(a) Five Lanes marching camp, Monmouthshire (NPRN 423454, ST 4488 9108)

This camp was discovered in a cropmark-rich field just to the north-west of Caerwent and east of the Five Lanes villa, on 29 June 2018 (figs 3 and 4). While the villa's location was first indicated by various tessellated pavements which came to light in the 19th century, the stone building itself was not documented from the air until 1995, when John Sorrell recorded indistinct parchmarks in pasture. Its location has subsequently been recorded from the air on several occasions, together with the cropmarks of an extensive block of field enclosures, apparently associated with the villa, which is enclosed in a separate angular enclosure. A separate rectangular enclosure, measuring 55 by 32 m, lies between the villa enclosure and the newly identified marching camp. A basic plot of the southernmost fields was reproduced by Edith Evans,Footnote ⁵ but this has now been superseded by the present plot. The characteristic outline of the camp was only recognised following desk-based analysis, which involved rebalancing the colour bands of the aerial photographs.

FIG. 3. Five Lanes villa (centre-left) and faint lines of the marching camp (centre-right): view from the south, 29 June 2018. (Crown Copyright RCAHMW, AP_2018_1643)

FIG. 4. Five Lanes marching camp: plot of cropmarks revealed on aerial photographs showing the villa and villa enclosure (lower left), the field system (centre) and the marching camp (right). (Crown Copyright RCAHMW, Ordnance Survey licence number: 100022206)

The new camp lies 220 m east of the villa. Only its western half is visible, together with the north-west and south-west corners and the north, south and west gates. Its eastern side is cut by a country lane and road junction, but it can be judged to measure 131 m² based on the complete western half. The locations of the gates indicate that the camp faced south and was of small size, enclosing just under 2 ha. This is the second small camp to be found close to Caerwent and only the third marching camp to be found in Monmouthshire, a region which must have seen extensive pre-Flavian military operations. The diminutive size of this camp and that on Killcrow Hill (0.98 ha) is of some interest when compared with that at Llancayo (20.2 ha) to the north of Usk.Footnote ⁶ This must surely reflect a different scale of operations on the lowlands of Gwent compared with the concentration of a large army tasked with pushing north and west up the Usk valley and beyond.

Of particular interest is the novel gate type at Five Lanes: gates provided with an external clavicula whose course is marked by a curving ditch. Examples of this are very rare, because for the most part camps with clavicular entrances have an internal clavicula with no accompanying ditch or, where there is no apparent protection for the entrance gap, the clavicula – whether internal or external – has been destroyed by agricultural operations. Whether the Five Lanes camp was also provided with an internal clavicula, though lacking a ditch (as per normal), is impossible to say. Rebecca Jones has already remarked upon the rarity of the Five Lanes gate type and it is the first instance of such in Wales.Footnote ⁷

(b) Three Cocks (Gwernyfed Park, Aberllynfi) marching camp, Powys (NPRN 423861, SO 17038 37127)

RCAHMW aerial reconnaissance on 19 July 2018 identified part of a previously unrecorded marching camp adjacent to a fort (NPRN 423858; see 1d below) in Gwernyfed Park, Three Cocks (see fig. 7). Only its north-eastern side, 64 m long, together with its north-west and north-east rounded corners, is visible, c. 23 m away from the south-west corner of the fort, suggesting a close association between the initial defence of the site and the subsequent construction of the fort. No further traces of the camp can be seen to the south-west, though the presence of a minor stream, c. 90 m south-west of the visible line of its rampart may suggest its maximum extent. Its siting is similar to the possible marching or construction camp associated with the fort at Cefn Brynich (NPRN 419242), which is sited only 25 m west of the fort defences, and of which only the north-west corner is visible.Footnote ⁸

(c) Carrow Hill fort, Monmouthshire (NPRN 423337, ST 4300 9034)

Discovered on 29 June 2018, but most clearly recorded under ripening crops on 6 July, this site occupies the summit of a low hill WSW of Caerwent and south of the course of the Roman road. It appears as the cropmark of a bivallate ditched enclosure, the inner one 50 m² (0.23 ha) and the outer 73 m² (0.5 ha) (figs 5 and 6). An annexe 52 m across is attached to its eastern side. The corners of both fort and annexe are beautifully rounded in military fashion. Although there is no clear indication of an entrance, such may have lain to the north where the ditch system is obscured by a modern lane. The ditch system, with an interspace of c. 12 m between the inner and outer ditch, resembles that of pre-Flavian military installations such as Hod HillFootnote ⁹ and Great CastertonFootnote ¹⁰, where a ‘killing zone’ has been fashioned, with the outer ditch being of Punic form. In terms of size, the nearest comparators are small forts of Flavian date in Scotland such as Inverquharity (0.36 ha),Footnote ¹¹ Gatehouse of Fleet (0.35 ha)Footnote ¹² and Mollins (0.4 ha), the last also possessing an annexe.Footnote ¹³

FIG. 5. Carrow Hill fort: plot of cropmarks revealed on aerial photographs showing the fort, the annexe (to right) and the plough-levelled barrows on the southern side. (Crown Copyright RCAHMW, Ordnance Survey licence number: 100022206)

FIG. 6. Carrow Hill fort: seen from the north on 6 July 2018. (Crown Copyright RCAHMW, AP_2018_2977)

The fort's southern defences abut a pair of plough-levelled round barrows (NPRN 423338); the eastern smaller one, measuring 30 m in diameter, has a ‘keyhole’-shaped cutting into its centre from the north, potentially evidence for its modification by the Romans to form an oven or kiln.

This discovery is particularly significant insofar as, though of small size, it is the first plan of an auxiliary fort to be recorded in the Vale of Gwent. Its location 1 km south of the course of the Roman road is indicative of an early, and probably pre-Flavian date. Carrow Hill may possibly be linked with two other broadly contemporary sites: Chepstow,Footnote ¹⁴ c. 13 km to the east, where a fort probably exists on the basis of early cremation burials and an associated Claudian coin, and Coed-y-Caerau,Footnote ¹⁵ another probable early fort c. 5.5 km to the north-west, which encloses 1 ha and is situated on high ground overlooking the Usk valley and the site of the legionary fortress at Caerleon. Though originally included as a fort in the 1969 edition of The Roman Frontier in Wales, Coed-y-Caerau was subsequently dismissed on the basis of its apparent integration into the defences of a complex Iron Age hillfort;Footnote ¹⁶ further consideration leads to the conclusion that its Roman military credentials should be reinstated.

(d) Three Cocks (Gwernyfed Park, Aberllynfi) fort, Powys (NPRN 423858, SO 1710 3724)

Partial cropmarks of the long-suspected fort under pasture at Gwernyfed Park were finally revealed on 19 July 2018 (fig. 7). It is situated on level ground on the floor of the valley in close proximity to where the river Dulas flows into the Wye, in an area occupied by the playing fields of Gwernyfed Park School and modern buildings. Roman iron-smelting furnaces of first-century date were excavated in this locality in 1951.Footnote ¹⁷

FIG. 7. Three Cocks fort and marching camp: plot of incomplete parchmarks revealed on aerial photographs in grassland at Gwernyfed Park. The star marks the site of furnaces excavated in 1950. (Crown Copyright RCAHMW, Ordnance Survey licence number: 100022206)

Despite the severity of the drought, with the underlying geology showing through the pasture, only fragments of the fort's perimeter defences were revealed and nothing of any internal buildings. Given the presence of parchmarks at the southern end of the fort relating to modern drains and a semicircular hardstanding – presumably for athletics – it is likely the parkland here has been graded or landscaped.

The rounded bivallate south-west corner of the fort is clearly marked, together with part of the bivallate defences on the north-eastern side and a single ditch along the south-eastern side, showing that it measures 157 m² (externally) and encloses c. 2.5 ha. The north-west corner has been lost under modern housing. The fort entirely encloses the site of the previously excavated iron-working area (NPRN 309372), though it is unclear whether this forms part of the fort's interior or post-dates it. The newly identified marching camp lies adjacent on the south-western side (NPRN 423861; see 1b above).

Another possible fort has been located c. 2 km to the north at Glasbury, while a large campaign fort, with a smaller fort within, lies c. 8 km to the north-east at Clyro.Footnote ¹⁸ The course of a putative Roman road lies c. 375 m to the north. Parchmarks c. 300 m to the south-east of the Three Cocks fort, also discovered on 19 July 2018, show the line of the road (NPRN 423860) approaching the fort; a projected alignment shows that it would pass its southern side to within 115 m of the south gate, suggesting a T-junction would once have existed.

Davies and Driver have already suggested that an early Roman communication route linked the complex at Clyro and the newly discovered fort at Cefn-Brynich.Footnote ¹⁹ The discovery of another fort at Three Cocks, together with a stretch of possible road, adds further credibility to this suggestion. Whether the Three Cocks fort belongs to the same pre-Flavian era as Cefn-Brynich is unclear. On the basis of the associated ceramics, the excavators suggested a Flavian date for the furnaces and associated occupation deposits.Footnote ²⁰ Given the apparent bunching of sites between Three Cocks and Clyro, however, it is unlikely that they were exact contemporaries, and must, therefore, represent differing episodes in pre-Flavian and early Flavian operations in this part of Wales.

(e) Great House or Graig-Olway Farm, near Usk, Gwent; possible fort (NPRN 86829, S0 38840 00350)

The site occupies elevated ground at 60 m above Ordnance Datum overlooking the floodplain of the river Usk and the site of the legionary fortress from the east. The site is positioned towards the leading western edge of a low escarpment so as to command an aspect over the Usk valley at this point. Cropmarks show two closely spaced, parallel ditches defining a square enclosure 155 m across, enclosing 2.3 ha, with neatly rounded corners of what is probably a fort, orientated south-west–north-east (fig. 8).

FIG. 8. Great House or Graig-Olway Farm, possible fort: looking east with the south-western portion of the enclosure visible, 6 July 2018. (Crown Copyright RCAHMW, AP_2018_2917)

Cropmarks have been recorded from the air by RCAHMW since 1989, and particularly clearly in 2005 by Sorrell. Although the site has previously been noted under ‘Trostrey’,Footnote ²¹ where its size is calculated as c. 2 ha, the aerial views taken on 6 July 2018 clearly show the south-western portion of the enclosure and the character of its ditches. If the site is indeed that of a fort (and a prehistoric date has also been suggested) then it is most likely a precursor to the establishment of the legionary fortress at Usk under Nero and could well represent one of those forts founded in Silurian territory under Ostorius Scapula or his successor Didius Gallus sometime between a.d. 48 and 57.

(f) Pen y Gaer fort, Cwmdu, Powys (NPRN 92190, SO 1586 2195)

Although this site was first identified in the early 19th century,Footnote ²² the most recent summary of the fort and its associated extramural activity in 2010 is still surprisingly sketchy,Footnote ²³ this despite some small-scale excavations in 1966 and 2007 and an extensive extramural geophysical survey in 2006.Footnote ²⁴ While further details have since emerged from continued excavations south of the fort in 2011–12,Footnote ²⁵ the overall picture has now been radically transformed by a remarkable series of aerial views obtained as a result of parching on 19 July 2018, shortly before the end of the drought (fig. 9). The parchmarks were subsequently overflown with a drone by Walters on 26 July in a timely survey commissioned by the Brecon Beacons National Park; this provided new photogrammetric imagery of the fort and external structures, despite a slight loss of detail in the parchmarks caused by intervening rain.

FIG. 9. Pen y Gaer fort: general view from the east, taken at the height of the drought on 19 July 2018 and showing the newly identified annexe and courtyard buildings in the foreground. (Crown Copyright RCAHMW, AP_2018_5567)

Details of the fort plan are now much clearer than hitherto (fig. 10), demonstrating that it measures 152 m east–west by 114 m north–south and encloses c. 1.6 ha (3.9 acres). It can be shown to have faced east and to have been provided at some stage with stone defences and stone, or at least stone-footed, internal buildings; these latter being manifest for the first time. Two buildings in the north-eastern corner of the praetentura show up particularly well and represent barrack blocks arranged per scamna. While the eastern one is 9.6 m wide, including a corridor or veranda, 1.6 m wide, running the length of the western side, its full length is unfortunately obscured by buildings to the south. Inside each barracks is a series of rooms occupying the western portion of the block, both measuring c. 3.6 by 3 m. Within the retentura the partial outlines of at least two long buildings are visible to either side of the via decumana; that to its south is c. 48 m long and shows faint room divisions. These too appear to be barrack blocks, this time aligned per strigas.

FIG. 10. Pen y Gaer fort: plot of cropmarks revealed on aerial photographs showing the extent of vicus buildings revealed to the south of the fort and the extensive range of newly recorded extramural buildings and annexe to the east of the fort. (Crown Copyright RCAHMW, Ordnance Survey licence number: 100022206)

Far more significant, however, is the new detail that has emerged concerning the fort's extramural adjuncts. Prior to 2018, our knowledge of the fort's extramural settlement was entirely dependent upon geophysical surveys and some small-scale excavations. These had shown that a civilian vicus had been established in the area south of the fort, to the west of the modern lane, with evidence of several buildings on stone foundations. The 2018 aerial photographs now confirm the presence of extramural activity east of the modern lane extending south from the fort gate. Here faint traces of several stone-built rectangular buildings, one c. 20 m long and 7.7 m wide, with gable ends facing on to the lane can be seen, all presumably fronting a prolongation of the via principalis which would have functioned as the principal road within the vicus. In this respect, Pen y Gaer parallels the situation at many other Welsh sites and beyond, where civilian settlement typically straddles the roads extending out from one or other of the portae principales, with an apparent preference (as here) for the dextral, or right-hand side.Footnote ²⁶ Excavations at the site have shown that such buildings extended at least 100 m south of the fort and that there were also subsidiary streets set at right angles to the main road.

The 2018 photographs do not reveal any significant features along the northern and north-eastern sides of the fort, in areas where previous geophysical surveys had also proved unproductive. Significantly, however, along the eastern and south-eastern sides – which have not been tested by geophysics and where nothing has shown on previous aerial views – the situation has now been transformed by the discovery of a series of buildings that is clearly discrete from the rest of the civilian activity (fig. 11). Attached to the eastern side of the fort and to the south of the prolongation of the via praetoria, the new photographs apparently show a walled annexe extending for a distance of 56 m to the east, with a single ditch beyond the wall. A stone building (or buildings) lies within; though any interpretation of its character and function is complicated by the overlying vegetation, parallels can be found at other Welsh sites including both Caer Gai and Cefn Gaer, Pennal, both in Gwynedd, where a possible mansio and bath-house lay outside the porta praetoria.Footnote ²⁷

FIG. 11. Pen y Gaer: newly recorded extra-mural buildings seen in a drone view taken on 26 July 2018 (vertical view with north to the right), showing the fort annexe (centre-right) and the courtyard building complex (centre-left). (Courtesy of Mark Walters, Skywest Surveys)

Separated from the annexe by what appears to be a road following the course of the south-eastern defences lies a complex of at least two, possibly three, large structures, which at present have few clear parallels within the range of extramural buildings attached to Roman forts in Wales. A square courtyard building (measuring 28 by 25 m) with a range of rooms on the southern side is apparently linked to another courtyard building (measuring 42 by 33 m) to the south whose central yard is bounded by ranges of rooms on the western, southern and eastern sides. Parchmarks suggest quantities of collapsed rubble within the southern and eastern parts of the interior. The drone photographs taken by Walters suggest that these buildings may be part of a single complex. To the east, and set at a slightly oblique angle, is a rectangular, multi-roomed building on an east–west axis (measuring 34 by 15 m), apparently with a ‘corridor’ on its southern side and one large and three smaller rooms at its eastern end.

This is not the place for a full discussion of this complex of buildings, though a couple of general points might be made. In their discussion of the internal layout of military vici, Burnham and Davies tentatively note the tendency for what would appear to have been official buildings to occupy discrete locations away from the main areas of civilian activity, which generally straddled the main roads extending out from the fort.Footnote ²⁸ This is certainly the case, for instance, at Brecon Gaer, where at least three buildings lay behind the main road frontages on the northern side of the fort.Footnote ²⁹ Recently, this point has been explored further by Paul Bidwell and colleagues in their discussion of the annexe and related features at Slack within the wider Romano-British context.Footnote ³⁰ A second point of interest lies in the fact that small-scale excavations within the vicus at Pen y Gaer have shown that some of the buildings continued to be occupied into the Antonine period, with the implication that the fort was presumably garrisoned, though how extensively remains uncertain. The explanation for this relatively long occupation, when compared with the well-attested evacuation of many Welsh forts in the earlier second century, must surely lie in its location on the strategic road linking the fortress at Caerleon with the forts at Abergavenny, which also has an Antonine phase, and ultimately Brecon Gaer in the upper reaches of the Usk valley, which continued to be occupied to the third century and probably beyond. This raises interesting questions about other sites in Wales where there is evidence for seemingly official-looking complexes and for a similar tail of activity extending on into the second century. These and other matters will need further consideration in the future.

(a) Wyndcliff villa, Gwent (NPRN 410589, ST 52766 97583)

This site, known since the late 19th century, has been previously classified as either a villa or a temple. Part of a bronze statuette of Mars, a Bath-Stone roof finial and its hilltop location favoured the latter, as did a comparison with Lydney which lies within the north-western part of a large univallate ditched enclosure, a likely hillfort enclosing c. 5 ha and overlooking the river Severn to the south. The 2018 drought revealed cropmarks of a major later prehistoric enclosure at Wet Wood (NPRN 424030), 700 m downslope and to the west of the villa, possibly a lowland Silurian seat of power in south-eastern Wales and significant in its proximity to the later villa on the hill above.

Wyndcliff villa has shown up well during previous drought summers; a clear aerial view by Sorrell has been published by Evans along with a basic sketch plan,Footnote ³⁵ while a subsequent good general view of the villa and encircling enclosure was noted in July 2013.Footnote ³⁶ The photographs taken on 19 July 2018 reveal the layout of the stone building with exceptional clarity, allowing a new site plan to be produced (figs 12 and 13), while also showing that it is manifestly a villa, not a temple. It comprised a central rectangular block of four or five rooms measuring 24 by 7 m overall, with two distinct and symmetrical square rooms measuring 7 by 7 m located one at each end. Two narrow wings (that on the northern side measuring 4.3 by 2.3 m internally), linked by a porticus, extend 3 m along the eastern (uphill) side. There is also a corridor 3 m wide to its rear. The central block is more and distinctly marked than either of the wings or rear corridor, which appear to spring from it, perhaps suggesting they are later additions. The absence of other ancillary stone buildings is a feature which the structure shares with other known Welsh villas such as New Mill Farm (see below) and Abermagwr, Ceredigion.

FIG. 12. Wyndcliff villa: detailed aerial view taken on 19 July 2018 from the north-east, showing the central block of rooms, the two shallow wings linked by a porticus in the foreground and the rear corridor beyond. (Crown Copyright RCAHMW, AP_2018_5672)

FIG. 13. Wyndcliff villa and defended enclosure: plot of cropmarks revealed on aerial photographs taken on 19 July 2018. (Crown Copyright RCAHMW, Ordnance Survey licence number: 100022206)

(b) New Mill Farm villa, Gwent (SS 9115 6978)

The site of this villa was first recorded in 1976 by the Glamorgan-Gwent Archaeological Trust following ploughing. The most prominent feature is a rubbly mound, c. 65 m long by 25 m wide and standing c. 0.5 m high, aligned north-east–south-west. A smaller mound at its south-western extremity is interpreted as representing another part of the building. A scatter of mortared rubble included sandstone roofing slabs, brick and tile, opus signinum, possible tesserae and pottery of first- to third-century date.Footnote ³⁷ Geophysical survey was undertaken in October 1999,Footnote ³⁸ while the first aerial views were obtained in 1996, revealing the circular enclosure.

High-resistance anomalies on the geophysical survey reveal a rectilinear area measuring 30 by 16 m, open-ended on the south and aligned east–west. It is divided into several rooms, without any discernible wings or stone ancillary structures. The aerial photographs taken on 6 July 2018 show the site as a ripening cropmark with considerable clarity, together with the vestiges of the villa footings (fig. 14). It sits within the south-eastern part of a double-ditched concentric enclosure, c. 116 m in diameter and enclosing 0.9 ha, which has a broad inner ditch and a far narrower outer ditch. The original gateway lies on the eastern side, defined by simple terminals. A narrower ditch running from the southern terminal to the interior may have been designed to enclose the villa on its southern and eastern sides, and thus appears to represent a modification of the pre-existing defences. The villa clearly also partly overlies the inner ditch along the eastern side of the enclosure, which was presumably infilled as part of modifications prior to the construction of the building. What has been termed late prehistoric pottery, found in 1976, may indicate a pre-Roman Iron Age origin for the enclosure ditches.

FIG. 14. New Mill Farm, Vale of Glamorgan: striking cropmarks of the villa (rectangular footings, centre-left) and the encircling prehistoric enclosure crossed by a diagonal farm track, seen from the north-west on 6 July 2018. (Crown Copyright RCAHMW, AP_2018_2797)

(c) Stoop Hill villa, Caldicot, Monmouthshire (NPRN 268154, ST 48326 87360)

Set on a low rise overlooking the Caldicot Levels, next to the M4 and the Severn Bridge, this enigmatic enclosure site has been known since its detection from the air in the 1950s. In addition to some photographs in the Cambridge University Collection (CUCAP), there are others in the John Sorrell Collection and the RCAHMW archives. Its date has been much debated, though a Romano-British origin has long been favoured; this may be confirmed by the photographs taken on 6 July 2018, which reveal a rectangular stone building on a north-west–south-east alignment set within a rectangular enclosure (fig. 15).

FIG. 15. Stoop Hill villa: seen from the east on 6 July 2018 and showing the bivallate rectangular villa enclosure and the footings of the villa (left-centre). The carriageway of the M4 approach to the Second Severn Crossing can be seen at lower left. (Crown Copyright RCAHMW, AP_2018_2992)

The photographs show three sides of a sharp-angled, bivallate rectangular enclosure, measuring 123 m north-east–south-west by c. 100 m north-west–south-east, with a narrow outer ditch – possibly a palisade or fence footing – and a broad inner ditch, 4–5 m wide. The inner ditch cannot be traced on the south-western side where the enclosure may be open save for the narrow perimeter ditch. The southern corner is truncated by the line of the M4 motorway approaching the Second Severn Crossing. It is possible that the enclosure's south-eastern side is fossilised – or buried – by the current field boundary. There are faint indications of a possible smaller, inner bivallate enclosure, c. 60 m², with the footings of the villa on its south-western edge.

The villa measures 19 m north-west–south-east by 9 m and is represented by a simple rectangular parchmark, with no visible internal room divisions. However, a parched wall line extending north-east, away from its eastern corner, suggests the presence of a wing on the building's north-eastern side, facing onto the open space of the villa enclosure.

(d) Walton Wood, Llys y Fran, Pembrokeshire (NPRN 423780, SN 0335 2400)

The cropmarks of a rectilinear enclosure complex, situated on a gentle east-facing slope, were first seen from the air in 1989 by Terry James of the Dyfed Archaeological Trust (DAT); though only partly visible, he described them as a series of paddocks, possibly with various buildings in the vicinity, all probably of Roman date (DAT PRN 14377). On 19 July 2018 they were shown much more clearly to comprise a series of rectangular enclosures extending over an area of c. 170 by 100 m (fig. 16). The best-defined element is a broad-ditched enclosure with sharp angles measuring c. 87 by 62 m, attached to an adjacent enclosure of similar character but set at an angle. Extending away from these are at least two larger enclosures with narrow ditches and sharp angles. Subsequent Cadw-funded geophysical survey by DAT in July 2019 clarified the picture: two overlapping, sharp-angled rectangular enclosures sit at the heart of a wider complex of angular field enclosures, with the visible footings of several roundhouses within.Footnote ³⁹

FIG. 16. Walton Wood, Llys y Fran, Pembrokeshire: cropmarks of the enclosure complex, seen from the north-east on 19 July 2018. (Crown Copyright RCAHMW, AP_2018_5125)

In a north Pembrokeshire landscape of ‘raths’ and curvilinear cropmarks of later prehistoric defended enclosures, this complex of sharp-angled enclosures in a prominent location is difficult to parallel. Some later prehistoric or Romano-British defended enclosures do have a square-built character, including Penpedwast, north Pembrokeshire, which has been the subject of a recent geophysical survey,Footnote ⁴⁰ but these are rarely associated with conjoined or multiple sharp-angled enclosures. Nearby hut groups and field systems of a similar date on Bernard's Well Mountain and at Fagwyr Fran have a sinuous, accreted character.Footnote ⁴¹ The survey would certainly suggest a complex Romano-British farmstead, if not a villa at Walton Wood, with parallels in the sharp-angled rectangular or polygonal villa enclosures now known at Abermagwr, Ceredigion, and elsewhere.Footnote ⁴²

(e) Cwmysgyfarnog, near Llandeilo, Carmarthenshire (NPRN 42342, SN 5674 2325)

Sited at the head of a west-facing slope overlooking the Afon Dulas, c. 3 km north of Dryslwyn Castle, is a typically enigmatic enclosure site discovered on 9 July 2018 (fig. 17). The enclosure is univallate with sharp angles and a broad ditch; it measures c. 60 by 30 m and is aligned generally east–west. There is a likely entrance gap in the northern side, but no clearly visible internal structures. While difficult to date from morphology alone, the enclosure is most readily paralleled with other Romano-British enclosures now known to contain villas. Subsequent Cadw-funded geophysical survey by DAT in July 2019 confirmed an entrance on the enclosure's northern side, but generally returned less information than the parchmarks displayed.Footnote ⁴³

FIG. 17. Cwmysgyfarnog, near Llandeilo, Carmarthenshire: view of the sharp-angled enclosure from the north-east on 9 July 2018. (Crown Copyright RCAHMW, AP_2018_3552)

(f) Glan y Mor (NPRN 409821, SH 4541 6023) and Plas Farm (NPRN 423651, SH 4565 6154), Bontnewydd

On the southern shore of the Menai Strait, west of Segontium (Caernarfon), is the sheltered inlet of Foryd Bay, fringed by good-quality farmland. Aerial reconnaissance here in 2006 identified a series of rectangular structures at Glan y Mor, on the eastern shore of the inlet. The cropmarks showed again with greater clarity on 10 July 2018 (fig. 19), when they were further extended by the identification of an early field system running north-west–south-east between Plas Farm and the shores of the Menai Strait; the fields are rectangular, with at least one visible ditched hollow way.

FIG. 19. Glan y Mor rectangular coastal structures, west of Caernarfon, on 10 July 2018. The structures are showing on a parched ridge of land bordering a coastal inlet with one rectangle visible at lower centre and the second group visible centre-left. (Crown Copyright RCAHMW, AP_2018_4035)

The rectangular structures at Glan y Mor are most directly comparable with the buildings and ditched enclosures identified in the south-eastern part of Tai Cochion. Two groups are visible (at SH 4541 6022 and SH 4540 6034), on slightly raised ground bordering the inlet. To the south is a rectangular ditched structure or possible building, with an internal subdivision, measuring 25 by 13 m. To the north is a pair of almost identical structures; both groups are linked by a linear ditch or boundary.

(g) Glanllynnau, Llanystumdwy (NPRN 423653, SH 4597 3773)

At Glanllynnau, west of Criccieth, on the southern shore of the Llyn peninsula, aerial reconnaissance on 4 July 2018 recorded extensive cropmarks and parchmarks of an early field system running for 600 m east of Glanllynnau Farm in an area of low-lying coastal fields dotted with small ponds and saltmarsh. The cropmarks incorporate trackways and at least one rectangular enclosure measuring c. 40 by 35 m (at SH 461 377), which has internal subdivisions and is surrounded by pitting. It appears to be an enclosed settlement, but has very narrow ditches. While different from the remains noted on the Menai Strait to the north, this may represent a further potential area for Romano-British coastal settlement in north-western Wales.

(a) Carmarthen to Kidwelly, via Idole (SN 413 157) and Llandyfaelog (SN 416 114) (NPRNs 415842, 423818, 423819, 423820, 423833)

The most important discovery in this class is the identification of intermittent stretches of a Roman road heading south from Carmarthen towards the Kidwelly area (fig. 20), on a line more than 1 km east of that previously proposed by James of DAT during aerial survey in 1986, following initial aerial observations in 1976 (RR60d, PRN 7459).Footnote ⁴⁶

FIG. 20. Map showing Roman forts (black squares) and known or postulated Roman roads in south Carmarthenshire, including the newly identified course of RR60d between Carmarthen fort and the potential fort or port at Kidwelly. Confirmed sections of the new road-line are shown as solid black lines. Potential southerly routes for the road between Kidwelly and Loughor are shown as dotted lines. (Crown Copyright RCAHMW)

The newly recorded road line can be described thus: a stretch, 2.5 km long, can be traced south of Carmarthen following a straight section of minor road at Idole (SN 413 157), before being continued by a lane; it has then been recognised on LiDAR by Bryn Gethin, as an agger through fields (NPRN 415842), and from the air as an agger with flanking quarry pits continuing south to Bwlch-y-gwynt Farm (NPRN 423818). After a gap, the road re-emerges as a striking parchmark, 500 m long, flanked by quarry pits, south of Llandyfaelog village at Nantllan (NPRN 423819, SN 416 114; fig. 21); after another break a terraced hollow way and pits emerge beneath the line of the modern A484 road north of Kings Wood (NPRN 423820, at SN 417 097). The last evidence for the road is approaching Kidwelly itself, where the straight line of a former lane emerges from the roundabout and modern bypass north of the town at Llys y Gorlan (lane at SN 414 077), with vestiges of a spread agger, c. 370 m long, visible on LiDAR heading south towards Millands Farm (NPRN 423833).

FIG. 21. Newly identified road agger at Nantllan (running diagonally left to right; NPRN 423819, SN 416 114) with flanking quarry pits and an area of intensive quarrying on the knoll beyond the farm buildings. (Crown Copyright RCAHMW, AP_2018_5029)

No further traces of the road nor indeed a military installation were seen in the undeveloped portions of Kidwelly, a medieval castle borough and port sited on the river Gwendraeth. A few Roman finds have been reported to the Portable Antiquities Scheme, however, principally from the area north-west of the castle (including a denarius of Antonius Pius, a.d. 151–2 (ID no. 362857) and a first-century Polden-Hill-type bow brooch (ID no. 63592)), while a counterfeit denarius of Julia Titi (ID no. 96984) was recovered by a metal-detectorist from the newly identified road-line west of Millands Farm (SN 411 073).

The road seems to have taken a much more southerly course than was formerly believed, thus linking the Flavian fort at Carmarthen and that at Loughor to the south-east (cf. fig. 20). The implications of this discovery are clear: an estuarine fort must exist at or near the town of Kidwelly. Could it underlie the medieval castle, as is the case at Loughor? This illustrates the importance that the Flavian army placed upon seaborne supply in the maintenance of its garrison bases along the coasts of south Wales (viz. Cardiff, Neath and Loughor) and raises the strong possibility of another to the west in Milford Haven. A road linking the putative fort at or near Kidwelly to that at Loughor remains to be discovered, the most direct route being by way of Trimsaran, thence to Llanelli and the bridging of the river Loughor.

(b) Ffynnon-las (Pumsaint), Cynwyl Gaeo (NPRN 407779, SN 6579 3949)

The line of RR62c connecting Llandovery fort to Pumsaint (and the mines at Dolaucothi) and beyond to Llanio has been well documented on the ground, despite its sometimes circuitous route.Footnote ⁴⁷ While various stretches continued to show well in 2018, that at Ffynnon-las also revealed an enigmatic complex of additional features only 100 m away on the opposite side of the modern A482 (fig. 22). Winter aerial reconnaissance here in November 2005 had already identified the earthwork remains of several possible relict water channels,Footnote ⁴⁸ running downslope towards the valley floor only c. 800 m south of the recorded limit of the workings associated with the mines at Dolaucothi.Footnote ⁴⁹ Interestingly, the name ‘Ffynnon’ implies a well or spring and several small water channels issue from the hillslope at this point.

FIG. 22. Ffynnnon-las (Pumsaint) road (parchmark, foreground) and potential industrial remains of scouring channels (parched, right) on 29 June 2018. (Crown Copyright RCAHMW, AP_2018_1828)

These apparent water channels showed with much greater clarity in photographs taken on 29 June 2018. They issue from a block of woodland upslope of Ffynnon-las, running downslope for c. 200 m to the west and fanning out. Two broad, main channels are evident, but these are heavily braided upslope with multiple channels visible. Key to the potential artificial character of these channels, and only revealed in the 2018 drought, is the presence of distinct blocks or grids of more narrow, multiple parallel channels on ‘islands’ in between and north of the main channels. These are reminiscent of areas associated either with sluicing, as one might expect with the working of surface or alluvial deposits, or with the settling of extracted material during processing. Given their proximity to the Dolaucothi complex, these features might represent an extension of the workings beyond the trenches at Cwrt-y-cilion, onto the south-western flank of Allt Ogofau. Unfortunately, it did not prove possible to gain access to clarify these features on the ground, so their precise character, interpretation and dating must remain uncertain; so too must the source of any associated water supply, though Stephen Briggs has recently drawn attention to what might be evidence for hydraulic activity on the hillside above the farm, represented on the 1868 Gardner Wilkinson survey of the mines.Footnote ⁵⁰