Full dataset build · Evolution of the_long_walk.html · Companion to Correction #6

The Long Walk, interactive: network, ghosts & reconvergence

A real, pannable, zoomable map — not the fixed schematic — scrubbing through deep time. This build demonstrates three new mechanics in one place: a pulsing Green Sahara corridor that grows and shrinks with the actual humid-period record, African origin nodes that appear on their own dated evidence instead of one fixed pin, and "ghost branches" — lineages that fade to grey and vanish to a faint marker when they die out, with a new bright marker lighting the same ground later when someone else arrives.
COMPLETE BUILD — the entire researched dataset is now on one scrubbable timeline: 552 dated anchors across five evidence tiers, 111 researched migration routes, three forced routes, and one text lineage — including the Siberian spine that carries the Ancient North Eurasian line to the Arctic — drawing progressively across their own real date windows, 15 gold leader walk-back lines, and 6 African corridor connectors linking the origin nodes so the story no longer starts abruptly with the southern migration. The timeline reaches back 1.8 million years, so the contested Gibraltar hominin claims and the pre-sapiens Luzon evidence sit at real dated positions rather than floating outside it. Scrub or hit Play below.
Layers
The camera only ever zooms out and never pans. Pause to move the map yourself — the fog eases back while paused so you can read what is underneath.

Evidence tier — every one of the 552 anchors carries one

Roughly 20% of this map is flagged contested — and that is the point, not a defect. The aim is to show how well each claim is supported, not to assert what happened. A contested anchor is drawn as prominently as a solid one, in a different colour, so you can see the shape of the disagreement instead of having it quietly resolved for you.

Type to search all anchors by label and site name. Use the up and down arrow keys to move through results and Enter to open one on the map. Opening a result moves the timeline to that anchor's earliest dated evidence.

~350,000 YEARS AGO Before the Network
Nothing yet — the stem populations that will eventually diverge into every branch this project tracks haven't left evidence this far back.
Before UsSahul1999
Brightness is time, and only time. A mark is bright at its date and dims as the slider moves on, down to a faint version of the same shape. A dim mark is further in the past — it is not weaker evidence; evidence is the colour and shape, and the popup. Every mark stays clickable at every brightness. The three forced routes never dim. Origin node — a source, not a site Persisting branch (never ghosts — alive today) Ghost — fading, no living descendants of this line Faint marker — where a line died out Reconvergence — new arrival lighting the same ground Green Sahara corridor (pulses open/closed) Site — well-supported Site — contested / disputed Site — pre-sapiens hominin (a separate tier, not a weaker one) Claim — no verified evidence Tradition-only — the record does not confirm it either way Text witness — arc = how much of the text this object holds Text witness, unprovenanced — no findspot at all Named leader / walk-back (toggleable) Diffusion origin — where the thing came from Diffusion — actually adopted here Capability held Capability lost — the people are still there Fog — ground no one has reached Lost ground — people were here, and are gone Ice sheet The ice sheets are the crudest geometry on this map, and unlike the shorelines they are switched ON. Four sheets, 26 vertices between them — Laurentide 8, Cordilleran 6, Fennoscandian 6, Patagonian 6. They are bounding boxes, not margins, and the dates they carry are better than the shapes that carry them. Read them as "ice was roughly here", never as an edge. Real LGM margins are among the best-mapped features in Quaternary science and the replacement is identified — Batchelor et al. 2019 for the Northern Hemisphere — but that deposit carries no licence on the artefact, only on the paper describing it. This project does not ship data whose rights cannot be traced to the thing itself; it refused a basemap on exactly that ground. So the boxes stay, visibly labelled as boxes, until the licence is resolved rather than assumed. The Patagonian sheet has no licence-cleared source at all. Palaeo-shorelines — exposed continental shelf, land that is sea today. The coastlines are now sourced: extracted from De Groeve et al. 2022, the only global shoreline reconstruction with a real glacial-isostatic correction, drawn as land, in the basemap's own land colour, as De Groeve et al. reconstruct the coast at three dates: about 19,000, 12,900 and 9,550 years ago. From 26,000 years ago on they are shown by date: as the slider passes from one reconstructed date to the next, the sea takes back the ring of land between them. The source and this map's global sea-level curve disagree: in De Groeve's raster the shelves flood while the curve still reads the sea tens of metres lower — 18–38 m in Sunda, 22–36 m in Sahul, 15–33 m in the Persian Gulf, less at high latitude. That is observed in their raster and not explained in their paper, each shoreline popup gives the measured figure, and it is why the coasts are labelled by date rather than by depth. Shown for the five researched shelves: Beringia, Doggerland, Sunda, Sahul and the Persian Gulf. A model, not a survey, extracted at the source's full resolution: how much of each region's shelf survives simplification is stated region by region and coast by coast in every shoreline popup. Each region shows only the shelf inside its own box, so where a box cuts the shelf the land stops in a straight line: that edge is the limit of this map's regional extraction, not a coast, and no coastline is drawn along it. The layer is off by default until that is ruled on. From 26,000 years ago to the present this is a reconstruction, and its coastline is drawn solid. Before 26,000 years ago it is an APPROXIMATION, not a reconstruction, and its coastline is drawn dashed. No reconstruction exists there: De Groeve et al. 2022 begins at 26 ka and no glacial-isostatic-corrected global product reaches further back — the edge of the science, not a gap in the collecting. What is drawn instead is the same De Groeve shapes, selected by this map's own sea-level curve — each dated coast matched to the depth the curve reads at its date, a match this map makes and the source does not — not by any reconstruction of that date, and with no glacial-isostatic correction: the rise and fall of the land itself under the ice sheets is ignored, so it is most wrong at high latitude — here, Beringia. The curve is coarse there too: four points between 26,500 and 130,000 years ago, with straight lines across the 50,000 years from 120,000 to 70,000 and the 40,000 years from 70,000 to 30,000. The switch happens at exactly 26,000 years ago with no blending; the HUD says which side of it you are on, and a mark on the timeline shows where it falls. Each shelf is still drawn only inside its researched window, so the approximation reaches back 36,000 years for Beringia, 65,000 for Sahul, 70,000 for the Persian Gulf and 90,000 for Sunda; Doggerland, held by ice, is never approximated. Shoreline data: De Groeve, J. et al. (2022), Global Ecology and Biogeography 31(11): 2162–2171, doi:10.1111/geb.13573; raster via Figshare, licensed CC BY 4.0. Changes were made: thresholded at three dates, polygonised, simplified, and clipped to five regions.

What this prototype is actually showing

The Sahara corridor pulses instead of sitting on the map as a static line, because it wasn't a static line. A 2023 Nature Communications paleoclimate reconstruction counts 20 North African Humid Periods over the past 800,000 years, recurring roughly every 20,000–40,000 years. This prototype animates that on/off cycle directly, keyed to the specific dates already established in this project's own Correction 3 research: broadly viable 246,000–96,000 ya, with two final documented closing pulses around 78,000 and 67,000 ya before the corridor shuts for good. Scrub slowly through that range and watch the ring grow and shrink with each cycle, rather than reading it as one bridge crossed once.

The four origin nodes replace the single fixed Rift Valley pin from the old schematic build, each keyed to its own dated evidence: Jebel Irhoud, Morocco (~315,000 ya); the Rift/Horn, Ethiopia (~250,000 ya, this project's own long-running anchor); the Kalahari/Makgadikgadi, Botswana (~200,000 ya, the Chan et al. 2019 claim — shown CONTESTED, per Correction #6, not as a settled single origin); and the Ituri Forest, DRC (~170,000 ya). None of the four is "the" start. The Khoisan lineage's first split (~135,000 ya) appears as a persisting branch — it never ghosts, because it's a living population today — and a brief pulse around ~100,000 ya marks the merger that forms the ancestors of eastern and western African populations (Ragsdale et al. 2023).

Three ghost/reconvergence chains show the same visual grammar Shawn asked for, in two honest variants. Same exact ground: Jebel Faya's older occupation layer (~125,000–100,000 ya) fades to a ghost and leaves a faint marker — then the identical coordinates light back up around 65,000 ya when the Southern Route's successful pulse actually arrives. Dorset culture in the eastern Arctic does the same at the same coordinates around 1300 CE, replaced by Thule with "virtually no evidence of genetic or cultural interaction" between them — same spot, unrelated population. Same broader ground, not the same point: Tianyuan man's own lineage near Beijing (~40,000 ya) is a genuine dead end — nobody alive today descends from him — and fades to a ghost accordingly; the marker that lights up later, around 9,000 ya, sits nearby in the Yellow River valley and represents the Neolithic farmers whose ancestry actually became dominant across China. That's a different population arriving in the same region generations later, not the same line returning — worth keeping that distinction honest on-air rather than letting "reconvergence" imply a bloodline reunion it isn't.

Known limitation, stated plainly: the routes now carry the full researched multi-stage paths from human_spread_map_2.html, but the dedicated Kelp Highway, Southern Road and African Migrations maps still hold detail this one doesn't — landmark-level stops, then-and-now tables, and per-route discussion. This map answers "when, and in what order"; those answer "where, exactly, and what's there now." Route date windows are drawn from the project's own sourced research; where a window is genuinely disputed the underlying anchor carries the CONTESTED tag, but the line itself does not yet visually distinguish a contested window from a settled one. That's the next honest improvement to make here.