1 The Aanloop Molengat site, research history and techniques
1.1 Discovery, impact and preliminary surveynext section
The Aanloop Molengat site was discovered on July 8 1984, 2.5 nautical miles west of the isle of Texel (53° 3.58’ N / 4° 39.30’ E; WGS84), in the approaches (aanloop) to the Molengat gully that leads to the Texel tidal inlet (fig. 1). When the find was reported to the authorities by Texel-based diver, salver and maritime explorer C.J. Eelman, this started a process that proved critical to the development of archaeological heritage management in Dutch waters. Policy development had been ongoing for several years, but in the absence of clarity in the relative applicability of heritage and salvage legislations, finders considered themselves keepers whatever the nature of the find, but all the more so if valuable metals were involved, as in this case. Lavish and intriguing stamps on the lead and tin ingots found in the wreck convinced the discoverers of the unique historical character of their find. It created an opportunity to settle the issue in favour of heritage policies. A decision was taken to conduct a systematic excavation and there was a political decree to stop applying salvage legislation to heritage, and instead to deploy the protective regime of the Dutch Ancient Monuments legislation both at sea and on land (Maarleveld 1993; 2006). The discovery and subsequent fieldwork and research have thus been vital for the development of underwater archaeology in the Netherlands and for the principle of authorised excavations only. Fieldwork was undertaken in close co-operation with the discoverers, whose maritime expertise and equipment were engaged. Local supporters and a large body of volunteers, wide exposure in local and national media, together with local and international exhibitions lent an air of ‘action archaeology’ to the project (Tilley 1989; Sabloff 2008; Carver 2011), which had a significant impact on perceptions of diving and heritage. It was intended as an example.
Figure 1 The location of the Aanloop Molengat site in the high-energy zone at the entry of the Texel tidal inlet (drawing: Th. Maarleveld (RCE)).
Fieldwork started in July 1985 with a first assessment of the site and the collection of data needed for the further design of the project. This included establishing a provisional rectangular grid of baselines marked by measuring poles at 3 m intervals alongside and 4 m intervals at the ends, recording overall characteristics and dimensions of visible remains, and recording the stratigraphy and the extent of find layers through coring and by excavating a trial trench.
1.2 The site and its main characteristics
The site lies in 16 m of water in a dynamic sand-rich offshore area with significant changes in sediment cover. It is a high-energy zone that is subject to annual changes in the coastal slope just outside the ebb-tidal delta of the Texel inlet (Sha 1990). Seas build up easily due to long fetch for all directions between south, west and north-northeast. Nevertheless, the general appearance of the site is that of a consolidated ‘wreck’ mound in the sense of Frost’s (1962) description of amphora sites in the sediment-lean Mediterranean. This is due to the fact that remains rest on a hard glacial till, which they could not sink into and which was resistant to the scouring that normally occurs around a wreck site in a dynamic sandy environment. The ship sits almost upright and the lower hull is kept in place by the heavy cargo, whereas the sides have been destroyed by biological processes. The fact that the top of the wreck mound consisted of lighter cargo material, packages of leather that were only partly abraded and degraded on discovery probably means that the process of site formation has been an intermittent one, interrupted by long intervals of sand cover. It is assumed that the process of abrasion had just restarted when material was caught in trawling nets in 1984 and discovery ensued. It remains unclear how much of the lighter cargo is missing.
The upper works of the ship are absent. Its remains are likely to have spread over a considerable area. No secondary site was discovered in the vicinity. Apart from a number of heavy, cast-iron gun barrels, the excavation produced few artefacts other than cargo material.
The wreck mound extends over approximately 27 x 9 m and is oriented northwest-southeast. The associated find layer extends somewhat further and was eventually examined over an area of 33 x 13 m (fig. 2 HyperLink Drawing. During the first observations, the bales of leather, or rather the smoothly rounded surface of the abraded vertically placed sheets, were conspicuous at the top (fig. 3). They were firmly concreted to a layer of wrought-iron bars running parallel to them below. Immediately southeast of these, but still on the wrought-iron, rested a tier of broken barrels containing rolls of tin. Lead ingots could be observed below the iron bars. To the northwest of the leather packages, but also on top of the iron bars, rows of cases were in evidence through the cubic concretions of their contents, notably cast-iron cannon balls of various sizes.
A geological profile, cored perpendicular to the main axis of the wreck mound, shows a slight depression in the underlying till and a contaminated layer sharply wedging out away from the mound. It extends slightly more to the east than to the exposed west (fig. 4). A metre-wide trial trench excavated to check the extent of the find layer along the southwestern side produced a limited number of small finds. Hand-drawn profiles and descriptions show the top of the till to be irregularly eroded. It dates to the Drenthe phase of the Saalian glaciation (laagpakket van Gieten), also known as the Borkumriff Formation (Laban 1995). The contaminated find layer is well-sorted, with coarse shingle (and small finds) occurring only in its lower part (and pockets) (fig. 5). This is as much a product of natural reworking as of the repeated removal of topsoil before actual excavation could continue.
Figure 4 Cross-section on the basis of 1985 observations and coring (drawing: Th. Maarleveld adapted by P. Kleij (RCE)).
Figure 5 A sketch characterizing the find layer in profile in excavation away from the wreck-mound (from: Dive „report Seweryn 20-07-1991).
1.3 Logistics, objectives and choice of methods
All on-site work at Aanloop Molengat was characterised by the exposed location close to a lee shore and at two hours’ sailing distance from the nearest harbour. Underwater time was scarce because of the substantial risk of adverse conditions. In view of this, the decision was taken to make the best of the occasional optimal conditions and allow for step-by-step progress in a repeated hit-and-run strategy with a relatively small vessel. This meant accepting slow progress. The approach demanded great flexibility but was realistic in terms of staffing and direct expenditure. It tied in well with the objective of engaging the original discoverers and the local and diving communities, and with using SCUBA as the diving technique. It is a completely different approach than one would choose under time pressure or under development-led circumstances, where it would be appropriate to deploy a large support vessel or platform and be less weather dependent.
With the aim of analysing the packing and stacking of the cargo material, it was considered essential to document the position and orientation of each item in three dimensions. In view of long interruptions to the work, it was not feasible to establish stable, retraceable and reusable reference points on and away from the wreck mound as required for direct survey (Lundin 1973; Adams 1986). Methods of trilateration, or combined measuring and sketching are reliable but time-consuming (Maarleveld 1984). Tidal currents and height differences interfere with direct measuring over distances exceeding a couple of metres. Considering the occasional occurrence of relatively good underwater visibility, it was decided to structure documentation around photography instead. This would fit in well with the hit-and-run operational strategy, and photogrammetry was a developing field that showed great promise (Maarleveld & Vos 1989).
A steel frame of 32 x 11 m was lowered around the wreck mound at the start of the 1986 season (fig. 6). It was oriented along the measuring poles of the 1985 survey and served as an anchor for the shot lines at each corner, as well as for the floating crossbar along which the photogrammetric camera could be moved. The bar was inspired by the system that replaced a photo tower in the Madrague de Giens excavations (Gianfrotta & Pomey 1980), but needed to be much more robust to meet North Sea conditions. Moreover, flashlights were needed and it was decided not to rely on a single-lens camera, but to ensure the creation of a stereo-pair at each shot by using a double-lens camera (Hasselblad/Ocean Optics MC-70).
It was unclear how many layers would need to be recorded and correlated. The depth of the find layer had been established with a handheld Kyholm corer (Nørnberg & Christensen 1980). Probing into the wreck mound was not possible.
The vertical photo documentation included temporary reference points and scales and was supported by distance, size and height measurements relative to a datum point, taken with the underwater height meter (UHM) that had recently been developed (Botma & Maarleveld 1987). Although a pilot study (Hugen 1990) showed that a three-dimensional site model can be created from the documentation (fig. 7), the process of photogrammetric processing proved unduly cumbersome (Vos 2011 HyperlinkVOS). Even on land and in clear water, photogrammetry is not necessarily the most cost-effective documentation method (Reinders 1986). With software that integrates various ways of data capture and 3D modelling, it adds to the archaeological toolkit but is hardly ever a replacement (Green & Gainsford 2003; Sanders 2011). The main problem is that even in non-parametric approaches the accuracy of automated results is fully dependent on the accuracy of the input and the input that qualifies as archaeological documentation needs selective and consistent, time-consuming interpretation. Part of the disappointment with photogrammetry was that the amount of time, equipment and computer power needed for processing was not cost-effective and that outsourcing interpretative phases led to faulty interpretations.
From the start, the vertical photographic pairs were complemented with oblique photographs of each individual cargo item to be removed. Individual items were identified on vertical photographs according to the imaginary grid. This obviously was done on land and in order to reduce precious bottom time. A label and a labelled photo assignment for one or more oblique shots of the same item were then prepared along with a slate with its sketched position, to be taken down on the next occasion. Assignments to remove and lift an object followed, similarly with an identification sketch on a slate (fig. 8).
Figure 8 A typical sketch informing the order of removal in excavation (from: Dive report Vos 22-07-1992).
The find number of each object relates to the first photograph on which it appears. Although in principle this allows for quick reference, it leads to a complicated numbering system, the more so since more photographs were needed for full coverage than anticipated. Also, it was a challenge to keep track of each item’s correct number. A simpler serial numbering system would have been advisable. Such a simple system was effectively used in the registration of small finds that were collected when excavating the find layers away from the wreck mound, both in the 1985 trial trench and in the south and eastern sides, where finds were collected in approx. 2 x 2 m squares (sections A to M, see Hyperlink Drawing).
1.5 Organisation, safety and technical issues
The hit-and-run project ran in tandem with the excavation of the 16th-century merchantman Scheurrak SO 1 (Daalder et al. 1998; Manders 2003) under the assumption that activities offshore could be combined effectively with activities in the Wadden Sea during 14-week summer campaigns out of a fieldwork base on Texel. But calm weather is also required for excavating in the Wadden Sea. The most productive Aanloop Molengat seasons were 1987 and 1992, with 22 and 30 days of work on site. Apart from monitoring missions in 1998 and 2004, work on site was interrupted in 1994, only to be resumed for a two-week campaign in 1999, when a dedicated professional team had been made available (Chart 1). All earlier seasons involved a mixed team, including a gradually increasing core of professional diving archaeologists, a decreasing number of locally hired maritime service personnel and a large number of volunteers, mostly archaeology students and avocational archaeologists – more than a hundred people in total (Chart 2).
Chart 1 Fielddays per season.
Chart 2 Fieldwork man-days per season divided according to staff category. A total of 102 persons took part in the fieldwork operations, spending a total of 1141 man-days on site. 80 % (921 man-days), however, was realized by a core of 28 persons, who each worked more than 10 days on site. 15 of these were employed as staff, 4 were external, each of them engaged from the local maritime community and the 9 others were persevering volunteers.
The frame and its deployment were a solid investment, but proved immensely useful. It served to delimit the site and to provide orientation and a safe and stable working environment in an area with shifting sands and varying surface cover. The camera-navigation bar proved a cumbersome piece of equipment. It needed at least a full diving day to install or remove at the beginning or end of the season. Despite its robustness, it was still subject to swell. It needed a lot of air in its floats to keep it stable, which meant heavy work on the winches to move it around. It was the sort of equipment that assumes priority rather than enabling more important work (Keith 1990). Twice, after a storm, it went missing and was replaced by a balanced, single-diver-operated support for camera and flashlights. A well-trimmed diver obtained the same results with the two-lens camera, with far less trouble. Swell was still a problem, but a running current actually had a stabilising effect.
A safety booth was fitted to the frame, inspired by the ‘telephone booth’ of the Yassi Ada excavations (Bass 1968) (see fig. 6). As diving was organised in untendered SCUBA, it was considered an extra support. Although a fixed feature, it was missing when the site was relocated in 1987. Despite heavy welding and bolts, it had been torn from the frame by a trawl net. It is unclear how much the frame moved accordingly, if at all. Damage to the archaeological material seemed very limited. The booth was not replaced. A spare cylinder- or surface-supplied regulator took its place until a system of through-water communication was implemented for the team. The central safety measure was a strict procedure of planning, checks and monitoring of the diving operation.
It was not just the removal of backfill and overburden of sand, but also the cutting away and removing of heavy netting and other alien material that distracted from archaeological work. The hit-and-run strategy was unsatisfactory for these essential activities. The small support vessel Phileas Fogg did not have adequate dredging and lifting equipment, and valuable time was lost as a result. At a particularly critical moment in 1991 it was therefore decided to call on the assistance of the Terschelling-based Duikteam Ecuador and their well-tried diving vessel Ursus II, which had stronger equipment and a compressor.
1.6 Progress, consistency and an integrated site plan
Conditions were hardly ever as initially expected and support activities such as technical preparation and cleaning before a photo-documentation run took up almost as much bottom time as reference measuring, setting up temporary data points, excavation, and recording, labelling and removal of find material (Chart 3). Data gathering spanned a period of nearly fifteen years but the total time spent on-site at the bottom surface remained limited to slightly less than 970 hours, the equivalent of what a team of six can achieve on a landsite in a four-week campaign.
Chart 3 Bottom minutes per season according to category of activity. Of a grand total of 58170 bottom minutes, 27902 were used in supportive and technical activities, whereas 30268 minutes were used directly in excavation and data gathering. Recording made up for 9834 minutes in total.
It was only after the 1988 season that the first layer was removed. The excavation of find layers adjoining the wreck mound began simultaneously, producing sketches, profiles and small finds that added to the variety of the assemblage. During the 1991 season, disappointing visibility conditions and a sand dune on part of the site led to a reconsideration of the documentation strategy (Briefing report 12 July 1991). Interestingly, this was almost exactly halfway (455 hours had been spent underwater, with another 515 to come). It was decided to remain true to the original recording strategy after the calculation that switching to conventional documentation in sketches or trilateration would take another 400 underwater hours of expert recording, a luxury that could not be afforded, even though more expert staff were available than at the start. In the end, expert recording accounted for only 164 underwater hours in total. Nevertheless, it took until the productive season of 1992 before a second layer could be documented vertically in full. Most of the subsequent removal and lifting of cargo material was finished when the project was discontinued in 1994. The excavation of the southeastern side was finalised in 1999. Although the fieldwork took far longer than the five seasons originally planned, it stayed within the original project design and budget.
In hindsight, the photographic record proves well up to the questions to be resolved. There is enough redundancy in verticals and obliques, and with modern computers and software it has been relatively easy to join the verticals in a 2D mosaic (fig. 9). A more elaborate site plan was prepared independently of this, integrating all types of data gathered in photographs, sketches and direct measurements (see fig. 2). The aim to integrate all observations in a three-dimensional model was abandoned in favour of achieving this result, but it would still be possible to create one if sufficient reason arises.