The Dolaucothi Gold Mines (grid reference SN662399), also
known as the Ogofau Gold Mine, are Roman surface and deep
mines located in the valley of the River Cothi, near
Pumsaint, Carmarthenshire, Wales.
They are the only mines for Welsh gold outside those of the
Dolgellau gold-belt, and are a Scheduled Ancient Monument.
They are also the only known Roman gold mines in Britain,
although it does not exclude the likelihood that they
exploited other known sources in Devon, North Wales and
Scotland for example. The site is important for showing
advanced Roman technology.
Roman mining methods
Archaeology suggests that gold extraction on this site may
have started sometime in the Bronze Age, possibly by washing
of the gold-bearing gravels of the river Cothi, the most
elementary type of gold prospecting. Sextus Julius Frontinus
was sent into Roman Britain in 74 AD to succeed Quintus
Petillius Cerialis as governor of that island. He subdued
the Silures, Demetae and other hostile tribes of Roman
Wales, establishing a new base at Caerleon for Legio II
Augusta and a network of smaller Roman forts fifteen to
twenty kilometres apart for his Roman auxiliary units.
During his tenure, he probably established the fort at
Pumsaint in west Wales, largely to exploit the gold deposits
at Dolaucothi. Frontinus later restored the Aqueducts of
That gold occurred here is shown by the discovery of a hoard
of gold ornaments in the 18th century, such as the wheel
brooch shown at right. Other objects found in the hoard
included snake bracelets, so named because they were soft
enough to be coiled around the arm for display. All the
objects are now held in the British Museum, and displayed in
the Romano-British gallery. A sample of gold ore was found
at the site by Henry de la Beche in 1844, confirming the
presence of gold.
Roman mining methods
Evidence from the fort (known as Luentinum from details
given by Ptolemy) and vicus show that the Romans worked the
mine during the first and second centuries AD (from circa AD
78 until around AD 125), judging by the occupation of the
fort. However coarse ware and Samian ware pottery recovered
from a reservoir (Melin-y-Milwyr) within the mine complex
show that activity at the mines continued until the late 3rd
century at least. Since Ptolemy's map dates to about 150 AD,
it is likely then that it continued being worked until the
end of the third century if not beyond. The Romans made
extensive use of water carried by several aqueducts and
leats (the longest of which is about 7 miles from its source
in a gorge of the river) to prospect for the gold veins
hidden beneath the soil on the hillsides above the modern
village of Pumsaint. Small streams on Mynydd Mallaen, the
Annell and Gwenlais, were used initially to provide water
for prospecting, and there are several large tanks for
holding the water still visible above an isolated opencast
pit carved in the side of the hill north of the main site.
The larger aqueduct from the Cothi crosses this opencast,
proving the opencast to be earlier.
The water was stored in the tanks and then released
suddenly, the wave of water sweeping away the soil to reveal
the bedrock and any gold-bearing veins beneath. Pliny the
Elder gives a dramatic account in his Naturalis Historia of
the method, possibly derived from his experiences in Spain.
The method is known as hushing and survived in use until the
seventeenth century in Britain. A not dissimilar method is
used today in exploiting alluvial tin deposits, and is known
as hydraulic mining. A smaller scale version of the same
method is placer mining, and both may have been used to work
alluvial placer deposits next to the river Cothi itself,
judging by a large aqueduct which tapped the river a mile or
so upstream, and enters the site at a low level compared
with the other known aqueducts on the site. The water supply
of the aqueducts was also used for washing crushed gold ore,
and also possibly driving stamping mills for comminution of
the ore (Lewis and Jones, 1969).
Small tank (A) near north opencast
One of the first aqueducts was built at a high level on the
east slope of Allt Cwmhenog and tapped a small stream about
two miles away. There is a large tank at its end, where it
sweeps around the brow of the hill onto the west side of the
ridge. A gold vein must have been discovered here, because
there is a large opencast below the tank. Yet the larger and
longer aqueduct (with a gradient of 1 in 800) taps the river
Cothi about 7 miles to the north-east and traverses the same
opencast, so must be later in date.
By contrast, several tanks found on the site did not show a
vein, so were abandoned. The tank shown at right occurs not
far from the north opencast and was probably intended to
find the limits of the deposit located in the adjacent
opencast (Tank A in the schematic diagram below). It clearly
didn't find the vein, and was thus abandoned. The water
supply may have been obtained from a small leat run from a
stream up the main Cothi valley before the much larger
aqueduct was constructed.
Prospecting was successful and several opencasts are visible
below the large tanks built along its length. The only
exception is the final and very large tank, below which are
two reservoirs. It is likely that this complex was used for
washing powdered ore to collect the gold dust.
More leats and tanks can be found below the line of the main
aqueduct, some of which are shown on the map of the site.
They surround the lip of the very large opencast and the
tank shown at right is one which was built on the main
aqueduct. It was successful in finding a vein, judging by
the opencast below, but must have been modified later to
feed a washing table built to the left-hand side (near the
figure in the picture), probably to wash the crushed ore
from the same opencast working. It is labelled Tank C in the
schematic diagram. Similar tanks occur below as the Romans
followed the large vein down to the road and the main
opencast. Most of the opencast workings must therefore be
Roman in origin, since one of the aqueducts has been
confirmed by carbon 14 dating as to predate all modern
workings. Just by the road itself the Carreg Pumsaint has
been erected in the space beside a large mound, now thought
to be a dump of waste material from mining activities.
The existing ponds above and below the minor road from
Pumsaint to Caeo, were probably part of a cascade for
washing ore, the upper tank having yielded large quantities
of Roman pottery from ca 78 to at least 300 AD (Lewis, 1977;
Burnham 2004). The upper pool is known as Melin-y-Milwyr, or
the soldier's mill, an intriguing name that implies that
watermills may have been used here during the Roman period.
Alternatively, it may have been a sequence of washing tables
for the crushed gold ore. A large-scale mill complex is
known from Barbegal in southern France, where no less than
16 mills (in two lines of 8 each) were built into the side
of a hill and supplied with water from a single aqueduct.
There were two lines of parallel overshot mills, the outflow
from one feeding the next below. The mill supplied flour to
the region. Moreover, Roman engineers used sequences of
reverse overshot water-wheels to dewater mines, and the deep
workings at Dolaucothi produced a fragment of such a wheel
during the 1930s when deep mining operations were resumed.
Sequences of such wheels increased the lift, and one
extensive sequence of 16 wheels was found in old Roman mine
workings on the Rio Tinto river in the 1920s. The wheels
were arranged in pairs and could lift water about 80 feet
from the bottom of the mine there.
The tank at the head of the small road from Pumsaint to Caio
was thought to be modern since it still holds water.
However, when the level of the water was low in 1970, it
yielded large quantities of Roman pottery which show that it
is of Roman origin and built early during their exploitation
of the mines. The section shows that it was connected to a
smaller tank just below the modern road by a drystone
culvert in a cascade. The lower tank also holds water but is
in an advanced state of eutrophication. The collection of
fragments included Samian ware and coarse ware from nearly
100 separate pots, and must have fallen into the reservoir
when the mines were in full operation. Analysis of the
pottery fragments showed a distribution of ages from the
late first century AD through to the end of the fourth
century. Since the fort and fortlet under the present
village of Pumsaint ends in the middle of the second
century, it shows that mining continued for a long time
after the military evacuation.
It implies that there is a large mining settlement in the
vicinity of the village of Pumsaint which has yet to be
The exact function of the cascade is related to the methods
of extracting the final traces of gold from the crushed ore.
There were probably washing tables between the two tanks so
that a gentle stream of water could be used to wash the ore
on the rough surface of the tables, the finer gold being
caught in the rougher parts of the tables, and removed at
the end of the process. The cascade would probably have been
built towards the end of the first century when underground
mining commenced following opencast development.
This site yields some of the earliest evidence anywhere for
the Roman use of water-powered trip-hammers to crush ore
(Burnham 1997). The ore was probably crushed on the famous
Carreg Pumsaint, a block of stone erected many years ago
after the Romans had left the site. There are parallels with
similar stones at other ancient Roman mines in Europe, and
the hollows in the block were formed by a trip hammer
probably worked by a water wheel or a 'water lever'. Such a
water-powered hammer would have been moved regularly as each
hollow became too deep, so producing the series of
overlapping oval hollows in its surfaces. The hammer head
must have been of substantial size judging by the width of
the hollows shown in the drawing. The stone is the only
example so far discovered at the site, but is not unique,
and Burnham refers to others of similar shape from Spain. As
one side of the stone became worn, it was simply turned to
reveal another side, so the block could be re-used several
times. When found years after the Romans had left, in the
Dark ages, it gave rise to the legend of the five saints,
who left the impression of their heads in the stone after
being found asleep by the devil.
They followed the veins with shafts and tunnels underground,
some of which still exist on the site. The remains of Roman
dewatering machines were found during the 1930s when the
mines were re-opened briefly.
Drainage wheel from Rio Tinto mines
Sequence of wheels found at Rio Tinto
The most interesting discovery included part of a reverse
overshot water-wheel which is now in the National Museum of
Wales. It was found with burnt timbers, suggesting that
fire-setting was used to help break up the hard quartz in
which the gold was trapped. A similar but larger wheel was
rediscovered during mine operations at Rio Tinto in Spain,
and is now in the British Museum, where it is displayed
prominently in the Roman gallery. The Spanish example
included a sequence of no less than 16 reverse overshot
water-wheels, each pair of wheels feeding water to the next
set in the sequence. Each wheel would have been worked like
a treadwheel, from the side rather than at the top, but it
would have been a hard and lonely activity for the miners
working these wheels lifting water from the mine bottom.
Since the fragment of a reverse overshot water-wheel was
found 80 feet below any known adit or stope, it must have
been part of a similar sequence at Dolaucothi to that in
Spain. Gold mining was sophisticated and technologically
advanced at Dolaucothi, suggesting that the Roman army
itself pioneered exploitation at the site. The construction
of such dewatering machines is described by the Roman
engineer Vitruvius writing in 25 BC, and their use for
irrigation and lifting water in thermae was widespread.
At another part of the mine, on Pen-lan-wen, water would
have been in short supply; a siphon could have transferred
water from the main aqueduct or one of its tanks, but
remains unproven. The vein carries along the hill for some
considerable distance, and has been trenched out. This
method involved excavating the vein vertically down while
keeping the top open. However, ventilation becomes a problem
when fire-setting is used, so three long adits were driven
in from the hillside to the north. They are much wider than
normal galleries, suggesting that their primary purpose was
to allow circulation of air through the trench and permit
safe fire-setting. The upper two adits are still open to the
trench, but the lowest one is currently blocked.
Although there is nothing directly comparable with
Dolaucothi in Britain in terms of the extensive hydraulic
systems, there are many other know Roman mines in Britain,
some of which seem to show traces of hydraulic activity.
They include the extensive remains of lead mining at
Charterhouse in the Mendips, Halkyn in Flintshire, and many
areas in the Pennines. Dolaucothi is most directly
comparable with gold mines in the Carpathian mountains of
Transylvania in modern Romania, at Rosia Montana, and with
the Roman gold mines in north-west Spain, such as the very
much larger site of alluvial mining at Las Medulas and
Montefurado. The Romans may have used slave labour taken
from the local area to work the mine, although the army
itself was probably most directly involved, especially for
their engineering skills in surveying and building
aqueducts, reservoirs and water tanks or cisterns.
Coin of Septimius Severus
There is some evidence that some of the gold was worked at
the site, judging by the finished brooch shown above, as
well as other finished gold products.
A part engraved jewel has also been found in the vicinity.
Such activities would have needed skilled, not slave labour.
No workshops or furnaces have yet been found, but it is
likely that both existed on site. Ingots of gold would have
been easier to transport than dust or nuggets, although a
high-temperature refractory furnace will have been needed to
melt the gold, which has a melting point of 1064 �C. Pliny
mentions such special furnaces in his Naturalis Historia. A
workshop will have been vital for building and maintaining
mining equipment such as the drainage wheels, flumes for
washing tables, shuttering for aqueducts, crushing equipment
and pit-props. Official mints would have produced gold
coins, a key component of Roman currency. After the military
occupation the mine may have been taken over by
Romano-British civilian contractors some time after 125 AD,
although the final history of the site has yet to be
Following the Roman departure from Britain in the 5th
century, the mine lay abandoned for centuries. There was a
revival in the 19th century and attempts to make successful
ventures at the site in the early 20th century, but they
were abandoned before the first world war. In the 1930s a
shaft was sunk to 430 feet in an attempt to locate new
seams. Falling into disrepair and unsafe due to flooding at
its lower levels, the mine finally closed in 1938. It was
during this period that ancient underground workings were
found, and the fragment of the dewatering mill discovered
within. The extensive surface remains, especially the traces
of hydraulic mining, were to be discovered only in the 1970s
by intensive fieldwork and surveying. Although there is yet
no comparable site in Britain, it is likely that field work
will locate other mines, simply by tracing the remains of
aqueducts and reservoirs, and often, if not usually, aided
by aerial photography. Physical remains like tanks and
aqueducts are often recognised by the shadows cast by the
structures in oblique lighting conditions. Thus Tank A was
first seen in early morning light when the sun's rays cast
an oblique light across the hill (Allt Cwmhenog) on which
the structure is situated.
Other local mines
The lead mines of Nantymwyn near Rhandirymwyn village some 8
miles away to the north may also have been first worked by
the Romans, judging by hushing tanks and aqueducts found
there in the 1970s both from fieldwork and aerial
photographs. They occur at the top of the mountain called
Pen-cerrig-mwyn, and the veins were followed underground by
several tunnels leading to the workings. Inside, the veins
have been removed and debris carefully stacked within the
stope. The workings lie far above the later modern mines and
processing plant (now derelict). The later mine was once the
largest lead mine in Wales.
Other local sites
There are Roman forts at Llandovery and as of 2003, in
The United Kingdom's National Trust has owned and run the
site since 1941. Manchester and Cardiff Universities were
active in exploring the extensive remains in the 1960s and
70s and Lampeter University is now closely involved with the
archaeology of the site. The National Trust organises guided
tours for visitors, showing them the mine and the Roman
* Davies O., Roman Mines in Europe, Clarendon Press
* Jones G. D. B., I. J. Blakey, and E. C. F. MacPherson,
Dolaucothi: the Roman aqueduct, Bulletin of the Board of
Celtic Studies 19 (1960): 71-84 and plates III-V.
* Lewis, P. R. and G. D. B. Jones, The Dolaucothi gold
mines, I: the surface evidence, The Antiquaries Journal, 49,
no. 2 (1969): 244-72.
* Lewis, P. R. and G. D. B. Jones, Roman gold-mining in
north-west Spain, Journal of Roman Studies 60 (1970):
* Jones, R. F. J. and Bird, D. G., Roman gold-mining in
north-west Spain, II: Workings on the Rio Duerna, Journal of
Roman Studies 62 (1972): 59-74.
* Lewis, P. R., The Ogofau Roman gold mines at Dolaucothi,
The National Trust Year Book 1976-77 (1977).
* Annels, A and Burnham, BC, The Dolaucothi Gold Mines,
University of Wales, Cardiff, 3rd Ed (1995).
* Burnham, Barry C. 'Roman Mining at Dolaucothi: the
Implications of the 1991-3 Excavations near the Carreg
Pumsaint', Britannia 28 (1997), 325-336
* Hodge, A.T. (2001). Roman Aqueducts & Water Supply, 2nd
ed. London: Duckworth.
* Burnham, BC and H, Dolaucothi-Pumsaint: Survey and
Excavation at a Roman Gold-mining complex (1987-1999), Oxbow