Before engines, a ship went where the wind let it. Officers wrote down their position, the wind and the weather every day, and thousands of their logbooks survive. Drawn as lines, 6,524 stretches of real voyages from British, Dutch, Spanish and French logbooks show the highways of the age of sail: out with the trade winds, round in a great loop home, and east along the roaring forties.
From the 1750s onwards almost every officer on an ocean-going European ship kept a logbook. Each day at noon it recorded where the ship was, worked out from the course, speed and the occasional sight of the sun, and how the wind blew, because the wind decided how far the ship had drifted. Between 2001 and 2003 a European project called CLIWOC, the Climatological Database for the World's Oceans, went into the archives of Britain, the Netherlands, Spain and France and typed up 260,631 of those daily entries from 1,740 logbooks, for climate scientists. They also happen to be the best open record of where sailing ships actually went.
The map below draws every one of them. Each line is a passage: an unbroken run of daily positions in one logbook, from the Channel to the Cape, Cádiz to Havana, Batavia to St Helena. Colour shows the ship's nation. The arrows are the logbook winds themselves, averaged on a 5° grid. Slide through the decades, pick a nation or a kind of operator, search for a ship, or click a line to see whose it was.
Lines: CLIWOC v2.1 daily logbook positions (noon positions, mostly by dead reckoning), cut into passages where the log is silent for more than 10 days. Arrows: mean logbook wind direction in each 5° square with at least 25 observations; length shows how steady it was, colour how strong. Today's shipping: World Bank and IMF AIS density, 2015 to 2021, summed over 2° squares. Base map tiles © Esri.
The lines make the first rule of the age of sail plain: you could not sail straight to where you were going and back the same way. Ships bound south from Europe ran down the coast of Africa past Madeira and the Canaries, picked up the north-east trade winds, and crossed the equator well out in the Atlantic, at a median of 23° W, to avoid the calms of the Gulf of Guinea. Coming home they could not beat back against those same trades, so they swung out west in a great arc, the Portuguese volta do mar or "turn of the sea", until they reached the westerlies near the Azores and ran home before them.
The logbooks measure it. Of the 1,139 crossings of 30° N by ships sailing between Europe and the South Atlantic, outbound ships crossed at a median of 18° W and homebound ships at 36° W, 18° further west. At 38° N, the latitude of the Azores, 66% of homebound ships passed between 35° W and 20° W; only 2% of outbound ships did.
Where ships crossed 30° N
Ships on passages between Europe and the South Atlantic, by the longitude where their daily track crossed 30° N, in 5° bins. Hover a bar.
In 1611 the Dutch captain Hendrik Brouwer found that a ship leaving the Cape of Good Hope could run east in the strong westerlies near 40° S and then turn north for Java, instead of creeping up the coast of Africa against the monsoon. The Dutch East India Company soon made his route compulsory. The logbooks show it still in use two centuries later. Of 323 Dutch crossings of 75° E heading east in the southern Indian Ocean, the median was at 38° S and 100% were south of 30° S. On voyages to Batavia the median Dutch ship got as far south as 40° S before turning north.
British ships used the forties too, but many were bound for India rather than Java and peeled off northward sooner: at 90° E the median British eastbound crossing was at 15° S, against 34° S for the Dutch. The logbook winds in the forties blow from a westerly point 70% of the time, with a mean force of 9.3 m/s, nearly twice that of the equatorial calms.
How far south ships crossed 75° E
Eastbound crossings of 75° E in the southern hemisphere, by latitude, in 2.5° bins.
Because each log is dated, the time between leaving one place and reaching another can be read straight off it. I counted a voyage from a ship's last position in home waters (the North Sea, the Channel and its approaches) to its first position at the destination, allowing calls on the way. 219 voyages reached Batavia, nearly all of them Dutch: the median took 102 days, and the quartiles were 94 and 119. The fastest was the ADMIRAAL JAN EVERTSEN in 1842, in 80 days.
Earlier voyages took far longer. The 15 voyages that left before 1820 had a median of 176 days, against 101 days from 1820 on. Part of that is the long stop at the Cape that Company ships made, but even the run from the Cape to Batavia fell from a median of 71 days to 45. The early numbers rest on few voyages, so treat them as a sketch.
Voyage times on the great routes
Days from the last position at the start to the first position at the destination. Dot: median; bar: middle half of voyages; whisker: fastest and slowest. Number of voyages in brackets.
The Spanish mail packets, the correos marítimos, ran from La Coruña to Puerto Rico in a median of 38 days, downwind all the way in the trades. Getting home from Havana took 81. British ships reached the Caribbean from home waters in 38 days and India in 133.
Officers recorded the wind in words and compass points, in English, Dutch, Spanish and French, and the CLIWOC team translated them into degrees and forces. 241,792 daily entries (93%) give a direction. Averaged in 5° squares they draw the planet's wind belts as a meteorologist would: easterly trade winds either side of the equator, westerlies further out, and calms in between. In the Atlantic, 73% of winds between 10° and 25° N came from between north and east, the north-east trades, and 68% between 5° and 25° S came from between east and south. Between the equator and 5° N, 7.9% of entries were calm or variable, 2.9 times the rate in the trade-wind belts: the doldrums.
The Atlantic's wind belts, from logbooks and from today's climate
Mean west-to-east part of the wind by 5° latitude band, 70° W to 20° E. Negative: from the east (trades). Positive: from the west (westerlies). Logbooks: average of daily wind directions. Modern: NCEP/NCAR reanalysis monthly means for the same places and months.
To check them, I set every logbook wind against the modern long-term average (the NCEP/NCAR reanalysis, 1991 to 2020) for the same place and calendar month. In the 318 squares where today's average wind is steady, the logbook mean points within 45° of it in 97% and within 30° in 93%, with a median difference of 9°. A single day's wind is noisier: 65% of daily entries fall within 45° of the average. Across the 39 latitude bands of the Atlantic and Indian Oceans, the logbook and modern west-to-east components correlate at 0.93. The northern Indian Ocean shows the monsoon flip: from December to February 65% of winds there came from the north-east quadrant, from June to August 81% from the south-west.
Each nation's logbooks trace its own empire. Joining the places each ship sailed between (passing over island calls such as Madeira and St Helena), the busiest route in the record is between Europe and Batavia, with 252 legs, 236 of them Dutch, followed by Europe and the Cape (242). The Dutch also sailed to Suriname and the Gold Coast: many of those logbooks belong to the Middelburgsche Commercie Compagnie of Zeeland, whose ships carried enslaved Africans to the Caribbean, and their lines trace that triangle. Spanish logbooks run from La Coruña and Cádiz to Puerto Rico, Havana and the River Plate. British logbooks, most of them from the Royal Navy, reach the Cape, the Caribbean and Ceylon.
The most sailed routes for each nation
Legs between named places (both directions together), counted from consecutive port visits in each logbook. Click a route to see it on the map.
This is a map of surviving, archived and digitised logbooks, not of all shipping. CLIWOC worked in British, Dutch, Spanish and French archives, so Portuguese, Danish, Swedish and American ships are almost absent (8 records between them), as is most merchant shipping, whose logs were rarely kept. Navies dominate: 1,030 of the 1,973 records (52%) are from naval ships. Each nation is lopsided in time too. 97% of the Spanish positions date from 1760 to 1799; 67% of the Dutch from 1820 or later, many of them logbooks held by KNMI, the Dutch weather service; and 73% of the British from before 1800. The thinnest decade is the 1810s, the years of the Napoleonic Wars, with 8,192 positions.
Daily positions by decade and nation
All 259,522 positions, by the decade of the logbook date. A handful before 1750 and after 1849 are included in the end bars.
Some famous voyages are in it. Cook's Endeavour is there from 1768, and his Resolution, whose log reaches 71° S on 30 January 1774, the farthest south anyone had then sailed. So are La Pérouse's Boussole, lost with all hands in 1788, and the Racehorse, which reached 80.8° N on 27 July 1773 on a Royal Navy attempt to sail towards the North Pole. And the end of the age is there too: 19 records come from steamships, the first of them the Dutch navy's Curaçao in 1827, on one of the first Atlantic crossings made largely under steam.
The positions themselves carry errors. Most were worked out by dead reckoning, from course and speed, and longitude at sea was uncertain until chronometers spread at the end of the eighteenth century. I removed 63 lone positions that jump hundreds of kilometres away and straight back, but smaller errors remain. A day here is one entry, usually at noon: the lines join noon positions with straight segments, not the ship's real course.
Steam, the Suez Canal (opened in 1869) and the Panama Canal (1914) freed ships from the wind, and the map of the sea changed. Comparing the logbook record with today's ship traffic, measured from AIS positions by the World Bank and IMF and summed over 2° squares, the two agree only loosely (rank correlation 0.22). 5.9% of the logbook record crosses the southern Indian Ocean between 35° and 50° S, against about 0.0% of today's traffic. The Red Sea and the Gulf of Aden carry 1.3% of today's traffic but only 0.01% of the logbook record, and the Strait of Malacca 0.04%. 31% of the busiest tenth of today's squares saw no logbook ship at all. One caution: the logbook record is a sample, so 'no logbook ship' does not mean no ship sailed there.
Share of the logbook record and of today's traffic
Logbooks: passages crossing each 2° square. Today: AIS ship density index, 2015 to 2021. Both as a share of their world total.
The logbooks still know the winds: in 97% of squares their average points the way today's does.
Data. CLIWOC version 2.1, as a GeoJSON of ship tracks prepared by Francisco Correoso (kikocorreoso/cliwoc_vis on GitHub) from the CLIWOC21 GeoPackage, which keeps each logbook entry's position, date, wind direction and force, and the ship, nationality, company, ship type and archive. The file holds 1,973 records (CLIWOC ship and voyage identifiers, from 1,740 logbooks) and 260,631 positions from 1662 to 1855; 100% fall in CLIWOC's period of 1750 to 1854.
Passages. I sorted each record's positions by date, kept one a day, dropped lone spikes, and cut the record wherever the log was silent for more than 10 days or the next position implied more than 450 km a day. That gives 6,524 passages with at least 3 positions. A passage can include days at anchor. Operators come from CLIWOC's company field (Royal Navy, East India Company, VOC, the Dutch navy and admiralties and so on); for Spanish ships, which have no company, I used the ship type: packet boats and mail frigates are counted as mail packets, ships of the line, frigates, corvettes and brigs as navy.
Routes and voyage times. I drew circles and boxes round 42 ports and anchorages and called a ship 'at' a place when its position fell inside. A leg runs from a ship's last position at one place to its first position at the next place, with no silence over 30 days; a voyage for timing may include calls on the way but no silence over 45 days. Crossings of a line of latitude or longitude are interpolated between daily positions. Times run from when the log last puts the ship in port or home waters, so they are a little shorter than door-to-door times.
Winds. CLIWOC's directions are where the wind blew from, with codes for calm and variable winds. In each 5° square I averaged unit vectors (so a strong gale and a light breeze count the same), and compared them with NCEP/NCAR Reanalysis 1 monthly means (1991 to 2020) interpolated to each entry's place and month. The reanalysis is a modern climate model's reconstruction, so differences mix logbook errors, sampling and real change since the eighteenth century.
Today's shipping. The World Bank and IMF Global Shipping Traffic Density raster, in the overview version that samples single pixels, summed over 2° squares, where it is a rough but fair index. It depends on where AIS receivers and satellites pick ships up, so treat it as a guide.