The Morning the Sky Caught Fire: 1 September 1859
On a quiet Thursday morning a country astronomer sketches a five-minute flash on the Sun — and seventeen hours later the world's telegraph network is throwing sparks, the aurora is burning over Cuba, and a number is set that we are still, statistically, waiting to meet again.
17 min read 11 sources
It is a little before noon on Thursday, 1 September 1859, and Richard Carrington is alone in the observatory dome behind his house at Redhill, south of London, doing the patient thing he does every clear morning: projecting the Sun’s image through his telescope onto a screen and tracing the sunspots in pencil. He has been watching one enormous group of spots for days. At eleven hours, eighteen minutes Greenwich mean time — a moment he will fix afterward to “not 15 seconds” —
He runs to fetch someone to corroborate it. By the time he is back, barely a minute later, the light is fading. Five minutes after it began, it is gone. Carrington sketches what he saw, notes the time with the obsessive precision of a man who knows no one will believe him otherwise, and has no way of knowing that the morning’s quiet anomaly is already on its way toward him through the dark. His written account, read weeks later to the Royal Astronomical Society, has the careful understatement of a man reporting something he cannot quite explain:
While I was directing the telescope, two patches of intensely bright and white light broke out… I hastily ran to call some one to witness the exhibition with me, and on returning within sixty seconds, was mortified to find that it was already much changed and enfeebled.
That single sentence is the birth certificate of an entire science. Carrington had no theory for what he saw, no instrument that could have predicted it, and the honesty to record exactly how fast it slipped away. The flare he could not explain would, within a day, write itself across the whole planet.
- Mexico ~23° N: 84
- United States ~40° N — telegraph chaos: 52
- British North America ~55° N: 30
- Cuba ~21° N — aurora overhead: 92
- Jamaica ~18° N: 88
- Colombia ~4° N — within sight of the equator: 100
- France ~47° N: 44
- England (Carrington / Kew) ~51° N: 36
- German states ~51° N: 42
- southern Japan ~33° N: 60
- Queensland ~23° S: 80
- New Zealand ~42° S: 70
What Carrington witnessed was a white-light flare: a sudden release of magnetic energy in the Sun’s atmosphere, bright enough to outshine the photosphere beneath it. Flares of that class hurl a coronal mass ejection — a billion tons of magnetized plasma — outward at millions of kilometres an hour. Most miss the Earth entirely; the Sun is a small target in a large sky. This one did not miss. And critically, the days before had already cleared its path: an earlier eruption had swept the solar wind ahead of it, so the great cloud raced through unobstructed space and reached Earth in
The physics of the hit is worth slowing down for, because it is the same physics that threatens the modern grid. A coronal mass ejection is not just fast plasma; it carries its own magnetic field, frozen into the cloud. When that field points southward — opposite to the Earth’s — the two fields connect, and the magnetosphere is peeled open like a flower. Solar particles pour in and are swept into a vast electric current encircling the planet, the ring current, tens of thousands of kilometres out. It is that ring, flowing the wrong way, that drags the Earth’s surface field downward — and the deeper the drag, the bigger the storm. Tsurutani’s reconstruction put the 1859 cloud’s southward field at roughly 90 nanotesla, several times the value that drives a serious modern storm.
The needle that ran off the paper
While Carrington slept that night, the instruments did the watching. At the Kew Observatory outside London, a self-recording magnetometer traced the Earth’s magnetic field onto a slowly turning drum of photographic paper — a steady line, the planet’s quiet pulse. As the plasma cloud slammed into the magnetosphere in the small hours of 2 September, that line convulsed. Half a world away, at the Colaba Observatory in Bombay, the horizontal-field trace plunged by roughly
The number that storm physicists care about is the Dst index — the disturbance-storm-time index, a measure in nanotesla of how far a planet-girdling ring of charged particles has dented the Earth’s field. A bad modern storm reaches −250 nT. May 2024’s superstorm that lit auroras over Puerto Rico bottomed out near −412 nT. Carrington’s storm is estimated at somewhere between −800 and −1,760 nanotesla.3From the Colaba depression Tsurutani et al. inferred a peak Dst near −1,760 nT, driven by an estimated southward interplanetary field of ~90 nT. Other reconstructions favour the lower end of the −800 to −1,760 nT range; the uncertainty is real and is discussed in the methodology note. Either way it dwarfs anything in the instrumental era. Tsurutani et al. 2003 The lower end of that range is already off the bottom of the modern severity chart. The upper end is in a category we have, mercifully, never had to name.
The wires come alive
It is the morning of 2 September 1859, and the telegraph network — the technological nervous system of the age — is turning against its operators, and then, impossibly, running on the storm itself.
Event 1 of 2: 1 Sep 1859, 11:18 GMT — flare
11:18 GMT — flare
Carrington and, independently, Richard Hodgson see the white-light flare at Redhill.
~04:00 GMT — impact
The CME reaches Earth after a ~17.6-hour transit; magnetometers convulse worldwide.
The lines begin to throw sparks. In Pittsburgh, the telegraph manager E. W. Culgan reports currents so strong that the platinum contacts are
And then it burns people. In Washington, the operator
Then they disconnect the batteries — and the line keeps working. At the American Telegraph Company’s Boston office the operators give up on their own power supply entirely and discover they can send messages to Portland, Maine,
Boston: “Please cut off your battery, and let us see if we cannot work with the auroral current alone.” Portland: “I will do so. Will you do the same?” Boston: “I have already done so. We are working with the aid of the aurora alone.”
It worked, and worked better: the Boston operator, Mr. Milliken, then sent private dispatches “much more satisfactorily than when the batteries were on,” and the line ran that way for more than two hours.5George Prescott’s 1860 History, Theory and Practice of the Electric Telegraph preserves the dispatch verbatim, noting the operators worked “steadier when the batteries were off than when they were attached.” The storm was a cleaner power source than the technology of the age. Prescott 1860 The most advanced communications system on Earth was, for those two hours, powered by nothing but the storm.
The genius of the moment is also its warning. A 19th-century telegraph wire is a thin copper thread; the currents the storm induced in it were enough to set paper on fire and weld relays. The principle is brutal and scale-free: a changing magnetic field drives a current in any long conductor beneath it, and the longer the conductor, the larger the current. In 1859 the longest conductors humanity had strung were telegraph lines a few hundred miles long. They threw sparks. The question that has haunted every grid engineer since is what the same sky would do to a conductor that is now a continent wide.
The night it looked like dawn
Above all of it, the sky was on fire — and not over the poles, where aurora belongs. The aurora is the visible signature of the same storm that ran the telegraph: charged particles funnelled down the magnetic field lines into the upper atmosphere, lighting the air like a colossal neon tube. Normally it crowns the Arctic. On the nights of 1–2 September 1859 it burned south across the entire inhabited world, and the newspapers that printed it the next morning had no scale for what they had seen.
In the northern cities, where aurora was at least familiar, the difference was its sheer brilliance. The display was bright enough to read by: the New York Times of 3 September reported that “at about one o’clock ordinary print could be read by light” of the aurora alone.6Period newspaper reports of the 1859 aurora are collected in the Green & Boardsen / SolarStorms.org archive. In the Rocky Mountain mining camps, miners rose and began cooking breakfast, certain dawn had broken. NYT, 3 Sept 1859 Closer to the coast, the color itself turned alarming — not the soft green of an ordinary aurora but a deep, oxidized red — and the New York Herald reached for the only word that fit:
The northern portion of the heavens assumed an almost blood red appearance, while here and there long streaks of light shot up from the horizon to the zenith.
In Abbeville, South Carolina, a crew of
What turns this from a spectacle into something unprecedented is how far south the light reached. Aurora over Boston is rare; aurora over the tropics had, until that night, essentially never been recorded. Yet the reports came in from the Caribbean and beyond — the New Orleans Daily Picayune gathering accounts from across the Gulf:
All our exchanges, from the northern coast of the Island of Cuba… come to us with glowing descriptions of the recent Aurora Borealis, which appears to have been as bright in the tropics as in the northern zones.
The light reached Colombia, within a few degrees of the equator — places that had never seen the northern lights and would not see them again for generations. The map above traces that descent: from the expected northern stage down across the United States and Europe, then over Cuba, Mexico, and the equatorial coast of South America, a single storm draped across nearly the whole inhabited world in one night.
Carrington, when he read the telegraph reports and the magnetometer traces over the following days, did something careful and consequential: he refused to overclaim. He noted the flare, noted its timing, noted the geomagnetic storm that followed seventeen hours later, and stopped short of asserting the one had caused the other — “one swallow does not make a summer,” he wrote. But the coincidence was too clean to ignore, and it founded a science. The discipline we now call space weather begins with a man being scrupulous about a five-minute observation he could barely believe he had made.
The arithmetic of the next one
Here the narrative has to turn, because the storm did not end in 1859 — it merely went quiet, and the bill it left was deferred to a civilization that had not yet been built. That civilization is ours, and it has done something the Victorians never imagined: it has strung the entire continent with the longest conductors in human history and made everything depend on them.
A modern high-voltage power grid is, electromagnetically, a 1859 telegraph wire scaled up by a factor of millions — thousands of kilometres of interconnected line, terminating in the one component that does not improvise its way through a surge. The transformers that step voltage up and down are the grid’s vital organs; a great geomagnetic storm drives a slow, quasi-DC current through them that saturates their cores, overheats them, and can destroy them outright. They are custom-built, weigh hundreds of tons, and have lead times measured in months to years. You cannot keep a continent’s worth in a warehouse.
The probability question was answered, with discipline, by the physicist Pete Riley in 2012. He took the historical record of great storms, fit a power law to how often events of a given magnitude occur, and extrapolated to the Carrington threshold.
The headline number is twelve percent. Riley’s power-law extrapolation puts the probability of a Carrington-magnitude storm —
And the bill is continental. The 2008 U.S. National Academies study warned that a severe storm could cascade through the interdependent systems modern life rests on — power, water, finance, communications.8The report treats space weather as a systemic, infrastructure-level hazard: lose the grid and you lose, in sequence, water pumping, fuel distribution, refrigeration, and the financial system — the cascade, not the blackout, is the threat. National Academies 2008 Lloyd’s of London, modelling it as an insurer in 2013, put concrete numbers on the worst case:
The numbers describe a chain, not an event. The storm itself lasts a day. The vulnerability is one component: the extra-high-voltage transformer, the multi-hundred-ton machine that steps power between the long-distance grid and the cities it feeds. A geomagnetic storm drives a slow, near-DC current up the grounded neutral of these transformers; that current saturates the iron core, so the machine overheats, hums, and — in the worst case — cooks its own windings.10The National Academies report identifies the large extra-high-voltage transformer as the critical single point of failure: there is no domestic stockpile, units are custom-built abroad with lead times of a year or more, and a storm could damage many at once — turning a one-day event into a multi-year recovery. National Academies 2008 What lasts is the silence after: a transformer that takes a year to replace is a city without power for a year, and a city without power is, in cascading order, a city without water pressure, without fuel pumps, without refrigeration, without functioning hospitals or banks. The 1859 storm fell on a world that ran on horses and gaslight, and so it left behind little more than scorched telegraph paper and the most beautiful skies in living memory. Dropped on a world that runs on the grid, it would fall on everything at once.
The shot we did not have to take
This is not a hypothetical drawn only from models. On 23 July 2012, the Sun fired a coronal mass ejection of Carrington magnitude straight across the Earth’s orbit — and missed, because the planet had been at that spot in its orbit nine days before. We have the measurements only by luck: the cloud struck NASA’s STEREO-A spacecraft, which happened to be sitting in its path and was equipped to record it.11The July 2012 CME hit STEREO-A in interplanetary space, where the weak ambient field meant the spacecraft survived to report a blow Earth never felt. Analysts later estimated that, had it struck Earth, the storm would have reached a Dst near −1,200 nT — squarely in Carrington territory. NASA 2014 Had it arrived a week earlier, analysts estimate the geomagnetic storm would have hit a Dst near −1,200 nanotesla, comparable to 1859 and roughly twice the severity of the 1989 storm that blacked out Québec. The physicist who led the study, Daniel Baker, did not hedge:
If the eruption had occurred only one week earlier, Earth would have been in the line of fire.
That near-miss is exactly what Riley’s twelve percent looks like in practice: not a distant abstraction, but a live round that passed through the space the Earth had just vacated. The same NASA analysis cites Riley’s estimate directly — a roughly 12% chance of a Carrington-class hit in the following decade.12NASA’s account of the 2012 event closes by quoting Riley’s figure, framing the near-miss as a reminder that the probability is not zero and the consequences are civilizational. NASA 2014
What stands between us and the next one
The honest answer is: warning, and not much else. The one real defense the modern world has built is foreknowledge. A satellite called
The clock that has been running since Carrington put down his pencil
This is the inheritance of that quiet Thursday morning. A solar storm is not a freak; it is weather, and like weather it recurs. The Sun has thrown Carrington-class storms before and will throw them again, and Riley’s twelve percent is not a prophecy of doom but a statement of standing odds — odds we re-roll every decade, against a target we keep building larger.
Scroll the rail below and the playhead walks that long arc — from a five-minute flash one Thursday morning to the modern machinery built to see the next one coming. Drag it, or click any date, to jump.
- 1 Sep 1859Carrington sees the flare
At 11:18 GMT on a clear Thursday, Richard Carrington is tracing sunspots at Redhill when two patches of brilliant white light erupt across one spot. He fixes the time to within fifteen seconds; by 11:23 the light is gone — a flare lasting roughly five minutes.
It is the first solar flare ever recorded by human eyes, and Carrington has no theory for what he has seen. (Carrington, MNRAS 20, 1859)
- 2 Sep 1859The storm hits
A coronal mass ejection — a billion tons of magnetized plasma — crosses the 150-million-km gap in about 17.6 hours, roughly four times a normal storm's speed, because an earlier eruption had swept its path clear. Telegraph lines throw sparks, shock their operators, and then run on the storm's own current alone.
Overhead, the aurora burns south past the United States and Europe to Cuba, Mexico, and Colombia — within a few degrees of the equator. (Tsurutani et al. 2003)
- 1 Jul 2003Tsurutani reconstruction
More than a century later, Tsurutani and colleagues reduce the surviving magnetograms — chiefly the Colaba (Bombay) record, which plunged about −1,600 nT — and infer a peak Dst near −1,760 nT, driven by an interplanetary field of roughly 90 nT pointing south.
For comparison, a severe modern storm reaches −250 nT and May 2024's superstorm bottomed near −412 nT. The 1859 storm remains the benchmark against which every great storm is still measured. (Tsurutani et al. 2003)
- 1 Jan 2008National Academies warning
The U.S. National Academies study reframes space weather as an infrastructure-level hazard. The danger is not the blackout but the cascade: lose the grid and you lose, in sequence, water pumping, fuel distribution, refrigeration, and the financial system.
The single point of failure it identifies is the large extra-high-voltage transformer — custom-built, weighing hundreds of tons, with no domestic stockpile and lead times of a year or more. (National Academies 2008)
- 23 Feb 2012Riley: ~12% per decade
Pete Riley fits a power law to the historical distribution of storm intensities and extrapolates to the Carrington threshold (Dst below about −850 nT). The result: roughly a 12% probability of a Carrington-class event in any given decade.
Not in a lifetime — in a decade. It is the most-cited figure in the field, and an admittedly uncertain extreme-value extrapolation, but it is not small. (Riley 2012)
- 23 Jul 2012The near-miss
Five months after Riley's paper, the Sun fires a Carrington-magnitude CME straight across Earth's orbit — and misses, because the planet had left that spot nine days earlier. The cloud strikes NASA's STEREO-A spacecraft instead, which happened to be in its path and equipped to record it.
Had it arrived a week earlier, analysts estimate the storm would have reached a Dst near −1,200 nT. "If the eruption had occurred only one week earlier," said the study's lead, Daniel Baker, "Earth would have been in the line of fire." (NASA 2014)
- 1 May 2013Lloyd's risk model
Lloyd's of London, modelling the scenario as an insurer, puts numbers on the worst case: 20 to 40 million people in the U.S. without power, for durations from sixteen days to one or two years, at a total economic cost of $0.6–2.6 trillion.
The duration tail hinges on the transformers — weeks if the grid only trips, years if the largest units are destroyed and must be rebuilt and shipped. That tail, measured in years, is what makes the risk uniquely dangerous. (Lloyd's / AER 2013)
- 11 Feb 2015DSCOVR at L1
The Deep Space Climate Observatory launches toward the L1 Lagrange point, a million miles sunward, where it samples the solar wind — and the all-important direction of its magnetic field — before it reaches us. That gives NOAA's forecasters 15 to 60 minutes of warning before a storm strikes the magnetosphere.
It is not much, but it is enough for a grid operator to shed load and protect transformers — if the warning is believed and acted on. That hour is the whole of our advantage over the Boston operators of 1859. (NOAA SWPC)
We have already lived through the near-miss, and we measured it only by accident. The flare Carrington sketched in pencil in 1859 is still out there, in a sense — not as a one-off curiosity but as a recurring number on the Sun’s books, a roll of the dice it will keep making whether we watch or not. The Boston operators learned what the storm was while it burned their paper. We have bought ourselves an hour’s warning and a single statistic, and the work of the next century is to make both of them count. Somewhere on that quiet star, the next one is already loaded.
Methodology. Every figure and quotation traces to a cited source. Carrington’s flare — the 11:18 GMT timing, the “two patches of intensely bright and white light,” the five-minute duration, and the verbatim pull-quote (“I hastily ran to call some one… mortified to find that it was already much changed and enfeebled”) — is his own account in MNRAS 20 (1859), via the NASA ADS record. The ~17.6-hour CME transit, the southward-field / ring-current mechanism, the Colaba −1,600 nT magnetogram depression, and the inferred peak Dst near −1,760 nT (southward field ~90 nT) are from Tsurutani et al. (2003, JGR 108 A7); the broader −800 to −1,760 nT range reflects genuine scientific disagreement over reconstructing a 19th-century storm from sparse instruments, so I cite the range, not a single value. The telegraph anecdotes (Culgan’s “streams of fire,” Royce’s shock, combusting paper, the Abbeville bricklayers) are from the HISTORY account (Klein 2012); the verbatim Boston–Portland dispatch (“cut off your battery… work with the auroral current alone”), the operator Mr. Milliken, the “much more satisfactorily than when the batteries were on” line, and the two-hours figure are from Prescott’s 1860 History, Theory and Practice of the Electric Telegraph. The period-newspaper aurora quotations — the New York Times (“ordinary print could be read by light”), the New York Herald (“blood red appearance… from the horizon to the zenith”), the New Orleans Daily Picayune (Cuba/tropics), and the Cincinnati Daily Commercial (“positively awful… red glare”) — are transcribed verbatim in Green & Boardsen’s NASA NTRS report, Eyewitness Reports of the Great Auroral Storm of 1859; aurora latitudes (Cuba, Colombia, Hawaii) are corroborated by the Wikipedia summary. The modern-storm comparison figures (May 2024 ≈ −412 nT; a severe modern storm ≈ −250 nT; the G1–G5 scale and its G5 “collapse of power grids” language) are NOAA SWPC’s published space-weather scales. The ~12%-per-decade recurrence is Riley (2012, Space Weather 10, S02012), a power-law fit at a Dst < −850 nT threshold — much-cited but an admittedly uncertain extreme-value extrapolation. The societal-cascade framing and the extra-high-voltage-transformer single-point-of-failure are the National Academies (2008) report; the $0.6–2.6 trillion cost, 20–40 million people, and 16-day-to-2-year restoration are the Lloyd’s/AER (2013) grid study. The July 2012 near-miss — the 23 July date, the STEREO-A measurement, the estimated Dst −1,200 nT, and Daniel Baker’s “one week earlier” quote — is NASA’s 2014 account (Phillips). DSCOVR’s L1 position and the 15–60-minute warning window are NOAA SWPC; the timeline’s DSCOVR entry uses its 11 Feb 2015 launch. No name, number, date, or quotation in this piece is invented; where a quantity is uncertain, the range and its source are stated rather than a false-precise single figure.