May 2024 Geomagnetic Storm

• May 2024 geomagnetic storm is an extreme solar–terrestrial disturbance driven by successive Earth-directed CMEs and X-class flares from AR 13664/3664.
• The storm’s structured three-step main phase produced record Dst depressions (up to –518 nT) and significant impacts on the magnetosphere, ionosphere, thermosphere, and satellite operations.
• CME–CME interactions and mesoscale magnetic variability highlighted forecasting challenges, emphasizing the need for integrated Sun-to-Earth models for better space weather prediction.

The May 2024 geomagnetic storm was an extreme solar–terrestrial disturbance centered on 10–11 May 2024 and driven by successive Earth-directed coronal mass ejections (CMEs) and X-class flares from the AR 13664/3664 complex. It is consistently characterized as the strongest geomagnetic storm since 2003, although its exact historical ranking depends on the index and comparison set being used. The event combined shock compression, a sheath-dominated multi-step main phase, later magnetic-cloud forcing, and unusually broad impacts across the magnetosphere, ionosphere, thermosphere, radiation belts, satellite-drag environment, and ground technological systems (Hajra et al., 2024, Liu et al., 2024).

1. Event definition and storm chronology

A fast forward shock reached near-Earth space at about 16:37–16:39 UT on 10 May 2024, and the sudden impulse or sudden storm commencement followed at about 17:05–17:15 UT. One overview reported an SI+ amplitude of +88 nT, while flash observations noted an SSC of about 78 nT at Kakioka and amplitudes up to about 130 nT at low-latitude dusk-sector stations. The storm main phase then developed in three steps over about 9 hours. In the first step, SYM-H reached −183 nT under sheath $B_s \approx 40$ nT; in the second, SYM-H reached −354 nT as sheath $B_s$ deepened to about 43 nT and a magnetosonic compression wave raised $|B|$ to about 71 nT; in the third and strongest step, SYM-H reached −518 nT as $B_s$ intensified to about 48 nT. Dst fell to about −412 nT near 02–03 UT on 11 May, and $Kp$ reached 9. The recovery phase persisted for about 2.8 days and included further southward-field magnetic-cloud intervals and additional shocks on 11 and 12 May (Hajra et al., 2024, Hayakawa et al., 2024).

These timings established the event as both a shock-compression storm and a prolonged solar-wind coupling episode. The convective electric field $E_y = V_{sw}|B_s|$ reached about 28.7, 31.4, and 35.0 mV m$^{-1}$ across the three main-phase steps, consistent with strong ring-current injection and with the rapid Dst depression reported across multiple studies. A common misconception is that the storm was a single-step disturbance; the published chronology instead shows a distinctly structured main phase, with different interplanetary substructures contributing to successive intensifications (Hajra et al., 2024).

2. Solar source region and eruptive drivers

The source region was AR 13664, corresponding to NOAA AR 3664, later interacting with AR 13668. It evolved into a compact, flare-productive, magnetically complex system with a long, strongly sheared polarity inversion line. One study reported total unsigned flux rising from just below $4 \times 10^{22}$ Mx to about $13.53 \times 10^{22}$ Mx, with a flux-emergence rate changing from $0.348 \pm 0.001 \times 10^{22}$ Mx day$B_s$0 to $B_s$1 Mx day$B_s$2. Another measured a six-hour averaged emergence rate peaking at about $B_s$3 Mx hr$B_s$4 on 8 May, with sustained values of order $B_s$5 Mx hr$B_s$6 across 7–9 May. The region’s magnetic free energy surpassed $B_s$7 erg on 7 May, and different flare censuses reported 12 X-class flares during 8–15 May and 23 X-class flares across the first nearside, farside, and second nearside transits (Romano et al., 2024, Wang et al., 2024, Jarolim et al., 2024, Hayakawa et al., 2024).

The region was also historically extreme in its photospheric metrics. The deprojected active-pixel area $B_s$8 reached 4394.67 $B_s$9Hem, placing it at the 99.95 percentile over 1874 May–2024 June, and the total unsigned line-of-sight flux $|B|$0 Mx ranked at the 99.10 percentile over 1996 April–2024 June. Five SHARP parameters—$|B|$1, $|B|$2, $|B|$3, $|B|$4, and $|B|$5—ranked at the 100.00 percentile over 2010 May–2024 June. NF2 nonlinear-force-free extrapolations showed steep rises in total magnetic energy and free magnetic energy after 7 May, with flare-related energy depletions and topological reconfigurations closely corresponding to AIA brightenings (Jaswal et al., 2024, Jarolim et al., 2024).

Two complementary physical descriptions recur in the literature. One emphasizes an emergence–convergence–shear sequence: repeated bipole emergence at nearly the same latitude and longitude, converging motions near the PIL up to about 0.5 km s$|B|$6, and shear up to about 1.0 km s$|B|$7 over about 200 arcsec for more than 2 days. The other emphasizes collisional shearing at two collisional PIL systems, with seven halo CMEs distributed primarily along those two PILs. This suggests that both large-scale compaction and localized collisional shearing were central to the buildup of non-potential magnetic energy and to the production of homologous eruptions (Romano et al., 2024, Wang et al., 2024).

3. Heliospheric propagation and compound ejecta

A defining feature of the event was CME–CME interaction. Different reconstructions identify four, five, six, or ten geoeffective or interacting CMEs, depending on source association and on whether later flank or glancing structures are counted. This suggests that the May 2024 storm is better described as a hierarchy of interacting ejecta than as a single magnetic cloud. Across these studies, the common result is pileup, overtaking, compression, and the formation of complex ejecta at 1 AU (Rodkin et al., 13 Jan 2025, Khuntia et al., 4 Apr 2025, Soni et al., 27 Feb 2026).

Quantitative reconstructions place several interactions well inside 1 AU. Using GCS and FRIS, one study estimated CME1–CME2 interaction at about 144 $|B|$8, CME3–CME4 at about 54 $|B|$9, and CME5–CME6 at about 110 $B_s$0. A LASCO–IPS–DBM study, focusing on the first four halo CMEs, found predicted L1 arrivals separated by roughly 8 h, 4 h, and 1 h, favoring interaction and compression. Near Earth, in situ analyses alternatively described six magnetic ejecta and three interaction regions within a large compound structure, or two main complex ejecta separated by a later shock on 12 May. These descriptions differ in segmentation but agree that the solar wind driver was a compressed, only partially merged, multi-ejecta system (Khuntia et al., 4 Apr 2025, Rodkin et al., 13 Jan 2025, Liu et al., 2024).

The “perfect storm” interpretation rests on the fact that none of the near-Sun CME speeds was exceptionally extreme on its own, yet the interactions preserved and amplified magnetic field strength. At Earth, the first complex ejecta reached $B_s$1 nT and $B_s$2 nT in one analysis, while another measured a near-record $B_s$3 nT and $B_s$4 nT during the compound interval. This is a central reason the event is treated as a benchmark case for interaction-aware heliospheric modeling rather than for isolated-CME forecasting (Liu et al., 2024, Abunina et al., 14 Jan 2025).

4. Magnetospheric, ionospheric, and ground-current response

The shock and sheath produced exceptional dayside compression. THEMIS crossings placed the magnetopause at about 8.24–8.51 $B_s$5 near 17:05 UT on 10 May and later at about 5.04 $B_s$6 and 5.37 $B_s$7 near 19:12–19:19 UT. A separate soft-X-ray simulation focused on the dense CME current sheet beginning near 22:30 UT reproduced a minimum subsolar standoff of about 4 $B_s$8 at about 22:34 UT, with the two cusps appearing as nearly parallel emission ridges before moving poleward for about 10 minutes after IMF reversal. Auroral and FAC responses were comparably extreme: IMAGE $B_s$9 reached −2632 nT at 22:35 UT during a supersubstorm, AMPERE indicated about 10-fold enhancement of polar E-region Birkeland currents, and Region-2 currents expanded equatorward to about 50° geomagnetic latitude. Using 80 vetted reports, a visual auroral oval boundary was conservatively reconstructed to 29.8° invariant latitude in the Southern Hemisphere (Ng et al., 3 Dec 2025, Hayakawa et al., 2024, Hajra et al., 2024).

The event also produced one of the largest Forbush decreases in the modern neutron-monitor era and a confirmed ground-level enhancement. One overview reported cosmic-ray decreases of about 17% at Dome C and about 11% at Oulu relative to quiet values. A dedicated neutron-monitor study found a 10 GV Forbush amplitude of 15.7%, a minimum hourly decrement of −4.4% h$Kp$0, a very small first-harmonic anisotropy with $Kp$1 and $Kp$2, and a storm-time magnetospheric correction of up to about 4% in the raw counts. GLE 74 was confirmed during roughly 02–10 UT on 11 May (Abunina et al., 14 Jan 2025, Hayakawa et al., 2024).

Ionospheric effects varied strongly by latitude. Over the Galápagos, vertical TEC decreased by approximately 30 TECU during the main phase and then gradually increased on 12–13 May, a negative storm response attributed to rapid recombination, disturbance dynamo, prompt penetration electric fields, and EIA disruption. Over Türkiye, regional mean vTEC dropped from daytime values of about 50 TECU to about 12–15 TECU on 11 May, corresponding to a 65–75% reduction and lagging $Kp$3 by about 2 h. Global TEC analysis also reported storm-enhanced density and a tongue of ionization during the main phase, followed by a broad negative storm in recovery. Ground technological coupling was strong but spatially heterogeneous: in the United States, a 47-site GIC compilation showed that site-specific modeling at four TVA substations achieved correlation coefficient greater than 0.8 and prediction efficiency between 0.4 and 0.7, while empirical regressions found that both the standard deviation and the maximum magnitude of GIC increased with geomagnetic latitude and MT-derived $Kp$4 scaling (López et al., 6 Feb 2025, Eraydın et al., 1 Apr 2026, Wilkerson et al., 9 Jul 2025).

5. Thermosphere, radiation belts, and near-Earth operational effects

The neutral upper atmosphere experienced both storm-time heating and an unusual post-storm undershoot. Swarm-A/B/C densities normalized to 490 km showed the expected enhancement during the main phase, followed by a rare thermospheric overcooling signature in the northern polar region on 12 May: Swarm-C reached −23% relative to 9 May, Swarm-A reached −22%, and Swarm-B reached up to −11.37% in the morning sector and −11.15% on the nightside. TIMED/SABER NO 5.3 $Kp$5m emission increased by a factor of 8–10 on 11 May relative to 8–9 May, peaked at 11.84 erg cm$Kp$6 s$Kp$7, and produced daily global radiated power of $Kp$8 W. In the notation used in that study, the cooling proxy is $Kp$9, while the direct drag consequence for spacecraft follows $E_y = V_{sw}|B_s|$0 (Ranjan et al., 2024).

These atmospheric perturbations translated directly into orbital effects. NRLMSISE-00 at 400 km indicated density increases up to about 6 times within about 12 hours of storm onset, and a TLE case study of KANOPUS-V 3 showed altitude-decay rates rising from about 38 m day$E_y = V_{sw}|B_s|$1 pre-storm to about 180 m day$E_y = V_{sw}|B_s|$2 during the storm, implying a local density ratio of about 4.7 under the short-window assumption that $E_y = V_{sw}|B_s|$3. Thousands of satellites then executed orbit-raising maneuvers nearly simultaneously, dominated by Starlink, while non-maneuverable debris and rocket bodies underwent substantial net altitude losses in the 400–700 km band (Parker et al., 2024).

Radiation-belt restructuring was likewise persistent. CALET on the ISS observed a long-lasting “electron storage ring” in the slot region immediately after the main phase. Count-rate enhancements extended from $E_y = V_{sw}|B_s|$4 to 3.2, breached the nominal slot-region barrier near $E_y = V_{sw}|B_s|$5, and were detected in the $E_y = V_{sw}|B_s|$6, $E_y = V_{sw}|B_s|$7, and $E_y = V_{sw}|B_s|$8 MeV channels. Global decay fits over $E_y = V_{sw}|B_s|$9 gave lifetimes of $^{-1}$0 days, $^{-1}$1 days, and $^{-1}$2 days for those three thresholds, respectively, and the $^{-1}$3 MeV component persisted for more than five months. CALET also reported storm-driven changes in the South Atlantic Anomaly, including sharpening of its eastern edge and a months-long recovery of a slot-region proton deficit (Ficklin et al., 3 Feb 2026).

6. Historical significance, forecasting, and interpretation

By $^{-1}$4 nT, one flash report identified the storm as the sixth-largest event since 1957; by $^{-1}$5-$^{-1}$6 nT, another overview described it as the second largest storm of the space age. This apparent inconsistency reflects different indices, cadences, and historical baselines rather than disagreement about the event’s severity. A recurring theme in the literature is that May 2024 was not merely a large storm, but a “perfect storm” sequence in which successive eruptions preconditioned the heliosphere, piled up at 1 AU, and preserved strong southward fields over many hours (Hayakawa et al., 2024, Hajra et al., 2024, Liu et al., 2024).

Forecasting performance exposed several bottlenecks. NOAA SWPC ap forecasts underpredicted the initial surge by roughly 100–300 ap units even at one-day lead. A best-performing EUHFORIA run predicted arrival with about 2 h lead and storm strength with about 70% accuracy, while the authors identified CME input specification—especially counts, geometry, tilt, and interaction history—as the major bottleneck. At the same time, STEREO-A at 0.956 AU and 12.6° west acted as a fortuitous sub-L1 monitor: it observed the shock 2.57 h before L1, consistent with $^{-1}$7, and mapped beacon data produced a modeled minimum $^{-1}$8-$^{-1}$9 of −478.5 nT, underestimating the observed minimum by only 8% (Parker et al., 2024, Soni et al., 27 Feb 2026, Weiler et al., 2024).

A further interpretive result is that mesoscale structure mattered even over small separations. For the first complex ejecta, Wind-based modeling yielded a minimum $4 \times 10^{22}$0 of −378 nT for the near-Earth flank, whereas the STEREO-A flank would likely have produced about −494 nT. This indicates that a longitudinal offset of only 12.6° sampled materially different magnetic content and geo-effectiveness. A plausible implication is that future extreme-storm forecasting will require not only better arrival-time prediction, but also better characterization of mesoscale magnetic structure, CME–CME reconnection, and downstream consequences such as NO radiative cooling and post-storm density undershoot. In that sense, the May 2024 storm now functions as a benchmark event for integrated Sun-to-Earth modeling and for multi-point operational monitoring (Liu et al., 2024, Ranjan et al., 2024, Soni et al., 27 Feb 2026).