Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands

CATENA-02758; No of Pages 14 Catena xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Catena journal homepage: www.elsevier.com/locate/catena

Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands Jan G.M. Verhagen a,b,c,⁎, Sjoerd J. Kluiving b,c, Emiel Anker a, Liz van Leeuwen a, Maarten A. Prins a a b c

Faculty of Earth and Life Sciences, Cluster Earth and Climate, VU University Amsterdam, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands Faculty of Humanities, Department of Archaeology, VU University Amsterdam, De Boelelaan 1105, 1081 HV Amsterdam, The Netherlands Research Institute for the heritage and history of the Cultural Landscape and Urban Environment (CLUE+), VU University Amsterdam, De Boelelaan 1105, 1081 HV Amsterdam, The Netherlands

a r t i c l e

i n f o

Article history: Received 16 October 2015 Received in revised form 5 March 2016 Accepted 24 March 2016 Available online xxxx Keywords: Geoarchaeology Roman period River geomorphology Waterworks Castellum

a b s t r a c t Romans who settled in the Low Countries at the northern margin of their empire were practicing diverse systems of water management to maintain economic and above all strategic stability. In the early Roman period (12 BC–AD 70) they created a shipping route from the Rhine towards the north by digging canals and constructing dams, such as the Dam of Drusus, accompanied by the adjacent Roman fortress of Carvium (Herwen). This dam was situated at the bifurcation point of the Rhine and Waal river branches and was designed to channel more water into the Rhine. All these engineering feats were undertaken in order to control the northern part of Germania via the Wadden Sea and the German rivers Ems, Weser and Elbe. By the middle Roman period (AD 70–270) the Romans had cancelled their efforts to subdue Germania and this is a period when the Rhine is known as the limes (Roman state border). The research area described in this paper is situated near Herwen in the eastern part of the Rhine–Meuse delta system. The area has a dynamic late Holocene erosional and depositional history, close to the river system's equilibrium point. In order to reconstruct the former landscape and to investigate whether evidence of Roman waterworks could be detected, geoarchaeological coring campaigns were carried out to gain insight into the sedimentology, chronology, stratigraphy and geoarchaeology of the region. Results indicate that Pleistocene sediments are only preserved in the western part of the research area, but further east then previously known. Dating of gullies and levees has confirmed Roman and potentially pre-Roman fluvial activity closer to the Roman fortress of Carvium then was previously known. Four newly discovered residual gullies provide a greater insight into the character of the Roman landscape than hitherto known. The largest of the newly identified gullies may be instrumental in finding the location of the Dam of Drusus, however, much depends on the question as to whether the gully represents an actual former stream channel or simply a crevasse and this cannot be ascertained on the current evidence. Nevertheless the results of this study reinforce the assumption that the Roman castellum was situated on the apex of the Insula Batavorum and close to the Dam of Drusus at the bifurcation of the Rhine and Waal. © 2016 Elsevier B.V. All rights reserved.

1. Introduction The successful administration of the vast Roman Empire was mainly due to the ease of transport and trade. Transport overland was made possible by the construction of a network of good roads. However, much larger quantities of goods and people could be transported over water leading the Romans to construct a network of canals linking to rivers and seas. Canals are known from different parts of the Roman Empire (Smith, 1977; Wikander, 2000; Grewe, 2008). The locations of some of these channels are well known from archaeological research, such as the “antique Suez Canal” (Schörner, 2000), but others such as the Fossa Drusiana (Canal of Drusus) in the Netherlands, have no ⁎ Corresponding author at: Ooyselandweg 5, 6905 DT Zevenaar, The Netherlands. E-mail address: [email protected] (J.G.M. Verhagen).

known remains. This difference is due to the fact that the Egyptian canal was constructed through a large area without rivers, while the latter channel was dug in the Rhine–Meuse delta and might have been eroded by the actively migrating rivers of this region. The Roman presence in the Rhine–Meuse delta in the Netherlands and a small part of Germany is archaeologically known from 19 BC when a legionary fortress was built at Nijmegen (Kemmers, 2006, 13– 57). During the early Roman period until about AD 40, there was an offensive phase, in which the Romans tried to subdue parts of Germania north of the delta (Van Es, 1981, 28–36; Polak and Kooistra, 2013, 440–447). To this end, they undertook military expeditions with fleet units through the Flevum and the Wadden Sea to the Ems, Weser and Elbe (Fig. 1). Around AD 40 they abandoned their efforts to subdue the local population and the Rhine became the northern border of the Roman Empire

http://dx.doi.org/10.1016/j.catena.2016.03.027 0341-8162/© 2016 Elsevier B.V. All rights reserved.

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

2

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

Fig. 1. Possible sailing routes to the north, created under the leadership of Drusus. Map background: Vos et al., 2011, p. 59 (late Iron Age). Inset: Present-day NW European river network with routes of the fleet campaigns of Drusus, Tiberius and Germanicus between 12 BC and AD 16 over the Flevum and Wadden Sea to the estuaries of the Ems, Weser and Elbe in Germania.

(Van Es, 1981, 36–37). This border area, known as the limes, is considered not only as a controlled boundary with Germania, but also as an especially designed and secured transport infrastructure corridor (Whittaker, 1994; Roymans and Derks, 2011, 16–19). From that time onwards, there is a consolidation phase of the Roman presence in the lowlands, after the Batavian revolt (AD 69–70) marked by increasing prosperity and development, a phase also known as the “Pax Romana”. From the last quarter of the 2nd century the military and political tide began to turn and there were alternating losses and recovery of Roman power until in the 5th century AD, after which the Roman Empire collapsed in this area (Van Es, 1981, 44–59). According to written records, the first Roman waterworks built in the Netherlands were those of Drusus, consisting of a dam (Tacitus, Annals XIII, 53) and one or more canals, the latter being “a work of unprecedented proportions” (Suetonius, Vita divi Claudii, I, 2–4). It is generally assumed that these works played a crucial role in the Roman military campaigns in northern Germania. In this way the Roman troops from the Rhine region could reach the northern estuaries of the Ems, Weser and Elbe on their ships via the Flevum (Lake Flevo) and Wadden Sea, without sailing across the North Sea. From the 16th century, when interest in the classics revived, until the 1990's a number of archaeological theories were proposed for the location of a canal (Fossa Drusiana) but none could provide any physical evidence (cf. Vollgraff, 1938; Willems, 1980; idem 1981/1984; Ritterling, 1906; Norlind, 1912; Harbers and Mulder, 1981; Holwerda, 1925; Huisman, 1995). Therefore, the question remained as to where in the Rhine–Meuse delta did the Romans dig the Drusus canal(s)? As we will see below, the position of the former Moles Drusiana (Dam of Drusus) is fairly well known, at the bifurcation point of the Rhine and

Waal. Because it is supposed that there has been a relationship between the Dam and the Canal(s) of Drusus, both serving the aim to provide a water transport route, our approach was to choose our (first) geoarchaeological research area near the location of the dam. The Fossa Corbulonis (Canal of Corbulo) in the western Netherlands was built according to tradition in or shortly after 47 BC, “in order to avoid the uncertainty of a trip across the sea” (Tacitus, Annals XI, 20). Remnants of this canal, which connected the Rhine at Leiden and the Helinium (Maas–Waal estuary, south of Naaldwijk), have been recorded at various locations (De Kort, 2013), revealing a 9 to 14 m wide canal, which was partly lined with wood. This wood has been dated by dendrochronology with a felling date of AD 50 (Jansma, 1995, 129). 2. Background The need to create an artificial shipping route by Drusus was driven by the character of the natural fluvial topography at the end of prehistory. Over the past few decades palaeogeographical research of the Rhine–Meuse delta has made considerable progress in providing a good chronological overview of the development of the various river branches of the delta (Cohen et al., 2012). This corpus of palaeoenvironmental information has allowed many older theories about the location of the Drusus canal(s) to be examined and refuted. Records of classical authors do not describe the purpose of the waterworks of Drusus directly, but this can be deduced. Their role was to provide a shipping connection of the Rhine to the Wadden Sea, in order to allow fleets to sail to northern Germania without having to go across the North Sea (Tacitus, Annales II, 8). Such a trip into the

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

North Sea was dangerous and had several logistical consequences, such as transfer to other types of ship, availability of these ships at the places where they were needed and prolongation of journey times. At the end of prehistory, there was no suitable natural connection between the Rhine and the Wadden Sea. The IJssel branch of the delta, from Doesburg to the north (for place locations see Fig. 3) did not yet exist (Makaske et al., 2008; Cohen et al., 2009). At that time the IJssel flowing from the southeast passed the contemporary settlements Doesburg, Rheden and Velp to Arnhem, where it merged with the Rhine. The contemporary IJssel to the north originated by a natural avulsion between the 5th and 9th centuries AD, after which the river segment east of Doesburg was described as the Old IJssel (Oude IJssel). The comparison of historical insights with more recent palaeoenvironmental research has allowed several older theories to be eliminated. Three options still remain as to where the Romans created a connection: the Vecht, the Gelderse Vallei or in the area of the later IJssel (Fig. 1: R1, R2 and R3). A crucial part of the waterworks of Drusus concerns the dam, the construction of which began in the period 12–9 BC. Its purpose can be deduced from the text of Tacitus about the destruction of the dam during the Batavian revolt (Historiae V, 19): it was meant to channel more water through the Rhine branch at the expense of the Waal branch. There has been wide agreement that the dam had to be located at the former bifurcation of the Rhine and Waal, but the location of this bifurcation has been disputed (Dederich, 1854; Sebus, 1919; Ramaer, 1928; De Waele, 1931; Hardenberg, 1935; Hettema, 1938, 34–39.). More information was retrieved by the discovery in AD 1938 at Herwen of a Roman tombstone bearing the place name “Carvio ad molem”, translated as “at Herwen near the dam” (Vollgraff, 1938).

3

During sand and gravel extraction between the 1930's and 1950's, fragments of Roman buildings and military paraphernalia have been retrieved to such an extent that it is assumed that these artefacts were associated with the remains of a castellum (medium size Roman fortification), which was located to protect the dam (Vollgraff, 1939; Vollgraff and Roes, 1942; Bechert and Willems, 1995, 64–65). Because these remains were not in situ but reworked by the meandering river Waal the precise location of the former castellum can only be determined to within a few hundred metres (Fig. 2). In addition the site is identified with the place Carvo, which is listed both on the Tabula Peutingeriana as in the Itinerarium Antonini (Verhagen, 2014). Although the general location of the remains of the castellum Carvium is fairly well known, the position of the Dam of Drusus was still unclear. There was a high probability that the remains of the dam were also displaced by the actively eroding migrating river. Presumably, the dam was not constructed of natural stone, which was used for buildings in this area only since the latter part of the first century AD (Van Es, 1981). Rather, the structure was constructed of wood and earth, which is less easily to be encountered in the sedimentary record. Furthermore, although well preserved, wood is easily eroded and can be transported large distances away from the point of construction, in contrast to stone, which because of its weight is quickly deposited in the channel. The remains of the castellum are located around 1.2 km from the Herwen channel belt (Rhine branch) (Fig. 2), but from comparison with the lower Rhine limes one would expect it to be situated like others right at the banks of the river Rhine. The research area is situated in the eastern part of the Rhine–Meuse delta system. The valley was formed after the course of the Rhine and Meuse was altered from a south–north direction to a more east–west

Fig. 2. Geomorphology of the area around the bifurcation point of the Rhine and Waal, as known before this study was undertaken. The beige zone is the problematic zone/area of research.

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

4

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

direction by the formation of ice-pushed ridges during the Saalian glaciation (200,000–130,000 BC). Since the Late Saalian the valley has been infilled with deposits of the Rhine and Meuse (Berendsen, 2005, 228–234). The lithostratigraphic deposits of the valley are assigned to the Kreftenheye Formation, Wijchen Member and Echteld Formation. The Kreftenheye Formation is 5–20 m thick and consists of moderately fine to very coarse sands, deposited by braided rivers up until the early Holocene (9000 BC). Overlying it the Wijchen Member comprises sandy and silty clays (Berendsen, 2005, 124), a generally 40–60 cm thick layer, deposited in the early Holocene (9000–7000 BC), during the transition from a braided to meandering river system. The uppermost Echteld Formation (Berendsen, 2005, 124–125) has formed over the remainder of the Holocene by meandering and more linear anastomosing rivers with both lateral and vertical accretionary deposits, residual channel, natural levee, crevasse-splay and floodplain deposits all recognized (Berendsen and Stouthamer, 2001, 43–44). The thickness of the Echteld Formation increases from about 1 m in the eastern part of the delta to over 20 m in the western part with lithological variation reflecting local facies type. Geomorphologically the Echteld Formation manifests itself in the many channel belts of the delta system (Cohen et al., 2012). The bifurcation of Lower-Rhine and Waal from later prehistory corresponds to the delta branching between the Herwen channel belt and the Waal channel belt. The latter is lying in the older channel belt of a previous phase of the undivided Rhine, while the Herwen channel belt is interpreted as an avulsion of the Rhine in later prehistory (Cohen et al., 2012). This avulsion took over the name of the main river, while half a century BC the remaining river branch already carried the name Waal (Caesar, De Bello Gallico IV, 10).

For this research, an inventory of archaeological and historical data was assembled of palaeochannel remains in the Herwen and (part of the) Waal channel belts (Fig. 3). The obtained channel dates were all medieval and in most cases terminus ante quem dates for river activity. This means that the ages of the channel remains might be older than the corresponding archaeological or historical date of human activities on the levees, possibly from the medieval period, and in some cases may be of Roman or Iron Age. The area of the Rhine-Waal bifurcation is also located around the contemporary zone of transition of erosion (incising river upstream of the area) and sedimentation and avulsions (downstream of the area); called terraces crossing by Berendsen and Stouthamer (2001). In prehistoric times, rising sea level caused the terraces crossing to move gradually from the western Netherlands to the region of Arnhem-Nijmegen-Emmerich (Berendsen and Stouthamer, 2001). A complicating factor in the current situation is that over the past 4000 years this process has also been influenced by human factors, such as increased sedimentation resulting from prehistoric agriculture and deforestation further upstream (Erkens, 2009) and engineering interventions in the river such as straightening of channels at bends (Middelkoop et al., 2014). Given the supposed relationship between the Dam and Channel (s) of Drusus the assumptions for this study area are: a. The Roman castellum should have been situated at the apex of the Insula Batavorum; classical authors mentioned the area between the Rhine and Waal branches and the North Sea. b. The Roman bifurcation point of the river should have been located (somewhat) upstream from the location of the castellum.

Fig. 3. Dates of channel remains and meander loops in the channel belt of Herwen and Meinerswijk based on archaeological and historical data.

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

The aim of this paper was to determinate whether traces of Roman waterworks could be identified and investigated by reconstructing the landscape using a palaeogeographical and geoarchaeological approach. More specifically this paper aims at: 1. Identifying where near the presumed castellum Carvium the prehistoric landscape has been preserved and where it has been eroded by former branches of the Rhine and Waal. 2. Identifying any former gullies of the Herwen channel belt that are situated closer to the castellum location than is indicated by previous data. In this paper we have addressed the above stated research questions using a multi-stage coring campaign, in order to collect suitable sediment samples from relevant stratigraphic horizons within both the gullies and other key alluvial contexts within the landscape. The results of this study will help provide a secure chrono and lithostratigraphic framework for future archaeological investigations of the regional landscape. 3. Methods Geoarchaeological fieldwork consisted of reconstructing core profiles along several field axes (Fig. 4) to support a comprehensive analysis with respect to research objectives 1 and 2. In total 72 cores were hand-drilled along 7 axes by a team of 6 people in 7 coring campaigns between 2012 and 2015. Six field axes (P-V) form an extended transect from the area between the Herwen and Waal channel

5

belts into the Herwen channel belt. The field axe Q-W is a short transect into the Waal channel belt. The coring equipment comprised a 7 cm Edelman screw-type hand auger and a 3 cm gauge. The cores reached depths generally of between 2 and 4 m, but in some places penetrated up to 8.0 m below the surface. Spacing between individual cores in the field ranged from 50 to 150 m, reflecting the need to cover large distances, whilst minimizing the possibility of missing channels and related geological structures. Where necessary, local spacing between cores was as little as 20 or 25 m. The locations of the boreholes were determined using a commercial handheld GPS (Garmin eTrex Vista HCx), with a spatial deviation of approximately 3 m. Field description of the core sediments is based on the NEN 5104 method (Bosch, 2000) distinguishing: depth, colour, texture, CaCO3 content, sedimentary structures, presence of admixtures (organic material, shells, pebbles, anthropogenic materials) and oxidation/reduction phenomena. Sandy textures in the field were recorded with a sand ruler. Calcium content, which can help distinguishing younger and older formations, was estimated on all samples using a 10% HCl solution. Regular sediment sampling was applied to 19 of the cores for further laboratory analysis. Sediment analysis was performed at the Sediment Analysis Laboratory of the VU University Amsterdam. Grain size analysis was applied after pre-treatment of the samples by adding a solution of H2O2 and heating to dissolve organic matter and subsequently a solution of HCl to dissolve carbonates. Finally, Na4P2O7 was added and heated to improve grain dispersal before laser-diffraction measurements in a Sympatec Helos Laser KR with a Quixel dispersing system

Fig. 4. Positions of the cores along the different field axes (black dots). Red dots: cores with data for end-member analysis. Grey dots: unpublished results. Map background: Topographical map 2009, scale 1:25,000, by the Dutch Topografische Dienst, map number 117. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

6

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

Fig. 5. Grain size distribution of end-members (cf. Erkens et al., 2013).

fitted with a 1.6 mm sieve, meaning that gravel-sized sediment was not measured. Furthermore, thermogravimetric analysis with a Leco TGA 701 was applied to quantify moisture, organic matter and carbonate content within each sediment sample. Statistical analyses and comparisons were achieved through endmember modelling of the grain-size data set (cf. Weltje, 1997; Weltje and Prins, 2003, 2007), which aims at separating the varying grain size distributions and identifying a limited number of end-members that best represent the dataset. The results can be used to distinguish between lithological units, related to specific sediment sources and/or depositional mechanisms. Units represent a group of layers with a common development characteristic or other distinctive features, i.e. lithology or textural trend (Erkens et al., 2013). Chronological control on sedimentation was achieved through archaeological spot dating of pottery finds with respect to local typologies (by the first author) and radiocarbon analyses. Dating of the pottery has taken into account the taphonomic processes associated with fluvial environments, which may lead to fragments being reworked on multiple occasions by both aqueous processes as well as bioturbation. Later we will discuss the impact of such processes leading in most cases to a terminus post quem dating of sediments. Five core samples were subjected to 14C, 13C radiocarbon dating using Accelerator Mass Spectrometry (AMS) analysis, performed in the laboratory of Beta Analytic Inc. in Miami, Florida. To avoid deviations in age results, samples have been pre-treated by the laboratory to remove carbonates and mobile humic acids. In order to analyse the spatial evolution of residual gullies being recorded through coring, we created a Digital Elevation Model (DEM) using LIDAR data for the research area (Fig. 8B, source: www.ahn. geodan.nl/ahn, AHN1 (1998), view 6 June 2014, viewer set between 10.0 and 13.0 m above sea level). 4. Results On the basis of data collected during coring fieldwork a number of distinct lithological facies have been identified. Laboratory analysis of samples of the sediment sequences was limited to a representative portion of the cores of the first two coring campaigns, fairly well spread

along the core transects. The purpose of the next campaigns was densification of cores, especially where possible residual gullies appeared to be present in the substrate. In total, 319 sediment samples from 19 cores (Fig. 4, red dots) have been used for grain size analysis. In the application of end-member analysis, the mathematical algorithm yielded four different endmembers (Fig. 5, Table 1); all have unimodal grain size distributions and represent 92% of the total distribution. Differences in end-members can be interpreted as reflecting factors such as sediment source or depositional mechanism, i.e. mode of transport and energy levels. Based on the data from the field description and the laboratory analyses, including the end-member analysis of part of the cores, the lithological facies were clustered into 15 lithogenetic sedimentary units, 9 of which are supported by available grain size/ end-member and TGA-data. These units and their characteristics are presented in Table 2. A combination of the core descriptions together with their characteristics and the associated units are presented in schematic profiles along the extended transect (Figs. 6A–C) and the short transect adjacent to the Waal channel (Fig. 7). Samples of the sedimentary sequences of only a representative part of the cores were analysed in the laboratory (Figs. 6 and 7, core nos. presented in red). For this reason the core descriptions (field data) – describing lithology, inclusions, colour, textural trends (increasing or decreasing median grain size with depth), iron oxide, calcium and organic content – are considered to be the primary data for the recognition of lithological units. The general trend in the sediment colour in the cores was a downward one from dark yellowish brown (10YR3/4) to brown (10YR4/3) to greyish brown (10YR5/2) to brownish grey (10YR6/3). In some cores the colour changed further down to grey (5Y5/1) and dark grey (5Y4/1). Concretions that were found in several layers consisted of manganese and/or iron oxide, calcium carbonate and/or organic matter. For the samples that were also analysed in the laboratory, slight differences in texture and organic content between the field data and laboratory data are tolerated due to the large difference in measurement precision of the two datasets. In some of the cores archaeological remains were found, mostly comprising pottery fragments, which were dated to prehistory (Bronze Age/Iron Age), Roman, Medieval and post-Medieval periods up until the 17th century AD. The age of these potsherds, not described here, was used to provide additional chronological control on the lithostratigraphy (Table 2). Results of AMS radiocarbon dating are presented in Table 3. The positions of these samples are marked on Fig. 6. In the schematic profile of units in the long transect (Figs. 6A–C) two trends can be recognized. Firstly a lateral trend to an older lithological history to the west and a younger lithological history to the east. Secondly a vertical trend of transition from fluvial and in-channel deposits to overbank and floodplain deposits. The lowermost deposits of units A through E contain no datable material (including archaeology), so their age can be determined only relatively. There is one exception: at core nr. D10 radiocarbon sample S3 from a small peat layer dated the deposition of the bed load sand in which it was included at AD 350–425 (Table 3). The sediments above units A through E are characterized by overbank and floodplain units, which were dated by artefacts in the cores. Several units include gullies that incise into the underlying deposits. Some of these gullies were dated by AMS radiocarbon analysis

Table 1 End-member characteristics. End-member

Clay (%) [b8 μm]

Silt (%) [8–63 μm]

Sand (%) [N63 μm]

Peak

Possible interpretation

EM1 EM2 EM3 EM4

0 4.42 2.16 48.62

2.44 0.66 35.87 48.31

97.56 94.92 61.96 3.07

Coarse/very coarse sand (500–2000 μm) Middle coarse sand (250–500 μm) Very fine sand (63–125 μm) Very fine/fine silt (8–32 μm)

High energy (fluvial, alluvial) Medium-high energy (fluvial, alluvial) Low-medium energy (fluvial floodplain, alluvial) Low energy (fluvial floodplain, marsh, lake)

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

7

Table 2 Summary of the characteristics of the lithogenetic sedimentary units distinguished in the field description and further laboratory analysis. Lithology, grain size, carbonate content and admixtures were coded according to the NEN 5104 classification with a 1 for low/trace values, 2 for medium values and 3 for high values. Unit Lithology (NEN code)

Colour (Munsell code)

Common grain size (sand)

End-member distribution (%) 1 2 3 4

Carbonate content

Organic content

Interpretation Archaeological Chronology Iron indicators oxide content

A

Sand (Zs1-Zs2)

(Pale) brown to dark grey (10YR 5/3, 10YR 6/3, 10YR 4/1)

36, 56, 2,

1

1

1

Fluvial

Pleistocene

A1

Sandy clay or clayey sand (Kz1-Kz3, Zk) Silty sand (Zs1-Zs4) and clay / silt layers Silty sand (Zs1-Zs4)

Grey(ish brown) to (very pale) brown (10YR 5/2, 10YR 5/1, 10YR 7/3, 10YR4/3)

Medium fine to very coarse (105 to 1400 (-2000) μm) Medium fine to very coarse (150-800 μm)

13, 53, 13, 21 1(-3)

1-2

1-2

Fluvial, floodplain

Early Holocene

Medium fine to very coarse and gravelly

X

1-3

1-2

1-2

Channel / point bar deposits

Late prehistory / early Roman Age

Very fine to very coarse

X

1-3

1

1 Channel and (stains) overbank sediments

Late prehistory / late Roman Age

Moderately fine to very coarse (150-2000 μm)

19, 41, 21, 20 1-3

1

1 Channel (stains) deposits / point bar

Middle Ages

Very fine to moderately coarse (75-420 μm)

X

2

1 Overbank (stains) deposits

Rubble fragm.

Middle Ages

2

1-2

Pottery

Younger than or contemporary with Iron Age

1 -2

2 Channel-fill (stains) deposits

B

C

D

Sand (Zs1-Zs2), with gravel (g2)

E

Silty sand (Zs1-Zs4) with gravel (g2) and clay layers Silty clay (Ks1-Ks4) and locally sandy clay (Kz1) Clay (Ks1-Ks2)

F

F1

G

G1

H1

H2

J

Silty sand layers admixture (Zs1-Zs3) Silty clay (Ks2-Ks4)

Silty clay (Ks1-Ks4) with some silty sand layers Silty clay (Ks3-Ks4) with sand layers Silty clay (Ks3-Ks4) with thin sand layers Silty clay (Ks3-Ks4), Including sand layers

K

Poorly silty sand (Zs1)

K1

Poorly to strongly silty sand (Zs1) with silty to sandy clay layers

Light brownish grey, (light) yellowish grey and grey to brown and light brownish yellow (Light yellowish) grey to (light (yellowish)) brown, reddish brown and (light) brownish yellow (Very dark) grey and reddish / green grey to (pale) brown (10YR 3/1, 10YR 5/3 10YR 6/1-3, 5YR 5/2, 5BG 5/1) (Light brownish) grey and pinkish grey to (yellowish) brown (5BG 3/1, 10YR5/3, 10YR 6/2, 7,5YR 6/2, 10YR 5/4, 10YR 6/1) Dark grey to ((dark) yellowish) brown and (very) pale brown (10YR 5/1 - 10YR 5/4, 10YR 4/1-4, 10YR7/3) Dark grey to pale brown, black (peat) (10YR6/3, 10YR4/1, 2.5Y2.5/1) Light brownish grey to brown (10YR6/2, 10YR5/3)

Dark brown, light yellowish brown and light brownish yellow to greyish brown and light brownish grey (Dark yellowish) brown to reddish brown and reddish grey (7.5YR 5/3, 7.5YR5/4, 10YR4/3, 10YR 4/6 ) (Light brownish / yellowish) grey to greyish brown, (yellowish) brown and light brownish yellow Grey and light brownish grey to pale brown and (light) brownish yellow

6

1-3

1,

9,

23, 67 3

9,

42, 24, 25 1-3

Medium fine to medium coarse (105 -420 μm) Moderately fine to medium coarse (150-600 μm) Moderately to medium coarse (210-300/600 μm)

1 and 3 1 alternating

Bronze Age / Late Neolithic (RAAP)

1

3,

14, 35, 48 1-3

1-2

1-2

6,

17, 30, 47 2-3

3

1 Fluvial, (stains) floodplain

3

1

1 Channel fill (stains) deposits

Late prehistory / early Roman Age (C14 sample)

3

1-2

1 Channel fill (stains) deposits

(early) Roman Age

2,

29, 32, 37 3

1-3

1-2 Fluvial, (stains) floodplain

Rubble fragm.

Post-embankment

7,

25, 3,

3

1

Ceramic fragm.

Medieval

3

1

3 Channel (stains) deposits / point bar 2-3 Channel deposits / point bar

X

X

Dark grey to dark (greyish) brown, strong brown and yellowish / reddish brown (10YR3/3, 10YR4/2 , 10YR 4/1, 10YR5/4, 5YR4/3, 7.5YR 4/6) Light yellowish brown to Moderately fine brown (10YR6/4, 10YR4/3) to very coarse (150-2000 μm) Very pale brown to greyish Moderately fine brown (10YR8/2, 10YR5/2) to very coarse (150-1400 μm)

Fluvial, floodplain

X

2

Fluvial, floodplain

Pottery

(Iron Age,) Roman and Medieval period

Rubble fragm.

Medieval period

Medieval

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

8

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

9

Fig. 6. A–C. Coring profile of the extended transect with lithogenetic units correlated between individual cores. Please pay attention to the fact that the horizontal scale between 1500 m and 2400 m is finer than outside of this profile segment.

of samples. The following section describes the specific characteristics of these units in more detail. The main lithology of unit A comprises fine to very coarse sand with varying amounts of silt, including fining upward sequences only being present in the western part of the profile. Unit A1 comprises a clay with a variable sand content and areas of locally clayey sand. Where A1 is present it is always on top of unit A, having a greater representation of EM3 and EM4 in the end-member distribution (Table 2). Spatially these two units form the basal layers in the cored transects. The unit is interpreted as the Kreftenheye Formation, while the upper clay layer is interpreted as the Wijchen Member (Busschers and Weerts, 2003). Unit B comprises medium fine to very coarse sand with some sandy silt and (very) strongly silty clay layers. The amount of silt in the sand increases upward and also laterally from west to east, indicating fining trends in both upward and eastward directions. The unit can be interpreted as bedload and point bar deposits of the Echteld Formation (Weerts and Busschers, 2003). The sand in the western half of this unit (up to 1900 m) has a low to medium carbonate content and becomes more coarse and gravelly downwards. In contrast the sand in the eastern half has a higher carbonate content and a well-sorted grain size distribution. Cores in the western half contain several alternating sand and sandy clay layers indicating that unit B could be part of a crevasse splay. Unit C comprises poorly silty to very strongly silty sand. The base of this sand unit contains less silt and up to 10% gravel. In general, the amount of silt increases upward. The most distinguishing characteristic of this unit in relation to the other sand units (A, B, D and E) is the very high stratigraphic position of this unit (just above 12 m above sea level). In one core near the eastern edge of unit C (F4, 1635 m) at 185– 230 cm depth a B-horizon with washed-in iron (organic Podsol-like soil) was recognized. Further upwards (175–185 cm deep) a buried/ covered A-horizon with some Fe-spots was found. The occurrence of

plant and organic remains in unit C coincide with the appearance of tiny clay layers. The observed soil formation features in this part of unit C (e.g. iron oxide stains, remains of organic soil horizons), are indicative of long stable periods which allowed pedogenesis.

Fig. 7. Coring profile of the short transect with lithogenetic units correlated between individual cores. For legend see Fig. 6.

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

10

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

Table 3 AMS 14C, 13C radiocarbon dating results of five organic samples. NAP = Normaal Amsterdams Peil = sea level. Sample Core nr./position (m above sea level)

Material

Beta analytic ref. nr.

Measured radiocarbon age

Calibrated age (2 sigma)

Unit

S1 S2

B1 9.75 + NAP B1 9.45 + NAP

Wood remains N180 μm Bulk organic fraction b180

411053 411054

1630 +/− 30 BP 2180 +/− 30 BP

420–570 AD 345–55 BC

S3 S4 S5

D10 8.27–8.32 + NAP E5 8.70–8.85 + NAP C25 bis 8.28–8.33 + NAP

μm Plant remains N180 μm Plant remains N180 μm Twig wood undetermined

G1, gyttja layer in residual gully G1, strongly silty clay layer in residual gully

417139 417141 417142

1700 +/− 30 BP 1610 +/− 30 BP 1910 +/− 30 BP

340–425 AD 420–570 AD 90 BC–55 AD

D, peat layer in bed load sand G, peat layer in residual gully H1, clayey/silty filling in residual gully

The western part of unit D consists of poorly sorted sand. The amount of silt increases upwards, indicating a fining upwards sequence. To the east, moderately silty sand is interbedded with peat and clay layers. In core A12 we observed fragments of undetermined ceramic building debris. The eastern part of unit D consists of sand with major gravel and minor silt admixtures, and shows complete fining upward cycles as well as coarsening upward sequences. The unit is interpreted as a channel deposit. The EM analysis shows EM 1 + 2 = 60%, testifying to high energy levels (Table 2). The unit can be interpreted as belonging to the Echteld Formation. Unit E is a clayey sand unit with a fine to moderate coarse grain size. The unit shows a fining upward sequence of variable amounts of silt with a high clay content at the top. The origin of the unit can be interpreted as overbank deposits. Unit F consists predominantly of clay but with increasing silt content with depth, locally even a sandy clay at the base. Several pottery fragments were found in settlement layers and could be dated to the Bronze Age/Iron Age. Unit F has clearly incised into the Pleistocene substrate eroding locally units A1 and A. The largest incision is that of subunit F1, including a single peat layer. Some of the deepest clayey layers contain plant remains. Comparing EM characteristics F1 indeed has a much coarser EM assemblage (Table 2). The main lithology of unit G is a moderately to very strongly silty clay, which is laterally uninterrupted in the long transect. Small pebbles (0.2 to 1 cm diameter) are present in the upper part of the unit, especially where the top of the unit is at or near the surface. Anthropogenic markers, such as charcoal, pottery and brick, were occasionally found, while shells are present throughout the unit. A subunit G1 is distinguished east of 2600 m in the core transect. The lithology of this unit is silty clay with a coarsening upward trend. Locally some poorly to moderately silty sand layers are present. Brick fragments were also observed. A deep and wide gully incision characterizes unit H, in which two phases can be recognized. Unit H1 comprises a poorly to very strongly silty clay unit with a fining upward trend. Laterally (to the east), the amount of silt is variable with no particular trend, but with a few sand layers observed at the base. The unit shows a sequence of layers of sand which can be interpreted as channel fill deposits, with a younger incision of unit H2 on top. Unit H2 comprises a moderately to very silty clay unit with a slightly coarsening upward trend. The sediments can be interpreted as channel fill deposits, where H2 is a younger incision, partly eroding into H1. The main lithology of unit J consists of strongly to very strongly silty clay, which exhibits in the Ossenwaard (up from 3150 m) a fining upward trend. The unit is found in a long stretch but with a discontinuous lateral extent. Anthropogenic markers and pebbles are found irregularly throughout the unit, which is interpreted as post embankment sediment. The base layer of this unit, lying on top of unit G, often consists of poorly silty sand, which can be seen as recent overbank sediment of the nearby Old Waal (Oude Waal) meander. Here the clay above this sand is often strongly sandy. The lithology of unit K, which is only present in the short transect, consists of poorly silty sand with a fining upward trend. Pebbles are

present throughout the unit in medium to high amounts. In core A25 in the top of unit K a red-baked heavily eroded and not datable piece of ceramic was found. K1 is a sub-unit, consisting of poorly to strongly silty sand with several moderately silty clay and sandy clay layers. 5. Discussion 5.1. Chronological control On the basis of sedimentological descriptions and agestratigraphical relationship interpreted from the various profiles, the following sequence of events is proposed. The basal units A and A1, although they yielded no datable material, are correlated with the Late-Pleistocene Kreftenheye Formation and associated early Holocene (9000–7000 BC) flood basin clay (Wijchen Member) that are generally encountered in the Eastern Rhine–Meuse delta (Busschers and Weerts, 2003). All other described units are younger and belong to the Echteld Formation. The interpreted bedload and point bar deposits of unit B can be divided in two parts, suggesting a high-energy deposit in the western part, potentially a crevasse splay, while the well-sorted sandy nature of the eastern part suggests another potential stream channel in this unit. Assuming that the adjacent units H2 and H1 are residual gullies of this stream channel, this suggests that the eastern part can be dated to Later Prehistory or Early Roman Age, based on AMS radiocarbon dating of a sample (Table 3, S5) from unit H1; otherwise the eastern part should be older. Likewise the western part of unit B is relatively dated to the 5th/6th centuries AD by an AMS sample (Table 3, S4) from a peat layer in a gully-like part of unit G (near 1600 m), but it may also be older. Possibly contemporary with but most likely older than unit B is unit F, floodplain (and potential overbank) deposits that are dated to or before the Iron Age, based on settlement layers with Bronze Age/Iron Age pottery, but also Neolithic finds (personal communication P. Schut, 2013, Amersfoort) included in this unit. Sub-unit F1 represents a gully that crosses the field axe (Fig. 6, between 250 and 650 m) and already has been interpreted as dating to the Bronze Age/Late Neolithic (3000–800 BC; Polman and De Boer, 2000). As demonstrated, unit C is distinguished from the other sand units by a higher elevation and the interpreted remains of an organic Podsol-like soil. If the adjacent unit H1 is interpreted as a residual gully related to this unit, this should imply that unit C (at least its lower layers at this side) is datable to Later Prehistory or Early Roman times. However, unit C should be older if it has been eroded by unit H1. Furthermore, it is possible that the upper layers of unit C are overbank deposits, and when these are related to unit D the overall age of unit C could be (late) prehistory to late Roman period (see below). We have no lithological or chronological evidence to relate it to the late-Pleistocene Kreftenheye Formation. The eastern side of unit C appears to have been eroded by the channel and point bar deposits of unit D. Near this edge in the basal part of unit D a residual gully is recorded. This gully has been dated by three AMS radiocarbon samples (Table 3, S1, S2 and S3). Sample S3 dates a

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

small peat layer under a small clay layer in the sediment of the bed load (Fig. 6, core D10) about AD 340–425 and sample S1 dates the gyttja to about a century later (AD 420–570). The resulting age of sample S2 (345–55 BC), from a strongly silty clay layer between these two, is rejected, because it comprises analysis of a bulk organic sample; it is likely that S2 comprises reworked fine organic remains from elsewhere. In an eastward direction, unit D is interpreted as a point bar gradually overlain by unit E, which can be seen as overbank deposits, but could also represent point bar deposits. Units D and E extend into the medieval Lower-Rhine system east of Herwen. We interpret a boundary to be present at the immediate east side of the Polderdike (Fig. 6: 3150 m). The recent Old Rhine (Oude Rijn) can be seen as a point bar developing from this location to the Erfkamerlingschap; the position in which the Lower-Rhine was fixated after the digging of the Pannerden Canal in AD 1707, when it got the name Old Rhine. But east of the Erfkamerlingschap an older point bar system is present (Fig. 2). Cohen et al. (2012) (nr. 635: Steinward channel; Fig. 3, grey channel) suggest that this point-bar east of Erfkamerlingschap dates from the very early Holocene. We assume however that in the early Middle Ages here the river already migrated over a long distance from the area west of the Polderdike to the Steinward, and that this area later has been eroded partially by the younger point bar of the Old Rhine/Erfkamerlingschap. Units H1 and H2 constitute a deep gully system that has eroded older layers down from a level in the base of unit G and also may have deposited sandy sediments beside the gully. The coring profile of Fig. 6 suggests that the sand layers in the gully fill of H1 are cut off by unit H2 and therefore we assume that unit H2 is somewhat younger than unit H1; the latter was dated to Later Prehistory or the Early Roman period (Table 3, S5). The association of units A through F is overlain by the floodplain sediments of unit G. It is dated to the Roman and Medieval periods by Roman and Medieval pottery fragments found at certain levels in this unit. We have taken into account that fragments of pottery can be reworked by fluvial processes, and so strictly the presence of potsherds leads to a terminus post quem date for the layer. However, reworked pottery fragments are mostly deposited together with coarse sand and gravel, while unit G consists of clay. Bioturbation is also a possible taphonomic problem, but it is estimated not to have had a big influence in this case, also because of the relative age of the underlying units, facilitating human occupation on the overbank sediments after prehistory. Eastward of the erosional remnant of unit C (Fig. 6, 2600 m), sub-unit G1 has developed in the Medieval period, based on the dating of the underlying bed load of unit D. A top layer in the part of the core transect that is located river side of the dikes is unit J, the post embankment unit. It is not recognized in the cores located on the landward side of the dikes. Locally this unit has been dug away for dike raising and brick production. Although the first dikes were built around AD 1200, they have been raised periodically into the 20th century. Until the 18th century AD the water reached over the dikes regularly, so formation of unit G may have continued into the 18th century. The Pannerdense dike (Fig. 6, 540 m) and Herwense dike (2360 m) were built in AD 1771–1772, as replacing dikes surrounding the intruded large meander of the Old Waal, implying that in the transect part between the two dikes unit J mainly has been formed from the 18th century. From 1600 m to the east (Fig. 6) the entire late Pleistocene Kreftenheye unit (A) has been eroded by several (potentially migrating) channels. This implies that all units, except A and A1, belong to the Echteld Formation. Along the eastern part of the long transect and the south-western part of the short transect (Fig. 7) the Kreftenheye unit is replaced by sandy units B, C, D, E and K. Units D and E have been formed from the end of the Roman period into the Middle Ages and both have the character of point bar systems. For units B and C it cannot be excluded that they have a later prehistoric origin, although it is possible that they have a lithogenetic relationship to the gullies in that part of the research area.

11

5.2. Palaeogeography around the Roman period Coring results revealed three (residual) gullies that were not known previously (Fig. 8, gullies A, B and C). Gully A (Fig. 6, 1550–1700 m), is a deep incision belonging to unit G. Gully B corresponds to units H1 and H2 (Fig. 6, 2100–2400 m) and gully C is present in core B1 as a narrow residual gully superimposed on the sandy channel fill of unit D (Fig. 6, 2580–2700 m). Outside of the cross sectional area gully D was observed, based on the DEM image of this area (Fig. 8B). Gully D cuts off gullies B and C and is therefore younger than the latter ones. Although less distinct, the course of gully C may be also traced on the basis of the DEM image. This indicates that both C and D can be seen as former river channels. Through additional coring the character of gully D was determined and information about the orientation and course of the gullies B and C was gathered (unpublished results, locations of the cores are represented in Fig. 8A). Through this work gully B, corresponding with units H1 and H2, was mapped over a distance of approximately 700 m, orientated from south-east to north-west. Some elements of contemporary relief can be related to the edge of this gully. However, to the north-west, no relief, which could be related to the continuation of this trench, is visible in the field. Thus, it is not certain that this feature has ever been a true river channel, and it may represent a crevasse channel. The course of gully A is also not yet known. No continuation of the feature can be derived from the relief in the field to the north of the cores and therefore this may also represent a crevasse channel. Generally we can conclude that in the research area the prehistoric landscape has been preserved from the west as far as gully A (Fig. 8). The area between the Roman Rhine and Waal branches and the North Sea is known as the Insula Batavorum (Isle of the Batavians) (Tacitus, Historiae V 19, 2–3) and now we can state with more certainty that the place with the remains of the castellum Carvium has lain at the apex of the Insula Batavorum (Fig. 9). Based on the present data set it is possible to identify former gullies of the Herwen channel belt that are situated closer to the castellum location than the ones known previously. Four gullies have been interpreted that were unknown up until this research (Fig. 8). The largest of these gullies is B and has two phases (units H1 and H2), the oldest of which can be dated to the last centuries of prehistory or the early Roman period. On either side of this gully two additional gullies dated to around or shortly after the end of Roman period were found (A and C). A question is whether these gullies are related to the Herwen channel belt or maybe to the Waal channel belt (Fig. 9). In our view gully C and units D and E can be related to the Herwen channel belt, because these have the character of a point bar developing into the Herwen channel belt (the medieval Lower-Rhine). With respect to the origins of gullies A and B we are unsure whether they can be seen as regular river gullies, because no traces of continuing residual gullies can be seen in the contemporary surface north of the present transect lines. If they have a still unknown continuation to the north/north-west, they can be related to the Herwen channel belt, but if they are crevasse systems, they might have come from the Waal channel belt, discharging to the north into the flood basin. We can state that no direct traces of Roman waterworks have been recognized in the lithological profiles. However, the question remains as to whether there is a relationship between gully B and the waterworks of Drusus, especially the moles Drusiana, because the gully and the building of the dam date from the same period. Several options exist, dependent of the exact dating of the gully, which is not yet known securely. If the gully should appear to be older than 12 BC then it is possible that this gully is the river Rhine branch of the period before the works of Drusus (Cohen et al., 2012, nr 65). Then the construction of the dam should have caused more water to flow into this branch.

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

12

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

Fig. 8. A. Overview of the position of gullies A, B, C and D as plotted during this project. B. Digital Elevation Model image of the area of the gullies A through D. Legend is shown in metres above NAP (= sea level).

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

13

Fig. 9. Geomorphology of the area around the bifurcation point of the Rhine and Waal, on the basis of the results of this project (cf. Fig. 2). The oval represents the area in which the Dam of Drusus could have been located as concluded by this research. Yellow: Intact prehistoric landscape Grey: Prehistoric landscape eroded by the Rhine/Waal Beige: Zone with gully remains of about the Roman period. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

If the gully should appear to be not older than 12 BC, we can distinguish two possibilities. The first possibility is that the gully can be seen as an artificial connection near the dam beyond the splitting point of the Rhine and Waal, some kind of bypass of the river Rhine branch, already suggested by Van Tol (1988). In this case it cannot be seen as a purely artificial canal, because then a width of 10 to 15 m would have been sufficient, like the width of the Canal of Corbulo (De Kort, 2013). If it originated as a dug channel, then the enhanced water flow caused by the dam should have widened the channel considerably. The second possibility is that the gully is a crevasse and represents an unintended side effect of the Dam of Drusus, possibly appearing in times of high water levels. In addition, an option utilizing an already existing Rhine branch is still possible. So a causal relation between gully B and the Drusus dam is a possibility. From an archaeological viewpoint, an interesting question is whether the castellum Carvium was located at the banks of the river Waal, those of the river Rhine, or at the banks of both. Its position would reflect that for defensive reasons castella were located along the Roman border and at the safe side of it. However, this was the case during the period of the limes, from approximately AD 45 and we must realize that the castellum Carvium was built already in the offensive phase, being not younger than AD 10. On the basis of the site of the castellum remains we cannot exclude the possibility that the castellum was situated at the edge of the Waal channel, but much depends on the question of which of the above options corresponds with

reality. Without further investigations we cannot provide a more definitive answer. 6. Conclusions Coring campaigns near Herwen have revealed that Pleistocene sediments are only preserved in the western part of the area of research, but further eastwards than was previously known. We have more evidence now about the fact that the castellum Carvium was located at the apex of the Insula Batavorum. Core results combined with DEM analysis also revealed four gullies that were unknown until now. Chronological control of these gullies has revealed that the largest of the four dates from the late prehistoric or early Roman period. Following this we can conclude that the Herwen channel belt in the Roman period was located closer to the castellum remains than previously known. We can state that no direct traces of Roman waterworks have been recognized in the lithostratigraphic profiles. However, a relationship between gully B and the waterworks of Drusus, especially the dam, is plausible because the gullies and the building of the dam date from the same period. The character of the large gully B is not yet fully understood, but there are three possibilities: 1. This gully is the river Rhine branch of the period before the works of Drusus. The construction of the dam would have caused more water to flow into this branch.

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027

14

J.G.M. Verhagen et al. / Catena xxx (2016) xxx–xxx

2. The gully can be seen as an artificial connection near the dam, some kind of bypass of the river Rhine branch. 3. The gully is a crevasse and represents an unintended side effect of the Dam of Drusus, possibly occurring in times of high water levels. There also remains the possibility of a combination of the above options. Gully B cannot be seen as a completely artificial channel, because a width of 10 to 15 m would have been sufficient. If it originated as a dug channel, then the increased water flow caused by the dam construction would have widened the channel considerably. Nevertheless our observations reinforce the assumption that the castellum was situated close to the Dam of Drusus and the Roman period bifurcation of the Rhine and Waal. This study shows that it is difficult to identify the remains of Roman channels in a delta region principally because of the migration of its rivers. In areas where Roman canals were built in zones with no later fluvial erosion, archaeological remains of canal building are preserved in situ and can be excavated. Where this is not possible, indirect methods have to be used, combining historical and geoarchaeological data to elucidate past histories. We envision that this project has provided an exemplar for studies elsewhere around the globe. Acknowledgements The authors are grateful to Feike Miedema, Eckhart Heunks, Jurgen de Kramer, Jos de Moor and members of the AWN (Dutch association of volunteers in Archaeology) for cooperation in field work, to Nico Willemse and Laura Boukje Stelwagen (RAAP, NL) for their help with the core profile presentation, to Andy Howard (Landscape & Research Management, UK) for correcting the English in this paper and to two anonymous reviewers that have greatly improved the quality of our paper. We thank also the province of Gelderland, the VU University Amsterdam and the municipality of Rijnwaarden (NL) making this study possible. References Bechert, T., Willems, W.J.H., 1995. Die römische Reichsgrenze zwischen Mosel uind Nordseeküste Stuttgart/Nijmegen. Berendsen, H.J.A., 2005. Fysisch-geografisch onderzoek: thema's en methoden Gorkum. Berendsen, H.J.A., Stouthamer, E., 2001. Palaeogeographic development of the Rhine– Meuse delta The Netherlands. Bosch, J.H.A., 2000. Standaard Boor Beschrijvingsmethode, Versie 5.1. Nederlands Instituut voor Toegepaste Geowetenschappen TNO, Zwolle. Busschers, F.S., Weerts, H.J.T., 2003. Beschrijving lithostratigrafische eenheid — Kreftenheye. Nederlands Instituut voor Toegepaste Geowetenschappen TNO, Utrecht. Cohen, K.M., Stouthamer, E., Hoek, W.Z., Berendsen, H.J.A., Kempen, H.F.J., 2009. Zand in Banen — Zanddieptekaarten van het rivierengebied en het IJsseldal in de provincies Gelderland en Overijssel, Provincie Gelderland. derde geheel herziene druk, Arnhem. Cohen, K.M., Stouthamer, E., Pierik, H.J., Geurts, A.H., 2012. Rhine–Meuse Delta Studies' Digital Basemap for Delta Evolution and Palaeogeography. Dept. Physical Geography, Utrecht University Digital dataset http://persistent-identifier.nl/?identifier=urn: nbn:nl:ui:13-nqjn-zl. De Kort, J.W., 2013. Het kanaal van Corbulo‐ Onderzoek naar een Romeinse waterweg in de gemeente Leidschendam-Voorburg tussen 1989 en 2010. Westerheem. 62, pp. 233–243. De Waele, F.J., 1931. Bouwsteenen voor een geschiedenis van Nijmegen. I. Noviomagus Batavorum (Romeinsch Nijmegen) Nijmegen. Dederich, A., 1854. Geschichte der Römer und der Deutschen am Niederrhein Emmerich. Erkens, G., 2009. Sediment dynamics in the Rhine catchment. quantification of fluvial response to climate change and human impact. Netherlands Geographical Studies. 388 Utrecht. Erkens, G., Toonen, W.H.J., Cohen, K.M., Prins, M.A., 2013. Unravelling mixed sediment signals in the floodplains of the Rhine catchment using end-member modelling of grain size distributions. Proceedings of the 10th International Conference on Fluvial Sedimentology, pp. 109–110 Leeds.

Grewe, K., 2008. Tunnels and canals. In: Oleson, J.P. (Ed.), The Oxford Handbook of Engineering and Technology in the Classical World. Oxford (Oxford University Press), pp. 319–336. Harbers, P., Mulder, J.R., 1981. Een poging tot reconstructie van het Rijnstelsel in het oostelijk rivierengebied tijdens het Holoceen, in het bijzonder in de Romeinse tijd. KNAG Geografisch Tijdschrift, Nieuwe Reeks, 15/5, pp. 404–421. Hardenberg, H., 1935. De Rijnverdeling in den Romeinschen tijd en in de vroege middeleeuwen. Bijdragen en mededeelingen Gelre. 38, pp. 29–49. Hettema Jr., H., 1938. De Nederlandsche wateren en plaatsen in den Romeinschen tijd. Nijhoff, Den Haag. Holwerda, J.H., 1925. Nederlands Vroegste Geschiedenis Amsterdam. Huisman, K., 1995. De Drususgrachten: een nieuwe hypothese. Westerheem. 44, pp. 188–194. Jansma, E., 1995. RemembeRINGS. The Development and Application of Local and Regional Chronologies of Oak for the Purposes of Archaeological and Historical Research in the Netherlands (Dissertation University of Amsterdam), Amersfoort (Nederlandse Archeologische Rapporten 19). Kemmers, F., 2006. Coins for a legion. An analysis of the coin finds of the Augustan legionary fortress and Flavian canabae legionis at Nijmegen (unpubl. diss.) Univ. Nijmegen. Makaske, B., Maas, G.J., van Smeerdijk, D.G., 2008. The age and origin of the Gelderse IJssel. Neth. J. Geosci. — Geol. Mijnb. 87 (4), 323–337. Middelkoop, H., van der Perk, M., Cohen, K.M., Stouthamer, E., Hoek, W.Z., Kleinhans, M.G., Erkens, G., Jansma, E., 2014. Tracing the Anthropocene in the Rhine–Meuse delta. Geophy. Res. Abstr. 16. Norlind, A., 1912. Die geografische Entwicklung des Rheindeltas bis an das Jahr 1500 Lund/Amsterdam. Polak, M., Kooistra, L.I., 2013. A sustainable frontier? The establishment of the Roman frontier in the Rhine delta. Part 1: from the end of the iron age to the death of Tiberius (c. 50 BC–AD 37). Jahrbuch des Römisch-Germanischen Zentralmuseumns Mainz. 60, pp. 354–458. Polman, S.P., De Boer, G.H., 2000. Drie dorpen poldxer: gemeente Rijnwaarden: een aanvullende archeologische inventarisatie. RAAP-rapport 491 Amsterdam. Ramaer, J.C., 1928. Het Nederlandsche alluvium in den Romeinschen tijd en de middeleeuwen. Tijdschr. Kon. Ned. Aardrijksk. Genootschap tweede serie, 45, 202– 235 en 593–628. Ritterling, 1906. Zur Geschichte des Römischen heeres in Galliën unter Augustus. Bonner Jahrbücher. 114, pp. 159–188. Roymans, N.G.A.M., Derks, A.M.J., 2011. Villa Landscapes in the Roman North. Economy, culture and lifestyles, Amsterdam (AAS 17). Schörner, H., 2000. Künstliche Schiffahrtskanäle in der Antike. Skyllis. Der sogenannte antike Suez-Kanal 3, pp. 28–43. Sebus, J.H., 1919. De Drususdam en zijn verband met de Drususgracht. Tijdschr. Kon. Ned. Aardrijksk. Genootschap. 36, pp. 686–696 tweede serie. Smith, N.A.F., 1977. Roman canals. Transactions of the Newcomen Society. 49, pp. 75–86. Van Es, W.A., 1981. De Romeinen in Nederland. Fibula-van Dishoeck, Haarlem. Van Tol, T., 1988. . Een oud taboe doorbroken: de moles te Carvium en de steenmassa te Herwen. Westerheem. 37, pp. 293–305. Verhagen, J.G.M., 2014. Using distances to identify Roman places in Itineraria — a case study on the lower Rhine limes. Archäologisches Korrespondenzblatt. 44, pp. 543–562. Vollgraff, C.W., 1938. De dijk van Drusus. Mededeelingen der koninklijke Nederlandsche akademie van wetenschappen afd. letterkunde. Nieuwe Reeks. 1, pp. 555–577. Vollgraff, C.W., 1939. De moles van Drusus. Mededeelingen der koninklijke Nederlandsche akademie van wetenschappen afd. letterkunde. Nieuwe Reeks. 2, pp. 141–143. Vollgraff, C.W., Roes, A., 1942. Nieuwe oudheidkundige vondsten uit Lobith. Mededeelingen der koninklijke Nederlandsche akademie van wetenschappen afd. letterkunde. Nieuwe Reeks. 5, pp. 283–329. Vos, P.C., Bazelmans, J., Weerts, H.J.T., van der Meulen, M.J. (Eds.), 2011. Atlas van Nederland in het Holoceen Amsterdam. Weerts, H.J.T., Busschers, F.S., 2003. Beschrijving lithostratigrafische eenheid — Echteld. Nederlands Instituut voor Toegepaste Geowetenschappen TNO, Utrecht. Weltje, G.J., 1997. End-member modelling of compositional data: numerical-statistical algorithms for solving the explicit mixing problem. J. Math. Geol. 29, 503–549. Weltje, G.J., Prins, M.A., 2003. Muddled or mixed? Inferring palaeoclimate from size distributions of deep-sea clastics. Sediment. Geol. 162, 39–62. Weltje, G.J., Prins, M.A., 2007. Genetically meaningful decomposition of grain-size distributions. Sediment. Geol. 202, 409–424. Whittaker, C.R., 1994. Frontiers of the Roman Empire. A social and economical study Baltimore. Wikander, C., 2000. Canals. In: Wikander, Örjan (Ed.), Handbook of Ancient Water Technology, Technology and Change in History. 2, pp. 321–330 Leiden (Brill). Willems, W.J.H., 1980. Arnhem-Meinerswijk, een nieuw castellum aan de Rijn. Westerheem. 29, pp. 334–347. Willems, W.J.H., 1981/1984. Romans and Batavians, a regional study in the Dutch eastern river area. Berichten van de Rijksdienst voor het Oudheidkundig Bodemonderzoek. 31, pp. 7–218 idem 34, 42–341 (dissertation UvA Amsterdam 1986).

Please cite this article as: Verhagen, J.G.M., et al., Geoarchaeological prospection for Roman waterworks near the late Holocene Rhine-Waal delta bifurcation, the Netherlands, Catena (2016), http://dx.doi.org/10.1016/j.catena.2016.03.027