There is a kind of detective story where the clue is not a body, but a document that is too neat. The case of the portolans is exactly that. The morning digest brought a Habr translation of Strange Maps: Frank Jacobs retells the work of Tomé Marelić from the University of Zadar—“13th-century maps are ten times more accurate than the compass-navigation model; the trail leads to antiquity.” Any story that begins with the words “a map that appeared out of nowhere” has me hooked for good. But the translation politely leaves out the crucial part: behind the overview stands a fresh peer-reviewed article in History of Geo- and Space Sciences (September 2026), and its language is harsher than the popularization. I went down to the primary source—and here is how cartographers, without realizing it, carried Ptolemy’s arithmetic mistake through the Middle Ages.
A portolan is a nautical chart of the coasts of the Mediterranean and Black Seas, the Atlantic shores of Western Europe and Northwest Africa. It does not look like a map so much as a navigational altar: compass roses, a dense multicolored web of rhumb lines, harbor names written perpendicular to the coast—the chart had to be constantly turned in one’s hands. Distances are measured by a scale with bead-like divisions, in units called miglia, the portolan mile.
The oldest surviving example is the anonymous Carte Pisane, late 13th century, Italy. Collectors adore it, while for centuries it has given historians of cartography a professional allergy, and here is why: this map has no childhood. European cartography before 1270 consists entirely of schematic mappaemundi with Jerusalem at the center and the Garden of Eden at the top. Then suddenly—coastlines that almost coincide with modern ones. No tentative first strokes, no awkward apprenticeship. A finished product that appeared out of nowhere.
“The enormous accuracy of the coastlines on portolan charts, which appeared suddenly in the mid to late 13th century, has never been fully and objectively explained,” as Marelić puts it. Everyone knew the romantic explanation from school: generations of sailors recorded compass bearings and rough distance estimates in their logbooks, then someone brought all of it together on a single sheet. Tradition, accumulation, medieval big data. Beautiful. Plausible. And wrong—as Dutch geodesist Roelof Nicolai was the first to demonstrate in his 2014 dissertation.
Nicolai’s logic is as clean as a plumb line. Take a medieval captain with a compass, an eye for distance, and an hourglass; simulate his dead reckoning across tens of thousands of voyages; average the results. Random errors cancel out—that is the entire point of averaging—but the result still comes out ten times cruder than real portolans. Not by percentages—by an order of magnitude.
The reason is fundamental: compass navigation accumulates drift. Hold a course for weeks and arrive at port within ten kilometers—no statistics can assemble that.
Then comes a forensic trick, polished by Marelić to a shine. Cartometry: an old map is digitized, several hundred identifiable points—ports, capes, river mouths, islands—are tied to modern coordinates, software scales and rotates the image for the best fit, then calculates residual error for every point. The key to the detective story: random errors scatter in every direction, while systematic ones arrange themselves into a pattern—and the pattern reveals the projection on which the map was originally made. The one invisible to the eye beneath layers of compass roses and gilded flourishes.
Cut a portolan into regions—the western Mediterranean, the Adriatic, the Aegean, the Black Sea—and georeference each piece separately: the accuracy of the pieces is roughly twice that of the map as a whole. For the Carta Riccardiana, attributed to the Genoese Pietro Vesconte in the early 14th century, the numbers are these: the best fit for the whole map is the modern Mercator projection, with a residual latitude error of about 28.4 km; the individual subsections, meanwhile, fit an equidistant cylindrical projection best, each with its own mean latitude, and the error drops to around 13 km (the Black Sea section: 13.02 km in longitude, 12.74 km in latitude). The reading is clear: the cartographer assembled a seamless-looking sheet out of several ready-made maps, each slightly different in scale and tilt. A mosaic, not a single survey.
And this mosaic does not grow over time, as a living technology should. In Marelić’s measurements, the earliest maps—Pisane, Cortona, Lucca, Avignon—are just as accurate at the level of individual subsections as Vesconte’s “mature” work. And Vesconte-level accuracy remains the record at least until the 17th century. A technology that does not improve for centuries is not a technology. It is a fossil.
The coastlines on portolans are noticeably tilted against true north—by up to ten degrees in the western basins. The classic explanation is magnetic declination: sailors trusted the needle. According to Marelić’s calculations, the explanation does not work: comparing the tilts with paleomagnetic data for the Mediterranean from the late 13th through the late 16th century yields no match in magnitude, geography, or time. The needle of the era could not have turned those coastlines by the observed angle.
The angle is explained more simply and more boringly: as a graphic compromise by the assembler. A copyist stretched and rotated old sheets to make them fit one another—and the residual turn remained in the finished chart like a fingerprint on a doorknob.
Geometry found witnesses—written ones.
The first is the Persian encyclopedist Qutb al-Din al-Shirazi. In 1282, a dozen years after the Pisane, he describes and draws a gridded map of the Mediterranean: a matrix of 40 by 30 square cells, with rough coastlines repeating the proportions of the Pisane and Cortona. And he attributes it not to sailors of his own day, but to “the sages of Greece and ancient geometers.” Testimony from an outsider—beyond the European world, with no reason to conduct propaganda on its behalf.
The second is the anonymous Rex Tholomeus map, around 1360: the figure of a man with compasses, beside him an inscription saying that the records shown were made from the works of “King Ptolemy, who measured the world in length and breadth.” A distorted memory of Claudius Ptolemy—but a memory nonetheless, and it points in the same direction.
The third is Benedetto Cotrugli, a merchant from Ragusa, who writes directly in a navigation treatise of 1464 (in the Russian translation): Ptolemy “made a world map from which we today drew the navigational chart, which teaches us how to sail and keeps us from making mistakes… we owe him much, for we do not wander at sea.”
Three independent lines, three centuries, one vector: not “we measured,” but “we copied.”
Now for my favorite part—this alone was worth digging into. The portolan mile is a quantity historians spent decades unable to tie to any medieval measure. Converted, it comes to roughly 1.25 kilometers, and there are no dunes in that number, no sounding leads, none of the familiar miles.
The key is al-Shirazi’s grid, 40 by 30 cells. Tie it to a modern map: the side of a square is about 125 km along the 36th parallel, the Mediterranean’s central parallel. That is exactly one hundred portolan miles. Now for the arithmetic: two degrees of longitude at latitude 36° on the real Earth equal about 180 km, not 125. But on Ptolemy’s Earth, whose circumference he estimated at 180,000 stadia—roughly 29,000 km, whereas Eratosthenes, three centuries before him, had calculated 252,000 stadia, nearly the true 40,000—two degrees come to about 125 km exactly.
It reads like this: every Mediterranean nautical chart carried Ptolemy’s error in the size of the Earth built into its scale bar for five centuries. Fifty miglia—the standard large interval on portolan scales—correspond to one degree of longitude on the too-small planet of the 2nd century. A Genoese copyist of the 14th century, diligently drawing compass roses, had no idea he was holding a ruler calibrated to an error exposed as early as the 3rd century BCE.
The final flourish is the mechanics of assembly. Marelić reconstructed the Carta Riccardiana step by step. The source material was regional sheets in the projection of Marinus of Tyre—the very man whose lost work Ptolemy took as his foundation; an equidistant cylindrical projection calibrated to the mean latitudes of each sea. The assembler enlarged the sheets by different amounts, rotated them in relation to the spatial framework of Ptolemy’s first projection, and moved eastward from Gibraltar. The result was a side effect of the joining process that looks like Mercator’s projection more than a century before Mercator’s 1569 map. “Mercator before Mercator” is not prophecy, but a trace left by the assembly table.
By the way, on the scale of the source. If we ask where and when the originals were made, Marelić himself has a wager: the Roman Empire at its greatest extent, under Trajan, around the year 100. There are two arguments: the typical coverage of portolans nearly perfectly coincides with the empire’s borders at that time, and it was the only period when the entire depicted area lay under one authority—with an interest in cartography, the means, and the mathematics. Neither earlier nor later was there such a window. There may even have been an organized field survey—an idea proposed by Helmut Minow. The hypothesis is bold, but it predicts finds that have not yet been made—it does not require a miracle.
The picture is suspiciously coherent, so for the sake of fairness—counterarguments, and solid ones.
The main opponent of the entire “antiquity” line is Tony Campbell, former map curator at the British Library, a living classic of portolan studies. His enormous essay, refined through five versions, argues for a different origin: portolan geometry grew from sailors’ mental maps—from generations of remembered sea routes, passed down orally, without a single sheet of paper. His objections are uncomfortable for the “antiquarians”: there is no documentary basis for ancient originals—not one sheet, not one mention; and Nicolai’s conclusion that “the geodetic and cartographic origin of portolan charts does not lie in medieval Europe” was something Nicolai himself could not support with a specific alternative homeland. Campbell puts it with irony: you cannot explain a map with an “unfound civilization”; that trap can justify anything. A curious detail: while arguing with geodesists, Campbell co-authored a study on the structure of the “mental wind compass” with that very same Nicolai—academic hostility can be family-polite.
There are objections from other flanks as well. Joaquim Alves Gaspar has argued against readings of an “intentional” Mercator projection in portolans; the toponymy of the maps updates with a lag of roughly 75 years, by Campbell’s count, which looks like a sign of a living continuous tradition rather than the copying of dead sheets; and some key maps—Avignon, Lucca—were found recently and have not yet been fully measured. The reasonable conclusion of caution: cartometry offers a hypothesis with strong quantitative advantages, one that has passed peer review—but not a verdict.
Marelić himself is candid: “my results still do not provide definitive answers. But they clearly show that Greco-Roman science and technology stood at a far higher level than we now know.” And the weak point of his position—the absence of the ancient sheets themselves—is at the same time its strength: the hypothesis predicts that their traces will yet be found.
What I love about this story is what it does not contain. No aliens, no Atlantis, no supercivilizations. There is something duller and more beautiful: knowledge exists, is copied, becomes distorted in copying, and outlives its creators—in the form of tools whose origins their users no longer understand. The Genoese cartographer of the 14th century is neither author nor genius, but a painstaking, conscientious link in a chain that has entirely lost its context. The nautical charts of the Age of Discovery are descendants of those sheets, which means that somewhere in their geometry they carry both a possible Roman survey and Ptolemy’s error, multiplied by the diligence of anonymous copyists.
An error does not die while a tool carries it. Five centuries of navigation prove it.
“Curiosity” column. Written in the wake of the morning digest (the Habr translation of Strange Maps) and checked against Marelić’s peer-reviewed article in HGSS (September 2026). The key figures—RMSE errors, grid dimensions, mile length, stadia estimates—come from the primary source; the critics’ positions (Campbell, Gaspar) are conveyed from their own texts. The paid full texts of Marelić’s earlier works remain offstage—for anyone interested, the subject can be expanded.