Sunday, 15 November 2026

Where the ice stopped: the Bristol Channel at the Last Glacial Maximum

This post is the companion to the short film Where the ice stopped, which animates the BRITICE-CHRONO model of the last ice sheet in the Celtic Sea and Bristol Channel from 31,000 to 15,000 years ago. The measurements behind it, with sources, follow.

Five published reconstructions of the last British–Irish Ice Sheet give the position of the ice margin in the Celtic Sea and Bristol Channel between 31,000 and 15,000 years ago. Their deposited GIS data are drawn together and measured in an interactive map: Bristol Channel LGM ice limits. No reconstruction carries Irish Sea ice across the north coast of Cornwall, Devon or Somerset. One margin is drawn along that coast. The others end between 4 and 24 km from it at their closest.

Open the interactive map: Bristol Channel LGM ice limits.


The reconstructions

  • BRITICE-CHRONO empirical reconstruction (Clark et al. 2022): minimum, optimum and maximum ice extents at thousand-year steps from 31 to 15 ka.
  • BRITICE-CHRONO model reconstruction (same paper): an ice-sheet model nudged towards the empirical limits, deposited as grids of about 2.5 km.
  • DATED-1 (Hughes et al. 2016): minimum, most-credible and maximum extents for the Eurasian ice sheets, with a census date of 1 January 2013.
  • Batchelor et al. 2019 (Nature Communications 10, 3713): a single best-estimate LGM outline within a Northern Hemisphere compilation.
  • PaleoMIST 1.0 (Gowan et al. 2021): global margins at 2,500-year steps.

These are three source families, not five independent tests. PaleoMIST takes its Eurasian margins from the DATED-1 most-credible lines, with intermediate slices interpolated. Batchelor et al. compile published outlines. BRITICE-CHRONO supplies both an empirical reconstruction and a model fitted to it. Reconstructions published only as figures, without deposited geometry, are not included.

Distance from the modern north coast

The coast was followed for 316 km from Land's End to Brean Down and sampled every 2 km. For each margin, the distance from each coastal point to the nearest ice was computed, using the union of all time-slices so that each margin is taken at its farthest reach.

Margin Closest approach Median distance offshore
BRITICE-CHRONO maximum boundat the coast1.6 km
BRITICE-CHRONO model, grounded ice3.7 km14.6 km
BRITICE-CHRONO optimum8.6 km21.7 km
Batchelor et al. 201914.8 km30.1 km
DATED-1 most credible18.5 km42.7 km
PaleoMIST 1.023.6 km75.3 km

The BRITICE-CHRONO maximum bound lies within 5 km of the shore along 89% of the coast measured. It overlaps 53 km² of land, 0.4% of the peninsula, as coastal slivers no more than 1.8 km deep at the sampled points. The optimum of the same reconstruction lies 8.6 to 41.5 km offshore.

The model grows ice on the South-West peninsula, 546 km² summed over all time-slices, but as separate upland ice on Dartmoor and Exmoor. It is not connected to the Irish Sea Ice Stream in any slice.

Lundy, Scilly and Fremington

Lundy separates the margins. The island lies inside the BRITICE-CHRONO maximum bound and inside the model's grounded ice, and 6.0 km outside the optimum. Rolfe et al. (2012) report 10Be exposure ages of 31.4–48.8 ka from the island and read them as glaciation before the LGM. Carr, Hiemstra and Owen (2017) argue that Lundy was not ice-covered in the Devensian.

The Isles of Scilly sit on the BRITICE-CHRONO empirical margin, within 0.1 km in all three versions. Smedley et al. (2017) date the ice limit on the northern islands to about 26 ka.

Fremington lies outside every margin in every time-slice. The nearest is the BRITICE-CHRONO maximum bound, 12.6 km away.

Sea level

Distances are measured to the modern coast because that is where the onshore evidence begins, and because it is the line the maximum bound follows. It was not the shoreline of the time. The shoreline of each date is taken from the glacio-isostatic palaeotopography published with the BRITICE-CHRONO reconstruction.

In that model the shoreline lies a median 47 km from the modern north coast at 31 ka. Crustal depression under the ice brings it to a median 19 km at 25 ka, and it withdraws to 52 km by 19 ka. At no date does it come within 5 km of the coast measured. A uniform lowering of sea level by 120 m would not reproduce this pattern.

A margin that follows the modern shoreline for most of 316 km therefore follows a line that was not the shoreline at any date in the reconstruction.

The same palaeotopography shows how much of each ice front that now lies under the sea stood on dry land. At 26 ka the figure is 37% for the BRITICE-CHRONO optimum, with the remainder in water of 48 m median depth, and 47% for the model. For the DATED-1 most-credible margin at 23 ka it is 5%.

Limits

  • Line-work in regional and hemispheric compilations is drawn at small scale. Differences of a few kilometres between margins, and between a margin and the coast, are within drafting precision.
  • The palaeo-shoreline is an output of a glacio-isostatic model tied to the BRITICE-CHRONO ice history. Beside the DATED-1, Batchelor and PaleoMIST margins it is indicative only.
  • The chronologies differ. DATED-1 places the Celtic Sea lobe in its 23 ka slice; BRITICE-CHRONO places the greatest extent in this area at 26 ka.
  • In the animation, flow direction is inferred from the modelled ice-surface slope. The deposited data give flow speed but not direction.

Notes on the film. The animation shows the BRITICE-CHRONO model reconstruction (Clark et al. 2022), deposited as thousand-year grids of ice thickness, surface elevation and surface speed at about 2.5 km; frames between slices are interpolated. Flow lines are particles moved down the slope of the smoothed modelled ice surface at a rate set by the modelled surface speed: the deposited data give speed, not direction. Land and sea at each date are from the same study's palaeotopography; the thin line is the modern coast. Relief: NOAA ETOPO 2022. Music: original.

Sources

  • Clark, C.D. et al. (2022). Growth and retreat of the last British–Irish Ice Sheet, 31 000 to 15 000 years ago: the BRITICE-CHRONO reconstruction. Boreas 51(4), 699–758. doi:10.1111/bor.12594. Data: doi:10.1594/PANGAEA.945729.
  • Hughes, A.L.C., Gyllencreutz, R., Lohne, Ø.S., Mangerud, J., Svendsen, J.I. (2016). The last Eurasian ice sheets – a chronological database and time-slice reconstruction, DATED-1. Boreas 45(1), 1–45. doi:10.1111/bor.12142. Data: doi:10.1594/PANGAEA.848117.
  • Batchelor, C.L. et al. (2019). The configuration of Northern Hemisphere ice sheets through the Quaternary. Nature Communications 10, 3713. doi:10.1038/s41467-019-11601-2. Data: osf.io/7jen3.
  • Gowan, E.J. et al. (2021). A new global ice sheet reconstruction for the past 80 000 years. Nature Communications 12, 1199. doi:10.1038/s41467-021-21469-w. Data: doi:10.1594/PANGAEA.905800.
  • Smedley, R.K. et al. (2017). New age constraints for the limit of the British–Irish Ice Sheet on the Isles of Scilly. Journal of Quaternary Science 32(1), 48–62. doi:10.1002/jqs.2922.
  • Rolfe, C.J., Hughes, P.D., Fenton, C.R., Schnabel, C., Xu, S., Brown, A.G. (2012). Paired 26Al and 10Be exposure ages from Lundy. Quaternary Science Reviews 43, 61–73. doi:10.1016/j.quascirev.2012.04.003.
  • Carr, S.J., Hiemstra, J.F., Owen, G. (2017). Landscape evolution of Lundy Island: challenging the proposed MIS 3 glaciation of SW Britain. Proceedings of the Geologists' Association 128(5–6), 722–741. doi:10.1016/j.pgeola.2017.06.005.
  • Base map: NOAA ETOPO 2022 relief; GSHHG 2.3.7 shoreline (Wessel and Smith 1996).

Map, measurements and code outputs: github.com/TimDaw37/bristol-channel-lgm-ice. Measurements made on 15 November 2026 from the data as deposited.

Saturday, 14 November 2026

Great Britain in relief: open LiDAR maps to download

A shaded relief map of the whole of Great Britain and the surrounding sea floor, built from the same open LiDAR terrain models and EMODnet bathymetry that drive LiDAR Lens. It comes as ready-made images, as an interactive map that exports framed and credited images, and as base-map tiles for web maps. A second version lays BGS bedrock geology over the relief. Everything is free to reuse under CC BY 4.0.

Shaded relief map of Great Britain with a green-to-brown elevation tint, showing the Scottish Highlands, the Lake District, the Pennines, Snowdonia and the chalk and limestone scarps of southern England, with the sea floor shaded in blue by depth
Great Britain, shaded relief and sea floor, 700 m grid, British National Grid. Click for the 3500 px version.

Downloads

Version1000 px
700 m grid
3500 px
200 m grid
7000 px
100 m grid
ReliefJPGJPG · PNGJPG · PNG
Grey hillshadeJPGJPG · PNGJPG · PNG
Elevation tintJPGJPG · PNGJPG · PNG
Bedrock geologyJPGJPG · PNGJPG · PNG

The 7000 px versions print at about A1 (200 dpi) or A2 (300 dpi). Vertical exaggeration is 3.5×, 3× and 2.5× from the smallest size to the largest, enough to keep lowland scarps visible at each scale. Title, scale bar and data credits are printed on every image.

Map of Great Britain showing bedrock geology by period, coloured to the international chronostratigraphic chart and draped over shaded relief
Bedrock geology (BGS Geology 625k) by period, on shaded relief. Igneous intrusions are coloured by composition.

Make your own map

The interactive map uses the same data and rendering. It offers four styles (relief, grey hillshade, elevation tint, geology) and four colour schemes, with a multi-directional or single sun, adjustable sun direction and height, and adjustable vertical exaggeration. Inland water can be switched on or off. Any frame can be exported, as on screen or in A-series, square or 16:9, at up to 7200 px, with an editable title and the required credits. Tapping the map gives the grid reference, height and, in geology mode, the bedrock period. A link opens the same spot in LiDAR Lens at full 1 m resolution.

For web maps

Web Mercator tiles (zoom 5 to 11, sea floor included, Ireland and France transparent) are available in relief, grey and geology versions for Leaflet, MapLibre or OpenLayers:

<script src="https://timdaw37.github.io/gb-relief/gbrelief.js"></script>
GBRelief.layer('relief').addTo(map);   // or 'grey', 'geology'

Template: https://timdaw37.github.io/gb-relief/tiles/{style}/{z}/{x}/{y}.webp. Full details are in the repository.

Data and method

Terrain. The terrain comes from 1 m open LiDAR digital terrain models. England is covered by the Environment Agency LiDAR Composite DTM. Wales is covered by the Welsh Government LiDAR DTM, via DataMapWales. Scotland is covered by Scottish Government, SEPA, HES and OIC LiDAR, via the Scottish Remote Sensing Portal. The data were resampled to about 90 m and reprojected to the British National Grid at 100 m. Parts of the Highlands and Islands and the Isle of Man have no open LiDAR. There the terrain is GEM-15 v1, Google's experimental 15 m bare-earth model (Batchu et al. 2026). Against the Scottish LiDAR its median per-tile offset is −0.09 m, so the two join without a visible step at this scale.

Sea floor. Offshore, the surface is the EMODnet Digital Bathymetry DTM 2024 (mean-sea-level version, about 115 m cells, CC BY 4.0), as prepared for LiDAR Lens, and is shaded in the same way as the land under a blue depth tint. The overdeepened glacial basins of the Inner Hebrides and the Firth of Lorn, the linear tidal sandbanks off East Anglia, and the continental slope into the Rockall Trough (deeper than 1300 m at the north-west corner of the frame) all show. EMODnet compiles surveys of different ages and quality and fills the gaps between them, so straight lines and steps can be survey artefacts rather than landforms. Depths are relative to mean sea level, not Ordnance Datum, and are not for navigation.

Coast and water. The Environment Agency composite includes LiDAR returns over the sea and in estuaries, at about −4 to +2 m. Left in, they produce false land, such as a strip joining Lundy to the Devon coast. Land is therefore clipped to mean high water using OS Open Zoomstack, which also supplies inland water bodies and foreshore.

Shading. Slopes are taken from central differences. The light is a weighted four-direction sun: 315° at weight 0.5, 270° and 0° at 0.2 each, and 225° at 0.1, all at 40° elevation. This brings out ridges and scarps that run north-west to south-east, parallel to the main light, which a single north-west light leaves flat. The elevation tint runs from green lowland to brown upland, with pale summits above 1100 m.

Geology. The bedrock is BGS Geology 625k (DiGMapGB-625), coloured by period with the International Commission on Stratigraphy chart colours. Felsic and mafic intrusions are shown separately. At 1:625,000 the geological boundaries are generalised and coarser than the relief beneath them.

Resolution. At these scales the 1 m LiDAR adds little beyond consistency and the absence of interpolation artefacts. A national relief map at 100 m built from OS Terrain 50 would differ little. The LiDAR matters at field scale, which is what LiDAR Lens shows.

Licence and credit

The maps and tiles are free to use under CC BY 4.0. Credit "GB Relief, Tim Daw (sarsen.org)" and keep the data credits:

LiDAR DTMs: Environment Agency (England); Welsh Government, DataMapWales (Wales); Scottish Government, SEPA, HES, OIC, Scottish Remote Sensing Portal (Scotland). Open Government Licence v3.0. GEM-15 v1 (experimental), produced by Google, CC BY 4.0. Sea floor: EMODnet Digital Bathymetry (DTM 2024), CC BY 4.0, not for navigation. Contains OS data © Crown copyright and database right 2026. Geology versions: contains British Geological Survey materials © UKRI, Open Government Licence v3.0.

EMODnet Bathymetry Consortium 2024. EMODnet Digital Bathymetry (DTM 2024). doi:10.12770/cf51df64-56f9-4a99-b1aa-36b8d7b743a1.

Batchu, V.V., Neumann, M., Li, T., Allen, B., Zhang, J. and Van Arsdale, C. 2026. GEM-15: Global surface and terrain elevation models (15 meters) from sub-meter optical stereo imagery. EarthArXiv preprint, not peer reviewed. doi:10.31223/X55N75.

Thursday, 12 November 2026

LiDAR Lens goes outdoors: field mode for walkers

 


LiDAR Lens now has a second face. Alongside the full research view (now called Lab) there is a stripped-down Field mode built for a phone held in one hand on a footpath: where am I, where is the earthwork, which way do I walk, and what is it when I get there.

Open it at lidarlens.sarsen.org/?ui=field, or switch between Field and Lab with the toggle at the top of the screen in either view.

What Field mode shows

The screen carries a search box (place, postcode or grid reference), zoom buttons and a single control panel at the bottom. Everything else in Lab — the fifteen visualisations, sliders, data layers and analytical tools — is hidden. The LiDAR fills the whole screen rather than sitting inside the lens circle.

Four view buttons cover what matters outdoors:

  • Hillshade — single low sun from the north-west, the conventional reading of earthworks.
  • VAT — the Combined VAT visualisation (Kokalj & Somrak 2019), blending hillshade, slope, openness and sky-view; better for low banks and ditches that a single sun misses.
  • Old map — the Ordnance Survey six-inch map (1888–1913) over the aerial photograph, useful for field boundaries, lost tracks and antiquarian annotations.
  • Aerial — the aerial photograph alone, for orientation against hedges, gates and buildings.

Coverage is the whole of Great Britain where 1 m LiDAR exists: the Environment Agency composite DTM for England, the Welsh Government DTM for Wales and the Scottish Remote Sensing Portal data for Scotland, all Open Government Licence and fetched live.

Finding a monument

  1. Set a target. Tap ◎ Target, then tap the feature on the map — or search for a place or grid reference. The target's grid reference appears in the panel.
  2. Allow location. The phone's GPS position is shown on the map and updated continuously. With a fix, the panel gives the distance and compass bearing from you to the target.
  3. Use the arrow. By default the arrow is drawn on a north-up map. Tap Use compass and the arrow turns with the phone, so it points the way to walk. On an iPhone this needs permission for motion and orientation, asked for the first time the button is pressed.
  4. Recentre. Panning the map stops it following you. Recentre returns to your position (framing the target too when it is close) and resumes following.
  5. Arrival. Within about 25 m of the target the panel reports that you are there; the threshold widens to as much as 50 m when the GPS fix is poor.
  6. What is it? Lists the heritage records within 30 m of the target, or of your position if no target is set: Historic England (National Heritage List and Aerial Investigation and Mapping) in England, Cadw and Coflein in Wales, and Historic Environment Scotland and Canmore in Scotland, with links to each record and Creative Commons photographs from Geograph where they exist. Measure this feature from the DTM gives its dimensions from the LiDAR.
  7. Done. Clears the target and leaves the map where it is, for looking at the LiDAR around you.

The target is stored in the page address, so a link copied from the browser opens with the same target set — a simple way to send someone a barrow to find.

Putting it on your phone's home screen

Field mode is set up as an installable web app. Added to the home screen it gets its own icon, is named LiDAR Field, opens straight into Field mode and runs full screen without the browser's address bar.

iPhone and iPad (Safari)

  1. Open lidarlens.sarsen.org/?ui=field in Safari.
  2. Tap the Share button (the square with an upward arrow).
  3. Scroll down and tap Add to Home Screen.
  4. Keep the name LiDAR Field or change it, and tap Add.

Recent versions of iOS also offer Add to Home Screen from the share menu in Chrome and Edge.

Android (Chrome)

  1. Open lidarlens.sarsen.org/?ui=field in Chrome.
  2. Tap the ⋮ menu at the top right.
  3. Tap Add to home screen (shown as Install app on some phones) and confirm.

Location permission. If the panel reads Location blocked, the browser has refused GPS access to the site. On iPhone: Settings → Privacy & Security → Location Services → Safari Websites, set to While Using the App with Precise Location on. On Android: in Chrome, tap the icon to the left of the address, then Permissions → Location → Allow. Tap Retry in the panel afterwards. The map and target work without location; only your position and the distance and bearing are lost.

Before you go

  • Signal. LiDAR tiles and records are fetched as you look and kept in the browser's cache, so ground already viewed reloads instantly. Browsing the walk at home on the phone over Wi-Fi is the best preparation for a valley with no signal. This is a cache, not a guaranteed offline map: an area not viewed beforehand will not load without a connection.
  • Access. Field mode does not show rights of way. Check the route in Lab beforehand, where the public footpaths, bridleways and byways layer and CRoW access land can be switched on. Many features on the LiDAR lie on private land.
  • Bare earth. The model is a terrain model with trees, hedges and buildings removed. A bank clear on screen may be under scrub on the ground.
  • Scheduled monuments. Looking is fine; metal detecting, digging or driving on them is not.
  • GPS accuracy. A phone in open country is typically good to a few metres, worse under trees or beside steep slopes. Use the target to get close, then the ground itself.

The code is open source on GitHub. Faults can be reported from the Report a fault link in Lab, which fills in the browser details automatically.


Saturday, 7 November 2026

Alfred and the Danes: an animated campaign map, 865–899

The Danish wars of 865–899 are documented almost entirely by one source tradition: the Anglo-Saxon Chronicle, Asser's Life of King Alfred (which follows the Chronicle closely for these years) and Æthelweard's Latin chronicle. Together they give a year-by-year itinerary of winter camps, sieges, named battles and treaties. They do not give routes, force sizes or, apart from Asser's account of Ashdown, tactical detail. Several of the named places cannot be located with confidence.

A new interactive map, Alfred and the Danes, 865–899, sets that itinerary out in 34 steps on a terrain base, from the arrival of the Great Army in East Anglia to Alfred's death in October 899. Each step is keyed to its annal and, where relevant, to Asser's chapter.

What the map shows

Danish movements are drawn in red and West Saxon movements in blue. Movements by sea, or where the starting point is not recorded, are dashed. Markers distinguish camps, battles, sieges, fortifications and treaties, and are coloured by which side held the field. The sources give places, not routes, so each line simply joins the recorded places in order and is schematic.

The background is built from:

  • OS Terrain 50 relief, with a tint deepening from 40 m to 260 m OD;
  • a mask of land below 6 m OD, as a rough indication of former wetland in the Somerset Levels and the Fens;
  • named main rivers from OS Open Rivers;
  • Roman roads from the Itiner-e dataset (static version 2024), drawn solid where the course is conjectured and dashed where it is hypothetical.

The Burghal Hidage burhs and the boundary in the Treaty of Alfred and Guthrum can be switched on. The coastline is modern. It does indicate though the 9th-century Wantsum Channel, the Limen estuary at Appledore and the earlier Fen edge.

Uncertain sites

Six of the named places are disputed: Ashdown, Meretun, Cynuit, Egbert's Stone, Iley and Ethandun. For each, the map draws the routes to a working identification, rings it, and shows the other candidates as open dots. Selecting another candidate redraws the routes. The candidates and their main proponents are:

  • Ashdown (871). Æscesdun was a name for the Berkshire Downs as a whole. Candidates are Kingstanding Hill near Moulsford, Lowbury Hill, Thorn Hill near Compton and, on antiquarian grounds, White Horse Hill. Asser's single stunted thorn tree does not narrow the choice.
  • Meretun (871). Candidates are Martin (Hampshire), Marten (Wiltshire), Marden (Wiltshire), Merdon near Hursley, and Merton in Oxfordshire or Surrey. None is preferred in the literature, and the working point is for display only.
  • Cynuit (878). Asser names the fort; the Chronicle says only "in Devon". Candidates are Countisbury, Cannington Camp, Kenwith near Bideford and Castle Hill near Beaford. Historic England treats Cannington as the accepted site, although it lies in Somerset.
  • Egbert's Stone and Iley (878). The Chronicle places Egbert's Stone east of Selwood. Proposed sites include Kingston Deverill, Kingsettle Hill, Penselwood and Willoughby Hedge. Iley follows the English Place-Name Society's identification with Iley Oak, the lost meeting-place of Warminster and Heytesbury hundreds.
  • Ethandun (878). Edington, Wiltshire is the consensus identification, with the fighting placed on or below the escarpment near Bratton Camp. The Somerset Edington on the Poldens remains a minority alternative.

The Edington march

The 878 campaign gets its own day-by-day sequence at larger scale, because the Chronicle gives an unusually tight itinerary. Alfred leaves Athelney in the seventh week after Easter and rides to Egbert's Stone, where the men of Somerset, Wiltshire and part of Hampshire meet him. After one night he moves to Iley, and after a second he moves at dawn to Ethandun and fights. Asser adds the pursuit to the Danish fortification and the fourteen-day siege, which ends with hostages, oaths and Guthrum's baptism at Aller.

The panel computes straight-line distances for each leg from whichever candidates are selected. With Kingston Deverill and Eastleigh Wood as working sites, the legs are:

  • Athelney to Egbert's Stone: 50 km
  • Egbert's Stone to Iley: 8 km
  • Iley to Ethandun: 10 km
  • Ethandun to Chippenham: 21 km

The first leg ranges from about 40 km (Kingsettle Hill) to 55 km (Willoughby Hedge), depending on the candidate. These are minima, not road distances. On any of the eastern candidates, the march runs off the Levels, over the Selwood ridge and up onto the chalk, so the final approach to the Edington escarpment comes from the south-west across the Plain.


Chronology

The Chronicle's year in this period began in autumn or at Christmas, so modern dates for the early annals can fall up to a year before the annal number. The battle of Reading, for example, belongs to the turn of 870–871 under annal 871. For the 890s the map follows the corrected chronology; manuscript A numbers several of those annals a year later.

Use

The map steps or plays through the campaigns, and can be panned and zoomed. Each step has its own link, for example #step-878f for Ethandun. Layers for rivers, low ground, Roman roads, the schematic Ridgeway and Harrow Way, burhs and alternative sites are under Layers.

The code and data are on GitHub. Corrections to site identifications, and further candidates with references, are welcome.

Sources

Friday, 6 November 2026

Heritage at Risk 2026: still at risk, on the nation's own land


Historic England has just published its 2026 Heritage at Risk Register. It lists 4,991 entries: 197 were added this year and 85 came off for positive reasons. Those headline totals look like steady progress. The picture is very different when you look at land owned by the Crown, meaning the Crown Estate, the Duchy of Cornwall and other Crown bodies.

The Crown's list is getting longer

In 2025, 97 Crown-owned entries were on the register. In 2026, all 97 are still there, and none has been rescued. Ten more have been added, bringing the total to 107. Nine of the new ones are on the Isles of Scilly, where 46 of the islands' 50 at-risk entries are Crown land, and the tenth is Warren Farm Barn at Chadwell Heath.

These are not recent problems. Sixty-eight of the Crown's entries have been at risk since at least 2013. Historic England records 79 of the 107 as declining, 20 as stable and only 8 as improving.

On our doorstep: the Marlborough Downs

Six Crown-owned monuments on the Marlborough Downs, near us and near Avebury, are still on the list. They are barrows and a farmstead on Bishops Cannings Down, Horton Down and Beckhampton, and they range from Kitchen Barrow to the Horton Down barrow cemetery. All six have been at risk since at least 2013, and all six are declining. Across the wider Marlborough Downs, 36 sites are at risk, down from 38 last year.

The National Trust and Wiltshire

For comparison, 64 entries are on National Trust land, up from 61. All four new ones are on Lundy, and one, the Quarterwall cottages on Lundy, came off. Forty-two of the Trust's entries have been at risk since at least 2013.

Wiltshire and Swindon together have 144 entries at risk, down from 150. Eight came off, including the henge and barrows southwest of Fox Covert on the boundary of Bishops Cannings and Avebury, and the Liddington Warren long barrow. That is welcome news. One genuinely new entry was added, the medieval cross in St Mary's churchyard, Purton.

What should happen

The Crown is not a struggling private owner. It is the nation's own landholder, with professional land agents and considerable income. Yet it has not removed a single monument from the at-risk register this year, its list has grown by ten, and most of its entries are getting worse.

I would like to see a simple step taken: for every Crown and Duchy entry on the register, a published plan and a target date for getting it off.

You can explore the sites yourself on my Heritage at Risk map, which has a slider covering 2013 to 2026.

Figures come from Historic England's 2026 Heritage at Risk spreadsheet and map data. National Trust figures count register sites that overlap Trust land by at least 10%.

Thursday, 5 November 2026

The Hunt for the Altar Stone Source - the video

The Youtube video

Altar Stone diagnostic toggle map


The Altar Stone diagnostic toggle map is an interactive screen of candidate source areas for the Stonehenge Altar Stone. Each published diagnostic is a layer that can be switched on or off, so the effect of any one test on the list of surviving candidates can be seen directly. The code and data are in the GitHub repository.

What it plots

The map covers 85 candidate places: Ba/Rb geochemical clusters across Britain and Ireland, plus named sites with published measurements. An Old Red Sandstone underlay from the BGS Geology 625k dataset shows where the outcrop lies. Fourteen layers score each place:

  • L0–L9: the baseline screen, covering barium and Ba/Rb geochemistry, mineralogy, zircon ages, thermal maturity and related published criteria.
  • L10 and L11: a strontium ceiling and the Sr/Ca ratio, applied only where rock Sr has been measured.
  • U1 and U2: chrome spinel (CZi) and illite/smectite expandability, from Pirrie et al. (2026). Both start off.

How to read it

A place is a match only if it fails none of the layers switched on. Each layer has strict and relaxed settings where the evidence allows. Untested is not failed: a place with no data for a layer stays open on that layer. Presets switch one test or group off at a time (leave-one-out), which shows how much each test contributes, and the preset "CZi + I/S ON (Pirrie et al. 2026)" applies the new chrome spinel and clay results.

Several things are deliberately not scored: the ORS between the Midland Valley and the Welsh Borders, the Lanark Basin, Carboniferous sandstones such as the Clifton Rock, and the Sr–Nd isotope result, because the evidence is missing or has no comparison data. The Latheron–Niandt cool sliver is drawn as schematic only.

Limits

The map is a screening tool. A place that survives is untested or consistent with the evidence so far, not identified as the source, and exclusion by elimination covers only the ground that has been sampled. Corrections and additional data are welcome.

The Pirrie, Bevins, Ixer et al paper that prompted the latest update is discussed in a separate post.

Paper: Pirrie, D., Bevins, R.E., Ixer, R.A., Pearce, N.J.G., Turner, P., Power, M.R. et al. 2026. Searching for the origin of the Stonehenge Altar Stone: new sedimentological, petrographical, mineralogical, geochemical and isotopic constraints. Journal of the Geological Society 183, jgs2026-090. doi:10.1144/jgs2026-090