A Carrington-like geomagnetic storm observed in the 21st century

In September 1859 the Colaba observatory measured the most extreme geomagnetic disturbance ever recorded at low latitudes related to solar activity: the Carrington storm. This paper describes a geomagnetic disturbance case with a profile extraordinarily similar to the disturbance of the Carrington event at Colaba: the event on 29 October 2003 at Tihany magnetic observatory in Hungary. The analysis of the H-field at different locations during the"Carrington-like"event leads to a re-interpretation of the 1859 event. The major conclusions of the paper are the following: (a) the global Dst or SYM-H, as indices based on averaging, missed the largest geomagnetic disturbance in the 29 October 2003 event and might have missed the 1859 disturbance, since the large spike in the horizontal component (H) of terrestrial magnetic field depends strongly on magnetic local time (MLT); (b) the main cause of the large drop in H recorded at Colaba during the Carrington storm was not the ring current but field-aligned currents (FACs), and (c) the very local signatures of the H-spike imply that a Carrington-like event can occur more often than expected.


Introduction
Major disturbances of the magnetosphere, or geomagnetic storms, are a consequence of solar activity, and represent a serious hazard to our technology dependent society. On 2 September 1859, the Colaba observatory measured the most extreme geomagnetic disturbance ever recorded at low and mid latitudes. Although at that time the magnetic observatories used to produce Dst were not in operation, the Dst minimum value for this event was first estimated by Siscoe [1979] about −2000 nT. More recently,  and Cid et al. [2013] provided new estimations of −1760 nT and −685 nT, respectively. In any case, the so-called Carrington storm is ranked, according to the Dst index, as the most extreme geomagnetic storm ever recorded.
In October 2003, during the first Halloween storm, Tihany magnetic observatory (THY: MLat 45.87º, MLong 100.06º) in Hungary recorded a geomagnetic disturbance with an extraordinarily similar profile to that recorded at Colaba in 1859, but this time with a large dataset available from modern solar, interplanetary and terrestrial surface observatories. Although there is a large difference between the magnetic latitude of Figure 2 shows the terrestrial disturbance during the first hours of the storm event on 29 October 2003 (see Table 1 for geographical and magnetic coordinates of the observatories). Measurements of the horizontal magnetic field component recorded at mid-latitude (40º MLat) observatories spread in longitude appear in Figure 2a. Panels from top to bottom are: SUA, IRT, FRN, BSL, and SPT (increasing in MLong). Figure   2b shows, for the same interval as Figure 2a Regarding the SYM-H index, the first magnetic signature of the storm is a sudden commencement of approximately 80 nT (solid vertical line at 06:12 UT), which can be also observed in the Dst index. After the SC, SYM-H index decreases to a minimum value of -432 nT at 23:03 UT on 30 October (not shown in the figure) in a three-step profile. The SC signature is commonly related to an enhancement of the magnetopause current, which is expected to be observed across the globe. Therefore, we proceed to carefully analyse magnetic records at different locations.
First we compare the measurements of the horizontal component from a chain of midlatitude magnetic observatories spread in longitude in order to provide a sample all around the globe. The magnetic trace at IRT, located in the mid-latitude afternoon sector, shows an increase of just less than 70 nT at SC onset. However, larger enhancements of 120 nT appear at FRN and BSL at the SC. FRN and BSL, in the preand mid-night sector located, show a positive change which reaches 300 nT and 350 nT respectively, about one hour after SC. SPT and SUA, located in the dawn and prenoon sector, show different behaviour: instead of a positive disturbance, a rapid decrease appears at the time of the SC leading to negative depression development similar to the Carrington event and to that shown in Figure 1 for THY, herein after Hspike.
The intensity of the disturbance decreases rapidly when moving in longitude from dawn (where THY is located at the peak of the H-spike) (25 %, 5 degrees away from THY at SUA, and 40 %, 20º further away at SPT). These observations reveal that the extreme drop in H recorded at some observatories on 29 October 2003 was actually a very local drop in the dawn sector.
At low latitudes, we have checked the magnetic traces for this event at the four observatories involved in computing the Dst index (HER, KAK, HON and SJG). The SC appears at HON, KAK and SJG but is not noticeable at HER (in the dawn sector).
Then, the disturbance during the first two hours after the SC depends strongly on the sector at low latitudes as previously shown for mid-latitudes. The horizontal component shows a negative depression at HER reaching a minimum value of -330 nT, while the disturbance goes to positive values at SJG (in the after mid-night sector), reaching a maximum value of 210 nT at the time of the maximum disturbance at HER. KAK and HON show smaller disturbances, but also opposite (positive in HON and negative in KAK). The result of an average of all these measurements in the computation of the Dst index is the almost complete disappearance of the disturbance during these two hours after the SC. The higher temporal resolution of SYM-H index does not prevent from missing the disturbance of these first two hours after the SC, as this index is also computed by averaging disturbances at different longitudes.
From the third hour to the minimum value of the disturbance at 00:33 UT on 30 October, a slower decrease of the horizontal component is a common signature at all magnetic observatories analysed at low-and mid-latitude. Therefore, Dst and SYM-H indices are good indicators of the disturbance for this time.
Measurements by the DMSP satellite (F13) at 17:45 local time, reported by Mannucci et al. [2005], showed that the drift velocity at DMSP altitude (840 km October. These data reveal enhancement and prolonged auroral activity which was already observed using OI 6300 Å emission from Boston (48.3º MLat), which is approximately at 30º-40º equatorward in latitude of the normal auroral precipitation regions [Pallamraju and Chakrabarti, 2005]. NOAA Technical Memorandum OAR SEC-88 on Halloween storms reported that aurora sightings occurred from California to Houston to Florida. Tremendous aurora viewing was also reported from mid-Europe and even as far south as the Mediterranean countries (40º MLat).

Discussion
We report in this paper a geomagnetic disturbance case recorded on 29 October 2003 at THY magnetic observatory (hereafter, C03), whose temporal profile during two days is extraordinarily similar to the one recorded at Colaba in 1859 (hereafter, C59). Of course there are two remarkable differences among the Carrington event (C59) and the "Carrington-like" event (C03): (1) the intensity of the disturbance in C59 is double that in C03, and (2) the latitude of the magnetic observatory where the measurements were collected, which was very low for C59 (10º MLat) and middle for C03 (46º MLat).
However, those latitudes have a common feature: both were about 10º equatorward from the lower latitude reported of aurora sightings for the corresponding event, which in both cases was far from the typical edge of the auroral oval. imply that a Carrington-like event can occur more often than expected, as suggested by Kataoka [2013].
As previously shown for low-latitude magnetic observatories, longitudinal analysis of magnetic field measurements at mid-latitude observatories during C03 (Figure 2a) shows negative or positive depressions depending on the local time. These depressions happened simultaneously with auroral effects. The longitudinal variation at midlatitudes of the geomagnetic disturbance during C03 H-spike can be summarized as a day-night asymmetry peaking at dawn-noon sector and almost unnoticed in the noondusk sector, although the peak disturbance due to an asymmetric ring current effect is expected in the dusk sector [Li, 2009]. This result does not agree with Tsurutani et al. [2003] or Li et al. [2006] which pointed out that the H-spike at Colaba was related to the plasma injection into the nightside magnetosphere, which enhances the ring current.
However, the spike happened from  9 to 10 MLT, where other currents have a major contribution (Shi et al., 2008). Akasofu and Kamide [2005] pointed out that some intense storms tend to have a sharp forenoon decrease of the H component, which accompanies a greater disturbance at auroral latitudes. These signatures were previously related to FACs by Tsuji et al. [2012] for the geomagnetic storm on 7-8 September 2002 and by Yu et al. [2010] using MHD simulations. Love and Gannon [2010] also reported a large dawn-dusk asymmetry in low-latitude disturbances during the November 2003 storm and during the October 2003 storm. They proposed that these observations were the result of the superposition of magnetopause currents and partial ring current and, even FACs. Similar evidence of magnetic changes, confined in a narrow longitude range in the forenoon sector, and as a result with a little contribution to the Dst index, were provided by Akasofu and Kamide [2005] for some intense storms. Siscoe et al. [2006] and Green and Boarden [2006] already suggested that the H-spike in C59 might have a significant auroral-ionospheric component.
Boteler [2006] pointed out that the combination of the negative depression in H at Rome and the large positive Z excursion at Greenwich were the signatures of a large westward electrojet at 40º to 45º MLat, which will produce a considerable enlargement of the auroral oval during C59. However, he concluded that there was no evidence that it went far enough equatorward to contribute to the magnetic variation recorded at Colaba. More recently Cliver and Dietrich [2013], comparing C59 and the geomagnetic storm in May 1921, indicated that an auroral negative depression contributed to the negative H-spike in the Colaba trace during C59.
During C03, the availability of a large number of magnetic records spread through the globe has allowed us to go further in these suggestions regarding the auroral-ionosphere component in magnetograms and to propose a major role for FACs in C03 and by extrapolation during C59. Extending what Siscoe et al. [2006] pointed out for the effect of ionospheric currents involved in C59 to FACs, we conclude that the effect of FACs affecting a magnetogram at a latitude as low as 10º is also an exceptional aspect of the storm.
The H-spike at Colaba in C59 (considered up to now as a Dst-spike) was addressed by simulation assuming solar wind conditions with very intense ( 60 nT) and short time duration southern Bz and an additional extreme enhancement of the solar wind density (1800 cm -3 ) after the negative Bz [Li et al., 2006]. Similar profile shapes to the Li et al.
[2006] solar wind conditions have recently been seen in space plasma data [Kozyra et al., 2013] and intense short-duration southern fields (50 nT) were also recorded during 29 October 2003. However, extreme density values as those proposed by Li et al. [2006] have never been recorded before (as an example of extreme density event, we can cite the March 2001 ICME [Farrugia et al., 2006] which had densities just after the shock in excess of 100 cm -3 ). This extreme density value is a key-point in the simulation process as that is the only way to match the fast recovery of Colaba record assuming that the drop in H was due to the ring current. Li et al. [2006] modelled Dst as a sum of several terms, each one varying in time depending on a "source" term (which represents the external driving) and a "loss" term (which represents the decay rate of the field source).
In the simulation, the term representing the contribution of the main ring current dominates the decrease of Dst and a pressure dependent term dominates the fast recovery of Dst. This simulation can be considered as a "dynamically driven" approach [see Tsyganenko, 2013 and references therein]. In every dynamically driven approach the loss term depends on which current system is being considered. The TS05 magnetosphere model [Tsyganenko and Sitnov, 2005] provides interesting information regarding the dynamics and peak values of the total current corresponding to individual sources. Their relaxation/response timescales were found to differ significantly from each other, from as large as 30 hours for the symmetric ring current to only 50 min for the FACs. In C03 the time between the minimum value of H and the end of the spike was 45 min and in C59 was 50 min. Moreover, FACs peak at prenoon sector (9 MLT), which was the location of Colaba for C59 and THY for C03. All these results point to FACs as the main current involved in the large H-spikes of the "Carringtonlike" events.

Conclusions
The analysis of local magnetic records and magnetic indices during the event on 29 October 2003, when they are extrapolated to the Carrington event, led us to interesting implications. These are the following: 1) The Dst or SYM-H indices (commonly used to assess the severity of a storm) if calculated with multiple observatories as done today might have missed the large Hspike recorded at Colaba during C59.
2) The very local signatures of the H-spike imply that a Carrington-like event can occur more often than expected.
3) FACs played a major role in the H-spike in C03 and C59.
Acknowledgements. Geomagnetic field data have been obtained from INTERMAGNET magnetic observatories. The authors thank the Institutes that operate the observatories which provided data for this study; they also acknowledge the WDC for Geomagnetism for the magnetic field data and the SYM-H index. This research was supported by the contract "Estudio de la influencia de fenómenos relacionados con la Meteorología Espacial en las infraestructuras de REE" between Red Eléctrica de España and the Universidad de Alcalá and by grants PPII10-0183-7802 from the Junta de Comunidades de Castilla-La Mancha of Spain and AYA2013-47735-P from MINECO. The editor thanks two anonymous referees for their assistance in evaluating this paper.