How often is a completely unknown Roman legionary fortress found? Very rarely, we would say. Yet this now appears to be the case at Valkenburg, in the Netherlands, where a major military base has been discovered. And what a base it is! This article describes step by step how this important new find came about, thanks to research at a former naval airbase undertaken by two Dutch archaeological companies (ADC ArcheoProjecten and Archol).
Excavating Valkenburg
Valkenburg is already renowned for its Roman archaeology, thanks to a military base that was excavated by A E van Giffen during and just after the Second World War. This fort held an auxiliary unit made up of soldiers recruited or conscripted from the occupied territories, who supported the Roman citizens serving in the legions, which were housed in much larger legionary fortresses. Valkenburg fort was made famous by the fantastic remains of timber and stone structures, which were preserved in the ‘wetlands’ of the western Netherlands. On the basis of the Tabula Peutingeriana – a medieval map believed to be copied from a Roman original – the fort may have been called Praetorium Agrippinae. Whatever the case, plenty of archaeological research has taken place in the vicinity of this fort over the last few decades. The investigated areas have toponyms like Veldzicht (investigated in 1994 and 1996-1997), Marktveld (1985-1988), Woerd (1972 and 1988), and Weerdkampen (2019), and they all lie to the south of the fort site, near the modern village of Valkenburg. Not all of the data from these explorations have, though, been scrutinised to the last sherd or soil discoloration, and in some cases work has largely stalled.
In 2010-2011, numerous trial trenches were opened on the western side of Valkenburg village, within the former naval airbase, and the results were published that same year. Various Roman structures were encountered during this work. Most striking was the combination of a number of V-shaped ditches, a dam or road, and a possible tower. One of the wooden posts supporting this apparent tower was set on a large block of wood that can be dendrochronologically dated to AD 39±6 years: a spectacular result. As we will see, quite a bit of Roman military activity was under way in the wider region during that period, so the date – together with the other finds – prompted much speculation. We believe this speculation has now come to a definitive end: the remains of a Roman legionary fortress have been found at the former naval airbase. This allows the history of Roman Valkenburg, and indeed that of the early Roman Netherlands, to be rewritten. Along the way, a gratifyingly large number of loose ends left over from earlier work can also be satisfactorily tied up.
Following the 2010-2011 work, it was suggested that the structures belonged to a temporary camp established during the reign of the emperor Caligula (37-41). The ancient literature records that Caligula planned to campaign in Germania when he left for the Rhine front in AD 39. We do not know his exact destination, but Mainz seems a likely candidate. Be that as it may, Caligula spent the winter in Lyon (Lugdunum), after which he apparently had a change of heart and Britannia hove into sight as a choice candidate for conquest. Ancient authors such as Suetonius and Cassius Dio describe how this desire descended into a bizarre farce after Caligula lined up his troops along the shore of ‘Oceanus’. The emperor briefly sailed forth in a ship, before returning and ordering his troops to collect seashells. This booty, which Caligula claimed as spoils of his triumph over the sea, was eventually taken back to Rome.
There is no call to rehearse all the detail of this extraordinary escapade here, but archaeologists have not ruled out the possibility that this event took place on the Dutch coast near Katwijk. Recent decades have furnished a number of suggestive finds, including the regular recovery of both struck and unstruck coins of Caligula, as well as barrels from Bunnik-Vechten and Valkenburg that bear the emperor’s official title. The ancient historian Tacitus also writes about a member of the Cananefatian group in AD 69 called Brinno, whose father had scorned Caligula’s campaigns with impunity. As we know that the Cananefates later inhabited this part of the North Sea coast, it might again suggest Caligula’s presence in the vicinity. None of this can be considered hard proof, of course, and some scholars favour different settings, but these considerations certainly fuel speculation about the location of these historical events.
Valkenburg revisited
It was against this backdrop that follow-up research at the naval airbase began in 2020. The plan was to learn more about the V-shaped ditches, in order to determine the size of the possible temporary camp. Connecting the known stretches of ditch suggested a large area was enclosed, but questions remained about whether all of these elements belonged to a single fortification. Our excavation trenches were laid at right angles to the ditches, in order to secure good cross-sections across them. As work progressed, we occasionally encountered what seemed to be structures to the east of the ditches. One example was a set of posts creating a structure that covered an area roughly 10m by 13m, positioned parallel to the ditch. This was surprising, to put it mildly, as temporary camps typically comprise little more than a bank and ditch, with the soldiers living in tents.
Alarm bells really started ringing when cut timber beams about 3m long were found lying side by side and running for a length of 15m beside the ditch. Within this timber strip, we also encountered the heads of two posts that had been sunk vertically into the ground, with the post-holes for two further timbers lying just to the east. Together, these four elements formed a perfect square with sides 3.3m in length. Given the close association between the ditch and this structure, it must surely be a tower. Once this conclusion was reached, it became highly suggestive that our strip of wooden beams is virtually identical to another set excavated by Van Giffen in the Period 1 Valkenburg fort. He interpreted those timbers as a foundation platform for the rampart, the turf sods of which were still recognisable.
Given that our findings can be plausibly viewed as part of a Roman defensive system, the natural next step was to determine the size and nature of this installation. First of all, it is clear that it is not a temporary camp. Building a turf rampart on a beam foundation with wooden towers required too much effort for a fortification designed to be of fleeting use. This view is bolstered by some results from earlier excavations, including the discovery of two posts, which can now be recognised as another (half) tower, something that was previously suspected. In other places, stretches of more poorly preserved timber strips could be attributed to a rampart with some certainty. Indeed, when taking into account the possible tower found during the 2010 trial trench survey that gave the AD 39±6 date, there were three elements in a row over a distance of 180m, without a bend or corner in sight. When the average length-to-width ratio of a Roman fort along the Rhine in the Dutch part of Germania Inferior is considered (100m by 150m), it becomes apparent that something special is going on here.
Back to the drawing board
Armed with this knowledge, the trial-trench survey was reanalysed, which prompted the suspicion that the dated possible tower might be part of a gatehouse. With that, the hypothesis was set: we were looking at the western stretch of a Roman defensive system incorporating a gate, with two towers to its north. The broadly predictable nature of Roman defences means that there would have been towers to the south of the gateway as well, allowing a total length of over 360m to be reconstructed for this stretch of rampart. Support for this measurement came from the identification of the V-shaped ditch 170m south of the gateway, although no trace of the rampart was detected there. The mental step needed to envision a legionary fortress was quickly made, and a comparison with other such bases carried out to determine what sort of size might be anticipated.
It must be stressed that the average footprint of a legionary fortress differs by period and by place. The Augustan fortifications at Nijmegen in the Netherlands, for instance, as well as those along the Lippe River in Germany at Haltern, Oberaden, and Anreppen, are relatively irregular in shape. They also sometimes exceed 40ha in area, because they were designed to accommodate two legions at once. Other fortresses, including Inchtuthil in Scotland, Caerleon in Wales, Carnuntum in Austria, Novae in Bulgaria, and indeed Nijmegen, are significantly smaller, coming in at around 20ha.
Ensuring that the gatehouse hypothesis checked out was an essential step. Sure enough, re-excavation and extension of the original trial trench exposed the traces of a timber gatehouse consisting of two inwardly projecting towers and a double doorway; the entire structure measures about 24m in length from gatetower to gatetower. It seems most probable that this served as the porta decumana or rear gateway leading into the fortress. On the strength of this, measurements of at least 400m by 400m could be reconstructed for the fortress. Once again, support for this comes from reconsidering the results from some earlier excavations in Valkenburg, especially those at Marktveld.
If you draw a perpendicular line running east for 400m from our west rampart, you reach an area of Marktveld that has been excavated. Various fascinating discoveries were made during the archaeological work there, including a cemetery containing over 600 individuals, a small military post known as a fortlet, watchtowers, some farms, and two Roman roads. Among these features, the roads are most interesting for our purposes, as they crossed the site from north to south – much as the fortress rampart must have done – and ran parallel to the course of the Rhine. The latest road was built using oak, which can be dendrochronologically dated to AD 124-125, placing it within the emperor Hadrian’s reign (117-138). Rather than following a straight line this road curves; it had also been repaired, probably following damage caused by flooding.
The oldest road recognised at Marktveld was rather different in character. It ran in a straight line along the far eastern edge of the excavated area. According to the excavators, at least 45m of its route was represented by a timber bed, which they viewed as the possible remains of a foundation for a raised embankment the road ran along. The archaeologists noted, too, that this style of construction shared similarities with the Period 1 fort rampart examined by Van Giffen. This is indeed the case, and our work now suggests that this structure was not originally built as a foundation for the road. Instead, we believe that the Marktveld excavators unknowingly cut through what was originally the eastern rampart of the legionary base, before being later reused (c.AD 80-85) as an embankment for the oldest road known at Marktveld.
Various factors support this reading. Both the method of construction and the orientation of the Marktveld structure are identical to our western rampart. Indeed, even the dating of the apparent Marktveld rampart seems to be the same, as some of the wood has been dendrochronologically dated to AD 39-40. A case can also be made for two further ‘half’ towers having been excavated but not recognised; in both cases, the missing portions were destroyed by erosion caused by the Rhine. Such loss could also account for the absence of other features that cannot be identified in the excavation documentation, such as a gateway.
Even so, if it is accepted that the early Marktveld road started life as a rampart, more pieces of the puzzle fall into place. For example, there are two large military-style granaries at Marktveld, which have never been properly understood. Such storage buildings were usually placed securely within a defended perimeter, but the Marktveld granaries seemingly lay unprotected outside any known military installation. Even more curiously, granaries in the Valkenburg auxiliary fort are largely conspicuous by their absence. It has been calculated that the Marktveld storage buildings would have been sufficient to supply approximately 1,000 men with food for one year. Our legionary fortress hypothesis now places these granaries inside a defensive circuit. What is more, they are situated in the right place relative to where the gateway would have lain. Presumably more such structures would have existed – as the fortress at Inchtuthil, for example, had at least six and possibly eight granaries. The presence of a fortress makes the traces of some other buildings at Marktveld more explicable, too, especially a structure that bears a striking resemblance to a military barrack block.
Remains of internal structures have also been recognised during the current research at the former naval airfield, albeit with some difficulty. Nevertheless, we can already speak of pits and barrack- like structures, although their final analysis is still under way. As well as these structural traces, something can be said about the finds. The number from the current excavation is admittedly rather small, and once again study is ongoing, but some pre-Flavian (that is, pre-AD 69) material is certainly present. The existence of a fortress could also shed new light on finds of kit most commonly associated with legionary soldiers in the auxiliary fort at Valkenburg.
Many answers, but more questions
As always seems to be the case with archaeology, while our work appears to answer many questions, it raises even more. We cannot tackle all of them here, but we would like to outline some. One is that it is possible that we are dealing with two phases of ditch, with a break in activity between them. This potential second phase of ditch is less clearly V-shaped, and may date to the 2nd century. If so, did the installation remain in use as a military asset – though perhaps not always a legionary fortress – for some time? Two phases of ditch would certainly require some explanation when we only seem to have one phase of rampart, the gates and towers of which appear to have been intentionally demolished by the Romans. As part of the fortress interior was later used as a burial ground – something that could only have occurred after it was abandoned – we are hopeful that material from this cemetery will give us a date by which the military base must have been abandoned.
One recent area of clarity concerns the course of the north and south ramparts, as the fortress corners have now been located. Indeed, it is just possible that Van Giffen caught sight of the south rampart back during the Second World War, as he mentioned a ‘wooden path leading up to the airfield’. As for the northern side of the fortress, this seems to lie just behind a feature known as the Marktveld gully. While this lies in roughly the location where the fortress ditch would be expected, the gully seems to have held a greater significance. There were Late Iron Age finds in the feature, but afterwards it was transformed into a canal, which was then reconditioned in the Flavian era. Another breakthrough is a more precise dendrochronological date for the fortress, from a timber felled in the winter of AD 39/40.
Germania or Britannia, Gabinius or Corbulo?
Although there is not yet enough evidence to be certain about the historical context of the fortress, it may well have been established in the winter of AD 39/40. This could suggest a link with preparations for Caligula’s Germania offensive, or more likely his planned expedition to Britannia. In the latter case, we could even envision the base housing (some of) the troops that Caligula deployed along the shore of ‘Oceanus’. Ultimately, of course, there was no Caligulan invasion, but the emperor’s death did not mean the idea was shelved. Instead, plans to land Roman forces in Britain came to fruition just a few years later in AD 43, under the emperor Claudius. Our fortress was definitely in existence by then, and perhaps played a role in the Claudian invasion as a supply base from the area of the River Rhine. While Roman forces are generally believed to have departed from Boulogne, there is nothing that rules out some of the men sailing from Valkenburg-Katwijk as well, thereby approaching Britannia in a kind of pincer movement.
It is generally accepted that four legions were assembled for the AD 43 invasion. Of these, the Second Legion was based in Strasbourg, the Ninth in modern Croatia, the Fourteenth in Mainz, and the Twentieth in Neuss. It seems reasonable to accept that the main force will have sailed from Boulogne, after reaching their embarkation point via roads leading through central France. This is very likely for the Second and Fourteenth Legions, which were stationed near each other, while the Ninth could also have joined this force after traversing the Alpine passes. This leaves the Twentieth Legion, which would have had further to travel over land than over water. Considering the undisputed presence of military bases along the Rhine during this period, it may well have been identified as a conduit for some forces involved in the invasion of Britain. Therefore, it is at least possible that the Twentieth Legion sailed from the mouth of the Rhine. From there, the ships carrying the legion would have shadowed the coast of Holland and Belgium before finally crossing the Channel; this is a route that was frequently used in later Roman times to connect the Rhine area with Britain. While there is no hard evidence for this theory, we believe it is an interesting hypothesis, which also provides a plausible explanation for troop concentrations and perhaps troop movements in the Rhine area in the years before 43.
The invasion of Britain was not the only reason for legionary activity in the region during the Claudian period. We know from the ancient sources that in AD 41 a legion commanded by Gabinius retrieved the third and final lost eagle standard that had been taken by the Chaucians after the devastating Roman defeat in the Varus Battle (AD 9). Could Valkenburg have been the starting point for that expedition? This scenario seems less likely, but not entirely impossible.
Finally, the general Corbulo was operating with legions in the areas occupied by the Frisians (west and north Holland) and the Chaucians (north of Holland and north Germany), before being ordered back to the existing bases behind the Rhine by Claudius in AD 47. Here, too, a legionary base at Valkenburg could have played a role, and it may also have provided a home-base for soldiers engaged in digging the Corbulo Canal between the Rhine and Meuse rivers in the years AD 47-51. Indeed, could the canalisation of the Marktveld gully be linked to the Corbulo Canal project?
Naturally, a combination of several of these events is a possibility. It is also intriguing that the existence of a large military base at Valkenburg might help explain the name and the vignette associated with the auxiliary fort on the strength of the Tabula Peutingeriana: Praetorium Agrippinae. After all, ‘Praetorium’ could refer to a militarily important place, command centre, or residence, while ‘Agrippinae’ might refer to the mother of Caligula or – less likely – the wife of Claudius. Even so, the various problems associated with understanding this name cannot be considered solved yet.
Forts, fortresses, and the limes
Finally, something can be said about the possible relationship between the legionary fortress, the fort, and the later Roman frontier system known as the limes. It is unknown which legion(s) were stationed in the fortress, although the Twentieth is a good candidate for a short stay prior to its departure to Britain. When that happened, the legion would not have travelled alone, as it would have been accompanied by various units of auxiliary soldiers. The Fourteenth Legion, for example, is believed to have had attached units of Batavian auxiliaries. This prompts the question of which auxiliaries were associated with our legion at Valkenburg. Possibly they were Gallic auxiliaries, because the 3rd cohort of Gauls is known to have been in the Period 1 fort at Valkenburg. But since more than one auxiliary unit would be attached to a legion, does that mean more auxiliaries were stationed in forts near the fortress? Can an early military site still be claimed at Woerd, only 1,200m from Valkenburg fort and just 50m from the fortress? If so, only one of these sites – the Valkenburg fort – ultimately developed into a fully fledged limes fort, which would again fit with a number of units leaving in AD 43.
All of this emphasises how the results of our work in the naval airbase are forcing us to look at many longstanding theories in a new light. Given how much we thought we knew about Valkenburg, a plea must be made for a more integrated approach to understanding sites. This includes using all the ‘old research’ that too-often lies gathering dust on a shelf, as well as the finds recovered during previous work. Both remain so important when it comes to gaining scientific knowledge. As for fresh data, work is still under way on the finds from the most recent work at Valkenburg, making ours a story to be continued.
FURTHER READING R M van Dierendonck, D P Hallewas, and K E Waugh (eds) (1993) The Valkenburg Excavations 1985-1988: introduction and detail studies, Amersfoort (Nederlandse oudheden 15). F Kemmers (2004) ‘Caligula on the Lower Rhine: the coin finds from the Roman auxiliary fort of Albaniana (The Netherlands)’, in Revue Belge de Numismatique et de Sigillographie 150: 15-49. S L Wynia (1999) ‘Caius was here – the Emperor Caius’ preparations for the invasion of Britannia: new epigraphic evidence’, in H Sarfatij, W J H Verwers, and P J Woltering (eds), In Discussion with the Past: archaeological studies presented to W A van Es, Amersfoort, pp.145-148.
Acknowledgements Thanks to the excavation teams of ADC ArcheoProjecten and Archol BV, and to colleagues who gave their views on this legionary base, including J Chorus, Dr C van Driel-Murray, Dr H van Enckevort, E van Ginkel, E Graafstal, W Hessing, and Prof. Dr W A van Es. Naturally, this does not necessarily mean that they agree with all our ideas. For contact and further information, please email Dr Wouter K Vos: info@vosarcheo.nl.
All images: courtesy of Wouter Vos; the inset reconstruction is after Anne Johnson 1983.
