An overview of the impulsive geomagnetic field disturbances and power grid impacts associated with the violent Sun-Earth connection events of 29–31 October 2003 and a comparative evaluation with other contemporary storms

Abstract

[1] An overview is provided of the geomagnetic storms and impacts on electric power grids associated with the violent Sun-Earth events of October 2003. During the period from 29 to 31 October 2003, two large geomagnetic storms were observed, as measured by periods of high Kp, Ap, and Dst indices. In fact, these storms had Ap rankings of 6th and 16th all time. This ranking would suggest that the October 2003 storms would be significant with regard to geomagnetically induced currents (GIC) in power grids. However, the resulting geomagnetic storms were much lower in delta B and dB/dt intensity than other historically large geomagnetic storms. A variety of geomagnetic storm processes drove observed GIC. For example, ground observations indicated the presence of large dB/dt pulsations and GIC at North American midlatitude locations on 29 October 2003 that may be due to unusually intense Kelvin-Helmholtz shearing. Sustained disturbance conditions at low-latitude and equatorial latitude locations that are likely linked to ring current intensifications may be the source of sustained GIC at these locations and the cause of large power transformer failures. Comparative evaluations will be provided for the 29–31 October 2003 storms and other important and contemporary storms, such as those observed on 13–14 March 1989, 13–14 July 1982, and 15–16 July 2001. Rather than an index-based evaluation method, the comparative evaluations presented in this paper are based on comparisons of storm morphology. This approach provides a more meaningful comparison of geomagnetic field disturbance dynamics that are important to characterize large GIC threats to power grid infrastructures.

1. Introduction

[2] A number of recent publications have examined the nature of the solar events and resulting interplanetary coronal mass ejection (ICME) passages [Skoug et al., 2004; National Oceanic and Atmospheric Administration (NOAA), 2004; National Weather Service, 2004]. These provide some comparative perspective on the parameters of the solar activity, solar wind and ICME passages or encounters with the Earth's magnetosphere because of the intense solar activity during this period of time. These events were responsible for the relative size and importance of space environment disturbances observed at Earth. However, in assessing the potential for geomagnetically induced currents (GIC) and impacts they caused to power grid infrastructures, it is important to characterize storm morphologies that are in proximity to these exposed ground-based infrastructures. Planetary indices typically do not have the appropriate regional context to adequately describe the spatially complex and temporally dynamic variations in the geomagnetic field that can take place during a geomagnetic storm. Large impulsive geomagnetic field disturbances have been well understood as a concern for power grids in close proximity to these disturbance regions. Large GICs are most closely associated with geomagnetic field disturbances that have high rate of change; hence a high-cadence and region-specific analysis of dB/dt provides a generally scalable means of quantifying the relative level of GIC threat. These threats have traditionally been understood as associated with auroral electrojet intensifications at mid and high-latitude locations. However, both research and observational evidence has determined that the geomagnetic storm and associated GIC risks are broader and more complex than this traditional view [Kappenman, 2004]. Large GIC and associated power system impacts have also been observed for differing geomagnetic disturbance source regions and propagation processes. This includes the traditionally perceived impulsive disturbances originating from ionospheric electrojet intensifications and ground level propagation modes. However, large GICs have also been associated with impulsive geomagnetic field disturbances such as those during an SSC that will occur under sudden increases in solar wind dynamic pressure. Large GICs can be observed even at low-latitude and middle-latitude locations for brief periods of time during these events [Kappenman, 2003]. Recent observations also confirm that geomagnetic field disturbances usually associated with ring current intensification can be a source of large-magnitude and long-duration GIC in power grids at low and equatorial regions [Erinmez et al., 2002]. Observations, in particular from the 29–31 October 2003 storms, also suggest that Kelvin-Helmholtz shearing may be responsible for pulsations that can cause large GICs. The wide geographic extent of these disturbances implies GIC risks to power grids that have never considered the risk of GIC previously, largely because they were not at high-latitude locations in proximity to electrojet intensifications.

[3] Most of the world's power grid infrastructures generally are also located in regions of the world that contain the predominant population centers. In excess of 90% of the large high-voltage power grids of the world are located in geomagnetic latitudes between ±60°. To describe the impacts to these infrastructures, it is necessary to discretely evaluate the geomagnetic disturbance processes that can occur in these regions. Over this wide range of latitudes the dominant process in a given region may be different from that in another region or in a different event. This diversity of geomagnetic field disturbance processes also have pronounced geographic regions where they can cause the greatest potential for GIC and power system impacts. For example, large electrojet intensifications are generally of concern at high latitudes and can spread to midlatitude locations for storms with sufficient equatorward expansion of the auroral oval. On the other hand, ring current intensifications will generally have predominant influence at low-latitude and equatorial latitude locations. In general, planetary indices such as Kp, Ap, G, or Dst do not provide the needed resolution to adequately and consistently describe the environment conditions for GIC threats to power grids, and these indices also fail to provide a consistent means of historical comparison as well. A description based upon the morphologies of these disturbance processes will be utilized to describe the impacts observed from the 28–31 October 2003 storms. In addition, this morphology-based method will be utilized for more comprehensive comparisons of the 28–31 October 2003 storms with other contemporary and historically important storms.

2. Limitations of Geomagnetic Storm Indices for Characterizing GIC Impacts and in the Ranking of Historically Large Storms

[4] The October 2003 storms using index-based rankings were the sixth and sixteenth largest storms using the 24 hour running Ap over the ∼70 history of the K, Kp and Ap indices. The broad planetary indices are generally derived from observed local indices in designated auroral and subauroral locations around the Northern Hemisphere. The K indices, ranging from 1 to 9, are based upon a level of observed delta B variation over a three hour time window. This variation in threshold from K1 to K9 is also latitude specific and varies for each observatory. For example at the Boulder (BOU) observatory at a midlatitude location in North America, an observation of delta B of 500 nT over a 3 hour time window is the minimum threshold to rank the storm as being K9 intensity. The indices as they approach severe intensity levels do not provide a specific characterization of the possible dB/dt ranges which are important to characterize GIC levels. Figure 1 provides a graphical illustration of the limitations of the Boulder K index in characterizing dB/dt minimum and maximum ranges. As shown, as the K index increases, the possible range of dB/dt becomes less specific. At the highest intensity level of K9, the inherent limitation of the local indices is that they saturate and do so at relatively low delta B threshold levels. For example at Boulder for a K9 intensity, the minimum dB/dt is only ∼3 nT/min (assuming a very slow >500 nT change over a 180 min time window), while the maximum dB/dt can be any dB/dt in excess of 500 nT/min. For those concerned about GIC levels, this index design provides the worst possible set of characteristics in that it has only specific dB/dt ranges for very low storms that pose no impact threat, but the dB/dt ranges become increasing less specific as the K index increases. This has lead to situations where the infrastructure operator who has successfully survived a K9 storm with limited impacts is uncertain whether it was the misleading nature of the K index or operational mitigation measures undertook that lead to the limited impacts. Most power grid infrastructure operators do not attempt to do high-cadence magnetometer measurements of storms and therefore rely upon indices alone for knowledge of storm intensity. Figure 2 further illustrates the ambiguous nature of the K indices as they relate to maximum observed dB/dt during important storms. Figure 2 provides the observed maximum dB/dt at Boulder for all K6 to K9 storms in solar cycles 22 and 23 (through December 2003 of cycle 23). As shown, the highest dB/dt levels are not closely associated with the highest K index levels. As a case in point, the highest dB/dt over this period of time was observed for an interval of K8 intensity on 13 March 1989. In addition, there have also been a number of K7 and K6 events with dB/dt levels that exceeded those of most K9 disturbances at Boulder.

Figure 1

Boulder K index and relevant dB/dt maximum and minimum ranges as K level increases from K1 to K9. Because of the design limitations of the K index, the range of dB/dt becomes less specific as the storm intensity and K level increase. Note that at K9, because of saturation of the K index, the maximum dB/dt can approach infinity.

Figure 2

Observed maximum dB/dt at Boulder for all K6–K9 disturbances in solar cycles 22 and 23. As illustrated, storm events indicate that ambiguous dB/dt classification will result if only using a K index classification. Since GIC is tied to dB/dt, this illustrates the unreliability of the K index as a predictor of maximum GIC.

[5] The K index also fails to provide an unambiguous descriptor of the largest delta B or equatorward expansion of the electrojet currents. This is especially a problem for large storm events when these indices experience saturation at the K9 levels. These saturation problems spill over into planetary indices such as Kp and Ap which are derived from the local K index characterizations. Table 1 is the historic ranking using 24 hour running Ap of the top 30 geomagnetic storms (through December 2003) [NOAA, 2004]. Note that the 20 October 2003 storm, with Ap of 252, is only ∼13% smaller than the 13–14 March 1989 superstorm, which is ranked third all time. Similarly, the Oct 29, 2003 storm would be ∼10% larger than the storm on 13 July 1982, which is ranked tenth and ∼31% larger than the Bastille Day storm on 15 July 2000, which is ranked 30th. This ranking would imply that the 29 October 2003 storm would have extremes in observed delta B and dB/dt at important latitudes of concern for infrastructure operators. However as will be shown in latter sections of this paper, there are greater than factor of two differences between the extremes in observed dB/dt and delta B storm morphologies when comparing the 29 October 2003 with even smaller storms such as the July 1982 and July 2000 events. This apparent morphology conflict with indices comes about because of the presence of saturation for these large storms. For example, the October 2003 storms will be shown to have only slightly exceeded the K9 disturbance threshold, therefore for these storms only minor levels of saturation would have occurred. For the other storms that will have the morphologies comparatively evaluated, the K9 thresholds are extensively saturated. This excessive level of saturation is simply not accounted for in the planetary rankings used in the Kp, G, and Ap index calculations. Because this saturation masks the geomagnetic disturbance intensity levels, old Ap ratings cannot be reverse engineered to uncover the true nature of the environment extremes. As will be demonstrated in the morphology comparisons, the historic ranking using Ap as detailed in Table 1 may be essentially meaningless because of widely varying degrees of saturation that remain unknown for many of the storms listed in Table 1, because of lack of details on relevant geomagnetic disturbance environments. For very small intensity storms (for example K4 or smaller), the problem of saturation is not an issue and the window of possible dB/dt minimum and maximums are greatly reduced, as noted in Figure 1. Therefore the possible error is smaller in assessing GIC impacts on infrastructures for these classes of essentially nonstorms. However, when evaluating the design and performance of power networks, the design challenge is always one of countering the severe threats. Therefore a clear definition of the maximum threat environments cannot be readily derived from any historical K intensity or other derivative historical indices.

Table 1. Top 30 Geomagnetic Storms (Potsdam Running Ap) Since 1932^a

Rank	Intensity	Date
1	312	18 Sept 1941
2	293	12 Nov 1960
3	285	13 March 1989
4	277	23 March 1940
5	258	4 Oct 1960
6	252	29 Oct 2003
7	252	15 July 1959
8	251	31 March 1960
9	241	25 May 1967
10	229	13 July 1982
11	228	8 Feb 1986
12	226	29 March 1940
13	223	4 Aug 1972
14	222	5 July 1941
15	221	2 Sept 1957
16	220	30 Oct 2003
17	216	8 July 1958
18	215	28 March 1946
19	214	22 Sept 1946
20	212	1 March 1941
21	212	26 July 1946
22	203	19 Aug 1950
23	201	4 Sept 1982
24	199	7 Feb 1946
25	199	11 Feb 1958
26	196	12 May 1949
27	196	4 June 1991
28	195	24 March 1946
29	193	9 May 1992
30	192	15 July 2000

• a The 29 and 30 October 2003 storms and their respective historic Ap rankings are shown in bold.

[6] A number of recent publications have offered a defense of the value of indices and for detailed analysis of power system impacts using these indices [Oler, 2004a, 2004b; Gholipour et al., 2004; Forbes and St. Cyr, 2004; Thomsen, 2004]. While indices have a traditional role, it is unclear whether the authors understand the limitations as described above for analysis of impacts for storms that have exceeded the K9 threshold and therefore can exhibit widely varying degrees of saturation. Other indices such as Dst, while not specifically subject to saturation, will continue to exhibit design limitations due to the lack of temporal and spatial definition that is necessary to perform comprehensive assessments on a specific power grid, especially if that power grid is not located in an equatorial location.

3. Geomagnetic Field Disturbance Associated With Electrojet Intensification at Midlatitude Locations

[7] The impulsive geomagnetic field disturbances from this geomagnetic storm process are typically the most important source of GIC flows in exposed power grids at auroral to midlatitude locations around the world.

3.1. Electrojet Intensifications: ICME Passage on 29–30 October 2003

[8] In North America and western Europe, the most intense levels of geomagnetic field disturbance associated with electrojet intensification were observed around 0700 UT on 29 October. Figure 3 provides a contour map of delta B_h observed over these regions at time 0652 UT on 29 October. This time also appears to be one of the lowest-latitude observed equatorward expansions of the electrojet for the storm (Appendix A provides further details on the assimilative model used to develop the contour maps). A large region of decreased delta B_h was caused by the intensification of a westward electrojet intensification that extended from high-latitude portions of North America to midlatitude portions of western Europe. At very high latitudes of North America which are not in proximity to power grid infrastructures, the delta B_h reached an intensity of ∼4500 nT as observed at time 0645 UT at the IQA observatory. The disturbance intensity is somewhat reduced at time 0652 UT which coincides with the greatest equatorward expansion of the geomagnetic field disturbance conditions. At observatory latitudes that are at the northern perimeters of power grid networks, the intensity is significantly reduced from those observed at IQA. For instance at the PBQ station over eastern Canada and almost directly south of IQA, the delta B_h dropped to a peak of −2627 nT, while over western Europe, locations such as ESK (Scotland) and BFE (Denmark) reached peak delta B_h lows of −1957 nT and −1114 nT respectively. This electrojet intensification corresponds with the brief duration of IMF southward B_z during the sheath portion of the ICME passage. Another period of electrojet intensification started around 1900 UT and lasted into the early hours of 30 October, also coincident with a more sustained interval of southward IMF B_z during this same ICME passage.

Figure 3

Assimilative model contour map of observed delta B_h over the Northern Hemisphere at 0652 UT, 19 October 2003 (using a 500 nT contour interval).

3.2. Electrojet Intensifications: ICME Passage on 30–31 October 2003

[9] The second ICME passage initiated around 1600 UT on 30 October and lasted into the late hours of 31 October. This ICME had lower speed and smaller IMF B_t and B_z attributes compared to the initial ICME passage. Figures 4, 5, and 6, provides a series of contour maps of observed delta B_h disturbance conditions at respective times of 1956, 2004, and 2119 UT on 30 October. These illustrate the relatively large and dynamic nature of the region of intense westward electrojet intensification observed over portions of northern Europe. From a GIC perspective, the rate of change or dB/dt is the most important attribute of the impulsive geomagnetic field disturbance environment. During this period of time, the regional intensities in western Europe ranged from 600 to 700 nT/min in the Baltic region and northern Germany, with these peaks generally occurring between times 2000 to 2200 UT. It was this series of impulsive events that appears to have triggered the reported blackout in southern Sweden (city of Malmo) and associated transformer heating problems at a nearby nuclear plant.

Figure 4

Assimilative model contour map of observed delta B_h over the Northern Hemisphere at 1956 UT, 30 October 2003 (using a 500 nT contour interval).

Figure 5

Assimilative model contour map of observed delta B_h over the Northern Hemisphere at 2004 UT, 30 October 2003 (using a 500 nT contour interval).

Figure 6

Assimilative model contour map of observed delta B_h over the Northern Hemisphere 2119 UT, 30 October 2003 (using a 500 nT contour interval).

4. Long-Duration Pulsations and Suggested Association With Kelvin-Helmholtz Shearing

[10] On 29 October, there are clearly defined periods of delta B_h excursions associated with southward IMF B_z and associated electrojet intensifications. There are also long-duration periods of pulsations in the geomagnetic field that were the source of relatively large dB/dt and GIC observations that are not associated with southward IMF B_z, associated electrojet intensification morphology or with periods of local K9 or Kp9 indices. A morphology comparison based on an entirely different geomagnetic field disturbance process is necessary to explain the long-duration pulsations. A comparative examination of IMF B_z and observed B_h at the midlatitude station of Fredericksburg (FRD) in midlatitude North America and GIC in power grids in proximity to FRD best illustrates these occurrences. Figure 7 provides a time-history comparison of the observed IMF B_z, FRD delta B_h, FRD dB_h/dt and several plots of GIC observed in regional power grids in proximity to the FRD (Fredericksburg) observatory. As noted in both the observations of dB_h/dt and GIC, the largest magnitude events are not clearly associated with IMF southward B_z excursions or associated with the time coincident FRD delta B_h excursions due to electrojet intensifications.

Figure 7

Time-history comparison of observed geomagnetic storm disturbance conditions for (top to bottom) 29–31 October 2003, solar wind IMF B_z as observed by ACE satellite, FRD delta B_h, FRD dB_h/dt, and GIC as observed in power grid near FRD.

[11] Because of the extremely high speed of the solar wind (∼2000 km/sec) [Skoug et al., 2004], a considerable amount of buffeting is suspected to have occurred to the Earth's magnetosphere, which presumably would be especially pronounced in the morning dayside sector from the Kelvin-Helmholtz shearing process. While more investigation is warranted, this may be the most likely explanation for the sustained pulsations of the geomagnetic field observed at FRD and other midlatitude North American locations during the general time period from 1200 to 1900 UT on 29 October.

5. Morphology-Based Comparison of October 2003 Storms With Other Historic and Contemporary Storms: Electrojet-Driven Geomagnetic Field Disturbances

[12] A comparative evaluation of the 29–31 October 2003 storms with other historic and contemporary storms provides the most useful insights into the significance of these storms, especially in contrast to a comparative evaluation provided solely by indices such as Kp, G, Ap and Dst. As previously discussed, Table 1 provides a ranking using the Potsdam running Ap of the top 30 geomagnetic storms, over a period of approximately the past ∼70 years.

[13] When reviewing the synoptic maps of delta B_h or the specific observations at midlatitude stations, the differences between these storms are more substantial in the spatial extent and disturbance intensity than the Ap indices would suggest. Figure 8 provides a map of delta B_h contour at time 2200 UT on 13 March 1989. This indicates extensive regions of increased delta B_h over midlatitude portions of North America driven by an extensive region of eastward electrojet intensification over these midlatitude locations. Simultaneously a large and intense region of decreased delta B_h is occurring over northern Europe because of an extensive region of westward electrojet intensification. Figure 9 provides a synoptic map of conditions observed at time 2041 UT on 15 July 2000. One of the principal features of this synoptic map is the region of intense delta B_h increase over midlatitude portions of North America. These morphologies can be contrasted with those shown in 4, 5–6 for the 30 October 2003 storm conditions. For example when focused upon the northern European regions of decreased delta B_h, the 13 March 1989 storm at time 2200 UT has a ∼5 to 10° latitude greater equatorward expansion of the intense delta B_h contours. Meanwhile, over North America, the October 2003 storm has much more benign disturbance conditions than those observed because of a region of intense eastward electrojet driven delta B_h increases in March 1989. The 15 July 2000 storm indicated a morphology that resulted in a very intense region of increased delta B_h driven by an eastward electrojet over midlatitude regions of North America while only producing a relatively benign decrease in delta B_h over northern Europe. Therefore even though these synoptic maps of the storm features represent similar time frames with identical Kp rankings, the morphology features of these three storms are quite distinct.

Figure 8

Assimilative model contour map of observed delta B_h over the Northern Hemisphere at 2200 UT, 13 March 1989 (using a 500 nT contour interval).

Figure 9

Assimilative model contour map of observed delta B_h over the Northern Hemisphere at 2041 UT, 15 July 2000 (using a 500 nT contour interval).

[14] A more exact morphology comparison can be provided for these storms by comparing the observed delta B_h variations for these storms at several midlatitude locations. Figure 10 provides a plot of delta B_h observed at the FRD observatory (Fredericksburg) for the storms for 29–30 October 2003, 15–16 July 2000 and 13–14 March 1989. Since these storms are all plotted using the same “nT” scaling, the intensity differences are highlighted between the three storms. For example, during the 15 July 2000 storm, FRD had a delta B_h peak of +1460 nT, while reaching only +506 nT on 30 October 2003. Figure 11 provides a comparison of the observed dB_h/dt (in units of nT/min) at FRD for the October 2003 storms and the 13–14 March 1989 storm. Again, plotting both storms using the same scale, the intensity of the March 1989 storm at this location is more than a factor of two larger for the March 1989 storm than the October 2003 storms. Recalling the observations of GIC for the October 2003 storms provided in Figure 7, all things being equal (including the topology of power grid and orientation of geoelectric field), the levels of GIC for the March 1989 storm would have also been a factor of two or more larger than those observed in October 2003.

Figure 10

Comparison of observed delta B_h at FRD for geomagnetic storms on (top) 29–30 October 2003, (middle) 15–16 July 200, and (bottom) 13–14 March 1989.

Figure 11

Comparison of observed dB_h/dt at FRD for geomagnetic storms on (top) 29–30 October 2003 and (bottom) 13–14 March 1989.

[15] Similar comparisons can be made for the Scandinavian regions of northern Europe for several storms. Figure 12 provides a plot of the observed delta B_h at Brofelde (BFE) on 29–30 October 2003 and also for the storms on 13–14 March 1989 and on 13–14 July 1982. The most important comparisons occur during intervals of decreases in delta B_h associated with the proximity of westward electrojet intensifications. In the case of the October 2003 storms, the peak delta B_h of −1507 nT occurred at 2122 UT on October 30. In contrast, during the March 1989 storm, the delta B_h at BFE reached an intensity of −2943 nT, nearly twice as large as in October 2003. The 13–14 July 1982 storm reached very high delta B_h disturbance values and spanned a large geographic latitude region, as a result observations at the Lovo station (∼5° latitude north from BFE) are also provided in Figure 12. For this storm, the delta B_h intensity reached a peak of ∼−4000 nT at BFE and ∼−4700 nT at Lovo. These disturbances intensities are ∼three times larger than those observed during the October 2003 storms. Figure 13 provides a similar comparison for the dB_h/dt observations during these three storms. Again using the same plot scaling for all three storms, the intensity differences are obvious from the October 2003 storms compared with the March 1989 and July 1982 storms. On 30 October 2003 at time 2007 UT a regional power grid blackout was triggered by the geomagnetic storm in the southern Sweden city of Malmo. The BFE observatory which is nearby to Malmo recorded a peak dB/dt of 308 nT/min. Incidents of transformer heating in that region were also reported due to the storm. In contrast both the March 1989 and July 1982 storms reported peak dB/dt levels that were nearly 10 times larger, for example reaching a peak level of ∼2700 nT/min during the July 1982 storm at the Lovo observatory and ∼2000 nT/min at BFE during the March 1989 storm.

Figure 12

Comparison of observed delta B_h at BFE and LOVO for geomagnetic storms on (top to bottom) 29–30 October 2003 (BFE), 13–14 March 1989 (BFE), and 13–14 July 1982 (BFE and LOVO).

Figure 13

Comparison of observed dB_h/dt at BFE and LOVO for geomagnetic storms on (top) 29–30 October 2003 (BFE), (middle) 13–14 March 1989 (BFE), and (bottom) 13–14 July 1982 (LOVO).

6. Comparison of Impulsive Geomagnetic Field Disturbances Associated With SSC/Shock

[16] Impulsive geomagnetic field disturbances with high dB/dt intensity associated with the arrival of interplanetary pressure pulses at the leading edge of ICME arrival have been a source of large GIC observations even at low latitudes [Kappenman, 2003]. These are difficult to characterize on a planetary basis because of differing magnetospheric and ionospheric processes that produce impulsive geomagnetic field disturbances important to ground-based infrastructures such as power grids. At low and midlatitude locations several geomagnetic observatories collect geomagnetic field data at high cadence (1 s) which is sufficient to resolve the important spectral features of the impulsive disturbance. Observations at Kakioka (KAK) in central Japan provides one of the longest historical records available. This observatory also happened to be located in the dayside region of the planet at the time of the SSC onset at ∼0611 UT on 29 October 2003, which provides a good comparison for the ∼0341 UT onset on 24 March 1991, one of the largest SSC impulsive events on record for this observatory [Araki et al., 1997]. Figure 14 provides a time plot overlaying the observed SSC impulses for the 29 October 2003 event and the 24 March 1991 SSC impulse event. This comparison shows that the March 1991 SSC event is several times larger than that observed on October 2003.

Figure 14

Comparison of impulsive geomagnetic field disturbance at Kakioka for SSC events observed on 29 October 2003 and 24 March 1991.

7. Low-Latitude and Equatorial Region GIC and Geomagnetic Storm Comparisons

[17] In 2000, Metatech & Chubu Electric (located in the Nagoya/Chubu region of Japan) began a GIC monitoring campaign and undertook the development and validation of a GIC model of the network [Erinmez et al., 2002]. This monitoring/modeling campaign discovered the potential for large and long-duration GIC events that can occur in power grids even at low or equatorial latitudes. Because of Japan's middle- to low-latitude location (∼26° Mag. Lat. at Chubu), most power grids in central and southern Japan will not be frequently exposed to electrojet intensifications, as these occur at more northerly locations. In contrast, these locations will more often experience large GICs from SSC and equatorial region current intensifications [Erinmez et al., 2002]. These observations confirm prolonged durations and large magnitudes of GIC are possible at low-latitude locations. These prolonged GIC events are believed to be caused by intensification of the ring current in the magnetosphere, which has an equatorial location and it has been shown that GIC levels at these low-latitude locations roughly correspond to Dst intensity, though there are many important qualifiers to this correlation [Kappenman, 2004]. Table 2 provides a comparative ranking of the 10 largest geomagnetic storms as measured by the Dst index (over approximately the last 50 years) [NOAA, 2004]. In this comparison, the 30 October 2003 storm ranks as the 6th largest with a Dst of −401. This is in comparison to the largest being the −589 Dst reading in the early hours of 14 March 1989. A storm on 20 November 2003, related to the same active regions that triggered the 29–31 October 2003 storms had the 2nd largest Dst with a value of −465.

Table 2. Top 10 Dst Storms^a

Rank	Dst, nT	Date
1	−589	14 March 1989
2	−465^b	20 Nov 2003
3	−429	15 July 1959
4	−427	13 Sept 1957
5	−426	11 Feb 1958
6	−401^b	30 Oct 2003
7	−387	31 March 2001
8	−387	26 May 1967
9	−383	8 Nov 2004
10	−354	9 Nov 1991

• a Dst data from World Data Center for Geomagnetism [2005].
• b Provisional Dst, not the final value.

[18] Eskom, the electric utility that operates the extensive power grid in South Africa has determined that they have had as many as 15 large power transformers (400 kV operating voltage) permanently damaged from internal heating due to the storm events of October 2003 [Makhosi and Coetzee, 2004]. This power grid is exposed to comparable geomagnetic field disturbance environments to those described above for Japan, since the magnitude of its (southern) geomagnetic latitude is similar. Figure 15 provides a comparison of the observed delta B_h for the time period of 29–31 October 2003 at the Hermanus (HER) observatory located in South Africa. Also provided are the delta B_h observations from HER for the storm from 13–15 March 1989, with delta B_h on both plots using the same scale. Figure 16 provides a similar comparison at HER for the same two storms only contrasting the observed dB_h/dt for each storm. While the magnitude of the peak delta B_h and peak dB_h/dt is greater for the 13–15 March 1989 storm, it is quite apparent especially considering the dB_h/dt comparisons in Figure 16 that the 29–31 October 2003 storm has an extensive duration of dB/dt compared to that observed for the March 1989 storm. The accumulated dB/dt over the three day period of 29–31 October 2000 is ∼23% greater than that observed for the storm from 13–15 March 1989. Figure 17 provides a plot of HER dB_h/dt and the long-duration GIC observed in a transformer neutral in Africa for the storm events for 29–31 October 2003. These observations illustrate the close correlation between the observed GIC and dB_h/dt over the region.

Figure 15

Comparison of HER delta B_h on 29–31 October 2003 and 13–15 March 1989.

Figure 16

Comparison of HER dB_h/dt on 29–31 October 2003 and 13–15 March 1989.

Figure 17

Comparison of (top) HER dB_h/dt and (bottom) observed GIC in South Africa region during storms over 29–31 October 2003.

8. Conclusion

[19] The method of morphology-based comparisons provided in this paper illustrates the significant differences that can exist between storms, especially historically important storms with similar index-based characterizations. While the traditional K, Kp and Ap classifications would rank the 29–31 October 2003 storms as some of the largest and most significant of storms in history, these rankings do not hold up to scrutiny in many cases when examining the specifics of the geomagnetic field disturbance environments and morphology patterns of the storm relative to other recent and well-studied storms. For example when examining the impulsive geomagnetic field disturbances associated with electrojet intensifications, the comparisons have shown specific delta B and dB/dt environment differences that ranged from factor of 2 to nearly a factor of 10 differences between similar Kp and Ap ranked storms. In the case of other magnetospheric processes that can drive impulsive geomagnetic field disturbances, the Ap or Dst rankings did not provide meaningful guidance relative to impacts from GIC. The SSC impulsive disturbance appears to be indicated as considerably smaller than the known largest events. The impulsive disturbances driven by K-H shearing produced some of the largest dB/dt and GIC observations in some midlatitude locations for the October 2003 storms, yet this source of geomagnetic field disturbance is not at all captured by means of existing indices. These differences in specific environments would be a likely explanation for relatively modest power grid impacts during the October 2003 storms, even though these storms rank very high using index-based measurement methods.

[20] Because these October 2003 storms were significantly lower in both delta B and dB/dt intensity in ground-level geomagnetic field disturbance at midlatitude locations (<55° geomagnetic latitude) than comparable storms such as 13–14 March 1989, the resulting power system impacts were much lower in total than those observed with the March 1989 superstorm. On the other hand at equatorial locations, the duration of moderate intensity dB/dt exceeded that of the largest Dst storm of March 1989. It is uncertain, but this duration may be a primary mechanism for the internal heating related failures of large power transformers at these locations. While it is difficult to quantify, it is also probable that power systems worldwide benefited significantly from the heightened awareness of geomagnetic storms impacts and the nearly day ahead notifications that were made possible because of improved observations of solar activity in this solar cycle. This allowed utilities that were concerned about power system impacts to constructively engage in conservative operation modes prior to the storm onset, allowing more available reserves for voltage regulation contingencies that can be associated with storms. As a result, it is plausible that this could have contributed to lower overall total power system impacts as well. As described, particularly in the comparative evaluations, storms of much larger intensity than those observed in late October 2003 continue to pose the threat of significant power system impact potential.

Acknowledgments

[31] The author would like to thank the numerous suppliers of observed magnetometer data, including the U.S. Geological Survey, the NOAA National Geophysical Data Center, the Geological Survey of Canada, the 210 Longitude Chain of magnetic field observatories, the World Data Center for Geomagnetism in Kyoto, Japan, and the CANOPUS instrument array by the Canadian Space Agency.

Appendix A:: Overview of Assimilative Model to Develop and Map Contours of delta B_h Over the Northern Hemisphere

[21] The contour maps of ground level delta B_h shown in a number of the figures in this paper are mesoscale resolution depictions derived from a proprietary assimilative data model called Ground-Level Assimilative Model (GLAM). The maps of delta B_h are derived from the observations of geomagnetic field variation in 1-min cadence from available geomagnetic observatories in the northern hemisphere. Techniques to map these discrete observations into a continuous map of disturbance contours is generally derived from triangular interpolation techniques commonly used in geostatistical data mapping disciplines. Other researchers have previously performed similar multiobservatory analysis of geomagnetic storm morphologies. The first such effort was associated with the analysis of the large-scale disturbance patterns over North America during the August 1972 geomagnetic storm [Anderson et al., 1974]. Others have used similar multiobservatory data to map the complex spatial and temporal dynamics of the March 1989 superstorm [Boteler and Jansen van Beek, 1992; Kappenman, 2001].

[22] The specifics of the modeling approach used in the GLAM model can be summarized as follows.

[23] 1. Input data from available geomagnetic observatories (in North America latitude spacing ∼5°, longitude spacing more variable but less important to model output).

[24] 2. Horizontal component of geomagnetic data at each time step converted to vectors.

[25] 3. GLAM model utilizes a data fill technique based on coherent data characteristics to estimate into gridded regions not directly sampled.

[26] 4. Through data fill, a 2D gridded output can be estimated for cell dimensions of 2° latitude by 5° longitude.

[27] 5. The original observatory intensity data is fully preserved in the cell containing each observatory location.

[28] 6. Model will provide best estimate across large ocean areas, however with limited observed data, estimation uncertainty in those regions will increase versus continental regions of North America and western Europe which have closely spaced observatories.

[29] 7. Model will not prescribe a delta B value higher than any adjacent observed delta B values; therefore estimates of contour intensities are conservative and may at times understate total storm spatial intensity.

[30] Figure A1 provides a map summary of the observatories used for each of the modeled storms, which include the storms on 13–14 March 1989, 15–16 July 2000 and 29–31 October 2003.

Figure A1

Map of Northern Hemisphere geomagnetic observatories used in assimilative model of estimated delta B_h contours for geomagnetic storm conditions: (top) data used for 13–14 March 1989 storm, (middle) data used for 15–16 July 2000 storm, and (bottom) data used for 29–31 October 2003 storms.