A terrain model you can argue with. Open this to understand this post: https://sarsenroute.netlify.app/
Fifty of the fifty-two sarsens at Stonehenge are thought to come from West Woods, near Marlborough, about 25 km north of the monument, across the Vale of Pewsey. How they were moved is unknown, and several routes have been proposed.
This tool does not propose another one. It asks a narrower question:
What would a haulage route look like, if the only things that mattered were the shape of the ground and the water on it?
How to read it
The map shows the cheapest line from the sarsen ground to Stonehenge under four assumptions you control. Move a slider and the route changes, along with its length, climb, crossings and elevation profile.
There is no correct setting. The route is a function of the weightings, and the weightings are a hypothesis about what a hauling party feared most. The tool makes that dependence visible rather than hiding it behind a single line.
The dials
Climb penalty — the cost of gaining and losing height. Turning it up lengthens the route and flattens the gradients, and then does something less obvious: it drives the route into the valleys. Gentle grades are found along watercourses, so an effort-averse hauler ends up in the wet.
Wet ground penalty — the cost of level, low-lying ground: floodplain, marsh, the sort of place that carries a sledge badly in winter. Pushes the route up onto dry chalk.
Watercourse crossing cost — charged at every stream or river crossed. At zero the route crosses thirteen watercourses; at its highest setting, one. Dry ridge use rises from 17% of the route to 46%, at the price of about 5 km of extra length.
Climb pulls one way and water pulls the other. Which corridor emerges depends on whether you think a hauling party feared slopes more than bogs.
The descent prior
The fourth dial is different in kind. The first three are guesses about behaviour; this one carries evidence about a place.
Two crossings of the northern scarp are close enough in cost that the terrain alone cannot separate them. They are 2.8 km apart:
New Town / Workway (N2) has multiple roads and tracks on Andrews and Dury's map of 1773. There are usable tracks and a major road there today, confirmed on the ground.
Huish (N1) has only one irregular track down the face on 1773. It is a narrow holloway, on the ground; the hill is a steep isolated lump and tracks go around it.
The prior discounts N2 and surcharges N1, and only within a 600 m catchment on the scarp at each — nowhere else on the route. Three strengths, and it takes the strongest to settle the question. Across all 120 combinations of the other three dials:
| prior | descends at N1 | descends at N2 |
|---|---|---|
| off | 82 | 17 |
| moderate — N2 ×0.75, N1 ×1.25 | 51 | 54 |
| strong — N2 ×0.5, N1 ×1.5 | 13 | 105 |
With the prior off, the model prefers Huish in most settings. That is worth understanding rather than correcting away: a 50 m terrain model averages a scarp face, so a smooth unused hillside beats a deeply worn track beside it. The model was preferring the easier-looking slope precisely because nobody had ever used it.
At moderate strength it is close to a coin toss. Only at strong does the worn descent dominate.
A flip to N2 does not mean the sarsens went down that way. It means the model has stopped preferring an unused smooth face over a worn crossing 2.8 km west.
The elevation profile
Under the statistics is the long section of the current route, on a fixed vertical scale so settings can be compared directly. Two things are worth watching as you move the sliders: the depth of the dip in the middle, which is the Vale of Pewsey, and how abruptly the ground rises on either side of it. Those two shoulders are the whole difficulty of the journey. The rest is downland walking.
What is behind it
Importantly the model was built blind with no input from published suggested routes, including mine, or archaeology so, for instance, passing through Marden henge is purely judged on the terrain.
The terrain is OS Terrain 50 (Ordnance Survey, Open Government Licence), 50 m posts, over a frame of 22 × 32 km from south of Stonehenge to north of Marlborough. Watercourses, settlements and tracks are from OpenStreetMap (ODbL). The descent evidence comes from Environment Agency LIDAR Composite DTM at 1 m (OGL) over the two scarp crossings, and from Andrews and Dury's map of Wiltshire, 1773.
Routes are solved on a 40 m grid — about 441,000 cells and 3.5 million connections — with 360 combinations pre-computed so the sliders respond instantly.
The Kennet & Avon Canal and the Berks & Hants railway have been cut out of the elevation model. Their embankments are real earthworks and both run the length of the vale; left in, a route seeking dry raised ground finds the railway embankment and follows it.
Wet ground is inferred from shape: land that is level over a hectare and sits more than two metres below its surroundings. Flatness alone will not do, because in chalk country the hilltops are flat too. Checked against mapped geology, 72% of the ground this rule flags falls on clay, greensand or alluvium — units making up only 26% of the frame.
The source is a 1 km circle, entered at its edge, because the provenance is to an area of sarsen ground rather than a point.
What it is not
A least-cost path is not a route. It knows nothing of ownership, tradition, season, weather or alliance, any one of which could override every gradient here. It says what the ground permits, not what people did.
The paths step in 45° increments because they are solved on a grid, so they appear as staircases and their lengths are overstated by up to about 8%. They are deliberately not smoothed: smoothing would make them look more precise than the method is.
The terrain is modern. Palaeo-channels, the pre-drainage extent of the floodplain, and woodland cover — which would have mattered enormously — are all absent.
Anyone who knows this ground will find places where the line is wrong. That is expected. It is a floor to argue from, not a conclusion.
No comments:
Post a Comment
Comments welcome on fresh posts - you just need a Google account to do so.