A farming atlas

The first fields

Many beginnings, twelve thousand years of exchange. Follow dated crop and herd evidence alongside an illustrative spread model, then inspect the people behind one early Mediterranean community.

1500CE
Years ago–
Selected systems active–
Selected model coverage–

Estimated, not surveyed. Color shows the earliest selected crop system in an illustrative model. Pick an animal to inspect its separate layer. Dashed connections are schematic and have their own source notes below. Dark land falls outside the selected reconstruction; it is not proof of absence.

How the dates are estimated

Each crop system sends an illustrative wave across a half-degree land grid. A base speed of 1 km a year comes from a European farming-spread model; it is not a measured global rate. Authored barriers, regional adoption gates and maritime links modify that scenario. Colors show the earliest selected system, so switching a system off reveals other selected histories. Overlap is not a human ancestry estimate.

The barrier values are fitted, not measured. I adjusted them until the model landed near the dated sites in the field log, so the tuned fit shows only that the model is consistent with those dates. The held-out sites were added afterward and never used for tuning, so they are the fairer test, though there are few of them and they cover only Africa and North America. Places between dated sites are where the model does real work, so read those dates as rough.

What the model cannot explain

  • Some waits are written in by hand because no travel time explains them: Korea to Japan (about 1,650 years), New Guinea to the Bismarcks (about 6,300 years), and Samoa to the Society Islands (the Long Pause, about 1,750 years). Culture, boat technology and climate decided those, not distance.
  • The Crescent is drawn growing from one point on the upper Euphrates about 10,200 BCE so it reads as one region. That start is a drawing choice. Sites in the northern and southern Levant both date to about 9600 BCE, and the evidence does not show that one end came first.
  • Barrier edges and farming limits are hand-drawn along real features such as the Rockies front, the Fish River and the Siberian forest edge, then roughened so they don't look ruled. The wiggles carry no information.
  • The model has one speed per place and does not change through time. The green Sahara is approximated by giving herds a much lower cost than crops.
  • Crops and people are not the same thing. A wave here can be a migration, or local people adopting a crop. The Fertile Crescent branches are drawn by where their waves meet, which falls along the dry Jezira and the Zagros foothills. This is a geographic split, and it does not mean a sharp genetic border.
  • The Indus is not an origin here. Wheat, barley and herding reach it from the Zagros side. Mehrgarh was redated in 2025 to about 5100 BCE (Mutin et al.). Zebu cattle and cotton are local additions. Genomes from the Indus Valley show its Iranian-related ancestry split off before farming began (Shinde et al. 2019), so farming there may have spread more as know-how than as a mass migration.
  • In the Pacific, the map separates Melanesians, who carry New Guinea's crops as far as New Caledonia, from a Polynesian branch out of the southern Philippines. Real history was more mixed: the first Lapita people in Vanuatu and New Caledonia were tied to the same voyagers, and Papuan ancestry came in later.
  • The African layers distinguish Sahelian millet, eastern Sudanese sorghum and later regional farming associated with Bantu-speaking communities. Local rice, yams and other unresolved crop histories remain context records. Each layer has a narrower meaning than an entire population or fixed crop package.
  • Bones often can only be identified as sheep or goat, but the map treats them as two separate waves. Whether African cattle were also domesticated locally is still contested.
  • Records without usable calibrated dates remain accessible in the field log but cannot trigger timed markers or enter the date comparison. A cultivation interval, a diet change and a first occurrence retain different meanings.
Barriers and their costs
Sea crossings
Connection sources and uncertainty

Sources

  • Europe and the Near East: Gronenborn, Horejs, Börner & Ober 2021 (RGZM and ÖAI map).
  • East Asia and the rice spread: Fuller et al. 2011, Rice.
  • Early Crescent sites: dates as printed on the National Geographic map you supplied (Baptista and Healy, with sources including Kuijt, Schmidt, Willcox and Zeder).
  • Mehrgarh: Mutin et al. 2025, Scientific Reports.
  • Maize into eastern North America: Kistler et al. routes figure.
  • Africa, the Americas, Asia and the Pacific: dated sites from Codex's evidence package, each linked in the field log. Those were researched from papers and abstracts, and some date conventions still need checking against original tables.
  • Field log entries marked "from memory" come from the general literature and have not yet been checked against a paper.
  • Wave speed: Ammerman & Cavalli-Sforza 1971. Borders: Natural Earth.