Raymond Selkirk

Raymond Selkirk was a pilot who was engaged to help archaeologists take aerial photographs and went on to study archaeology himself.  This culminated in an important book (Selkirk, 1995) about Roman logistics, which mainly describes his efforts to find previously undiscovered Roman roads around the eastern end of Hadrian's Wall.

Selkirk argued that Roman-era barges navigated much farther up-river (well beyond tidal limits) than most archaeologists had realized. This follows from the inexorable mathematics of payload-to-fuel ratios: heavy goods can always be moved more cheaply on water than on land. So the Roman army moved most of its heavy equipment and supplies by water, and wheeled wagons or pack animals came into play only for short distances or if there was no good route by water.

He also wrote an earlier book (Selkirk, 1983) arguing that stonework beside the river Tees represented part of a Roman weir, not a bridge.  A Time Team TV episode found three distinct bridges there, probably of very different dates, but completely ignored the possible weir.  Selkirk's idea attracted fierce opposition from archaeologists, and a hatchet job by Anderson (1992), who wrote before Selkirk's main book and made a crucial mistake in not understanding that the surface of flowing water is not horizontal.  Selkirk also wrote a later book about his home town, plus a few other observations (Selkirk, 2001).  He died in 2006.

Selkirk's key insight is desperately simple: to be useful for transport, water needs to be deep enough to float a boat.  You don't need much depth, since a ton of cargo can float in six inches of water.  You don't need the water there permanently, just for long enough to float the boat.  And the precise technology does not matter: coracle, voyager canoe, fenland punt, or raft, whether rowed, towed, paddled, poled, or sailed.
Some archaeologists who resist his ideas have failed to realise that most inland waterways must be interrupted periodically in order to maintain their depth.  You do not need an elaborate lock to shift a boat from one water level to another, and anyone who studied in Cambridge will recognise this portage still in use today.

The reason why Selkirk is important to this website is that a large proportion of all Roman forts and civilian settlements are adjacent to rivers.  They are there for reasons of transport, more than for water supply.  Many of those rivers have been altered by humans at some point, but it is rarely possible to prove that those alterations happened in Roman times rather than later.  Names of ancient place names often reflect their waterside situations and one must be alert to the possibility of names based on engineering work (wharf, weir, bridge, etc) as much as on vegetation (reed, alder, corn, etc).

Anderson, J. D. (1992), Roman military supply in north-east England: An analysis of and an alternative to the Piercebridge Formula. Tempus Reparatum, Oxford.
Selkirk, R. (1983) The Piercebridge Formula. Stephens, Cambridge.
Selkirk, R. (1995) On the trail of the legions. Anglia Publishing, Ipswich.
Selkirk, R. (2001) Chester-le-Street and its place in history. Northern Archaeology Group, Washington.

Here follows a tidied-up version of an article posted online by Andy Davison, which we retrieved from Google's cache dated 17 Oct 2014 because it does not seem to be findable now.  It offers a useful summary of Selkirk's ideas for anyone who cannot obtain his books.


Raymond Selkirk

It may come as a surprise to most people that there were very large lakes in central County Durham and as several recently discovered Roman roads stop suddenly at the old shorelines, it looks as if the lakes were used by the Romans for inland water transport.

The former lakes stretch from north-west to south-west of Sedgefield and are now lush farmland known as Preston Carrs, Bradbury Carrs, Mordon Carrs, the Isle Carrs, Ricknall Carrs, Carrside Carrs and Nunstainton Grange Carrs.

The River Skerne traverses the beds of the old lakes from North to South before entering a ravine to the east of Newton Aycliffe. This ravine conducts the river to lower land north of Darlington. The Skerne joins the Tees at Croft-on-Tees.  The Aycliffe ravine is also used by the North Eastern Railway.

We know that the Romans preferred water transport to land transport because their records tell us that road transport was extremely expensive and was 58 times more costly than sea carriage and 16 times the price of river and canal freightage.

The Roman roads in Britain were intended for military police action and took little notice of gradients.  Marching men can cope with severe gradients and cavalry can dismount for climbs and descents.  In medieval times, when very limited local road transport was attempted after the long Dark Age, the medieval horse-drawn wagons followed what was left of Roman roads on level ground but at the hills had to make lots of zig-zags while the Roman road went straight up.  How therefore did the one-mile-per-hour Roman ox-wagons (on the flat) manage the terrible Roman road gradients?

The answers which are only now coming to light are that the Roman roads were for swift military action, postal couriers and occasional pack-animal traffic but all heavy goods travelled by sea, river, lake, canal and even via the smaller streams.  Take a modern horse and hitch it to a well-oiled cart and it will pull two tons on the level.  Hitch the same horse to a barge and the animal will pull a hundred tons.  Here lies our answer.  There was also much more water around in Roman times.  The process of land drying-up has continued over the last two thousand years and is now becoming a major problem for planners and water authorities.  In medieval times, there are records of barge basins in towns throughout Britain where now there is only a tiny stream with hardly sufficient water for a duck to paddle.  In Germany, Dr Martin Eckholdt has found that the Romans navigated canals which until recently were thought to be mere field drains.

On Northumberland's Roman Wall, it is known that the Romans used Greenlee Lough, Crag Lough and Broomlee Lough as reservoirs.  Because of the positions of the recently discovered aqueduct intakes, it is known that Crag Lough and Broomlee Lough are only half their Roman size.  The process continues: a 200 year old boat-house and jetty on Broomlee Lough are now fifty metres away from the shoreline.  Caw Lough which is shown on Armstrong's map of 1769, has disappeared completely.

For reasons which will become apparent shortly, our ancient County Durham lakes will be referred to henceforth as the Mordon Lake, Bradbury Lake, Bishop Middleham Lower Lake, Bishop Middleham Serpentine Lake and the Camp House Lake.  The River Skerne from the old and now dry bed of Bradbury Lake into the dry Mordon Lake via a narrow defile at Bradbury.  Old shorelines seem to indicate that the Bradbury Lake had a slightly higher level than the Mordon Lake with the shore of the Mordon Lake on the 75 metre contour line and the Bradbury Lake on the 80 metre line.  There was probably some kind of ancient lock in the narrow Bradbury cutting.

The Romans would therefore have navigated up to the Mordon and Bradbury lakes via the River Skerne from the River Tees which is known to have been well used by Roman barges.  With the Mordon and Bradbury lakes full of water, the 80 metre level of the Bradbury Lake would have flooded a finger through the Ferryhill Gap which would have enabled the Romans to navigate their barges through the ravine and to the River Wear via the Tursdale and Croxdale Becks.

Eighteenth century antiquarian Dr Cade mentioned a Roman site at Sunderland Bridge (near Croxdale).  This has been located from the air.  It is just above the confluence of the Croxdale Beck and the River Wear.  The Ferryhill Gap is now followed by the main line of the North-Eastern railway.  A lough, now partly dry, in the gap has necessitated a causeway for the railway.  The old 80 metre shoreline of the Bradbury Lake extends right up to Ferryhill Station.  Part of the suspected Roman canal still survives to the north-east of Ferryhill Station.

As the north-west finger of the lake approaches the Ferryhill Gap it passes Mainforth (MA) where Dr Cade reported a Roman camp on the hill and a causeway across the waterway.  The Romans used devices called piscanae which were storage reservoirs alongside canals.  Water was released into the navigation channels from time to time during dry spells.  Such a reservoir would have been useful in the Ferryhill Gap (FG).  The lough in the gap is suspected as the remnants of such a piscina.

A large swamp in the Ferryhill Gap (The Carrs) has been a continuation of the lough (crossed by the rail causeway, NZ 302330). At the southern end of the Ferryhill Gap a length of Roman canal survives and still contains water. This is not the only place in Britain where a modern railway has followed a Roman waterway. A suspected Roman water link, the Biggar Water (NT 050366) between the Rivers Tweed and Clyde has Roman fortifications at both ends. The railway follows the double-ended stream.

In Perthshire, the Pow Water (NN955224) runs in front of the line of Roman watch-towers along the Gask Ridge.  According to the Perth Library, 'pow' is the vernacular for a large ditch.  The ditch joins the Rivers Almond and Earn.  It crosses the same contour twice - a sure sign of the artificial.  There are Roman sites at both ends.  According to Abbey documents of 1244, a Roman canal existed between Kinnoul and Perth.  A modern railway, now dismantled, followed the line of the Pow Water.

Throughout their Empire, the Romans connected many upper reaches of rivers together.  Records tell us that the Romans contemplated connecting the Moselle and Saone so that cargoes could route up the Rhone and Saone from the Mediterranean and then into the Moselle and Rhine via the canal.  The canal would have had to cross a range of hills; this was evidently within Roman civil engineering capabilities.  The canal was not constructed due to political reasons.

Back on our Ferryhill suspected Roman canal, a small ancient dam at NZ 327303 (X) may have also held a side-pond (piscina) - Bishop Middleham Lower Lake (ML) at 81 metres.  This in turn could have been fed by the Bishop Middleham Serpentine Lake (SL) 83 metres.  The Serpentine Lake, now completely dry was ponded back by an ancient surviving dam (Y) just below Bishop Middleham Castle (NZ 329311). This lake was fed in turn by an 85 metre lake (CL) to the west of Camp House and the modern A177 which is on top of a Roman road. This upper reservoir was fed by the Stony Beck and also several springs.  An aqueduct of the simple ditch-type (AQ), part of which still exists (NZ338316), conducted water from the Camp House reservoir via Stony Bridge (NZ338317) to the head of the Serpentine Lake.  Modern quarry spoil has obliterated the lower part of this aqueduct.

As mentioned above, at the junction of the Ferryhill water link and the River Wear, just north of Croxdale Hall, a suspected Roman fort site (NZ275383) has been seen from the air and no doubt this protected the essential transportation water junction.  There have been many finds of Roman material at Croxdale.  Just north of the Ferryhill Gap, Surtees' map marks the findspots of ancient squared stones at an as yet unlocated site marked 'Castle Hill'; in the vicinity of Tursdale.

At the north end of the Ferryhill Gap, a Roman road crosses the suspected Roman canal (now a mere ditch) at NZ302334.  The spot height on the surface of the Roman road at its lowest position in the Gap where it would have crossed the Roman canal is 83 metres.  The route of the road is from South Church to Hartlepool via Kirk Merrington, Thrislington, Garmondsway and Trimdon.  This road has been excavated in several places by Robin Walton of Coxhoe.  A link to Binchester (Vinovia) is suspected from Kirk Merrington.  The very straight section at Stob Cross (NZ 316337) is lined-up exactly on the Roman bridge (NZ 204318) at Vinovia.  The stone and earth dam (Y) below Bishop Middleham Castle formed part of the medieval defences.  An arm of the Bradbury Lake flooded south of Island Farm (IF) all the way east to the A177 (Roman) road.  Island Farm to the east of Bishop Middleham Castle was at one time entirely surrounded by water.  The Bishop Middleham Castle Dam (Y) ponded-back water to the north of Island Farm.  There have been scores of Roman coins found on Island Farm on the highest point just north-west of the farmhouse.  The dam to the east of the castle has a sloping water face and at one time has had a stepped stone air face in the typical manner of Roman dam constructions.  Its shape has resembled (before the robbing of its stone-cladding), that of the Roman dam at Qasr Khubbaz on the Euphrates.

There were hundreds of dams in all the Roman provinces but few have been recognised in Britain.  Nevertheless, they are there and advancing skills at recognising Roman civil engineering techniques will increase the numbers found.  Several dozen Roman dams survive in Spain, France, North Africa and the Middle East.  Roman records tell us of others which have long since disappeared.  Nero had a cascade of three dams for ornamental purposes across the River Anio which adjoined the garden of his villa.  One of the dams was 129 feet high.  Tacitus tells us that the Roman army dammed a branch of the Rhine in order to deepen the water to flood a military canal.  There are more right under our noses.

With regard to the Bishop Middleham area, where the now dry Serpentine Lake met the A177, there are some strange earthworks (TP) on the north side of the Skerne and the west side of the A177 / Roman road.  These earthworks have been severely damaged by earthmoving machines but could be the remains of a Roman supply base.  The Bishop Middleham Serpentine Lake (83m) was at a higher level than the Mordon (75m) and Bradbury (80m) Lakes and this level was maintained by the Bishop Middleham ancient dam (Y).  It is interesting to note that the artificial lake ponded-back by the dam would have protected only one side of Bishop Middleham Castle whereas it would have completed a total water barrier around Island Farm where a Roman site is suspected.  This adds weight to the theory of a Roman origin for the dam.  No doubt the dam was incorporated into the later medieval defence system.  The Bishop Middleham Serpentine Lake was in turn fed from the aforementioned higher lake (CL) beside Camp House.  The Camp House Lake has been enclosed in a depression marked by the 85-metre contour line.  The only outlet of this lake was the suspected Roman (and later medieval) aqueduct (AQ).  The Camp House Lake was fed by the Stony Beck (dry at the time of writing) and several springs, also dry at present.

South-west of Sedgefield, there runs a Roman road to Great Stainton and Sadberge.  East-north-east of Mordon (NZ343269), a suspected Roman road is followed to the south-west by a bridle path (Peters Lane) to Harpington Hill Farm at Mordon.  In line with this Roman branch road to the shoreline of the former ancient lake, are earthworks believed to be of Roman origin (NZ326265).  A barn close by, recently demolished, contained reused Roman bricks and tiles.
br> The earthworks at Mordon (M) on the former lake shore are of Roman shape but immediately below them are what appear to be the remains of a rectangular barge basin.  The grass covered breakwater is still in place and the suspected entrance for vessels is at the north-west corner.  Parts of a suspected inner harbour wall are just visible in places.  Although these features were seen from our survey aircraft some years ago, perhaps we may be excused for failing to appreciate their true purpose when there was not a drop of standing water to be seen for miles.

It looks as if the Romans had an east-west ferry service operating across the lake.  This of course was in addition to the through south-north traffic from the Tees to the Wear and trade between all the Romano-British farms around the lake shore.

Ferry services across lakes were quite common and cheaper than circuitous road routes in Roman times.  Pliny the Younger who was the governor of the Roman province of Bithynia (Northern Turkey) c AD112, wrote to his Emperor Trajan describing a canal-construction project intended to cut down the cost of the transport of farm produce.  His letter states: There is a sizeable lake in the area of Nicomedia across which marble, farm produce, wood and timber are easily and cheaply conveyed by boats right up to the main road, from which, with great effort, and even greater expense, carts take them to the sea.  To connect the lake with the sea would require a lot of labour, but there is no shortage of that in this area.

In the case of the Mordon Lake, it looks as if the Romans would have no need of a canal to the sea as the lake was connected with the River Tees (Tisa F1.) to the south via the river Skerne.  To the north, a water link through the Ferryhill Gap made a connection with the river Wear (Vedra F1.)

The Romans preferred water transport to road travel even in Italy where the roads were excellent.  A Roman letter, still legible, recovered from Vindolanda, Northumberland, written by a Roman businessman, tells us that: If he doesn't get his cargo of hides soon, he will go bankrupt.  He adds that: the Roman roads in Britain are very poor and are not fit for wheeled transport!  Roman water logistics and transport have been neglected by generations of British historians and archaeologists, possibly because they assumed, incorrectly, that the Roman roads were the ancient equivalent of motorways and could not handle all necessary traffic.

In the Rev J Collinwood Bruce's Roman Wall, written in 1867, on page 76, he challenges any wagon driver of his day to attempt to take a laden wagon up the steep inclines of the Military Way of the Roman Wall.  No doubt the Roman regiment 'Barcarii Tigrisenses' (Bargemen of the River Tigris), who were stationed at South Shields, handled all the heavy cargoes up the Tyne.  The Roman civilian boatmen, the 'Utricularii', also assisted with waterborne logistics and their insignia was an inflated goatskin.  This is not an ancient form of lifejacket as has been suggested by an impractical historian, but a method of decreasing the draught of a barge or even a large punt when it encountered very shallow water. Dozens of pairs of inflated skins were strapped underneath the craft, almost lifting it out of the water like a hovercraft.

It seems that nearly all heavy cargoes in the Roman period were transported by water and the famous ox-wagons merely plied between jetty and fort, thus laying false clues for archaeologists who saw the wheel-ruts in the fort gateways and assumed that the oxen had made the whole cross-country journey.  The fact that the one-mile-per-hour oxen (on the flat) which worked five hours per day, couldn't take the road gradients, has passed without attempting to find an answer. That answer is WATER TRANSPORT. There are no Roman references anywhere to long distance cart transport.  When Roman Emperor Claudius travelled to Britain for the final battles of the Invasion, half of his journey across France was made by river and canal.  The Roman records unfortunately do not mention the names of the rivers and canals so used.

Our Mordon and Bradbury lakes were therefore not barriers to the Romans but presented several advantages.  The most important was the link for cheap cargoes by waterborne transport between the Tees and the Wear.  Another advantage was the easy transport of goods to destinations all around the lakes' shores, and many Romano-British settlements have been located on the former shoreline of the Mordon Lake which is about the 75m contour on modern OS maps.  No doubt the Romans fished extensively and also probably laid down beds of freshwater oysters.

In the centre of the former Mordon Lake are two farms with the curious names 'Great Isle' (NZ303269) and 'Little Isle' (NZ307270). These two farmhouses are on humps which rise above the 75m contour line.  They were built on former islands in the lake.  Great Isle Farm has an obvious harbour for small craft.  This is now a pasture in which horses graze.  An east-west road crosses Great Isle.  It is constructed of river-washed cobbles and these must have been brought a great distance, probably by Roman barge.  At the west side of the former island, the road stops suddenly at the 75m contour.

On the southern shore of the former Mordon Lake are farm buildings at High Grindon (NZ323243), which are on the site of a known deserted medieval village.  From the air, there is an obvious Romano-British farmstead (NZ320241) WSW of the great barn.  From Stainton Hill House (NZ323223), a bridle-path runs north to High Grindon.  This track lies on top of a recently identified Roman road.  From High Grindon which is perched on the ridge above the southern shore of the former lake, the Roman road is lost on the heavily cultivated hillside.  A railway line, now dismantled runs just above the southern shore of the dry lake.

To the east, there is a burial mound at Howe Hills Farm (NZ 328246) and this tumulus on an old shoreline suggests that the lake's edge in prehistoric times was the same as in the Roman period.

Much more research remains to be done.  The recently discovered Roman road to Mordon appears to have come from the Hartlepool direction via Three Gates Farm (NZ 455308).  Another recently discovered Roman road also heads for Three Gates Farm and this one turns eastwards from Sedgefield Church, under the A689 road beyond Beacon Hill and thence across farmland to Embleton and Three Gates.

Roman surveyors had tidy minds and although the very straight track (Peters Lane) to Mordon through Harpington Hill Farm ends suddenly at the suspected Mordon Roman fortifications and barge basin, the same alignment may have continued westward from the opposite side of the lake.  An investigation of the former western lake shore located a suspected Roman road exactly in line with the Mordon road.  If the road is genuine, then it is lined up on Staindrop (to the north-east of Barnard Castle.

It seems that the lakes have been drained bit-by-bit over the centuries as there are the remains of rig-and-furrow medieval agriculture here and there below the Roman-period shorelines.  Sometimes, the end of the rig and furrow marks the Roman shoreline.  Draining has continued into recent times and there is a Northumbrian Water pumping station at position 'Z' (NZ330300) to the south of Bishop Middleham.  The modern dam associated with the pumping station is to keep water out of the valley whereas the suspected Roman structures were to keep the water in.

This is merely an initial report and much more research needs to be carried out.  There is vast scope for students of archaeology, history, geology and geography, with uncluttered vision.  If we can find a water link between the Rivers Tees and Swale, and a likely route is via the Staindale Beck across the low plain near Appleton Wiske, to the River Wiske, then we will have completed our knowledge of a heavy goods Roman water connection all the way from the River Wear to Oxfordshire.

Paul Smith brilliantly discovered this map in an old Facebook post.