“We’ll just have to think our way back, step by step.” That’s what the Dutch geodesist Roelof Nicolai said in 2014 in a Strange Maps article. Nicolai had just spent his PhD years discovering that portolan charts — those gorgeous, wind-rose-covered coastline maps that suddenly appeared in Mediterranean ports in the second half of the 13th century — were geometrically too accurate to have been made the way everyone assumed. Something else was going on.
Twelve years on, we might have an answer. Tome Marelić, an assistant professor of geography at the University of Zadar in Croatia (and the current vice chair of the Working Group on the History of Cartography within the International Cartographic Association), has spent years doing exactly the step-by-step thinking Nicolai prescribed in 2014. Read together, his scientific articles amount to the most detailed forensic examination portolan charts have ever received. And the story they tell is more satisfying than even the more out-there theories (including “aliens did it”).
Mapping the Mediterranean
First, a quick refresher. A portolan chart is a nautical chart of the Mediterranean, western European, and northwest African coasts, criss-crossed with dense color-coded web-like networks often called “rhumb lines,” dotted with harbor names written perpendicular to the coast (so you had to keep spinning the chart to read it). The charts are calibrated with little beaded scale bars marking off distances in miglia, or so-called portolan miles whose roots remained enigmatic for decades until Marelić’s recently proposed explanation. (A Portolan mile is a bit shorter than a modern statute mile.)
The oldest surviving example, the anonymous Carte Pisane, appeared in Italy around 1270. Geometrically speaking, it looks like a finished product. There’s no awkward teenage phase of European cartography preceding it — no 13th-century equivalent of a flip phone before the smartphone. It just arrives, fully formed, with Mediterranean coastlines accurate enough to almost look like a map from today.
“The immense accuracy of coastline renderings on portolan charts, which, as far as is known, suddenly appeared in mid- to late-13th century, was never fully and objectively explained,” says Marelić.
Romantic and seemingly plausible — but wrong
The traditional explanation is romantic and, at first glance, plausible: Generations of sailors recorded their compass bearings and estimated distances in logbooks, and cartographers eventually compiled all that dead-reckoning data into a single chart. The trouble, as Nicolai discovered first, is that the numbers don’t work. Simulate a medieval ship captain’s best-effort compass-and-guesswork navigation, average enough of those readings together, and you get a chart ten times less accurate than the actual portolan chart.
This is where Marelić comes in. His method of choice is cartometric analysis — precise, computer-assisted measurements and error-computations. Take an old chart, digitize it, pin a few hundred identifiable points (ports, capes, river mouths, islands) to their true modern coordinates, and let the software calculate exactly how much the old map agrees with the modern map in a certain projection by scaling and rotating it until their geometrical best fit is obtained. The size of that remaining mismatch, expressed in measurable distance, tells you how accurate the chart really is, and even the exact magnitude and direction of errors for every pinned point. Random surveying errors should scatter every which way. Systemic errors, however, typically point to the map projection fabric in the background, which is sometimes, and especially in this case, invisible to the naked eye.
Marelić meticulously scrutinized that background fabric for years, stumbling upon some groundbreaking and intricate findings along the way.
Finding #1: Mosaics that defy the medieval origin narrative
Marelić confirmed and quantified that when you cut a portolan chart into regional bits — Western Mediterranean, Adriatic, Aegean, and so on — and georeference each bit separately, the accuracy roughly doubles compared to the map as a whole. That is what you’d expect if a cartographer had glued together several pre-existing charts of slightly different appearances into one seamless-looking whole.
Although portolan chart subsections, as he dubbed them, were detected by a few earlier scholars, Marelić originally computed that the earliest known pieces — such as the anonymous Pisane, Cortona, Lucca, and Avignon charts that are often treated as somewhat primitive — have portions of coastline that are nearly identical to those on allegedly more mature charts made by Genoese cartographer Pietro Vesconte in the early 14th century.
Even more intriguingly, his second article reveals that Vesconte’s charts remained the most accurate on a subsectional level for at least the next three centuries, while the third paper demonstrates that multi-sheet portolan atlases show the same pattern: images of seas that are geometrically frozen in time. Only this time, the atlases are mosaics within mosaics, with their subsections cut and sliced and arbitrarily encapsulated into sheets of varying map scales and coverages.
Tellingly, none of those authors seem to have fully understood the geometry of their own work, not even those who made portolan atlases, books that allow for more liberties to depict smaller regions than one-piece charts. Compare an atlas sheet to a portolan chart made by the very same cartographer, and even they don’t quite agree. These were not confident creators; they were copyists, doing their careful but imperfect best with material they didn’t fully understand.
Finding #2: The compass has an alibi
One of the more curious aspects of portolan charts is that their coastlines are all tilted roughly -10° off true north. The leading explanation has always been magnetic declination: Medieval sailors trusted their compasses, not realizing that magnetic north and true north aren’t quite the same thing — thus unknowingly infusing that error into every portolan chart.
It’s a neat story, but as Marelić found, it doesn’t hold water. He checked the theory against a paleomagnetic model of how Earth’s magnetic field actually behaved in the Mediterranean region between the late 13th and late 16th centuries. And the cartographic tilt doesn’t track with the magnetic reality of that era — not in degree, not in geography, not over time. Whatever initially rotated these coastlines, it wasn’t a 13th-century compass needle. In fact, his findings point to a completely different root cause: a late medieval graphical compromise between two map projections with significantly different appearances, both of which were developed about a millennium and a half before and long forgotten, but more on that later.
Finding #3: Medieval scholars point to Ancient Greece
Digging through historical sources, Marelić turned up three medieval witnesses who all, in their own words and independent of each other, say the same odd thing: that these charts came from an older time.
- Persian polymath Qutb al-Din al-Shirazi, writing in 1282 — a mere dozen years after the Carte Pisane surfaces — described a gridded crude map, geometrically almost identical to the Carte Pisane and the Cortona chart, that he attributes to the “sages of Greece and the ancient geometers.”
- A chart from around 1360 carries an inscription crediting a shadowy figure called Rex Tholomeus — likely a garbled mashup of “king” and “Ptolemy” — who supposedly “measured the world” long ago for being the author of its source data.
- A 1464 navigation manual by Ragusan merchant Benedetto Cotrugli says that Ptolemy made a world map from which “we have drawn a navigational chart” and that sailors of his day “owe him a lot because we do not wander [across the sea] today.”
Finding #4: The suspiciously specific mile
Every portolan chart uses the same basic unit of distance, the miglio, a mile generally reckoned at around 1.25 km. Nobody has ever satisfactorily explained where that particular number comes from. Marelić thinks he’s found it, and the culprit is, fittingly, a mistake.
Claudius Ptolemy, the 2nd-century astronomer whose Geography dominated Western cartography during the Renaissance, underestimated Earth’s circumference by about 29%. Measure out two degrees of longitude along the Mediterranean’s mid-latitude using Ptolemy’s too-small Earth, and you get almost exactly 125 kilometers — which just so happens to be 100 portolan miles, the length of one grid square on these charts.
In other words: The miglio isn’t a freshly measured medieval nautical mile at all. It’s a fossil, the shadow of an ancient astronomer’s arithmetic error, smuggled across more than a thousand years and built into every portolan chart’s ruler.
Finding #5: At last, a name
Marelić’s most recent paper is his boldest: an attempt to reconstruct, step by step, exactly how a specific chart — the Carta Riccardiana, likely drawn by Pietro Vesconte sometime between 1300 and 1325 — might actually have been assembled.
Marelić suggests that Vesconte based his work on older regional maps drawn in the style of Marinus of Tyre (a 1st-century Greek-Roman geographer whose lost atlas was a major influence on Ptolemy). According to Marelić, Vesconte tried to combine these separate maps into a single map by enlarging some parts, changing their angles and positions, and joining them together. The result was an imperfect attempt to imitate Ptolemy’s more advanced map projection: a single curved map centred on Gibraltar.
In parallel, this discovery identifies the root cause of the well-known “Mercator effect” — the appearance of a geometrical pattern across every portolan chart of the Mediterranean and Black Sea areas that resembles the Mercator projection, long before it was established in 1569. That effect is the result of collating local maps made in equidistant cylindrical projection – a simpler technique than Mercator, which nonetheless can resemble that later projection when stitched together. This resolves the apparent paradox of the “Mercator projection before Mercator,” which previously puzzled some scholars — even Marelić himself, in earlier stages of his research.
Lay all of Marelić’s studies end to end, and a single picture emerges. The portolan chart wasn’t a eureka moment of 13th-century Italian seamanship. It was a very good copy — or, more precisely, different copies — of cartographic knowledge that was already ancient before the Middle Ages even began, filtered through at least one confused but conscientious medieval hand, and finally dressed up with compass roses and scale bars to look shipshape for a new seafaring age.
Medieval sailors didn’t make these accurate maps. As they navigated the waters of their Mediterranean world, they were unwittingly standing on the shoulders of giants.
Back in 2014, Nicolai laughingly waved off the idea that lost civilizations had left behind the first portolan chart, the way you might snub a conspiracy theorist at a dinner party. Marelić’s five papers are, in a sense, the long-delayed main course for that dinner.
His findings require no spaceships, no lost super-civilization. Just Ancient knowledge, imperfectly transmitted by medieval copyists, intriguing enough to baffle modern humans.
Sometimes the most satisfying ending to a cold case isn’t a twist-and-reveal; it’s the meticulous piecing together of all the evidence hidden in the source material – one portolan mile at a time.
+ + +
We had the chance to ask Professor Marelić himself about the motivation for his research and the meaning of his discoveries. The following Q&A has been edited for length and clarity.
What first drew you to portolan charts?
“In short, it’s the mystery of their origin, which is one of the biggest unresolved questions in the history of cartography. They are out-of-place artefacts, the cartographic equivalent of the Antikythera mechanism.”
“I first learned about portolan charts via Charles Hapgood’s Maps of the Ancient Sea Kings when I was still in elementary school, back in the late 1990s. The subject was too complex for me to understand as a kid, but I was very impressed by the images of those old charts that were drawn so accurately.”
“The mystery lay dormant until my PhD studies decades later, when I cartometrically examined the accuracy of Adriatic Sea renderings on portolan charts. I realized that Hapgood’s theory of ancient lost civilisations as creators of the source maps is implausible. Many historians of cartography regard his book as pseudoscience. I don’t fully share that negative view: I believe his curiosity was pure, and I admire his pioneering audacity to use quantitative analysis on charts, even though they weren’t conducted with due diligence.”
“Then I started applying my analytical approach to the entire area covered by portolan charts, from the Atlantic to the Black Sea. And here we are, roughly four years and five peer-reviewed journal articles later.”
“The two most influential sources for my work are the doctoral dissertations of Scott Allen Loomer, from 1987, and Roel Nicolai, from 2014. Theirs are among the most extensive and intricate cartometric studies of portolan charts. What kept me going was the incompleteness of their results, and the occasional overcomplexity of their methods.”
“That incompleteness is very normal in scientific research, as well as understandable, since they didn’t have any historical evidence of pre-medieval origins. In contrast, I discovered Benedetto Cotrugli’s 1464 mention of an origin in classical Antiquity early on in my PhD research, and it fed my belief that there were other similar sources around. Luckily, that proved a good hunch.”
“Besides, I gradually came to think that the overcomplexity of some of the existing methods leads nowhere. I thought that if those charts were made by late medieval illustrators who didn’t understand map projections, some blatantly simple explanation for their geometry must be lurking somewhere in the dark. This suspicion eventually led me to what is probably the decoding of the portolan mile length and the mechanics of portolan chart assembly.”
Your work sits between geodesy, the history of science and the history of navigation. What did you have to teach yourself along the way?
“My pre-PhD degrees are in geography, a discipline that examines the environment and the world itself in a holistic manner. Also prior to my PhD, I developed decent GIS software skills and an understanding of various projected coordinate systems in the GIS environment and their applications. Combined with my curiosity about the history of technology and science, formed the foundation for this interdisciplinary research on portolan charts, which can be classified as digital humanities.”
Cotrugli, one of your key witnesses, was a Ragusan (i.e. from Dubrovnik), so basically a neighbor of yours. Did that geographic kinship add a personal touch to your work?
“Also known as Benedikt Kotruljević in Croatian, Cotrugli is most renowned for his book Of commerce and the perfect merchant, which contains the first description of double-entry bookkeeping in history. While it’s cool on a personal level to share a homeland with such a consequential figure, on a professional level, I interpret his writings as objectively as those of Qutb al-Din al-Shirazi, the Persian scholar whose records fall in the same camp as Cotrugli’s.”
Exactly what could a late-medieval navigator do with a portolan chart that they couldn’t do without one? In other words, What made these maps so valuable?
“The charts are a topologically fairly accurate representation, thus they give a good “big-picture” insight into Mediterranean sub-basins and other areas. So at the very least, they could have served as graphical confirmation of the topology that sailors had memorised.”
“However, this doesn’t work the other way around: the mental topologies memorised by navigators could not have produced charts of such accuracy on their own – the areas are too large, and the Earth’s curvature must be factored in. In other words: you’d need to start with a proper survey. And that survey could not be conducted with dead reckoning, because that method would lead to an accumulation of errors greater than those detected in portolan chart geometry.”
“Contrary to their – limited – usefulness, portolan charts have a number of features that could have made them death traps for navigators who trusted them blindly.”
“First, most of the charted area was not aligned with the magnetic declination prevailing when they were made. This limited their usefulness for sailing by compass, and that problem increased over time as the (easterly) declination decreased while the chart image remained unchanged.”
“Scales varied regionally, numerous transitions in coastline geometry reduced accuracy, and their generally small scale – often around 1:5,000,000 – made them poorly suited to navigating coasts dense with islands. Also, coastlines were crudely simplified, and especially the earliest charts contain no inland geographical information. The absence of isolated prominent landmarks like Mount Etna, which would have been visible from ships before the ports at its base, reduced navigational usefulness, and is a further argument against a medieval origin.”
In plain terms, what is the single most important thing your research has revealed about these charts that nobody knew before?
“If I had to choose one discovery, it would be the mechanics of portolan chart assembly proposed in my latest paper. But that would overlook the work of my earlier papers, because each added important pieces to the puzzle.”
“The first paper showed that the earliest anonymous charts, which are considered less advanced, contain coastline sections almost identical to those on Pietro Vesconte’s charts, which are considered more mature. This suggests that they all copied the same source material.”
“The second paper showed that Vesconte’s charts are the most accurate at the regional level over a period of three centuries. It also brought together, for the first time, three historical accounts claiming that portolan charts originated in classical antiquity.”
“The third paper provided the first cartometric analysis of portolan atlases. It showed that they are essentially portolan chart images arbitrarily split into physically separate pieces, and exposed extensive evidence against a medieval origin.”
“The fourth paper marked a shift towards positive evidence for an origin in Classical Antiquity. It quantitatively analysed Qutb al-Din al-Shirazi’s description of a map made by Greek sages and ancient geometers, identifying numerical evidence supporting his claim. It also proposed that the mysterious portolan mile was a late-medieval byproduct of treating Ptolemy’s inaccurate longitude intervals as distances and applying them to a relatively accurate geometric representation of the Mediterranean.”
“Finally, the fifth paper identified, for the first time, the geometrical fingerprints of Marinus of Tyre and an additional set of Ptolemy’s fingerprints in portolan chart geometry. It also proposed the simplest model yet for how portolan charts were assembled, one consistent with the historical development of map projections, and explaining the root cause of the so-called “Mercator effect” across the Mediterranean.”
Across several papers and years of work, what for you was the most surprising result?
“Two results spring to mind. The first one is the realisation that 50 portolan miles – a typical large interval on portolan chart scale bars – correspond to one degree of longitude along the Mediterranean’s mid-parallel, according to Ptolemy’s inaccurately perceived size of the Earth. The second is that typical portolan chart geometry, including the so-called “Mercator effect”, can be achieved by partially enlarging some source maps and rotating them by half a magnitude relative to the spatial layout on Ptolemy’s projection, moving eastward from Gibraltar.”
“Because these findings are based on simple and straightforward maths, reasoning or logic – whatever you want to call it – I couldn’t help but be surprised and amazed by the fact that no one had ever thought of it before. And, for that matter, that it hadn’t occurred to myself earlier.”
We may never be able to pin down the exact source with certainty, but if you had to put your money on it: where, when and with whom do you think the now-lost maps behind the portolan charts did first originate?
“Indeed! Even though the portolan chart origin story is increasingly coherent and less anachronistic as my research progresses – probably to the disappointment of those searching for more exotic origins in long-lost civilisations – my findings do still not provide definitive answers. What they do make clear, however, is that Greco-Roman science and technology were of a much higher level than we currently know, and that in itself is a fascinating conclusion, inviting more research.”
“If I was to bet on the most likely place and time of origin, I would say the Roman Empire at its greatest territorial extent, under Trajan, around 100 AD. Why? Two reasons. One is that the boundaries of the Empire at that time coincide almost perfectly with the typical coverage of portolan charts. The other is that this was the only time in history when the entire area covered by those charts was under a unified political authority with both an interest in mapping its territory, and the means and knowledge to do so to the same high standard of accuracy. Those conditions are missing from any period before or after.”
Strange Maps #1204
Got a strange map? Let me know at [email protected].
Follow Strange Maps on X and Facebook.
This article A medieval map mystery may finally be solved is featured on Big Think.