Revisiting Lomellina, 1516: The Hull Shape

ABSTRACT This paper is a reanalysis of the Lomellina shipwreck (1516), found in 1979 at Villefranche-sur-Mer and excavated between 1982 and 1990, and an attempt at reconstructing the hull shape in light of recent developments in computer graphics and hull reconstruction methodologies. After revising the excavation data, the authors propose a new reconstruction of this unique ship, using off-the-shelf 3D software (Rhinoceros), aiming at understanding its slim hull shape.


Introduction
The shipwreck was found in 1979 at Villefrance-sur-Mer, France, by Alain Visquis and excavated between 1982 and 1990 by Max Guérout, with the collaboration of Jean-Marie Gassend and Eric Rieth, and the support of the Groupe de Recherche en Archéologie Navale (GRAN). Lying at a depth of around 18 m, the site occupied an area of 35 × 12 m, corresponding to the forward bottom of the port side part of a large ship. The excavation lasted a decade, and produced an extensive body of data, most of it published . The excavation and publication were accompanied with models, exhibitions, and illustrations. The excavation started on the stern area of the ship, where only the port side remained.
The Villefranche-sur-Mer shipwreck has been almost surely identified through historical sources as Lomellina, a Genoese merchantman lost while at anchor during a storm, on 15 September 1516 ( Figure 1). There are very few Mediterranean shipwrecks from the early modern period, and Lomellina was extensively preserved, well excavated and documented, and produced a large body of available publications. Its excavation stands as one of very few examples of a 16th-century ship built in the Mediterranean.
Lost in 1516, the year Charles I (1500-1558) became king of Spain, Thomas More published Utopia, and Leonardo da Vinci moved to France with the Mona Lisa painting, Lomellina is a unique shipwreck in many ways. Firstly, it dates to a period of extraordinary development in seafaring skills and shipbuilding techniques. Secondly, it was built in the Mediterranean, the region that inspired the model for three-masted, oceangoing ship of the period of European expansion. Thirdly, it was extensively preserved, it was excavated over a decade and recorded in detail, and the primary data were published and are available online, in several web repositories, and collected at www.shiplib.org.
Given the paucity of information pertaining to the known Mediterranean shipbuilding traditions of the late medieval and early modern periods, we decided to revisit this shipwreck in light of recent developments in computer graphics and hull reconstruction methodologies. Moreover, this ship seems to have been built for speed, even if that choice entailed some loss of cargo capacity, and we decided to clarify certain aspects pertaining to its hull shape, hoping to shed some light into the decision of Lomellina's owner(s) to build such a beautiful hull and the modes of commerce that may have helped them making this decision, and the knowledge necessary to design and build it. Slim, with a wine-glassed midship section, and with strong and heavy scantlings, this ship seems different from the other known ships of its period. The sample of published Mediterranean shipwrecks from this period is small, and neither Mortella III (Cazenave de la Roche, 2020) nor Calvi 1 (Villié, 1989(Villié, , 1990(Villié, , 1991 have similar hull shapes. Italian influence in Spanish and Portuguese shipyards is relevant and welldocumented, and published Iberian hull shapes such as those of the Studland Bay shipwreck (1527), San Juan (1565), N. S. dos Mártires (1606) are never as slim as Lomellina's seems to be (Castro, 2005;Grenier et al., 2007;Thomsen, 2000). Later in the 16th century, even the English hired Italian shipwrights to try to improve the quality of their ships (Glasgow, 1970, p. 10).
This paper uses the extant bibliography about Italian shipbuilding of the period to help the authors develop a plausible reconstruction of Lomellina using off the shelf software (Rhinoceros). The present reconstruction is a working hypothesis that we can hopefully use to discuss and improve in later reconstructions. Reconstructing the past is an iterative process. We will analyse the excavation process and try to identify the main factors that may have hampered the recording of the original shape, we will propose a new approach to the fairing of the lines, based on the runs of both stringers and wales, and finally we will present a new set of lines, which turned out not to be radically different from the previous ones. We would like the present paper to be the first of a series of three dedicated to the Villefranche shipwreck, the first describing our attempt at reconstructing the ship shape from the archaeological timbers, the second de detailing the hull structure and what we know or can infer the entire structure from the excavation. Using the reconstructed structure to calculate the hull weights we would like to test the plausibility of our hypothesis through a basic hull analysis of our model, and a summary study of its intact stability. If possible, we would like to publish a third paper proposing a sail and rigging plan testing the plausibility of our hypothesis through a basic hull analysis of our reconstruction, and a summary study of its intact stability.
In the following section we will summarize the excavation process and explainbased on the extensive body of reports produced during the excavation and the following decadesthe methodology adopted and the potential environmental problems that could not be avoided, such as the degradation of distortion of the structural timbers over time.

In Situ Recording
The excavation of the Lomellina shipwreck site was a herculean job. As a general rule, the more extensive the preserved remains of a ship are, the more complicated it is to clean, tag, measure, and record each structural component. Extensive structures require large budgets, a conservation laboratory, and yearround curation of the artifact collection.
The first problem posed by the study of the Lomellina shipwreck was the size of the area occupied by the remains. The second problem concerned the recording of the preserved structures in three dimensions. The available resources made it possible to organize an annual excavation campaign lasting for about one month. Over nine field seasons the shipwreck was recorded in successive transverse sections with a width of about 4 m. The work started in the stern area of the ship, where only the port side of the hull remained and it was possible to study the planking, frames, and ceiling, to gradually approach the more complex structures of the decks, which needed to be recorded in three dimensions. At this pace it was expected that the excavation would last between eight and nine years, an estimation that proved to be accurate.

Topography
This duration of the work demanded the setting up of a geodesic network based on fixed reference points, strong enough to last throughout the planned decade of excavation. Even though the site was protected by a prohibition of diving and mooring in the area, it remained vulnerable because the Villefranche Bay is busy with both pleasure boats and cruise ships. For example, during the 1988 excavation campaign a pleasure motorboat dropped its anchor on the site during the night.
The network of datum points relied on a longitudinal baseline, an iron cable, set up along the eastern edge of the shipwreck. The baseline was 67.70 m in length and was marked with metric units. It was installed along the eastern edge of the shipwreck, which is roughly oriented north/south, and attached at each end to a concrete block anchored in the sediment. The baseline was equipped with a tensioning system installed on the north concrete block. The tensioning system, with a counterweight, made it possible to keep the baseline stretched in place at all time, Figure 1. Location of the shipwreck in southern France .
providing orientation and a reliable distance scale for measurement, regardless of eventual and momentary displacement of the line by a diver, by the implementation of underwater work, moving equipment, or any other disturbance.

Site Grid
A parallel line, located to the west of the shipwreck, was set up in relation with the baseline, connecting a series of lines that marked the working cross-sections. The grid obtained in this way made it possible to draw a first overall plan of the visible structures during the 1982 evaluation campaign, after the eastern edge of the wreck, between the hull and the first deck, was cleaned ( Figure 2).

Reference Point Network
During the 1983 excavation campaign, a larger network of datum points was set up by triangulation, ensuring a complete covering of the site. All parts of the shipwreck became accessible from multiple datum points, and every feature was carefully triangulated and plotted.

Cross Sections
Based on the fixed datum points, year after year Jean-Marie Gassend (Centre National de la Recherche Scientifique / Institut de Recherche sur l'Architecture Antique, CNRS/IRAA) undertook the rigorous recording of 15 cross-sections, which were adjusted on the horizontal plan in relation to the reference point network, and in a vertical plane in relation to a succession of longitudinal reference lines, horizontally laid, whose level was readjusted step by step ( Figure 3).
Transversal cross sections were thus recorded year after year, and carefully plotted on the site plan. The recording process evidenced probable deformations in the frames, some of which may have suffered some flattening, due to the weight of the overlaying structures ( Figure 4).

Reconstruction of the Hull Shape and Nautical Characteristics
Nautical archaeologists try to reconstruct ships from their archaeological remains, iconography, and coeval technical texts on shipbuilding. Lomellina was a large ship built in the first decades of the 16th century and this fact makes it an archaeological treasure. We know that the carvel-built merchant ships of the Mediterranean influenced the construction of the ships of the North Atlantic and the Baltic Sea and were the models for the ships of the European expansion of the 15th and 16th centuries, but we don't know much about them. As we have mentioned, few have been found and even fewer have been excavated and published. With the Calvi 1   (Villié, 1989(Villié, , 1990(Villié, , 1991 and Mortella 3 (Cazenave de la Roche, 2018, 2020) shipwrecks, Lomellina stands as a rare and exceptional example of 16thcentury Mediterranean ships.
The midship section was reconstructed by Guérout's team from the surviving structure ( Figure 5) and presented an interesting wine-glassed shape, which suggests a search for increased underwater area and better performance sailing upwind.
The interpretation process started during the excavation and guided the research questions. Axonometric drawings were produced and a first attempt at reconstructing the ship shape was developed (Figures 6-8). The site formation process produced a deformation of the waterlogged frames, which were lying under a heavy sediment layer for five centuries, and that was taken into account as the first tentative reconstructions were attempted.
The process of destruction of a ship from the moment it lands on the bottom until it becomes an archaeological site is complex. It depends on the structure of the ship and its progressive deterioration under the action of bacteria and xylophagous molluscs, on the nature of the bottom, and the environment in which it degrades: currents, waves, human actions like fishing or dredging. It is fair to hypothesize that Lomellina landed on the bottom inclined by about 45°on its port side. It undoubtedly sank a little into the sediment formed of silt and sand. According to Don Antonio Beatis (1913), who wrote the journal Voyage of the Cardinal of Aragon a little more than a year after the shipwreck, on 30 November 1517, the top of Lomellina's mainmast was still emerging 2 rods (about 2.5 m) above the water surface.
When the beams of the first and second deck give way, the port side, still at the top, undoubtedly bent outward despite the resistance of the transversal structures that constituted the decks: frames, beams and planking. The strength of the orlop has probably ensured the rigidity of the lower part of the hull, probably already embedded in the sediment. Another  factor in the deformation process may also have been the varying resistance of the sediment along the length of the vessel.
The first attempt at fairing the sections recorded and reconstruct the hull shape was done by Jean Thomas in 1991 ( Figure 9). His reconstruction gave the researchers a good grasp of the hull shape and suggested a shape markedly different from that of coeval merchantmen of the Atlantic, and in particular those built in the Iberian Peninsula. Lomellina was long and elegant, had a deep keel, and was probably a fast sailor.
The second reconstruction was developed by Roberto Greco, a professional industrial model maker and a former shipyard manager, who drew up a layout plan for Lomellina. The central part of the model is based on the data recorded in situ, but the hypothesis concerning the slenderness of the bow  retained by Roberto Greco, which is too low, is perhaps less consistent with the known data for the 16th century standard (Figures 10 and 11).
The third, virtual, reconstruction was part of a project initiated by the authors at the ShipLAB, Texas A&M University, with the help of several students, in an attempt to restore the forms and nautical characteristics of Lomellina, starting from the initial surveys carried out at Jean Marie Gassend, and directed by Max Guérout. Although we did not achieve a considerably different model, this third model is more elegant and has a higher length to beam ratio.

Units of Measurement
The units of measurement used in Italian shipyards in the beginning of the 16th century varied and obviously depended on the care taken when the gauges used by the shipwrights were cut and marked. For this project we have used the values generally adopted for the units of length in Genoa and in Venice (Lane, 1966;Pegolotti, 1936): . Genoa: Goa = 0.744 m; Palmo = 0.248 m.
Although we know that these units were not applied to the millimetre when shipwrights were sawing a large timber, it is necessary to use a set of reference values as the base of any attempt to reconstruct ships from their shipwreck remains.

Fundamental Dimensions
The fundamental dimensions defining the shape and size of a ship are generally ten: . the length of the keel; . the height and rake of the sternpost; . the height and rake of the stem; . the length from head-to-head, or length overall; . the beam at the master frame; . the depth of hold at the level of the main section; . the entries and the runs.
These data are usually considered when we want to compare ships and appear to be sufficient to determine  the proportions and tonnage of most ships from the early modern period. They are, however, insufficient to fully restore the hull shape of a vessel. In addition to the entries and runs, to reconstruct the shape of a hull it is necessary to know also: . the heights between the decks; . the longitudinal position of the master frame and the tailframes; . the dead rising of the floor timbers; . the sheer line; . the length of the upper deck, its profile from stem to stern, and its height at the master section level and on the posts.
In the following pages we present the known or estimated main dimensions of Lomellina, and the methodology adopted to reconstruct its hull shape and size.

Dimensions of the Keel
The length of the keel was only known after the ninth excavation season. It was measured with good accuracy and Guérout's team estimated its length on the ground, i.e. including the flat parts of the bow and stern knees attached to the keel timbers, at 33.38 m, plus or minus 2.5 cm. The known scantlings are presented on Table 1.

Keel Section
The keel was 28 cm sided and 42 cm moulded. The four elements of the keel were fastened with three flat horizontal scarves. Throughout the ship's structure, shipwrights used flat, hooked, and dovetail scarves in the assemblage of timbers.

Stern Post Section
At the stern, the keel section turned upwards, forming a knee that continued as the sternpost, which was 28 cm sided. Due to erosion, the moulded dimension  of the upper arm of the stern knee is unknown, as well as the length of its upward section.

Stem Section
Like with the sternpost, the forward section of the keel turned upwards, forming a bow knee. The length of the curved portion the keel, which forms the lower portion of the stem, was 2.30 m. The stem was 28 cm sided and 65 cm moulded. It was connected to the keel with a flat horizontal scarf.

Rake and Height of the Sternpost
The angle between the horizontal base line and the sternpost was close to 77°. To know the height of the runs we needed to know the height of the sternpost at the second deck. This height was deducted from a preserved knighthead. The discovery of the halyard knighthead, situated near the maststep, allowed the excavation team to estimate the height of the deck in the area of the halyard. It was found dismantled in three intact elements and after reassembly measured 8.22 m in height ( Figure 12). As the knighthead foot rested on the keelson, its upper part stood at a height of 9.34 m above the lower level of the keel. That is the value we get when we add the height of the knighthead (8.22 m) to the moulded dimensions of the keel (42 cm), the floor timbers (45 cm) and the keelson (25 cm).
The upper surface of the deck was 30 cm below the lower level of the knighthead's sheaves, 1.50 m below the upper surface of the halyard head, at 7.84 m above the lower face of the keel.
The height of the second deck at the posts was deduced from the sheer line of Theodoro de Nicolo's galleon, also at the second deck. In his drawing the sheer is pronounced and equal to the spacing between two bridges, i.e. approximately 2.50 m. This value gives us a height at the sternpost of approximately 10.34 m.

Rake and Height of the Stem
We traced an approximate shape of Lomellina's stem from Théodoro di Nicolo's -Instructione sul modo di fare galere (Lane, 1934)drawing ( Figure 13).
In our model, the height of the sternpost is 11.5 m, and the height of the stem is 10.70 m, considering that the stern is a little higher than the bow.

Runs
The runs were a complex surface that transitioned the section on the aft tail frame to the straight line of the stern post and aimed at ensuring a laminar flow along the sides of the vessel to optimize the rudder function. The height of the runs is often given in contracts and after stern panels became common, around 1500, this height was assumed to be the height of the point where the fashion pieces met the stern post. In Pre Theodoro's galleon this height is indicated as 13 piedi (4.51 m) and it is the point where the lower deck meets the sternpost.
We deduced the shape of the runs from the surviving sections, assuming Lomellina had a round tuck. It is difficult to reconstruct the ship's stern shape, and we decided to base our proposal on the shape of the Mataró model, made about 50 years earlier.
We believe that the sternpost extended 2.50 m abaft, measured from the after end of the keel to the vertical of the top of the sternpost, which we located 11.50 m high, for an inclination of the stern post of 77°. This value is more pronounced than the inclination of the sternpost of Theodoro's galleon, which is 70°, but was based on in situ measurements of the angle of the sternpost. Because the ship had a round tuck, the definition of the upper limit of the runs' surface is difficult to define.

Entries
As the bow was not completely preserved the method used to design the runs also allowed us to obtain an approximate value of the entries. We considered the stem to be an arc of a circle with its centre located at the level of the second deck and obtained 10.25 m for the rake of the stem. The examination of the data provided by the Genoese construction treatises (Gatti, 1975), allowed us to add some additional information. Most documents referring to shipbuilding in the region of Genoa present the main dimensions of the ships were expressed in goas (0.744 m) and palmi (0.248 m), whereas in Venice and Ragusa they were given in feet (0.347 m).
The dimensions obtained from Genoese contracts published by Luciana Gatti are shown in Table 2.
These measurements are approximated and indicative. Salme were a measurement of capacity equivalent to circa 0.24 tons. Thus, in contract n°2, the 3,000 salme are equivalent to 720 tons, and in contract n°14, the 2,000 salme are equivalent to 480 tons. In the contract for the construction of Santa Maria, however, the document mentions 2,300 salme or 750 botte, an equivalence that suggest a value for the salme of around 0.26 tons.

Length Overall
Overall length in Genoese contracts was referenced as roda in roda length, in English 'head-to-head' (Gatti, 1975). It was, with two known exceptions, expressed by a round number of goas. The length of the keel was only indicated in two of Gatti's contracts, respectively, numbers 14 and 22. This particularity has been noted by Furio Ciciliot (1998, pp. 27-30) who underlines: 'We could, for example, postulate the systematic use of the Genoese goa … to measure the length of the hull'.
For our reconstruction we considered Lomellina's length overall expressed as a round number of goas.  The length of the keel on the ground, measured in situ, was close to 33.38 m. The rake of the sternpost was estimated at 2.50 m on the point where the sternpost meets the transom or the main whale (Figure 22), in our case 11.5 m high for an angle of inclination of the sternpost of approximately 77°. The rake of the stem was estimated at 10.25 m. Based on these values we deduced an approximate value for Lomellina's overall length of: . Length of the horizontal keel = 33.38 m . Rake of the stern post = 2.50 m . Rake of the stem (estimated) = 10.25 m __________________________________ . Total = 46.13 m or 62.002 goas We assumed that the 'roda in roda' length of Lomellina could be 62 goas, or 46.12 m. In this hypothesis, the length of the entries would be 46.12 -33.38 -2.50 = 10.24 m. This approach gave a result very close to our evaluation and seems to suggest that the fundamental measurements calculated are consistent with those of the Genoese construction contracts available.

Beam at the Master Frame
Because the sections seem to have flattened through time and the reconstruction of their shape is difficult, we have rotated some of them around the turn of the bilge point to try to get a fair and plausible shape. The values obtained for the maximum beam vary between a minimum of 12.15 m and a maximum of 12.5 m, which correspond to length to beam ratios between 3.69 and 3.8 m, perhaps unexpected for a three-masted ship of the 16th century, mostly if we consider the known examples for the Iberian Peninsula and Northern Europe. Even when we consider the length to beam ratios indicated in Gatti's work (Table 1), we get a less slender hull than the one proposed for Lomellina (Table 3), which is only close to those of the lembo and the galeone, both oared vessels.
In any event, these measurements and proportions are not markedly different from those indicated by Pre Theodoro to the large galleon (Lane, 1934), or in a ship built in Ragusa in 1520 (Gatti, 1975), as indicated in Table 4.
If we compare these values with those of Iberian ships, we get the same sense: Lomellina was probably built for speed, even if this choice of shape sacrificed a little bit its cargo capacity. The most important body of data collected, to our knowledge, is that of maritime historian Carla Rahn-Phillips, who collected nearly 200 values of ship's dimensions from treatises, technical texts, and contracts, dating from the late 16th century to the early 18th century, and found length to beam ratios for cargo ships in the period 1575-1611 varying between 2.98 and 3.75 (Rahn-Phillips, pers. com., 2002). The main values obtained in the present reconstruction are indicated in Table 5.

Tailframes
The rising and narrowing of the pre-designed frames was deduced from the sections recorded, the systematic measurement of the floor timbers' rising on the centre line, and the narrowing on the turn of the bilge, considering the room-and-space. Based on these dimensions the excavation team placed the master frame on or near frame W59.  Floor timber W59 was assembled to two first futtocks on each side, and it marks a change in the direction of the assemblage of the first futtocks on the ship, two features that often indicate the presence of a master frame. Abaft W59 the futtocks are fastened to the after face of the floor timbers, and before to the fore face of the floor timbers. Based on these features, frame W59 was considered the master frame. Another floor timber, W70, was also assembled to two futtocks on each side. Its position in relation to the ceiling planking suggested that this frame was the stern tailframe.
In our model the master frame is therefore situated 20.9 m forward of the point where the keel meets the sternpost, and 4.2 m fore of the middle point of the keel. The ties between Portugal, Spain, and Genova are known, and it is interesting to note that this value is rather close to the one indicated in the Portuguese treatise Liuro da fabrica das naos, by Father Fernando Oliveira (1991, p. 116), who indicated that the master frame should be placed forward of the centre of the keel by a value equivalent to 1/8 of the keel length, in this case 33.38 / 8 = 4.17 m.

Hull Shape
As mentioned above, we see a certain similarity between Lomellina and the large Venetian galleon described by Pre Theodoro de Nicolo, which is a warship. Even though at the time warships and merchant ships seemed to have been morphologically similar, warships were expected to be built with sturdier scantlings. For instance, the recommendations made in 1540 by the Grand Master of the Order of Malta concerning the construction of the ship Santa Maria (National Library of Malta, AOM 417 f°s 239-242v) suggest that such a warship was indeed framed in a stronger way, when we compare its values with those of Lomellina's. The document mentions a ship of 750 botas or 2,300-2,500 salmas and that the planking from the keel to the main whale should be between 12 and 17 cm thick, and the frames 25 cm (1 palmo de canna) on a side. When we look at these values in Lomellina it looks like it was built as a merchant ship. The differences concern not only the scantling of the frames but also the extent of the overlaps between floors and futtocks. The document urges the shipwrights to be generous on the overlaps between timbers such as the frames and the whales.

Height of the Decks
As it was also mentioned above, the discovery of the knighthead of the main halyard allowed the team to deduce the height of the second deck at the level of the midship frame (Guérout et al., 1989, pp. 81-90). At the place of the knighthead, on the ship's centre line, the weather deck was located 7.42 m (2.45 + 2.45 + 2.52 m) above the upper surface of the keel ( Figure  12). The levels of the first deck and the orlop were deduced from the archaeological remains, considering a deck camber of about 35 cm amidships (2.8%), which makes the values plausible and compatible with the runs of the clamps, as we will see further on.

Load Waterline
An historical reference in the ship Grande Maîtresse inventory, dated to 5 September 1526, presents an indication about the position of the load waterline: 'the ship named Saincte Marie, or Grande Maîtresse, in all goodness, freshly careened, is lined with lead up to the wale, and above the water [nailed with bronze]'. The word 'redon' in the manuscript can be translated as 'first wale', and since this reference is about the hull lead sheathing, it probably refers to the lower wale (Guérout & Liou, 2001). We made a simulation of the load waterline values and obtained the drafts indicated in Table 6.
Referring to the sheathing, the 1986 field season report mentions: 'Several observations were taken to determine the height reached by the lead lining, and a trench was dug under the hull. The last strake lined with lead seems to be B26, from which a fragment taken does not bear any trace of doubling but characteristic square nailing marks' (Guérout & Liou, 2001, p. 9).
The first wale shown in sections 1 and 3 corresponds to strake B28, a feature that matches closely the description of Grande Maîtresse except for strake B27. The waterline was probably a little below the level of the first deck, and undoubtedly below the lower level of the main wale. This level is confirmed by the position of the bilge pump discharge pipe, which is located a little above the level of the first deck and must obviously be above load waterline ( Figure 14). We propose that the load waterline was located 4.257 m above the lower face of the keel at the lower wale's lowest point.

Deck Camber
To assess the height of the decks from the height of the stringers we defined a value for the camber, or transverse curvature of the decks, fixed at 2.8% of the beam at any point, as mentioned above. The height of decks is represented along the longitudinal axis of the ship, at the highest point of the deck beams, but the archaeological record only allows us to measure the hight of the decks on the clamps. The data relative to the height of the deck clamps do not appear, to our knowledge, in any document dating from the 16th century.

A Third Reconstruction of Lomellina
The third reconstruction of Lomellina proposed here was based on the values presented and justified above and is, as the previous two, a working hypothesis. It serves as a reflexion of the reconstruction of ship's hulls from their archaeological remains. Archaeology is an iterative process. As we learn from new case studies, we understand better the options we are faced with when trying to reconstruct a ship lost more than 500 years ago, collapsed, and flattened by its own weight combined with the weight of the overlaying sediment, then recorded manually, with the unavoidable errors committed by archaeologists, or resulting from the inherent inaccuracy of the tools.
The aim of this reconstruction attempt was to produce a set of lines drawings from the shapes of the surviving frames. The methodology employed in this attempt consisted of positioning the cross sections over the straight keel and try to deduce the runs of the wales and stringers placed at heights compatible with the position of the decks.
Only the eight sections still attached to the keelcross section CS6 to section CS13were used to produce a set of lines. Cross sections CS5, CS4, CS3, and CS2 were only used at the end of the reconstruction process, to test whether the aft part of the reconstructed ship matched the shape of the original frames.

Working with Rhinoceros Software
In this third reconstruction we used a traditional approach, designing a 3D model developed with Rhinoceros software and analysing its plausibility with a traditional set of lines drawings, composed of a sheer view, a half-breath view, and a body view, representing the bow and stern on each side of the centre line.
Lomellina's profile originally drawn by Max Guérout was modified several times, through an incremental and iterative process: the decks have been lowered so that the clamps appear in the desired position. The method used is described in the following pages, and encompassed ten iterative steps: (1) Aligning the sections recorded in situ along the centre line and spaced them according to the archaeological record; (2) Aligning the stringers and clamp (V20, V22, and S3) to try to get a best fit; this process was iterative, and entailed moving slightly both the sheer lines and sections; (3) Rotating the sections in order to level the points of the deck clamps, to get a better fit before modifying the shape of the frames; (4) Using the 'Bend' command to reshape the frames in order to get a fair set of sections in the body plan of our proposed lines drawings; (5) Plotting a set of waterlines from the faired sections; (6) Comparing the waterlines obtained with those obtained in the previous reconstructions; (7) Developing a stern shapea round tuckfrom the Mataró model (ca. 1460), which was extended circa 30% in length to match Lomellina's dimensions ( Figure 15); (8) Fairing of the waterlines through an iterative process, in dialogue with the best fit shapes of the sections; (9) Readjusting the position of the frames that were not attached to the keel and assess their shape against our hypothetical model; (10) Correcting the original sections one last time.
The present reconstruction is therefore essentially based on the nine cross sections where the frames were still attached to the keel. As mentioned above, the four cross sections where the frame was not attached to the keel, located at the stern, were only used at the end of the reconstruction process, to check if the shapes of the stern were plausible.
The first step of the reconstruction process consisted of modelling the frames and placing them over the keel in their original positions. The frames were 3D modelled from the 2D drawings taken during the excavation; they were modelled as single piecesnot as floors and futtocksfor two main reasons: because at this stage of our project the aim of the reconstruction process was to produce a set of lines, not a realistic 3D model, and because it was not always possible to clearly distinguish which futtocks were associated to each floor timber in the drawings. Stringers V20, V22, and S3 were also represented on their respective positions on each frame (Figure 16).
After laying the frames on the keel (Step 2) it was clear that the central part of the ship was severely distorted, in particular cross sections CS8 and CS9, which were the two closest frames to the probable position of the master frame ( Figure 17). Moreover, the sided dimension of the keel was about 28 cm, an insufficient surface to assess the inclination of the keel with high accuracy. In addition, the structural elements that make up the bottom structure had irregular dimensions and the differences were often levelled with wedges. Pitch was systematically poured to ensure that water would not percolate between the various elements, and its presence over the upper surface of the keel increased the difficulty of measuring its inclination due to the listing of the ship.
All the difficulties encountered during the recording of the ship sections determined the need to realign them by rotating each one in the transverse plane around the keel (Step 3).
In this process, the height of the decks and their longitudinal profile were of primary importance: they were our first boundary conditions and determined the reconstruction process to a large extent.
Step 3 entailed the alignment of stringers V20 and V22, and clamp S3 on each frame, where V20 and V22 were the stringers that ran respectively below and above the beams of the orlop deck (the faux pont) and S3 (serre) was the clamp upon which the beams of the first deck lay.
Ideally V22 should run right above the line of the orlop deck, which represents the top surface of the deck beams (the baux). V20 should run 10-15 cm below the same deck line, and S3 should run around 15 cm below the line of the first deck.
We used the drawing in Figure 18 as an initial guide to determine the shape and position of the deck lines, as well as the position of the stringers. The drawing was made by taking the heights of   the decks at the mainmast as fixed points. We considered the height of the first deck, measured from the upper surface of the keel, to be 2.45 m, the height of the second deck 4.9 m, and the height of the third deck, the weather deck, is 7.42 m. These are values measured on the site. As mentioned above, the height and rake of the posts were taken from Pre Theodoro's treatise (the stem is 12 m high and rakes 10.25 m while the stern is 11.5 m high and rakes 2.5 m).
The sheer line and deck lines and height of the posts were modified many times as the work proceeded, while the height of the decks at the mainmast remained fixed. The adopted camber of 35 cm implies that stringers V20, V22, and clamp S3 run 35 cm lower on the hull sides than the deck line, which is represented in the drawing above. In the drawing below the stringers are represented with a camber of 35 cm, while the red lines represent the deck lines at the centre of the ship, at their highest point.
Ideally, after rotation, stringer V20 and clamp S3 should run around 35 cm below the respective deck lines, as explained above. But looking at the profile view it seems that the first deck line is too high on frame CS12 in relation to the position of the top clamp S3. We believe that the bow frames may have been heavily distorted above stringers V20 and V22, since the deck knees were not preserved, and the futtocks lost their rigidity.
Steps 4-10 were iterative and aimed at obtaining the most plausible shapes of the archaeologically recorded frames that yielded a fair set of lines. As expected, the stringers did not match the position of the decks, in particular in the central part of the ship, which was probably the most deformed one. When the sections were recorded the original vertical line was difficult to establish, mainly due to the condition of the upper face of the keel, which was narrow and irregular. After rotating the frames, pivoting them on the keel, we got a fairer line.
It is also possible that the sheer of the decks was not as pronounced as in Theodoro's treatise and in some of the iconography we consulted. Since the frames could not be rotated further to raise the position of the stringers and clamp, and thus bring them closer to our theoretical deck lines, a rotation that would make the ship too narrow at the bow, we addressed this problem by slightly bending frame CS12 inwards, a solution that raised the position of clamp S3.
This operation (Step 4) helped us reshape the sections in order to obtain fair waterlines. In Figure 19 section CS12 was plotted against CS11, its closest frame towards the centre of the ship, and we can see that the upper part of frame CS12 is bent outwards and doesn't match the shape CS11 ( Figure 20).
As we plotted the first set of waterlines, we found the archaeological data much more in line with our theoretical fair hull shape and proceeded with the fairing process.
We were careful to implement the fairing process step by step, always in dialogue with the original section shapes in mind, and we averaged the values modified in the process to minimize the cumulative changes accrued in the iterative process. To obtain the final full set of lines we kept in mind both Roberto Greco's lines drawings ( Figure 10) and the set of lines drawings of the Mataró model stretched by 30% ( Figure 15). More specifically, we combined and adapted these two sets of lines with those obtained from Lomellina's frames.
Once we had a plausible set of lines for the lower hull, we tried to reconstruct the upper part of the frames. After looking at Pre Theodoro's treatise and at the iconography of contemporary ships, we decided to give the frames some tumblehome. The extent of the tumblehome that we gave is arbitrary. After fairing all the lines again and obtained a final set of lines. The result is shown below, on Figure 21.  The last operation was to place the frames that were not preserved down to the keel against our final lines drawings and see how close they matched our hypothetical model. The result was quite satisfactory (Figure 22).

Conclusions
The present paper is intended as a summary of what we know about the ship Lomellina and an attempted reconstruction of the hull shape based on the archaeological data, the iconography of the period, and what we can infer from a small number of technical texts from this period. It describes the excavation and recording process, discusses the difficulties encountered by the excavation team over the years, and proposes an interpretation of the discrepancies encountered when we tried to reconstruct the ship shape from the sections and longitudinal measurements.
We cannot emphasize enough the fact that this third reconstruction is just that: a third attempt at understanding Lomellina. With the data preserved on the shipwreck Max Guérout's team managed to produce three possible hull reconstructions, three hypotheses of what this amazing machine may have looked like.
The planned next steps will address other, perhaps more interesting questions, pertaining to the way this ship was designed, built, sailed, inhabited, and eventually lost. Archaeologists reconstruct the past from truncated remains of human activity through an iterative process and only further research will provide data to evaluate the reconstruction proposed in this paper: was the hull shape plausible, is the righting moment within known practical values, are the ballast and cargo necessary  to obtain the projected load waterline realistic, and so many other questions.
As mentioned in the abstract, we are planning to publish a second paper with the ship's hull structure reconstructed and the intact stability analysis, and a third paper proposing a sail and rigging plan. And we would like to reunite the tremendous amount of data already published from this excavation in a monography that complements the Archeonautica publication and makes justice to the tremendous effort that this archaeological project represents.
From the beginning, we did not consider applying arcs to the sections recorded in situ. Because we knew that the sections were deformed, we thought that to try to fit circular arcs to the transversal sections of the ship would probably land an extra biased look to our reconstructions. Pre Theodoro's documents suggest that the transversal moulds were designed with battens and offsets, and we assumed that the moulds used to cut Lomellina's frames were obtained in the same way. When we look at the central frames, which we believe to have been predesigned and cut from a small set of moulds, we find that it is not difficult to imagine a futtock arc with a radius of around 6.5 m as the best fit circular curve obtained in this reconstruction.
We would like to leave here our homage to Patrice Pomey, who constantly revisited his excavations and reanalyzed them in light of the new information and new research tools available in the present. We believe that one day nautical archaeologists will have access to the visual tools being developed in and around Hollywood to represent, measure, and evaluate imagined artifacts, environments, and situations, and that revisiting old data from early excavationsor even destructions perpetrated by treasure hunterswill yield new clues on the past and allow specialists to present our reconstructions in the present digital 3D paradigm, where recording, interpreting, representing, and reconstructing is much easier and produces more eloquent representations. Our follow up paper on the structure of Lomellina we will try to gather information about the conversion of the timber and the morphology of the trees utilized in the ship's construction.
We take this opportunity to indicate the extensive bibliography available about this shipwreck and invite the interested specialists to revisit it. There are almost no studies of Mediterranean round ships from the early modern period and Lomellina is a wonderful example of one of these amazing, inhabited machines.