Carrington Event & Space Weather

What happened in September 1859

On the morning of 1 September 1859, the English brewer-turned-astronomer Richard Carrington was sketching an unusually large sunspot group projected onto a screen behind his telescope. At about 11:18 GMT two brilliant beads of white light flared up within the group, brightened, and faded over roughly five minutes. A few miles away, Richard Hodgson saw the same thing. They had just made the first recorded observation of a solar flare — and, by luck, the largest one anyone has ever timed against its terrestrial aftermath.

Carrington noted that a small magnetic disturbance registered at Kew Observatory while the flare was visible. That was the flare's X-rays and energetic particles arriving at light-speed. The real blow came about 17.6 hours later. Near 04:50 GMT on 2 September, magnetometers around the world pinned their needles, the global magnetic field convulsed, and the night sky erupted in aurora. People in Cuba, Hawaii, Colombia, and Queensland — places that never see the northern or southern lights — read newspapers by their glow. Gold miners in the Rocky Mountains woke and made breakfast, thinking it was dawn.

And the telegraph network, the high-tech infrastructure of 1859, went haywire. Operators received electric shocks. Pylons threw sparks. Some lines were so saturated with induced current that operators disconnected their batteries entirely and kept sending messages on the geomagnetically induced current alone. It was the first time humanity's technology had been bitten by space weather — and the bite has only grown sharper as the technology has grown more electrical.

From flare to coronal mass ejection

The Carrington Event was really three Sun-driven phenomena stacked in sequence, each with its own messenger and travel time. Untangling them is the key to space-weather forecasting.

• The flare — light-speed, 8.3 minutes. A solar flare is a sudden release of magnetic energy in the corona via magnetic reconnection. Its electromagnetic radiation (X-rays, EUV, white light) reaches Earth in 8.3 minutes and ionizes the dayside upper atmosphere, disrupting HF radio almost instantly. This is what Carrington literally saw.
• The solar energetic particles — minutes to hours. Reconnection and the CME shock accelerate protons to a sizeable fraction of light speed. The fastest arrive within tens of minutes, posing a radiation hazard to astronauts, polar flights, and satellite electronics.
• The coronal mass ejection — hours to days. A CME is a billion-tonne cloud of magnetized plasma blasted off the Sun at hundreds to thousands of km/s. It is the CME's embedded magnetic field, plowing into Earth's magnetosphere, that drives the geomagnetic storm. The Carrington CME crossed 1 AU in 17.6 hours; a typical CME takes two to four days.

That 17.6-hour transit time is the headline number. The Sun-Earth distance is 1 AU = 1.496 × 10⁸ km. Crossing it in 17.6 hours = 63,360 s gives an average speed:

v_avg = 1.496 × 10⁸ km / 63,360 s ≈ 2,360 km/s

The actual launch speed was likely higher still — CMEs decelerate against the ambient solar wind. The favored explanation is that a CME a day or two earlier (a strong aurora is recorded on 28 August 1859) had already swept the path clear of slow solar wind, so the 1 September ejection plowed through with little drag, like a second car drafting the first. The pre-conditioned interplanetary medium is why this one arrived so devastatingly fast.

How a CME drives a geomagnetic storm

Earth's magnetosphere is a magnetic cavity carved out of the solar wind, with its dayside boundary (the magnetopause) normally near 10 Earth radii (≈ 64,000 km). A CME is dangerous when it carries a strong magnetic field whose direction is southward — antiparallel to Earth's own northward field at the dayside. When the two fields meet, they reconnect, peeling open the magnetosphere and letting solar-wind energy pour in.

The energy is loaded into the nightside magnetotail, then unloaded explosively into the inner magnetosphere. Charged particles surge westward around the planet at a few Earth radii, forming the ring current. By Ampère's law this westward current generates a magnetic field that, at the surface, points opposite to Earth's main field — depressing the horizontal field strength. That depression is exactly what the Dst index (disturbance storm-time) measures, averaged over low-latitude magnetometers:

Dst < 0   →  ring current intensifying (storm main phase)
Dst recovery → ring-current ions lost to charge exchange & outflow

Quiet:        Dst ≈ 0 to −20 nT
Moderate:     Dst ≈ −50 to −100 nT
Intense:      Dst < −100 nT
Great/severe: Dst < −250 nT
Carrington:   Dst ≈ −850 to −1,600 nT (reconstructed)

The Dessler–Parker–Sckopke relation ties the ring-current strength directly to the energy injected: the field depression is proportional to the total kinetic energy of the trapped ring-current particles divided by Earth's dipole energy. A −850 nT excursion therefore implies an enormous energy deposition — the reconstructed Colaba magnetogram from Bombay shows a horizontal-field drop of about 1,600 nT in roughly an hour, one of the steepest rates of change ever recorded.

Geomagnetically induced currents — the grid killer

The damage to technology is not done by the magnetic field itself but by its rate of change. Faraday's law of induction states

∇ × E = − ∂B/∂t

A magnetic field that swings by hundreds of nanotesla in minutes induces a horizontal electric field of a few volts per kilometre at Earth's surface (the exact value depends on ground conductivity — resistive cratonic shield rock makes it worse). That electric field drives geomagnetically induced currents (GICs) through any long grounded conductor: transmission lines, pipelines, railway signalling, undersea cables. Because the storm timescale (minutes) is enormous compared with the 50/60 Hz power-grid cycle (≈ 17 ms), the GIC behaves like a quasi-DC current superimposed on the AC.

In a high-voltage transformer, that DC offset is poison. It biases the magnetic core so that it saturates on one half of each AC cycle — half-cycle saturation. The consequences cascade:

• Heating. Stray flux escapes the saturated core into structural steel and windings, producing hot spots that can exceed 150 °C and char the insulation over minutes to hours.
• Harmonics. The distorted magnetizing current injects even and odd harmonics that can trip protective relays and capacitor banks.
• Reactive-power drain. A saturated transformer gulps reactive power (VARs), depressing voltage across the network. In the 1989 Québec storm, voltage collapse propagated across Hydro-Québec in about 90 seconds, blacking out 6 million people for up to 9 hours.

The strategic problem is the transformers themselves. The largest extra-high-voltage (EHV) transformers are bespoke, weigh hundreds of tonnes, and have manufacturing and shipping lead times of 12 to 24 months. Utilities do not keep spares for most of them. Lose a dozen in one storm and you cannot simply swap them in — which is why grid-scale recovery from a severe storm is measured in months, not days.

How Carrington compares to modern storms

Several severe storms in the instrumental era give us calibration points. None has approached Carrington, but each has done real damage — a reminder that we have simply been lucky.

Event	Date	Peak Dst	CME transit	Notable impact
Carrington Event	1–2 Sep 1859	≈ −850 to −1,600 nT	~17.6 h	Telegraphs sparked; aurora to ~23° geomag. lat.
New York Railroad storm	15 May 1921	≈ −900 nT (est.)	~13–20 h	Telegraph & rail fires; aurora to the tropics
Hydro-Québec storm	13 Mar 1989	≈ −589 nT	~52 h	9-hour blackout, 6M people; NJ transformer damaged
Bastille Day storm	14 Jul 2000	≈ −301 nT	~28 h	Satellite upsets; ASCA spacecraft lost
Halloween storms	28 Oct–4 Nov 2003	≈ −383 nT	~19 h	Sweden blackout; airline reroutes; satellite damage
23 July 2012 (missed Earth)	23 Jul 2012	Carrington-class (modeled)	~18.6 h to STEREO-A	Direct miss; measured by STEREO-A in situ
Gannon / Mother's Day storm	10–11 May 2024	≈ −412 nT	~?	Aurora worldwide; precision-ag GPS outages

The 1989 storm — at less than 70% of Carrington's likely Dst — was enough to take down a major grid in 90 seconds. The 2003 Halloween storms forced aircraft off polar routes and damaged satellites. The May 2024 "Gannon" storm, the strongest in two decades, painted auroras over the tropics and knocked out GPS-guided tractors across the US farm belt during planting season. Scaling those impacts up to Carrington intensity is what keeps grid operators awake.

The numbers: energy, mass, and timescales

Putting the Carrington Event on a quantitative footing makes its scale concrete.

Quantity	Value	Comparison
Flare class (estimated)	~X45 (≈ 5 × 10⁻³ W/m² in soft X-ray)	The largest directly measured was X28+ (Nov 2003)
Flare energy	~10³² erg (10²⁵ J)	≈ 2 billion megatons of TNT
CME mass	~10¹⁶ g (10¹³ kg)	About a billion tonnes of plasma
CME speed at launch	≳ 2,400 km/s	~6× normal solar wind (400 km/s)
Sun-Earth distance	1 AU = 1.496 × 10⁸ km	Light crosses it in 8.3 minutes
Peak |dB/dt|	~2,000–5,000 nT/min (modeled high-lat.)	Drives the worst GICs
Induced surface E-field	up to several V/km (resistive ground)	1989 Québec peaked near 2 V/km
Ice-core nitrate spike	contested as a proxy	¹⁰Be / ¹⁴C confirm rare extreme SPEs (e.g. AD 774)

The flare's ~10³² erg is comparable to the total solar luminosity (3.8 × 10³³ erg/s) integrated over a fraction of a second — a genuinely stellar-scale energy release, dwarfed only by superflares seen on other Sun-like stars by Kepler. Tree-ring ¹⁴C and ice-core ¹⁰Be records reveal even larger solar particle events in the deeper past: the AD 774–775 and AD 993 "Miyake events" show carbon-14 spikes implying proton fluences perhaps an order of magnitude beyond Carrington. The Sun can clearly do worse than 1859; we have simply never had electrical infrastructure standing in the way of its biggest punches.

Watching the Sun: forecasting and mitigation

Modern space-weather defense is a layered observing system feeding operational forecasts:

• Solar imaging. NASA's Solar Dynamics Observatory and ESA/NASA's SOHO watch the disk and corona continuously; coronagraphs catch CMEs leaving the Sun and let forecasters estimate speed and direction within hours.
• The L1 sentinels. DSCOVR and ACE sit at the Sun-Earth L1 Lagrange point, 1.5 million km upstream. They sample the solar wind's speed, density, and — critically — its magnetic-field direction (Bz) about 15–60 minutes before it hits Earth. A southward Bz is the red flag.
• Close encounters. The Parker Solar Probe (perihelion inside 10 solar radii) and Solar Orbiter sample the corona and inner solar wind in situ, improving the physics that goes into CME-propagation models like ENLIL.
• Operational response. Forecasts give grid operators hours of warning. They can reduce loading on vulnerable lines, postpone maintenance that takes spares offline, reconfigure the network to break long induction loops, and — increasingly — install series capacitors and neutral-blocking devices that physically stop GICs from entering transformer neutrals.

The hard limit is lead time. The flare and particles arrive in minutes to hours; the CME's magnetic structure — the part that matters most for the storm — is only reliably known when it reaches L1, giving under an hour of true warning of storm severity. That is why in-situ space weather monitoring and faster CME-prediction models are an active frontier, and why the Carrington benchmark drives planning at agencies from NOAA's Space Weather Prediction Center to national grid regulators.

How likely is the next one

Pete Riley's widely cited 2012 study fit a power law to the historical distribution of storm intensities and estimated the probability of a Carrington-class event striking Earth at roughly 12% per decade — about a coin flip's worth of risk over a human lifetime. That estimate carries large error bars, but the order of magnitude is robust, and it is corroborated by the 23 July 2012 near-miss.

On that date, a CME at least as fast and magnetized as Carrington's erupted from active region 1520 and crossed Earth's orbit. It missed only because Earth had occupied that longitude about nine days earlier; the storm instead struck the STEREO-A spacecraft, which measured it directly and confirmed solar-wind speeds of 2,000–3,000 km/s with a magnetic-field magnitude several times typical values. Had it left the Sun nine days earlier, 2012 would now be the textbook example instead of 1859. The lesson of that near-miss is blunt: Carrington-scale storms are not historical curiosities. They are a recurring feature of solar activity, and our exposure depends on the roll of orbital timing.

Common misconceptions and edge cases

• "The flare itself caused the blackouts." No. The flare's radiation disrupts radio and ionizes the dayside atmosphere, but the grid damage comes from the CME's geomagnetic storm, arriving up to a day later. Flare and CME are distinct events that often (not always) occur together.
• "A solar storm could fry every electronic device, like an EMP." No. The induced electric field is only a few volts per kilometre — harmless to a phone or laptop. The threat is to long grounded conductors where that field integrates into large currents: continent-spanning transmission lines, pipelines, cables. Your devices are at risk only secondhand, if the grid that powers them fails.
• "It would knock out satellites by frying their chips directly." Partly. Energetic particles do cause single-event upsets and total-dose damage, but the more universal satellite threat is atmospheric expansion: heating puffs up the thermosphere, increasing drag and dropping satellites' orbits. In Feb 2022 a modest storm doomed 38 newly launched Starlink satellites this way.
• "Low latitudes are safe." Less exposed, not immune. GICs peak at high geomagnetic latitudes under the auroral electrojet, but during a Carrington-class storm the auroral oval expands toward the equator — auroras reached ~23° geomagnetic latitude in 1859 — bringing induced currents to mid-latitude grids that are not built for them. Ground conductivity matters as much as latitude.
• "The Dst value is precisely known." No. The Dst index did not exist in 1859; Carrington-era estimates are reconstructions from a handful of surviving magnetograms (chiefly Colaba/Bombay), and credible values range from −850 to −1,600 nT. The uncertainty is itself a reason the event is hard to plan against.