thcentury (McNeill, 1992, pp. 246e247; Mikhail, 2011). However it isdocumented that Muhammad Ali also imported timbers from Italy(McNeill, 1992, pp. 246e247). Muhammad Ali was apparently un-happy with both the quantity and quality of wood from these lo-cations (Marsot Al-Sayyid, 1984, p. 221; McNeill, 1992, p. 247). Hisdesire for high-quality, accessible timber resources free fromOttoman control e which he had unsuccessfully attempted to ac-quire within Egypt e was, in fact, his primary motivation forinvading the Levant (including Akko) and southeast Anatolia in1831 (Kutluo�glu, 1998, p. 51; Marsot Al-Sayyid, 1984, p. 228;Mikhail, 2011, pp. 160e169; Rustum, 1936, pp. 63e64).

Akko 1’s wrecking site and hull construction indicate that it wasfriendly to, and was likely even built, in Egypt (Cvikel and Kahanov,2013). Like the Turkish and Italian timbers that Muhammad Alireportedly found unsatisfactory for his navy, the timber used tobuild Akko 1 was generally poor quality, cut from juvenile trees,and (based on the abrupt tree-ring growth suppression and re-leases in many of the samples) culled from a forest/forests thatexperienced frequent anthropogenic disturbance (possibly fromcoppicing and pollarding). If the timbers were in fact culled frommultiple areas in the Turkish and Italian forests, this may explainwhy the ship timbers do not dendrochronologically crossdate.

Even if the Akko 1 timbers were taken entirely from Turkishforests, it is still likely that the ship’s timbers were culled fromseveral different forest stands. Mikhail’s (2011) analysis of theOttoman and Egyptian archives demonstrated that the standardprocedure for timber exportation between Anatolia and Egypt wasthat: i) the Ottoman sultan agreed to grant an Egyptian request fortimber; ii) Anatolian timbers were cut from multiple forests in


Anatolia and then sent in bulk to the Imperial Dockyards in Istan-bul, or ports like Antalya and Alanya in southern Anatolia, whichserved as timber repositories; iii) the wood was then shipped to thedockyards in Alexandria; and iv) timbers were distributedthroughout Egypt as needed. Thus, wood used to make an entirefleet of Egyptian vessels might come from several locations withinAnatolia.

If ship construction concluded by 1830 (which our datingmodels indicate is entirely possible), Akko 1 might have even beenone of the six Egyptian armed brigs that sailed to Akko andparticipated in Ibrahim Pasha’s 1831 siege of the town (Cvikel and

Fig. 6. Probability distributions for Akko 1’s minimum construction date (MCD) at68.2% and 95.4% probabilities: (A) uses the north Aegean oak sapwood estimate; (B)alternative northwest Turkish sapwood estimate; and (C) Italian oak sapwoodestimate.

Kahanov, 2013). Considering that the ship was armed and sufferedconsiderable damage during its wrecking (Cvikel and Kahanov,2013), it is likely that it was wrecked during a naval battle. Thefact that no other naval campaign took place in the vicinity of Akkoafter 1840, combined with our dating model and analysis of thefinds, suggests that the Akko 1 was sunk as a direct result of the1840 bombardment of Akko by the Allied fleet, or by the explosionof the town’s main powder magazine.

If Akko 1 was a naval vessel in Akko on guarding and patrollingduties, she could have taken part and been sunk during the battle.She might also have been an auxiliary vessel friendly to Egyptianforces, like that which Captain Henry Codrington (commander ofthe British frigate HMS Talbot) observed anchoring in Akko harboron the eve of the bombardment in 1840 (Codrington, 1880, p. 182).If Akko 1 was the brig that Codrington observed, then she was anauxiliary vessel shipping arms and ammunition to the Egyptiansdefending Akko when she sank.

5. Conclusions

The Bayesian analysis model used in our study gives a muchnarrower possible range of dates for when Akko 1’s ship timberswere felled, for when the vessel was built, and for when the vesselwas wrecked than either single 14C dates or typological datesderived from the ship’s equipment or finds. This study is, to ourknowledge, the first time that such techniques have been used toprovide high-precision dates for a shipwreck found in the south-eastern Mediterranean. Our results show clear advantages forimplementing this type of analysis, particularly for dating historicalshipwrecks, and even buildings or archaeological sites, when theirspecific histories are not documented and direct dendrochrono-logical dating is not possible.

Previously, the broad range of dates from the 17th through 19thcenturies AD for the ship’s construction and wrecking availablefrom single-sample radiocarbon dates limited the ability to inter-pret the physical ship remains in the context of historical eventsduring a time period in which there were multiple political up-heavals, naval and military campaigns, and changing and expand-ing trade networks. However, with our dating model (and e asseems most likely e assuming that the timbers are from Anatolia/the Aegean and citing themodel excluding the two outliers), we areable to determine that the ship was constructed no earlier than AD1815e1837 at 68.2% probability (AD 1808e1856 at 95.4% proba-bility) and likely wrecked between AD 1823 and 1849 at 68.2%probability (AD 1814e1868 at 95.4% probability), or, allowing forthe level of manganese in the cannonballs, likely AD 1839e1855 at68.2% probability (AD 1839e1874 at 95.4% probability). If theanalyzed ship timbers came from either of the other two proposedregional oak populations, these dates vary by at most only a fewyears.

Combined with evidence from the small finds and botanicalanalysis of the timbers, we can then fit this information with thehistorical record to suggest that Akko 1 was an Egyptian brig builtfor Muhammad Ali’s navy during his political ascendency; possiblyplied the eastern Mediterranean during the First EgyptianeOttoman War in 1831e1833; and sank during the Alliedbombardment of Akko in 1840.

These data aid in further analysis of the Akko 1 shipwreck as anEgyptian naval brig. Even more critically, our study demonstratesthe advantages of using Bayesian analysis techniques combiningtree-ring analysis and 14C data, in interpreting and identifying otherhistorical-era shipwrecks (and shipwrecks from earlier periods)from the southeast Mediterranean. These methods provide thehigh-precision dates that are required to insert such important andrich ship finds into the region’s historical narrative.

Fig. 7. Probability distributions for dating the ship’s final voyage/wrecking from the dating model shown in Fig. 4: (A) uses the north Aegean oak sapwood estimate; (B) the northAegean oak sapwood estimate including the AD 1839 terminus post quem from Mentovich et al.’s (2010) cannonball analysis; (C) the alternative northwest Turkish sapwood es-timate; and (D) Italian oak sapwood estimate.


Acknowledgments

The underwater excavation and research of the Akko 1 ship-wreck were supported by Ron Marlar, the Yaacov Salomon Foun-dation, the Halpern Foundation, a Sir Maurice Hatter Fellowship,the Hecht Trust, the Jewish National Fund, anonymous donors, andthe President, Rector, Dean and Faculty of Humanities, University ofHaifa, to whom we are grateful.

Dendrochronological researchat theMalcolmandCarolynWienerLaboratory for Aegean andNear EasternDendrochronologyat CornellUniversity is supported by generous contributions by theMalcolmH.Wiener Foundation, the National Science Foundation, and numerousindividual patrons of the Aegean Dendrochronology Project.

We are also grateful to Cornell laboratory employees JenniferWatkins, Jessica Herlich, Kayla Altland, and Leann Canady, for theircontributions in the preparation and measurement of samples;Tomasz Wazny (University of Arizona) for his helpful comments;and Bernd Kromer (Heidelberg), Georges Bonani and Irka Hajdas(Zurich), and the team at the Oxford Radiocarbon Accelerator Unitfor providing the radiocarbon dates.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jas.2013.10.013.

References

Akkemik, U., Yaman, B., 2012. Wood Anatomy of Eastern Mediterranean Species.Verlag Kessel, Remagen-Oberwinter.

Alderson, R.C., 1843. Notes on acre and some of the coast defences in Syria. In:Papers on Subjects Connected with the Duties of the Corps of Royal Engineers,VI. John Weale, London.

Anderson, R.C., 1952. Naval Wars in the Levant 1559e1853. University Press ofLiverpool, Liverpool.

Ashkenazi, D., Cvikel, D., Iddan, N., Mentovich, E.D., Kahanov, Y., Levinstain, M., 2011.Archaeometallurgical study of the brass cases from the Akko 1 shipwreck.J. Archaeol. Sci. 38 (9), 2410e2419.

Bayliss, A., 2007. Bayesian buildings: an introduction for the numerically chal-lenged. Vernac. Archit. 38, 75e86.

Bayliss, A., Tyers, I., 2004. Interpreting radiocarbon dates using evidence from treerings. Radiocarbon 46 (2), 957e964.

Biger, G., Liphschitz, N., 1991. Regional dendrohistory and timber analysis: the useof wood in the buildings of nineteenth-century Jaffa. Mediter. Hist. Rev. 6 (1),86e104.

Brewer, P.W., Sturgeon, K., Madar, L., Manning, S.W., 2010. A new approach todendrochronological data management. Dendrochronologia 28, 131e134.

Bronk Ramsey, C., 1995. Radiocarbon calibration and analysis of stratigraphy: theOxCal program. Radiocarbon 37 (2), 425e430.

Bronk Ramsey, C., 2009a. Bayesian analysis of radiocarbon dates. Radiocarbon 51(1), 337e360.

Bronk Ramsey, C., 2009b. Dealing with outliers and offsets in radiocarbon dating.Radiocarbon 51 (3), 1023e1045.

Bronk Ramsey, C., van der Plicht, J., Weninger, B., 2001. ‘Wiggle matching’ radio-carbon dates. Radiocarbon 43 (2A), 381e389.

Codrington, H.J., 1880. In: Bourchier, Lady (Ed.), Selections from the Letters (Privateand Professional) of Sir Henry Codrington Admiral of the Fleet. Spottiswoodeand Co., London.

Cvikel, D., Kahanov, Y., 2006. The Dor 2002/2 shipwreck. Archaeol. Marit. Mediter. 3,79e98.

Cvikel, D., Kahanov, Y., 2009. The Akko 1 shipwreck, Israel: the first two seasons. Int.J. Naut. Archaeol. 38 (1), 38e57.

Cvikel, D., Kahanov, Y., 2013. The 19th-century Akko 1 shipwreck, Israel: hull-construction report. Int. J. Naut. Archaeol. 42 (1), 167e187.

Cvikel, D., Kahanov, Y., Goren, H., Boaretto, E., Raveh, K., 2008. Napoleon Bonaparte’sadventure in Tantura Lagoon: historical and archaeological evidence. Isr. Explor.J. 58 (2), 199e219.

Dee, M.W., Brock, F., Harris, S.A., Bronk Ramsey, C., Shortland, A.J., Higham, T.F.G.,Rowland, J.M., 2010. Investigating the likelihood of a reservoir offset in theradiocarbon record for ancient Egypt. J. Archaeol. Sci. 37 (4), 687e693.

Dodds, J., Moore, J., 2005. Building the Wooden Fighting Ship. Chatham, London.Dothan, M., 1993. Tell Acco. In: Stern, E. (Ed.), The New Encyclopedia of Archaeo-

logical Excavations in the Holy Land, vol. 1. Israel Exploration Society, Carta,Jerusalem, pp. 17e24.

http://dx.doi.org/10.1016/j.jas.2013.10.013http://dx.doi.org/10.1016/j.jas.2013.10.013http://refhub.elsevier.com/S0305-4403(13)00360-9/sref1http://refhub.elsevier.com/S0305-4403(13)00360-9/sref1http://refhub.elsevier.com/S0305-4403(13)00360-9/sref2http://refhub.elsevier.com/S0305-4403(13)00360-9/sref2http://refhub.elsevier.com/S0305-4403(13)00360-9/sref2http://refhub.elsevier.com/S0305-4403(13)00360-9/sref3http://refhub.elsevier.com/S0305-4403(13)00360-9/sref3http://refhub.elsevier.com/S0305-4403(13)00360-9/sref3http://refhub.elsevier.com/S0305-4403(13)00360-9/sref4http://refhub.elsevier.com/S0305-4403(13)00360-9/sref4http://refhub.elsevier.com/S0305-4403(13)00360-9/sref4http://refhub.elsevier.com/S0305-4403(13)00360-9/sref4http://refhub.elsevier.com/S0305-4403(13)00360-9/sref5http://refhub.elsevier.com/S0305-4403(13)00360-9/sref5http://refhub.elsevier.com/S0305-4403(13)00360-9/sref5http://refhub.elsevier.com/S0305-4403(13)00360-9/sref6http://refhub.elsevier.com/S0305-4403(13)00360-9/sref6http://refhub.elsevier.com/S0305-4403(13)00360-9/sref6http://refhub.elsevier.com/S0305-4403(13)00360-9/sref7http://refhub.elsevier.com/S0305-4403(13)00360-9/sref7http://refhub.elsevier.com/S0305-4403(13)00360-9/sref7http://refhub.elsevier.com/S0305-4403(13)00360-9/sref7http://refhub.elsevier.com/S0305-4403(13)00360-9/sref53http://refhub.elsevier.com/S0305-4403(13)00360-9/sref53http://refhub.elsevier.com/S0305-4403(13)00360-9/sref53http://refhub.elsevier.com/S0305-4403(13)00360-9/sref8http://refhub.elsevier.com/S0305-4403(13)00360-9/sref8http://refhub.elsevier.com/S0305-4403(13)00360-9/sref8http://refhub.elsevier.com/S0305-4403(13)00360-9/sref9http://refhub.elsevier.com/S0305-4403(13)00360-9/sref9http://refhub.elsevier.com/S0305-4403(13)00360-9/sref9http://refhub.elsevier.com/S0305-4403(13)00360-9/sref10http://refhub.elsevier.com/S0305-4403(13)00360-9/sref10http://refhub.elsevier.com/S0305-4403(13)00360-9/sref10http://refhub.elsevier.com/S0305-4403(13)00360-9/sref11http://refhub.elsevier.com/S0305-4403(13)00360-9/sref11http://refhub.elsevier.com/S0305-4403(13)00360-9/sref11http://refhub.elsevier.com/S0305-4403(13)00360-9/sref12http://refhub.elsevier.com/S0305-4403(13)00360-9/sref12http://refhub.elsevier.com/S0305-4403(13)00360-9/sref12http://refhub.elsevier.com/S0305-4403(13)00360-9/sref13http://refhub.elsevier.com/S0305-4403(13)00360-9/sref13http://refhub.elsevier.com/S0305-4403(13)00360-9/sref13http://refhub.elsevier.com/S0305-4403(13)00360-9/sref14http://refhub.elsevier.com/S0305-4403(13)00360-9/sref14http://refhub.elsevier.com/S0305-4403(13)00360-9/sref14http://refhub.elsevier.com/S0305-4403(13)00360-9/sref15http://refhub.elsevier.com/S0305-4403(13)00360-9/sref15http://refhub.elsevier.com/S0305-4403(13)00360-9/sref15http://refhub.elsevier.com/S0305-4403(13)00360-9/sref16http://refhub.elsevier.com/S0305-4403(13)00360-9/sref16http://refhub.elsevier.com/S0305-4403(13)00360-9/sref16http://refhub.elsevier.com/S0305-4403(13)00360-9/sref16http://refhub.elsevier.com/S0305-4403(13)00360-9/sref17http://refhub.elsevier.com/S0305-4403(13)00360-9/sref17http://refhub.elsevier.com/S0305-4403(13)00360-9/sref17http://refhub.elsevier.com/S0305-4403(13)00360-9/sref17http://refhub.elsevier.com/S0305-4403(13)00360-9/sref18http://refhub.elsevier.com/S0305-4403(13)00360-9/sref19http://refhub.elsevier.com/S0305-4403(13)00360-9/sref19http://refhub.elsevier.com/S0305-4403(13)00360-9/sref19http://refhub.elsevier.com/S0305-4403(13)00360-9/sref19


Ducousso, A., Bordacs, S., 2003. EUFORGEN Technical Guidelines for GeneticConservations and Use for: Pedunculate and Sessile Oaks (Quercus roburand Quercus petraea). International Plant Genetic Resources Institute, Rome,p. 6.

Elkin, D., Argüeso, A., Grosso, M., Murray, C., Vainstub, D., Bastida, R., Dellino-Musgrave, V., 2007. Archaeological research on HMS Swift: a British Sloop-of-War lost off Patagonia, southern Argentina, in 1770. Int. J. Naut. Archaeol. 36(1), 32e58.

Galimberti, M., Bronk Ramsey, C., Manning, S.W., 2004. Wiggle-match dating oftree-ring sequences. Radiocarbon 46 (2), 917e924.

Griggs, C.B., DeGaetano, A.T., Kuniholm, P.I., Newton, M.W., 2007. A regionalreconstruction of MayeJune precipitation in the north Aegean from oak tree-rings, AD 1089e1989. Int. J. Climatol. 27, 1075e1089.

Griggs, C.B., Kuniholm, P.I., Newton, M.W., Watkins, J.D., Manning, S.W., 2009.A 924-year regional oak tree-ring chronology for north Central Turkey. In:Manning, S.W., Bruce, M.J. (Eds.), Tree-Rings, Kings and Old World Archaeologyand Environment Papers Presented in Honor of Peter Ian Kuniholm. OxbowBooks, Oxford, pp. 71e79.

Haneca, K., �Cufar, K., Beeckman, H., 2009. Oaks, tree-rings and wooden culturalheritage: a review of the main characteristics and applications of oak dendro-chronology in Europe. J. Archaeol. Sci. 36 (1), 1e11.

Hillam, J., Morgan, R.A., Tyers, I., 1987. Sapwood estimates and the dating of shortring sequences. In: BAR International Series, vol. 333.

Hughes, M.K., Milsom, S.J., Leggett, P.A., 1981. Sapwood estimates in the interpre-tation of tree-ring dates. J. Archaeol. Sci. 8 (4), 381e390.

Kahanov, Y., Tresman, J.B., Mei-Bar, Y., Cvikel, D., Hillman, A., 2012. Akko 1 ship-wreck: the effect of cannon fire on the wooden hull. J. Archaeol. Sci. 39 (7),1993e2002.

Kutluo�glu, M.H., 1998. The Egyptian Question (1831e1841): the Expansionist Policyof Mehmed Alı Paşa in Syria and Asia Minor and the Reaction of the SublimePorte. Eren, Istanbul.

La Jonquière, C., 1900. L’expédition d’ _ggypte, 1798e1801. H. Charles-Lavauzelle,Paris.

Makhouly, N., Johns, C.N., 1941. Guide to Acre. Government of Palestine, Departmentof Antiquities, Jerusalem.

Manning, S.W., Kromer, B., 2012. Considerations of the scale of radiocarbon offsetsin the east Mediterranean, and considering a case for the latest (most recent)likely date for the Santorini eruption. Radiocarbon 54 (3e4), 449e474.

Manning, S.W., Kromer, B., Bronk Ramsey, C., Pearson, C.L., Talamo, S., Trano, N.,Watkins, J.D., 2010. 14C record and wiggle-match placement for the Anatolian(Gordion area) juniper tree-ring chronology w1729 to 751 cal BC, and typicalAegean/Anatolian (growing season related) regional 14C offset assessment.Radiocarbon 52 (4), 1571e1597.

Marsot Al-Sayyid, A.L., 1984. Egypt in the Reign of Muhammad Ali. CambridgeUniversity Press, Cambridge.

Masters, B., 2009. Acre. In: Ágoston, G., Masters, B. (Eds.), Encyclopedia of theOttoman Empire. Facts on File, New York, pp. 9e10.

McNeill, J.R., 1992. The Mountains of the Mediterranean World. Cambridge Uni-versity Press, Cambridge, UK.

Mentovich, E.D., Schreiber, D.S., Goren, Y., Kahanov, Y., Goren, H., Cvikel, D.,Ashkenazi, D., 2010. New insights regarding the Akko 1 shipwreck: a metal-lurgic and petrographic investigation of the cannonballs. J. Archaeol. Sci. 37(10), 2520e2528.

Mikhail, A., 2011. Nature and Empire in Ottoman Egypt: an Environmental History.Cambridge University Press, Cambridge, UK.

Miles, D.H., 2005. New Developments in the Interpretation of Dendrochronology asApplied to Oak Building Timbers (Ph.D. thesis). University of Oxford, Oxford.

Miles, D., 2006. Refinements in the interpretation of tree-ring dates for oak buildingtimbers in England and Wales. Vernac. Archit. 37, 84e96.

Millard, A., 2002. A Bayesian approach to sapwood estimates and felling dates indendrochronology. Archaeometry 44, 137e143.

Negev, A., Gibson, S. (Eds.), 2005. Archaeological Encyclopedia of the Holy Land.Continuum, New York.

Randerson, J.T., Enting, I.G., Schuur, E.A.G., Caldeira, K., Fung, I.Y., 2002. Seasonal andlatitudinal variability of troposphere D14CO2: post bomb contributions fromfossil fuels, oceans, the stratosphere, and the terrestrial biosphere. Glob. Bio-geochem. Cycles 16 (4), 59.1e59.19.

Reimer, P.J., Baillie, M.G.L., Bard, E., Bayliss, A., Beck, J.W., Blackwell, P.G., BronkRamsey, C., Buck, C.E., Burr, G.S., Edwards, R.L., Friedrich, M., Grootes, P.M.,Guilderson, T.P., Hajdas, I., Heaton, T.J., Hogg, A.G., Hughen, K.A., Kaiser, K.F.,Kromer, B., McCormac, F.G., Manning, S.W., Reimer, R.W., Richards, D.A.,Southon, J.R., Talamo, S., Turney, C.S.M., van der Plicht, J., Weyhenmeyer, C.E.,2009. IntCal09 and Marine09 radiocarbon age calibration curves, 0e50,000years cal BP. Radiocarbon 51 (4), 1111e1158.

Rustum, A.J., 1926. Notes on Akka and its defences under Ibrahim Pasha. In: Preparedfor the Archaeological Congress of Syria and Palestine. American Press, Beirut.

Rustum, A.J., 1936. The Royal Archives of Egypt and the Origins of the EgyptianExpedition to Syria, 1831e1841. American Press, Beirut.

Schweingruber, F.H., 1990. Anatomy of European Woods. Verlag Paul Haupt, Bern.Schweingruber, F.H., 1993. Trees and Wood in Dendrochronology: Morphological,

Anatomical, and Tree-Ring Analytical Characteristics of Trees Frequently Usedin Dendrochronology. Springer-Verlag, Berlin.

Suess, H.E., 1955. Radiocarbon content in modern wood. Science 122, 415e417.Tredgold, T., 1875. Elementary Principles of Carpentry. Revised and partly rewritten

by J.T. Hurst. Spon, London. Available from: http://www.archive.org/details/elementaryprinci00treduoft.

Turetsky, M.R., Manning, S.W., Wieder, R.K., 2004. Dating recent peat deposits.Wetlands 24, 324e356.

Wazny, T., 1990. Aufbau und Anwendung der Dendrochronologie für Eichenholz inPolen (Ph.D. thesis). Hamburg University, Hamburg.

http://refhub.elsevier.com/S0305-4403(13)00360-9/sref20http://refhub.elsevier.com/S0305-4403(13)00360-9/sref20http://refhub.elsevier.com/S0305-4403(13)00360-9/sref20http://refhub.elsevier.com/S0305-4403(13)00360-9/sref20http://refhub.elsevier.com/S0305-4403(13)00360-9/sref21http://refhub.elsevier.com/S0305-4403(13)00360-9/sref21http://refhub.elsevier.com/S0305-4403(13)00360-9/sref21http://refhub.elsevier.com/S0305-4403(13)00360-9/sref21http://refhub.elsevier.com/S0305-4403(13)00360-9/sref21http://refhub.elsevier.com/S0305-4403(13)00360-9/sref22http://refhub.elsevier.com/S0305-4403(13)00360-9/sref22http://refhub.elsevier.com/S0305-4403(13)00360-9/sref22http://refhub.elsevier.com/S0305-4403(13)00360-9/sref23http://refhub.elsevier.com/S0305-4403(13)00360-9/sref23http://refhub.elsevier.com/S0305-4403(13)00360-9/sref23http://refhub.elsevier.com/S0305-4403(13)00360-9/sref23http://refhub.elsevier.com/S0305-4403(13)00360-9/sref23http://refhub.elsevier.com/S0305-4403(13)00360-9/sref23http://refhub.elsevier.com/S0305-4403(13)00360-9/sref24http://refhub.elsevier.com/S0305-4403(13)00360-9/sref24http://refhub.elsevier.com/S0305-4403(13)00360-9/sref24http://refhub.elsevier.com/S0305-4403(13)00360-9/sref24http://refhub.elsevier.com/S0305-4403(13)00360-9/sref24http://refhub.elsevier.com/S0305-4403(13)00360-9/sref24http://refhub.elsevier.com/S0305-4403(13)00360-9/sref25http://refhub.elsevier.com/S0305-4403(13)00360-9/sref25http://refhub.elsevier.com/S0305-4403(13)00360-9/sref25http://refhub.elsevier.com/S0305-4403(13)00360-9/sref25http://refhub.elsevier.com/S0305-4403(13)00360-9/sref25http://refhub.elsevier.com/S0305-4403(13)00360-9/sref26http://refhub.elsevier.com/S0305-4403(13)00360-9/sref26http://refhub.elsevier.com/S0305-4403(13)00360-9/sref27http://refhub.elsevier.com/S0305-4403(13)00360-9/sref27http://refhub.elsevier.com/S0305-4403(13)00360-9/sref27http://refhub.elsevier.com/S0305-4403(13)00360-9/sref28http://refhub.elsevier.com/S0305-4403(13)00360-9/sref28http://refhub.elsevier.com/S0305-4403(13)00360-9/sref28http://refhub.elsevier.com/S0305-4403(13)00360-9/sref28http://refhub.elsevier.com/S0305-4403(13)00360-9/sref29http://refhub.elsevier.com/S0305-4403(13)00360-9/sref29http://refhub.elsevier.com/S0305-4403(13)00360-9/sref29http://refhub.elsevier.com/S0305-4403(13)00360-9/sref29http://refhub.elsevier.com/S0305-4403(13)00360-9/sref29http://refhub.elsevier.com/S0305-4403(13)00360-9/sref29http://refhub.elsevier.com/S0305-4403(13)00360-9/sref29http://refhub.elsevier.com/S0305-4403(13)00360-9/sref30http://refhub.elsevier.com/S0305-4403(13)00360-9/sref30http://refhub.elsevier.com/S0305-4403(13)00360-9/sref30http://refhub.elsevier.com/S0305-4403(13)00360-9/sref30http://refhub.elsevier.com/S0305-4403(13)00360-9/sref31http://refhub.elsevier.com/S0305-4403(13)00360-9/sref31http://refhub.elsevier.com/S0305-4403(13)00360-9/sref32http://refhub.elsevier.com/S0305-4403(13)00360-9/sref32http://refhub.elsevier.com/S0305-4403(13)00360-9/sref32http://refhub.elsevier.com/S0305-4403(13)00360-9/sref32http://refhub.elsevier.com/S0305-4403(13)00360-9/sref32http://refhub.elsevier.com/S0305-4403(13)00360-9/sref33http://refhub.elsevier.com/S0305-4403(13)00360-9/sref33http://refhub.elsevier.com/S0305-4403(13)00360-9/sref33http://refhub.elsevier.com/S0305-4403(13)00360-9/sref33http://refhub.elsevier.com/S0305-4403(13)00360-9/sref33http://refhub.elsevier.com/S0305-4403(13)00360-9/sref33http://refhub.elsevier.com/S0305-4403(13)00360-9/sref33http://refhub.elsevier.com/S0305-4403(13)00360-9/sref33http://refhub.elsevier.com/S0305-4403(13)00360-9/sref33http://refhub.elsevier.com/S0305-4403(13)00360-9/sref34http://refhub.elsevier.com/S0305-4403(13)00360-9/sref34http://refhub.elsevier.com/S0305-4403(13)00360-9/sref35http://refhub.elsevier.com/S0305-4403(13)00360-9/sref35http://refhub.elsevier.com/S0305-4403(13)00360-9/sref35http://refhub.elsevier.com/S0305-4403(13)00360-9/sref36http://refhub.elsevier.com/S0305-4403(13)00360-9/sref36http://refhub.elsevier.com/S0305-4403(13)00360-9/sref37http://refhub.elsevier.com/S0305-4403(13)00360-9/sref37http://refhub.elsevier.com/S0305-4403(13)00360-9/sref37http://refhub.elsevier.com/S0305-4403(13)00360-9/sref37http://refhub.elsevier.com/S0305-4403(13)00360-9/sref37http://refhub.elsevier.com/S0305-4403(13)00360-9/sref38http://refhub.elsevier.com/S0305-4403(13)00360-9/sref38http://refhub.elsevier.com/S0305-4403(13)00360-9/sref39http://refhub.elsevier.com/S0305-4403(13)00360-9/sref39http://refhub.elsevier.com/S0305-4403(13)00360-9/sref40http://refhub.elsevier.com/S0305-4403(13)00360-9/sref40http://refhub.elsevier.com/S0305-4403(13)00360-9/sref40http://refhub.elsevier.com/S0305-4403(13)00360-9/sref41http://refhub.elsevier.com/S0305-4403(13)00360-9/sref41http://refhub.elsevier.com/S0305-4403(13)00360-9/sref41http://refhub.elsevier.com/S0305-4403(13)00360-9/sref42http://refhub.elsevier.com/S0305-4403(13)00360-9/sref42http://refhub.elsevier.com/S0305-4403(13)00360-9/sref43http://refhub.elsevier.com/S0305-4403(13)00360-9/sref43http://refhub.elsevier.com/S0305-4403(13)00360-9/sref43http://refhub.elsevier.com/S0305-4403(13)00360-9/sref43http://refhub.elsevier.com/S0305-4403(13)00360-9/sref43http://refhub.elsevier.com/S0305-4403(13)00360-9/sref43http://refhub.elsevier.com/S0305-4403(13)00360-9/sref43http://refhub.elsevier.com/S0305-4403(13)00360-9/sref44http://refhub.elsevier.com/S0305-4403(13)00360-9/sref44http://refhub.elsevier.com/S0305-4403(13)00360-9/sref44http://refhub.elsevier.com/S0305-4403(13)00360-9/sref44http://refhub.elsevier.com/S0305-4403(13)00360-9/sref44http://refhub.elsevier.com/S0305-4403(13)00360-9/sref44http://refhub.elsevier.com/S0305-4403(13)00360-9/sref44http://refhub.elsevier.com/S0305-4403(13)00360-9/sref44http://refhub.elsevier.com/S0305-4403(13)00360-9/sref44http://refhub.elsevier.com/S0305-4403(13)00360-9/sref45http://refhub.elsevier.com/S0305-4403(13)00360-9/sref45http://refhub.elsevier.com/S0305-4403(13)00360-9/sref46http://refhub.elsevier.com/S0305-4403(13)00360-9/sref46http://refhub.elsevier.com/S0305-4403(13)00360-9/sref46http://refhub.elsevier.com/S0305-4403(13)00360-9/sref47http://refhub.elsevier.com/S0305-4403(13)00360-9/sref48http://refhub.elsevier.com/S0305-4403(13)00360-9/sref48http://refhub.elsevier.com/S0305-4403(13)00360-9/sref48http://refhub.elsevier.com/S0305-4403(13)00360-9/sref49http://refhub.elsevier.com/S0305-4403(13)00360-9/sref49http://www.archive.org/details/elementaryprinci00treduofthttp://www.archive.org/details/elementaryprinci00treduofthttp://refhub.elsevier.com/S0305-4403(13)00360-9/sref51http://refhub.elsevier.com/S0305-4403(13)00360-9/sref51http://refhub.elsevier.com/S0305-4403(13)00360-9/sref51http://refhub.elsevier.com/S0305-4403(13)00360-9/sref52http://refhub.elsevier.com/S0305-4403(13)00360-9/sref52

High-precision dating the Akko 1 shipwreck, Israel: wiggle-matching the life and death of a ship into the historical record1 Introduction1.1 Historical setting1.2 The Akko 1 shipwreck1.3 Bayesian modeling and 14C wiggle-matching

2 Material and methods2.1 Tree-ring analysis and radiocarbon wiggle-matching of the ship timbers2.2 Dating the ship's final voyage/wrecking2.3 Testing model robustness and possible regional 14C offsets

3 Results3.1 Tree-ring analysis and wiggle-matching the ship timbers3.2 Dating the ship's construction3.3 Dating the ship's final voyage/wrecking3.4 Testing model robustness and possible regional 14C offsets

4 Discussion5 ConclusionsAcknowledgmentsAppendix A Supplementary dataReferences

journal of archaeological science - cornell university · 2014. 1. 28. · 1516 ( masters, 2009, p....

Documents