Doel 1 was constructed out of oak wood. Only a few smaller parts, a treenail and the six examined fairings, were made of other species. The primary function of the fairings is to prevent obstacles or objects from colliding with and damaging the through-beams that protrude through the hull, by guiding them past the beam ends; hence the fairings have a tapered and streamlined shape. Their shape also justifies the use of the term ‘fairing’ instead of the more conventional term ‘fender’, which is used for ship elements that are designed to absorb a direct impact. All the preserved fairings of Doel 1 were consistently made of alder, and therefore this was probably not a random choice. We believe that not only the shape of the fairings protects the through-beams and hull, but also the deliberate choice of a softer hardwood is relevant here. Compared to oak, alder has a lower density [alder: (12%) 490–640 kg/m3, (green) 680–1000 kg/m3; oak: (12%) 500–970 kg/m3, (green) 900–1200 kg/m3], and is considered to be less hard (Janka hardness: alder 33–38 N/mm2; oak 50–66 N/mm2) (Wagenführ, 2007: 255–60, 276–78). Therefore, on the occasion of an unintended point impact, fairings made of alder are more likely to absorb the (kinetic) energy, causing them to splinter, and having to be replaced. Were the fairings made of oak, they would have the same density and strength properties compared to the hull planks or through-beam, guiding the released forces into the hull and increasing the risk of damage to the planking. Clearly, the former is the preferred scenario, rather than a leak caused by cracked hull planks.
Felling date = construction date?
For two of the examined ship timbers, tree-ring analysis was able to determine the exact felling date. The question remained whether this AD 1325/26 date can be interpreted as the felling date for the bulk of the ship timbers. Seven other dated timbers with incomplete sapwood do point towards this date when relating the probability density function of the expected number of sapwood rings to each one of them (Fig. 7). Therefore it seems very likely that all these timbers were collected at the same time for the construction of this ship, which probably took place in or shortly after AD 1326. Some time needed for transport could be added to the felling date. However, it is most likely that this can be ignored as the freshly cut stems could easily arrive at the shipyard within the year since transport over water (rafting or aboard ships) is not only the most economical way of bulk transport, but also the fastest. Time for seasoning of the wood does not seem to be applicable, as working dry oak is definitely a more laborious task compared to shaping unseasoned timber into planks or framing timbers (Darrah, 1982: 220–22). Furthermore, for Doel 1 the use of ‘green’ (unseasoned) timber can be inferred, as it can for many cog wrecks, from the frequent appearance of (repaired) cracks down the centre of wide planks. These cracks are usually located near the pith (central tissue in a tree-trunk) where (juvenile) wood has a higher density compared to the wood laid down during the mature stage of a tree's life. Therefore the wood around the pith is more susceptible to shrinking and splitting while drying. Presumably the cracks appeared on the hull planks after green timbers were fastened to the frames by treenails, and had time to season.
Only the keel plank, with a possible earlier felling date, could have been subjected to some pre-treatment before it was used as the backbone of this ship. An explanation for the potential earlier felling date of this tree could be that the wood was stored in water for several months or years, a process referred to as ‘ponding’. The main advantage of this technique is that wood can be stored for a long period without any loss of quality related to fungal deterioration or degradation by wood-boring insects. This ancient technique is also supposed to ameliorate the wood's workability and dimensional stability while drying (Boerhave-Beeckman, 1949: 420–24; Crumlin-Pederson, 1986: 142). Furthermore, it is possible that during water storage the natural durability of the wood is improved since starch and other carbohydrates are leached, making it less prone to wood-degrading fungi. Although the latter is certainly true for coniferous wood (Klaassen, 2010), it can be questioned if this holds for oak timber, since the heartwood of oak already has a high natural durability. During (water) storage, gradual release of internal mechanical stress, built up in the stem of a growing tree (growth strain), will occur. This results in wood that tends to warp less while drying and avoids the formation of large cracks. Relaxation of internal stress in timber is a natural process, and a much appreciated (side-) effect of ponding, as it results in dimensionally more stable timber (Boerhave-Beeckman, 1949: 420–24; Crumlin-Pedersen, 1986: 142). Selecting ponded oak timber, free of internal stress and unlikely to develop cracks or warp, for the keel plank, as the backbone of the ship, can therefore be considered as an act of craftsmanship.
For one of the repair planks, its earliest possible felling date is AD 1327. So the year in which this particular oak tree was cut down is certainly situated after AD 1327. Unfortunately, we do not know how much heartwood might have been trimmed off in the shaping of the plank. If we assume that only sapwood was trimmed off and no heartwood rings, the felling date is situated between AD 1327 and 1346 (95.4% C.I.), with AD 1339 having the highest probability (average number of SWR is 19). This demonstrates that this repair was performed after the completion of Doel 1. To gain more information about the potential life-span of this ship, however, we cannot make a more precise statement based on the dendrochronological date of this repair plank. Nevertheless, it is most likely that this repair was performed a number of years after the construction of the ship. Furthermore, from the archaeological recording of the ship remains and the dendrochronological dataset, it is clear that this ship already had suffered much damage and many major repairs had been performed before it was abandoned.
We now have good indications that Doel 1 was constructed with timbers originating from oak forests located in the vicinity of the rivers Elbe or Weser and their tributaries. As shown on the correlation map (Fig. 8), the highest correlation values were reached with site chronologies from Medingen and Truhen, both data from the University of Göttingen. As such, this provenance determination for Doel 1 is similar to that of the Bremen cog (Daly, 2009: 115–16). The lack of evident division between the planks and framing timbers, from a dendrochronological perspective, indicates that both types of timber were harvested in the same region. The dendrochronological data from the Bremen cog suggests that the examined through-beam could have a more southern provenance compared to some other ship timbers (Bauch et al., 1967: 290); however, this was not observed for Doel 1 where the tree-ring patterns of the beams have a similar provenance signal as the planks and framing timbers. Defining how broad and how far from the shipyard this region could be is a more difficult matter. At first sight, the low overall correlation between the tree-ring series (Fig. 7) suggests a widely dispersed catchment area for the timber source, with the tree-ring series representing trees that originate from different and distant forest sites. In general, one expects to find higher correlation values for oak trees growing in close proximity. This is especially true when tree growth is subjected to some limiting factor (Fritts, 1976; Schweingruber, 1996: 439–548). In such cases the climatological signal embedded in tree-ring series is stronger and, as a consequence, this results in high correlation values for the tree-ring patterns of neighbouring trees. However, this would mean that nearly all timbers from Doel 1 originate from different forest sites, which is very unlikely. On the other hand, favourable conditions for oak trees, as is the maritime-influenced climate in coastal NW Germany, not only stimulate vigorous growth (wide tree-rings) but will result in ring-width patterns that do not necessarily display a high agreement with the ring-width pattern of neighbouring trees. Therefore the relatively low correlation between the tree-rings series of the Doel 1 timbers does not necessarily reflect a wide region of provenance.
The distance between the timber source and the shipyard should be discussed taking into account bulk transport of wood along river systems. This has been reported since the Roman times (Domínguez-Delmás et al., in press) and there is abundant archaeological and documentary evidence for the transport of this bulky wood as rafts since the Middle Ages (Van Prooije, 1990; Houbrechts, 2008: 17–25; Eissing and Dittmar, 2011; Haneca and Debonne, 2012: 32). Furthermore, the transport of timber aboard ships cannot be discarded either. At (local) wood markets, huge amounts of timber were gathered and different assortments were compiled, blending timbers from diverse forest sites and woodlands. Buying timbers from such a wood market that covers a wide region (or entire river catchment) could result in a dendrochronologically diverse assortment of tree-ring series, each representing single forest sites with local (micro-) climates. However, documentary evidence from 16th to the 18th century suggests that the selection of timber for shipbuilding was done in the forest, especially for the framing elements (Albion, 1926; De Aranda y Anton, 1990; Domínguez-Delmás et al., 2013: 132). Timber was then brought directly to the shipyard. Therefore it seems unlikely that the timber used to build Doel 1 was purchased at a wood market. Besides, the consistency in felling dates of some structural timbers also supports the hypothesis that these timbers were selected and harvested all at once.
At least three of the examined repair planks clearly have a different provenance compared to the bulk of the original ship timbers (Fig. 10). These repair planks originate from oak trees that probably grew near the mouth of the river Vistula or along the coastal area of northern Poland (Gdansk region). Although the repairs could have been carried out at Gdansk harbour, another location cannot be ruled out, as it is known that timbers with a southern Baltic provenance have been exported and traded all over Europe from the early 14th century up to c. AD 1660 (Ważny and Eckstein, 1987: 511–13; Bonde et al., 1997: 202–3; Ważny, 2005). Furthermore, as these repair planks were radially split timbers, they have all the characteristics of so-called ‘wainscots’, a high-quality oak timber-product that could be obtained from trees with a fine grain and without major defects, which were one of the most appreciated assortments in the ‘Baltic timber trade’ (Ważny and Eckstein, 1987: 511–12; Haneca et al., 2005a: 267; 2005b: 290). Most likely, wainscots were widely available in the ports along the North Sea coast, visited by merchants and their ships. Besides, they could have been part of the cargo of Doel 1 as well, readily available for any kind of repair if needed. Indeed, recent discoveries of shipwrecks with cargoes of exactly such planking, at Skjernøysund, Norway (Auer and Maarleveld, 2013), dating to winter AD 1393–4 and of southern Baltic origin (Daly, 2011b), underline that very possibility. So the location where these repairs were carried out cannot be determined based on the timber provenance as demonstrated by dendrochronology.
Expertise in shipbuilding
Studying shipbuilding technology is our closest link to the shipwright. It provides insight into the available woodworking technology and the knowledge of the raw material used in the process of constructing a seagoing cargo-vessel. Clearly, the shipwright's expertise is closely related to the range of timber at his disposal. Selection and conversion of timber for specific elements in the construction are fundamental for the final quality and strength of the vessel. Looking at the orientation of the tree-ring pattern on a cross-section of the grain of the wood (the main direction of orientation of tissues and cells), one can deduce the orientation of the timber in the original trunk or branch, and reconstruct the conversion of the tree-trunk. For instance, the Y-shaped framing timbers of Doel 1 are made of forked trees of which one of the branches was cut off, or trees with a heavy side-branch. The remaining, naturally curved piece of timber was then hewn to its desired shape and size. Planking, on the other hand, was sawn. The orientation of the growth pattern on the cross-sections of a sample of 42 timbers (photographed for tree-ring analysis) was categorized according to Figure 13a. It was noted that the bulk of these timbers were flat-sawn, and did not include the pith (Fig. 13b). Only 12% of this subset of timbers was described as a central plank of a trunk, including the pith (type C), or sawn as a radial plank without pith or sapwood.
Figure 13. a) Schematic drawing of samples extracted from a split (left) and sawn (right) tree-trunk. Four types of planks are defined. (Drawing: Glenn Laeveren); b) Division of 42 hull planks according to the schema presented in 13a
Download figure to PowerPoint
How many planks were sawn from one parent log? As most of the planks are of type A, one could assume that at least two planks were sawn from one trunk, one on each side of the pith. Theoretically, more adjacent planks could originate from the same log, but this was not observed in the dendrochronological dataset of Doel 1 (see Table 1). This suggests that the diameter of the original logs was too small to yield more than one pair of planks with an appropriate width for the construction of the bottom and hull. Furthermore, a striking pattern was observed for the bottom of the ship, from strake A to D, where the two planks from the same tree-trunk had a mirrored arrangement on both sides of the keel plank (Fig. 12). This detail in construction is sometimes referred to as the ‘twin-sister layout’. The symmetrical layout, in terms of wood use, was abandoned on the higher strakes. Nevertheless, such a dedicated plan for the bottom of the Doel 1 ship echoes the observed layout on several other archaeological ship-finds, where planks converted from the same parent tree are clustered along the hull, for example the Magor Pill boat (Nayling, 1998) or the Bredfjed ship (Bill, 1997a; 1997b). However, this type of layout was mostly deduced from the similar patterns observed in the grain and location of knots on two planks. From a Roman boatbuilding tradition, some examples are also known, for example from the lake of Neuchâtel in Switzerland (Arnold, 1992a; 1992b), the Woerden 7 from the Netherlands (Vorst, 2005) or the Barland's Farm boat (Nayling and McGrail, 2004), as well as medieval examples, such as the OZ36 and Almere wijk 13 cogs from the Netherlands (Bill, 1997b; Vorst, pers. comm.). The strict symmetrical layout of the bottom planks also might suggest that timber arrived at the shipyard as logs, and that the planks were sawn on the spot. The two, fresh-sawn central planks of the trunk were shaped and assembled in flush-lying strakes, on starboard and portside of the central keel plank. Furthermore, the timber chosen to construct the central keel plank was the results of a deliberate selection. The construction of the bottom and the keel of the Doel 1 clearly illustrate the expertise and dedication of the shipwright(s) concerned.
The bottom of the ship was planned systematically in terms of wood use and symmetry, but was this also the case from a wood-technological point-of-view? As can be seen from Figure 13, most timbers did not have the pith included. This could be intentional since planks with the pith included tend to have a higher risk of developing cracks. Furthermore, this orientation of the planks also entails a tendency to warp (Fig. 13a). Despite the neatly worked-out symmetry of the bottom strakes, no attention was paid to the (systematic) orientation of the pith towards the inside or the outside of the ship (Fig. 13b). So it seems this wood-technological feature was not taken into account or considered to be of any importance.
In contrast to the hull planks, all eight repair planks examined were consistently made of radially split oak timbers, which suggests an explicit choice for this grade of timber. This type of wood conversion follows the grain of the wood and, on the condition that they originate from logs without defects and knots, results in timbers that are less sensitive to warping while drying (Crumlin-Pedersen, 1989: 30–1). Such planks can be considered as the most dimensionally stable timbers, and therefore also assured a higher level of ‘safety’ to the repair.
From the keelson we now have two pieces of timber of which the growth pattern was recorded. The two pieces are located on either side of the hole that was cut in the hull of Doel 1 at its discovery. Both patterns could be dated, and although belonging to the same piece of timber, they do not overlap (gap of 16 years). Does this mean that the keelson was made out of two pieces? Not necessarily, as both locations where the growth pattern was recorded lie at least 4.03 m apart, not including the gap which is estimated to be around 2 m wide. This could explain why the tree-ring series do not overlap. Although dendrochronology cannot demonstrate that the keelson was fashioned out of a single piece of wood, from a wood-technological point-of-view it might be considered as a likely choice by the shipwright, since strength properties of a keelson made of two connected pieces could be inferior. Considering the structural function of the keelson to support the mast and sail, a scarf not far aft of the mast-step seems improbable.
Knowledge gain and concluding remarks
From a dendrochronological point-of-view, the examined ship timbers of Doel 1 provides one of the most comprehensive datasets in nautical archaeology at the level of a single ship-find, alongside the Newport ship and the Mary Rose. In total, the tree-ring pattern of 110 ship timbers was recorded and analysed. For dating purposes, timbers with sapwood or bark were targeted in order to anchor the find in time with the highest level of precision. Only those timbers allow us to narrow down the felling date to a precise time interval, to the year or even to the season if the bark edge is preserved as well. This is the most common strategy in dendrochronological studies, and should always be implemented when examining archaeological ship-finds. A narrow selection of easily accessible timbers without sapwood or bark edge from one particular find should not be followed, as this could lead to a post quem date that strongly deviates from the true felling date. As can be seen from Figure 6, some timbers had an end-date for their tree-ring pattern in the interval AD 1191–1200, more than 130 years away from the actual construction date of Doel 1.
Regarding information gained, the repair planks, the keel plank and the planks from the bottom strakes have proved most interesting. Targeting repair planks that are not part of the initial construction process allows us to gain information about the life-span of the ship and, therefore, their sampling should be considered in similar projects on archaeological ship-finds. In this case one repair timber had its earliest possible felling date certainly after the construction date of Doel 1, but in cases where sapwood is preserved, such repairs could provide more detailed insight into the life-span of a ship. It was also observed that these repairs were performed with high-quality timber. Furthermore, in terms of wood grading, the tree-ring analysis of the keel plank also provided arguments for the selection of high-quality timbers for structurally important elements of the ship. For Doel 1, there were no indications that this was also the case for other structural timbers such as the stem and sternpost, but should be kept in mind for future research projects.
A common criterion in selecting timbers for dendrochronological analysis is the estimated number of rings that can be measured on a cross-section. Timbers with at least 60–80 rings are usually chosen for further analysis. However, some of the bottom planks of Doel 1 had less than 50 rings. They were measured anyway since a neat cross-section was already accessible, due to the sawing of the planks in 2000. Despite the limited length of the tree-ring series, these growth patterns had an added value as they helped to exemplify the consequent, symmetrical layout of the ship's bottom strakes. Where dendrochronology is solely used as a dating technique, short series will not be registered; but if a systematic approach in the layout of a ship is to be explored, timbers with a low number of tree-rings should also be selected and included in the dendrochronological dataset. Therefore, the need for a research-question driven sampling strategy is demonstrated here.
Finally, we want to stress that this dendrochronological project benefitted from the fact that many planks and timbers were already sawn into manageable pieces during salvage in 2000. Therefore, on many timbers a clean and easily accessible cross-section was available for the registration of the tree-ring pattern. The number of 110 timbers analysed might not necessarily be taken as a normal sample size required for a comprehensive dendrochronological study of the ship. The tree-ring pattern of many samples did not substantially add to the dating of the ship, or change the interpretation of constructional details, but it did supply a detailed picture of the timber procurement strategy that was used. A minimum of ten dated samples for each group of timbers (framing elements, planks) should be strived for in order to obtain qualitative dendro-provenancing results (see for example Domínguez-Delmás et al., 2013: 134). For example, this is in contrast to the dendrochronological-dating and -provenance analysis of the well-known Bremen cog, which is based on only three dated tree-ring series from the cross-beams (Liese and Bauch, 1965: 41; Bauch et al., 1967: 290). In order to obtain qualitative datasets, one should be able to take many samples from an archaeological ship-find. Sampling is often a synonym for sawing timbers in order to get access to the growth-rings. Taking cores from ship timbers is also a possibility; however, this often gives poor results with waterlogged wood as the cores often disintegrate (degradation of the wood), fall into pieces (presence of cracks) or, if sapwood is present, it is often extremely difficult to keep it intact and attached to the heartwood. Furthermore, in the case of Doel 1, due to the use of flat-sawn planks, taking cores that contain the maximum number of tree-rings is very hard due to the strong curvature and orientation of the growth-rings (Fig. 13a). Therefore, the registration of the tree-ring patterns on waterlogged archaeological material should preferably be performed on cross-sections in order to obtain a dendrochronological dataset with the highest quality standard and scientific value, despite the necessity of sawing through the individual ships' components. On the other hand, in the case of air-dried and archived ship timbers, novel techniques such as X-ray CT can provide a non-destructive alternative for obtaining access to the growth-ring pattern (Bill et al., 2012).