ROMAN ROAD Pen y Gwryd to Traeth Mawr


                Sources:   1795   Evans J - Map of North Wales (ca 1795)

1818   Ordnance survey drawing 2in to 1mile in British Library cat no 305.

                               1838   Ordnance survey 1in/mile 1st ed  as reprinted Harley/Oliver, Old Series Ordnance Survey Maps, vol 6 (ISBN 0903541440)

1891   Ordnance Survey 6” to mile plans, 1st edition 1891, as read

B        James Bransby – A description of Llanberis & the Snowdon district (1845) (Caernarfon public library)

GAS    refers to Gwynedd Archive Service, Caernarfon

GAT   refers to Gwynedd Archaeological trust records.

HH      Edmund Hyde Hall, A Description of Caernarvonshire (1809-1811) ed Emyr Gwynne Jones 1952 (Caernarfon public library)

J         Jenkins D E, Bedd Gelert; its facts, fairies and folklore (1899) (Caernarfon public library)

M       I. D. Margary, Roman Roads in Britain, vol 2, 1st ed 1957.

                                    P        Pritchard R T, The Porthdinllaen turnpike trust Trans Caerns Hist Soc vol 20 (1959) pp87-98     

                                    R        Royal Commission on Ancient & Historical Monuments (Wales) Caernarvonshire vol ii     

                                   W   Edmund Waddelove, The Roman Roads of North Wales, 1999. (ISBN 0950680311)

                                   WW          Williams W Observations on the Snowdon mountains (1802)  (Caernarfon public library)




A remarkably direct road, considering the terrain, appears to run from the Roman camp at Pen y Gwryd down to Llyn Gwynant and then through the hills to Cwm Nanmor (Nantmor on some maps and not the same location as the hamlet of that name), whether to meet up with M route 68, or go down to the Glaslyn estuary, is not clear. J p278 has a description which reflects this possible route, “the Roman bridle path from Dinorwic [he probably means Dinas Dinorwig near Llanddeiniolen, not the modern village of that name, since other C19th writers associate this Iron Age hillfort with the Romans] coming through the pass of Llanberis over Gorphwysfa, down past Cwm Dyli and through the land of Hafod y Rhisgl, until it crossed [Roman Bridge –see below]. It then kept under Gallt y Wenallt, along the brink of Llyn Gwynen, behind the hill between Bwlch Myrchan [also appearing on maps as Mwrchan, Mwydlehen, Mwyalchen] and the lake, passing Bwlch Myrchan, and fording the river right in front of that farm. Thence it passed Plas Gwynant and joined the Segontium road…beyond Blaen Nanmor. It need not puzzle anyone to find this path in Blaen Nanmor and Nant Gwynen, as the traces are well defined”.


There are 4 possible routes down into the Gwynant valley from Pen y Gwryd:


1)       The present A498 follows a twisting but regular gradient, rather narrow by modern standards, and is the final version of the Porthdinllaen to Capel Curig turnpike road, not there in 1818 but on map by 1838. The trust was created by an Act of 1803 under the initiative of William Madocks, as part of the proposal that persisted for over 50 years to develop Porthdinllaen on the Lleyn peninsula as a substitute to Holyhead for traffic to Ireland. No detailed trust records exist for this road, but various incidental references can be traced. Lack of finance was a persistent problem for the trust, which never achieved its ambitious goal, and never made a profit. Even after it was formed there are numerous references in quarter sessions records to “presentments” for its improvement, eg GAS XQS/1815/119  (ie 1815) which refers to the “common and ancient King’s Highway from Caernarfon to Llanrwst” being in a state of disrepair for 7000 yards between “Pont y Gwychyd” (presumably Pont y Gwryd) and Capel Curig; and again GAS XQS/1815/123 “part of the common highway in pa. Beddgelert, beginning at Llidiart Llyn y dinas and continuing towards Pont y Gwychyd, 10,000 yds. long and 18ft wide, is in a state of disrepair” although this second application was withdrawn. The bridge itself features in GAS XQS/1815/121, as do bridges at Pont haefod y risk over the river Afon y chyd risk – presumably the side stream at Hafod Rhisgl 657528 (GAS XQS/1815/122); a bridge over the river Hafod y Borth, commonly called Pont Shone Dinbych – apparently 605490 (GAS XQS/1815/120); and Beddgelert bridge itself (GAS XQS/1814/141). However, in relation to this part of the route we do have (GAS XD2/17258) a letter dated Sept 3 1832 by John Williams of Tremadoc, presumably either Secretary or Surveyor to the trust, he was later County Sheriff, commenting that he had been repairing the road from Capel Curig to Beddgelert. Mr. Faudrey and other Commissioners were anxious to make a new line to ascend “Gwryd hill” and he had been instructed to make a survey for the next Grand Meeting. Although this would be a great expense for their depleted funds he feels it to be necessary if enough were to be attracted to pay tolls. Clearly this document must refer to the modification which now constitutes the main road, and reflects the concern that the first attempt, particularly 6573 5455 below, was as HH observes a deterrence to potential traffic.


2)       A packhorse trail going straight down by the stream and marked on 1891 as “Ancient Trackway”, ie continuing the trackway similarly marked along Dyffryn Mymbyr from Capel Curig to Pen y Gwryd. R lxvi=GAT3877 refers, and this is the route mapped by Evans in 1795. It will be what WW p57 describes as “the road is very bad, circuitous, and winding, and absolutely impassable when the floods are violent after a fall of rain”, and again p73 “a road, or rather a mere right of passage, goes through [Capel Curig] from Bethcelert to Llanrwst; but it is exceeding rugged and uneven”. 1891 marks the northern end of the path from 6589 5547 to 6584 5530 as being on the W side of the stream; although some very slight traces are to be found there, there is a more clearly defined track above the E side, coming down to the valley bottom in a reverse curve to cross Nant Cynnyd at 6571 5502. Continuing S from there it can be very clearly seen from the main road above. The whole trail is mostly a hollow way, sometimes two or even three in parallel, up to 2m deep with a wet bottom even in dry weather conditions. There are no signs of deliberate construction. It is no longer used as a footpath, and is now very difficult to walk. For much of its course boulders have been piled up on one side or the other, whether to clear out the bottom, to funnel livestock, or as a marker, one can only guess. It crosses Afon Trawsnant at 6554 5470, marked as a ford on 1891 and still easy to negotiate in dry weather, and becomes a used footpath/RoW S of that point. After passing a cluster of prehistoric sites (J p288 et al, “Muriau’r Dre”, GAT803 & 883 etc) just N of Cwm Dyli power station, it again fords the stream at the valley bottom and continues on the E side of the water as a hollow way. This stretch is no longer a footpath/RoW and becomes steadily more indistinct as the meadows open out. Its course in the vicinity of “Roman Bridge” (see below) is no longer apparent. The 1891 references to “ancient trackway” reappear from time to time between Llyn Gwynant and Llyn Dinas, eg as the approach road to Bryngwynant youth hostel 6412 5141, and in the vicinity of Ysgoldy bridge 6350 5144 (GAT11680). This is a clapper bridge very similar in type to one at Castle Upon Alun in Glamorgan (Gwyndaf Breese, The Bridges of Wales (2001) pp 160 & 170), and would logically be a dry foot pedestrian alternative to the ford marked on the trackway in 1891 a little further downstream at 6328 5125, where the river is still two to three times its normal width. It must then have passed Hafod Y Llan farm and run in with its approach road to the present main road, there is no further field evidence for it. 


3)       The by road running to W of and below the present main road until it rejoins it by the N end of Llyn Gwynant. This is of a very substantial construction, still a public albeit gated road used by all terrain vehicles, branching from the present main road at 6592 5547. This is the original course of the turnpike, it is not on 1795 but was completed and in use by 1811 (HH p38), mapped as it is now in 1818, OSD 305 of which it forms the western boundary. The question must be considered whether it has been superimposed on an older route, as is proposed to have happened W of Capel Curig (W ch5), since there are suggestions of Roman surveying, eg the upstream zig zag at 6584 5480 (photo), and the gradient of the straight downhill pitch S of it at 6573 5455, which is what led to its later abandonment – cf. HH’s comments (p220), “the present line…even in its improved state extremely steep and fatiguing”, but there is no way of testing that hypothesis and the apparent conflict with route 4 (see below) is against it. At 6571 5386 it becomes the service road for Cwm Dyli hydroelectric power station built in 1906. The present bridge is on a concrete base and is of this date since it is identical to the bridge over Afon Glaslyn at the site itself, but there is clear evidence of another small upstream zigzag beside it to accommodate its predecessor, of which no physical trace remains. The road continues down the hill, finally running in the angle of the slope between valley bottom and steep hillside and closing with the present main road at 6507 5225, where it was modified between 1838 and 1891.


4)       There is a track descending into this valley not directly from Pen y Gwryd itself but diverging 6513 5531 from the old road leading up to Pen y Pass (see Llanberis Pass section). This presumably is what Jenkins refers to at p278; it is marked on 1795 and 1891 and referenced by R lxvi. It is now a RoW, there are hollow ways on it from time to time but its main characteristic is that it comes straight down the brow of hill between Afon Trawsnant and Nant Cynnyd, apparently with a zigzag on the steepest incline. After crossing the packhorse trail at right angles at 6557 5465 it fords Nant Cynnyd at 6562 5461, the mapped 1891 path/RoW then climbs the hill on a steep but viable course, but there appear to be no traces on the ground. In practice walkers use what appears to be a steep, somewhat degraded terraceway running on the E side of the wall line marked on the map from 6563 5467 through 6565 5441 to meet the early turnpike at 6568 5436. This wall is hardly more than an irregular revetment of the terrace, it is not a dry stone wall in the conventional sense. However, the line continues on up the hill beyond the modern track still as a wall/terraceway, although the terracing becomes progressively less distinct. Examination of the intersection between the two clearly shows that the construction of the turnpike terrace has interrupted the gradient of this line, indicating thereby that it is older. Although this latter section is not a RoW, walkers still use it to gain access to the prominent viewing point and car park on the main road above at 6582 5416.  However, in its present form the remnants of the wall end at a bluff at 6576 5417; to the E of the bluff at 6577 5411 there is some fragmentary evidence that could represent a way through, but further S the hillside is completely overgrown with bracken in summer conditions, so that nothing can be observed. More work needed here. The interest of this possible route is that it appears to be surveyed and laid out to stay on open higher ground as much as possible, crossing the valley of Nant Cynnyd at its narrowest point, and completely avoiding the  “valley bottom” characteristic of the packhorse trail. NB Roman era hut circle recorded  at 6551 5243 (RCAHMW 302895 and GAT4524). The reference at GAT7980 “65695182” to “Bont Du milestone” is a grid reference error, probably should be “65691822” at Bont Du near Barmouth.


Beside Llyn Gwynant itself the present main road runs at the lakeside with no obvious signs of any predecessor to it, and little or no scope for any course further up the valley side. However at 6456 5172, just S of the side stream Afon Cors-y-Celyn, a footpath/RoW leaves the road and strikes due S, initially as a modern access track, with a wall to the E of it. A little way up, there is a suggestion between this wall and the stream of a hollow way. At 6455 5159 the RoW leaves the track by some modern steps and goes directly up the hill, whilst the access track heads for the same destination by a series of zig zags, already mapped on 1838. On the RoW the first definite evidence this could be a road is where a small stream has evidently been culverted at 6455 5158, since the culvert is some 6-7m wide running skew. A slight scatter of metalling is then in evidence, followed by a hollow beside the stone wall on the E, mostly obscured by rhododendrons which have run wild, and form a significant feature of this side of the Gwynant valley for several miles. This is a very steep section, which would explain the creation of the later alternative. The zigzag comes back to the line at 6455 5147 before turning away on a further leg, but the original track/RoW continues straight up the hill, the zig zag coming back onto it for the final time at 6454 5135. From this point on the gradient eases and it becomes a well metalled track between stone walls, 4-5m wide (click for photo). Along here there is no evidence of recent use, although a derivative track leaves the line by turning E at a sharp angle at 6451 5129. Our road goes straight on, although at 6448 5119 the mapped RoW branches SW across a meadow (footpath on OS 6”) going past two old barns before entering the woodland as a stony track. At 6446 5115 we pass through a gateway, and the line becomes a terraceway, with a low stone revetment on the E side, and a wall now on the W side sitting on the outer edge of the terrace. The obvious road peters out at a crosswall at 6432 5092, but present day walkers still come this way and a rather twisting and obstructed footpath continues through the rhododendron thickets for 200m or so, finally breaking back onto the RoW through the wall on its E side at 6418 5076 (photo). On this section there are occasional hints of terrace in the undergrowth, but it is very difficult to tell, and what would appear to be this older line is not really detectable if walking in the opposite direction. GPS readings are absent or unreliable in this thickly wooded area, and have been crosschecked against OS mapping. There is a GAT9746 reference to “stepping stones” in this wood at 63625055, the SMR shows this is a Forest Enterprise reference to a 2nd edition OS description, now untraceable in the bushes.


To summarize, the ascent of the hillside has two distinct versions, one going straight up, the other easing the climb by various small detours. This seems typical of the variations that W has noted at other locations, and logic points to the direct version being the older, with very probably a packhorse trail as the later, version.


Where the two come together again the RoW is already a well metalled track about 3m wide, and this continues straight on S through the wood. This is currently wellgrown Sitka spruce, so there is no grass underfoot, and it is this lack of cover that has allowed the metalling to become exposed, as walkers to Bryn Gwynant youth hostel regularly use this section of the route. Where the path breaks out into open scrub, the turf cover resumes and the line immediately becomes less obvious. A ladder stile at 6414 5061 over a long since patched wall shows this has not been used as a road for a considerable period. Back in the wood the metalling is again exposed (link to photo), occasionally there are traces of edgestones on the lower side, and at one point two parallel ditches. At 6399 5026 another stile (this time with an overgrown gate beside it) brings the path out into the open, so the metalling disappears again. However the footpath does appear to faithfully follow the original line, and where a stream is crossed at 6389 4999 there does appear to be a small, collapsed, zigzag on the S side, although it is not very obvious. A barely perceptible terrace on the side of a small hillock at 6388 4968 leads to the crossing of the Afon Llynedno at 6382 4958, where there is a small weir with a paved ford on the upstream side. Finally the line becomes a lane between two walls again, running into the tarmac road which now joins it from Plas Gwynant to the NW at 6371 4944. That road now forms the northern link between Blaen Nanmor and Nant Gwynant, effectively replacing the disused section just described. J pp236-8 refers to a Daniel Fawdrey buying the Plas Gwynant estate in 1803 and making roads to improve the land, and since there are no other roads on the estate this would seem to refer to this minor road and/or a variation of it at the N end, which is now a footpath with a bridge. 


All this constitutes a near dead straight line through very broken country, and it is interesting to note that a transmission line from Cwm Dyli power station follows the same course albeit 200m or so away, reinforcing the impression of a carefully surveyed route. It has clearly been abandoned for a very long time, having lost its original purpose and been superseded by the valley bottom road to Beddgelert, a village of C6th origin (R p16) which it completely ignores.


For several miles SW the present minor road through Blaen Nanmor follows the same line, but has not yet been examined in detail. J p279 considered it to be Roman, including the lane coming up from Plas Gwynant. R vol 3, p115 refers to “remains of a stepped trackway” between 6353 4863 and 6362 4865 which is at right angles to this route, no further details. More research is needed here, and on the question of possible destination. It crosses “Sarn Helen” at Bwlchgwernog 612 453 and runs down to Traeth Mawr at 601 435. nb RCAHMW 23504 “old trackway” Llwbr Rhys Goch 5985 4514. There is a possible forward course through Ynys Fawr 600 424 and Ynys Ceiliog 597 404 where a track/RoW cross the next piece of marsh and over Bwlch Glas  597 398 (a curious name for this location). The continuation of this alignment brings us to the old crossing of Traeth Bach at Abergafren 601 379, W p and Ogilvy’s 1675 road from Welshpool to Caernarfon: Nb Ty’n y Berllan “Roman” settlement 6003 3935 (GAT 1192),  “ancient trackway” 6033 3891 RCAMHW 23632 and “trackway” 6000 3800 GAT 3887. Further work needed here. 




Bwlch Ehediad and “Roman Bridge”


Just above Nant Gwynant a  “county road” branches E from the early turnpike at 6562 5284, now a footpath which climbs the hill very steeply indeed as a narrow hollow, crossing the main road at 6572 5272, where it becomes nominally a bridlepath and continues the steep ascent to cross the skyline by a prominent tump of rock (link to photo) at Bwlch yr Rhediad or Ehediad (both forms appear on maps of different dates). Thereafter it crosses the headslopes of the Afon Cwm Edno by Sarn Diwaunydd (“stepping stones” ?mediaeval of GAT 8266) and gradually descends the hillside directly to another Roman Bridge at 7112 5158 (see GAT 7120 & W pp130-131 & references). This route graphically illustrates what a “county road” of 200 years ago could mean. It was mapped by Evans in 1795, described as a county road by HH fig24, is not shown on 1818 but reappears on 1838 as a path, and E of Sarn Diwaunydd as a road. It is difficult to imagine it would ever have been capable of sustaining wheeled traffic, and there is no evidence it is used even as a bridlepath now; the ascent is so steep, rough, and overgrown I doubt whether any rider would wish to attempt it today. Nonetheless there is some scattered metalling, both surface gravel and foundation stones, and the occasional culvert. Beyond the pass it is usually a very rough narrow hollow, and 1891 again entitles it “Ancient Trackway”. At the Sarn there is some suggestion of a rough embankment where it enters forestry land. Nonetheless it has few regular constructional characteristics, it seems a clear example of a route that “just growed” rather than had been constructed. Across the moor the modern walker needs the very comprehensive waymarking that Conwy council has provided. There are a number of archaeological sites in its vicinity, particularly Roman era round huts GAT 4519 and 4520 S of it in what is now conifer forest. The route branching S via Moel Goch clatter bridge (“post mediaeval” GAT 8489) appears to be secondary. By descending to the Lledr valley it does provide a means of communication, however primitive, between Nant Gwynant and Llanrwst/Conwy, but the creation of the turnpike, although a mile or two longer, would have made it superfluous. HH p220 “Before the present turnpike roads were made, [Beddgelert] parish was accessible only by rough and uneven bridle roads”. 


W of 6562 5284 there was formerly (1891) a footpath through Hafod Rhisgl across the meadows to the “Roman bridge” 6487 5273 (GAT 11719 = R, J p277 ff, click for photo). The RoW has now been diverted more to the S away from the farm, breaking the immediate connection. The bridge is a substantial clatter bridge, extended following a change of course in the river bed, and a line of RoWs can then be traced from that point S and then W through Bwlch Cwm Llan at 606 521, Pen ar Lon 584 525 to Rhyd Ddu, Drws y Coed (Bp52 “on the S side of Drws y Coed may be distinctly seen the traces of a still more ancient road, Mignedd, which forms a passage…to Cwm Marchnad” – see Segontium-Aberglaslyn section)), and ultimately the Llŷn peninsula. The name of Cwm Marchnad itself may be significant if it indicates an old market site, as B suggests. Further investigation will show to what extent this line on the map is represented by features on the ground, if so it would constitute a very direct long distance route from the peninsula to Llanrwst and points E, which suggests the possibility of a drovers’ road.



Copyright John Byde.    Revision date 30 December 2003