SPACE WEATHER, VOL. 4, S06006, doi:10.1029/2006SW000220, 2006
Solar energetic particle flux enhancement as an indicator of halo coronal mass ejection geoeffectiveness H. Gleisner1 and J. Watermann1 Received 4 January 2006; revised 27 February 2006; accepted 6 March 2006; published 28 June 2006.
[1] Nearly all strong magnetic storms are generated by halo coronal mass ejections (CMEs), but most frontside halo CMEs are not followed by strong storms. Hence additional information is required to discriminate highly geoeffective CMEs from those less geoeffective. There is a tendency for the strongest magnetic storms to be generated by the fastest CMEs, and in the absence of detailed information on the internal structure of an observed CME, the speed is often used to evaluate the risk that a strong storm will follow. We here report on the alternative use of solar energetic particle (SEP) flux as an indicator of the geoeffectiveness of halo CMEs, addressing the question of whether SEP flux is better in this respect than CME speed. On the basis of a list of 137 frontside halo CMEs, we have investigated the relations between CME speed, SEP flux, and strong magnetic storms. We find that enhancements of the 10 MeV SEP flux close to CME onset can be used to indicate whether a full halo CME will be followed by Dst below 100 nT within 18 to 72 hours (our definition of a CME --strong magnetic storm association). Ranking the CMEs by speed and by SEP flux enhancement shows that the latter indicator results in a better discrimination between highly geoeffective CMEs and those less geoeffective. The results suggest that SEP flux enhancements may provide a more efficient discrimination than CME speed for any choice of discriminating thresholds, resulting in lower rates of both false alarms and missed predictions. Citation: Gleisner, H., and J. Watermann (2006), Solar energetic particle flux enhancement as an indicator of halo coronal mass ejection geoeffectiveness, Space Weather, 4, S06006, doi:10.1029/2006SW000220.
1. Introduction [ 2 ] Coronal mass ejections (CMEs) are large-scale expulsions of plasma and magnetic fields from the Sun, which drive coronal material into interplanetary space and which may generate substantial disturbances in the ambient solar wind. CMEs are now known to be the main source of strong magnetic storms [e.g., Gosling et al., 1991; Webb et al., 2000; St. Cyr et al., 2000; Cane et al., 2000]. Hence the occurrence of a CME, and particularly a frontside halo CME with a significant Earthward velocity component, is an important predictor of strong magnetic storms. CME observations now play a central role in medium-range (1 to 4 days) geomagnetic forecasting [Luhmann, 1997]. [3] The occurrence of a large, frontside halo CME is by itself a rather inefficient predictor [e.g., Webb et al., 2000]. Defining a strong magnetic storm as a period of enhanced
1 Geomagnetism and Space Physics Program, Danish Meteorological Institute, Copenhagen, Denmark.
Copyright 2006 by the American Geophysical Union
geomagnetic activity when the Dst index falls below 100 nT, we find that nearly all strong magnetic storms are preceded by frontside halo CMEs, but that most of these CMEs are not followed by a strong storm. As we will see is this study, even a majority of full halo CMEs, that is, the CMEs most likely to strike a direct hit at Earth, are not followed by strong storms. Forecasts of magnetic storms solely on the basis of the occurrence of large, frontside halo CMEs would thus suffer from high rates of false alarms [e.g., St. Cyr et al., 2000]. Additional information is required to discriminate the highly geoeffective CMEs from those less geoeffective. [4] Indicators of CME geoeffectiveness, potentially providing the warranted discrimination, have been extensively discussed since it became apparent that CMEs play a central role in the generation of magnetic storms. A fundamental condition for a CME to be geoeffective is that material ejected from the Sun, often referred to as an interplanetary CME (ICME), or shocks driven by the ejected material should hit the Earth. This condition is potentially fulfilled for frontside halo CMEs where the
S06006
1 of 10
S06006
GLEISNER AND WATERMANN: SEPs AND GEOEFFECTIVENESS OF HALO CMEs
angular width and symmetry of the CME, as observed in white light coronagraph images, indicate that the CME is moving directly toward the Earth [Cane et al., 1987; St. Cyr et al., 2000]. However, even if an ICME reach Earth, it may still not be highly geoeffective. An additional requirement is that the interplanetary magnetic field (IMF) should have a southward component Bz of sufficient magnitude and duration. It has been noted that some interval of southward magnetic field is likely to occur within or ahead of most ICMEs [Webb et al., 2001], and that some level of enhanced geomagnetic activity is observed for most ICMEs that actually hit the Earth. Nevertheless, for accurate predictions of the geomagnetic activity, the preferred indicator of CME geoeffectiveness would be the detailed magnetic field configuration of an ICME. In addition to the magnetic field configuration, there are also other CME properties that influence the geoeffectiveness. Two properties that can be observed, and that can be obtained in near-real time, are the CME speed close to the Sun and the energetic particles generated by a CME. In the absence of direct observations of the magnetic fields, or accurate predictions of them, these observables may be regarded as indicators of CME geoeffectiveness in a statistical sense. [5] There is a tendency for the strongest magnetic storms to be generated by the fastest CMEs. In line with this, it has been suggested that CME speed to some extent can be used as an indicator of CME geoeffectiveness [Srivastava and Venkatakrishnan, 2002; Kim et al., 2005]. This is a concept often seen in operational space weather forecasting, where a ‘‘fast’’ halo CME is interpreted as a high risk for adverse space weather whereas a ‘‘slow’’ CME is interpreted as a smaller risk. Lindsay et al. [1999] found a relationship between CME speed and the magnitude of the total magnetic field in the associated ICME. Yurchyshyn et al. [2004] described similar relations between CME speed, the magnitude of Bz, and the geomagnetic Dst index, while Gonzalez et al. [2004] showed that the expansion speeds of halo CMEs associated with magnetic clouds are related to the peak Dst index of the resulting magnetic storms. The speeds used in these studies are the speeds of CME structures as projected onto the plane of the sky. Several studies [e.g., Gopalswamy et al., 2000; Cane et al., 2000] have reported on relationships between the CME’s projected and travel speeds which, even though accompanied by a large scatter of the data, give some credence to the use of projected speeds as a proxy for the true travel speed of CMEs. [6] The strongest magnetic storms are often accompanied by particle storms, strongly enhanced fluxes of energetic protons and ions referred to as solar energetic particles (SEPs). SEP events lasting for more than a day are primarily generated by strong shock waves propagating in the solar corona and in the interplanetary medium [e.g., Cane et al., 1987; Reames, 1999; Kahler, 2001]. They exhibit a close association with CMEs, and it is currently thought that the large, gradual, and long-lived SEP events are due to CME-driven shocks [Gosling, 1993; Reames,
S06006
1999]. Peak SEP flux intensities are correlated to CME speed [e.g., Kahler et al., 1978; Cane et al., 1998] but a broad scatter of the data shows that the SEP fluxes depend on other factors than the CME speed, such as preceding CMEs [Kahler and Vourlidas, 2005] or the presence of preexisting SEPs in the ambient medium [Kahler, 2001; Gopalswamy et al., 2004]. [7] The close relations between CMEs, CME ejecta, and energetic particles are demonstrated by the common use of particle signatures as a tool to probe the large-scale magnetic field structures of CME ejecta as they pass by the Earth [e.g., Richardson, 1997; Malandraki et al., 2005]. From the fact that faster CMEs tend to generate stronger magnetic storms, and from the reported relations between CME speed and SEP flux intensities, we could expect some sort of relation between SEP fluxes and the strength of magnetic storms. SEPs observed at 1 AU convey information on the shocks driven outward from the Sun by CMEs and it is not unreasonable to ask whether SEP flux characteristics somehow can be used to indicate CME geoeffectiveness. This particular space weather application of energetic particle observations have been reported a few times in the scientific literature. Dmitriev and Bakhareva [2000] tried to establish a connection between the intensity of 1 MeV protons and CME velocity in order to predict CME travel times, but their results appear to be inconclusive. Smith et al. [2004] used observations of 47 -65 keV ions, that is, an energy range more than 2 orders of magnitude lower than in the present study, to predict the arrival of interplanetary shocks hours before they arrive at Earth. They suggested that the combination of halo CME observations and 47 -- 65 keV ion enhancements exceeding a certain threshold can be used to improve predictions of magnetic storms. In a recent study, Gleisner and Watermann [2005] discussed the use of 10 MeV SEP fluxes as a tool to discriminate halo CMEs followed by strong magnetic storms from those not followed by strong storms, while Valtonen et al. [2005] demonstrated the feasibility of using 1 -- 110 MeV protons to evaluate the geoeffectiveness of halo CMEs. In the latter study, MeV particle fluxes were also used to derive a proxy for the transit times of interplanetary shock waves driven by CMEs. [8] In the present paper, we report on the use of 10 MeV SEP flux enhancements close to CME onset as an indicator of CME geoeffectiveness. On the basis of a list of 137 halo CMEs (all frontside halo CMEs for which alerts were issued by the LASCO team at the Naval Research Laboratories during the 4-year period 2001 to 2004) we have investigated the relations between halo CMEs, CME speed, SEP flux, and strong magnetic storms. From the conditional probabilities, and the related false alarm and miss rates, for strong magnetic storms following upon full halo CMEs, we address the question of whether SEP fluxes may provide an indication of CME geoeffectiveness and, more specifically, whether this indication is more efficient than that provided by CME speed. The results might help us develop methods to discriminate CMEs
2 of 10
GLEISNER AND WATERMANN: SEPs AND GEOEFFECTIVENESS OF HALO CMEs
S06006
Table 1. All Full Halo CMEs for Which Alerts Were Issued by the LASCO Team at the NRL From January 2001 to December 2004a Mail 83 85 85 86 87 91 92 93 96 99 101 102 104 105 106 107 108 109 110 117 119 120 125 131 132 133 135 137 139 143 146 147 158 160 163 166 167 168 173 174 176 177 183 185 187 195 201 202 210 212 213 214 215 217 222 225 226 227 233 234 238 240 241 248
Start 2001-01-10 2001-01-20 2001-01-20 2001-01-28 2001-02-02 2001-02-11 2001-02-15 2001-02-28 2001-03-19 2001-03-25 2001-03-28 2001-03-29 2001-04-05 2001-04-06 2001-04-09 2001-04-10 2001-04-11 2001-04-12 2001-04-26 2001-08-14 2001-08-25 2001-09-11 2001-09-24 2001-10-09 2001-10-19 2001-10-19 2001-10-22 2001-10-25 2001-11-04 2001-11-17 2001-11-22 2001-11-22 2002-02-12 2002-03-11 2002-04-17 2002-05-07 2002-05-16 2002-05-22 2002-07-15 2002-07-18 2002-07-23 2002-07-26 2002-08-16 2002-09-05 2002-10-25 2003-01-30 2003-06-15 2003-06-17 2003-10-23 2003-10-26 2003-10-28 2003-10-29 2003-11-02 2003-11-04 2003-11-11 2003-11-18 2003-11-20 2003-12-02 2004-01-20 2004-01-21 2004-03-31 2004-04-06 2004-04-08 2004-06-07
00:54 19:31 21:54 15:54 14:54 01:31 13:54 14:50 05:26 17:06 12:50 10:26 17:06 19:30 15:54 05:30 13:31 10:31 12:30 16:08 16:50 14:54 10:30 11:30 01:27 16:50 15:06 15:26 16:35 05:30 20:58 23:30 15:06 23:06 08:26 04:06 00:50 03:26 20:30 08:06 00:42 22:06 12:30 16:54 09:50 09:39 23:54 23:30 08:54 17:54 10:54 20:54 17:30 19:54 13:54 08:06 08:06 10:26 00:06 04:54 21:53 13:31 10:30 01:27
Longitude, deg
VCME, km/s
Dlog(F10)
Min(Dst), nT
43 40 40 59 85 57 16 12 7 22 25 19 50 31 4 9 27 40 23 10 34 29 27 17 18 29 18 20 18 42 67 55 37 43 37 28 14 56 1 15 72 30 17 30 19 2 78 61 55 42 9 6 56 88 62 18 8 66 5 25 12 15 11 0
642 673 1576 795 390 824 601 296 268 639 501 991 1046 1103 1086 1678 750 912 918 415 1130 703 2231 1041 468 705 1304 884 1620 1173 1246 1500 414 883 1192 611 548 1496 1078 1090 2170 737 1404 1642 789 567 2002 1779 1110 1432 2125 1948 1826 2381 1383 1175 547 1234 874 852 220 1075 915 596
0.18870 0.37219 0.40711 2.47578 0.14046 0.45732 0.18398 0.16490 0.13564 0.18524 0.20226 1.85020 0.01979 0.43775 1.15279 1.67620 1.00470 0.15942 0.21984 0.33805 0.30760 0.20819 3.54436 0.37439 1.50744 0.43607 2.25480 0.50569 3.81028 0.86161 3.97959 4.31780 0.18982 0.32312 2.18939 0.03037 0.26307 1.66883 0.31503 0.24592 0.38115 0.41068 0.18152 0.21877 0.17689 0.16208 0.26351 0.11824 0.75589 3.46907 3.49370 0.55548 1.96029 1.36250 0.06933 0.73309 1.29475 0.69713 0.19180 0.21170 0.14914 0.21087 0.18588 0.15088
24 38 38 39 5 50 13 19 165 97 358 358 51 54 256 256 146 66 33 9 19 58 101 74 166 166 101 160 277 36 213 213 21 9 151 24 24 108 13 27 15 16 48 170 64 50 145 78 67 180 401 401 89 32 66 472 241 54 149 149 88 34 34 29
S06006
Table 1. (continued) Mail 251 252 255 256 257 266 268 269 271 275 276 277 278 280 281 283 284
Start 2004-07-13 2004-07-13 2004-07-22 2004-07-23 2004-07-25 2004-09-12 2004-10-30 2004-11-01 2004-11-03 2004-11-07 2004-11-08 2004-11-09 2004-11-10 2004-12-03 2004-12-08 2004-12-30 2004-12-30
00:54 09:08 07:31 16:06 14:30 00:36 16:54 03:54 16:06 17:18 04:06 17:26 02:26 00:26 20:26 10:57 22:30
Longitude, deg
VCME, km/s
Dlog(F10)
Min(Dst), nT
45 51 13 4 30 49 28 49 38 17 20 51 49 2 7 53 46
406 571 913 990 1228 1326 670 550 1016 1770 520 1853 2900 1150 360 1132 1150
0.76776 0.14033 1.01260 0.43397 2.22398 0.29769 0.46506 2.54258 0.09780 3.03803 0.00700 0.44475 1.06297 0.22792 0.13798 0.16044 0.13257
14 24 129 150 182 57 32 27 24 296 296 264 132 28 53 57 57
a The CMEs are identified by the ID number of the corresponding alert mail sent by the NRL. We further list the start date and time (in UT) of the CME, the heliographic longitude, the CME speed VCME, the SEP flux enhancement Dlog(F10), and the minimum Dst index from 18 to 72 hours after the start of the CME. In Figure 1, the distributions of CME speed and SEP flux enhancements are shown for the 81 full halo CMEs.
followed by strong storms from those only followed by weaker geomagnetic activity.
2. Observational Data 2.1. Coronal Mass Ejections [9] CMEs are routinely detected from white light coronagraph images obtained by the LASCO instrument onboard the SOHO spacecraft [Brueckner et al., 1995]. The heliographic location of a CME is disclosed by EUV activity in the low solar corona, which is detected by the EIT instrument onboard SOHO. Through an effort by the LASCO team at the Naval Research Laboratories (NRL), data on halo CMEs are rapidly dispersed to the user community to give an early warning of possible space weather disturbances. The archived alert mails comprise a useful data set that is available at the NRL web site (http:// lasco-www.nrl.navy.mil/halocme.html), and the CME data referred to below are taken from this data set. [10] In this study we have used all frontside halo CMEs for which alerts were issued by the NRL during the 4-year period January 2001 to December 2004. Only those halo CMEs which were observed by LASCO, and for which the complementary EIT data show that the CME is directed toward the Earth, are included. As pointed out by St. Cyr [2005], the halo CMEs for which alerts are issued are selected under a tight time constraint. They are normally reported within 24 hours and events that are later considered halo-like are not added to the list of halo CMEs. For our purposes, to make a statistical comparison of CME speed and SEP flux as indicators of CME geoeffectiveness, this is not a serious limitation. Furthermore, since the CMEs are reported before they arrive at Earth, the data 3 of 10
S06006
GLEISNER AND WATERMANN: SEPs AND GEOEFFECTIVENESS OF HALO CMEs
Figure 1. Distribution of CME speed and 10 MeV SEP flux enhancement for 81 frontside full halo CMEs. The CMEs have been ranked from smallest to largest VCME and from smallest to largest Dlog(F10), and these variables are here plotted with rank order on the abscissa. set is completely unbiased with respect to the resulting geomagnetic activity. [11] Each CME in the list has been classified as ‘‘full halo’’ or ‘‘partial halo’’ depending on the angular width of the CME in the LASCO images. Eruptive events with an apparent width of about 360° are regarded as full halo CMEs while those wider than 120°, but less than 360°, are regarded as partial halo CMEs [St. Cyr, 2005]. For each CME, the speed referred to is a linear fit of visible CME
S06006
structures, seen in running difference images and measured at the position angle where the leading edge shows maximum speed. The start time of the CME, or the CME onset, is defined as the first appearance of the CME in the LASCO C2 images. The heliographic location of the CME has been determined from complementary EIT data, although in this study we only use it to establish that the observed halo CME is a frontside, rather than backside, event. Our final list of CMEs consists of 137 entries; 56 classified as partial halo CMEs and 81 classified as full halo CMEs. These 81 full halo CMEs are listed in Table 1 and their distributions of CME speed, VCME, and SEP flux enhancements, Dlog(F10), are shown in Figure 1. [12] For the purposes of this study, we also need a criterium of whether a CME is followed by a strong magnetic storm. Here, we regard a CME as being followed by a strong storm if we find Dst less than 100 nT within 18 to 72 hours after CME onset. We do not attempt to unambiguously identify the solar source for each individual storm. Two or more CMEs sometimes occur in close connection and it cannot be unambiguously determined which one, if any, caused the observed geomagnetic activity. Instead of introducing more complicated or subjective criteria of CME-storm associations in these ambiguous cases, we simply stick to the simple and very strict definition given above. Hence each CME is treated separately without considering any other CMEs. The sole exception to this is the definition of miss rates in section 5 where, in a few cases, the occurrence of nearly simultaneous CMEs are partly corrected for.
2.2. Solar Energetic Particle (SEP) Flux [13] SEP fluxes are observed by the SEM instrument onboard the series of GOES geostationary satellites. The data are distributed in real time through NOAA’s Space
Figure 2. Definition of SEP flux enhancement Dlog(F) as the logarithm of the ratio between the maximum SEP flux Fpost observed during the 12 hours following CME onset and the minimum SEP flux Fpre during the 6 hours preceding CME onset. CME onset is here defined as the first detection of the CME in the C2 images from LASCO at a distance of about 2 solar radii from the Sun’s surface. 4 of 10
S06006
GLEISNER AND WATERMANN: SEPs AND GEOEFFECTIVENESS OF HALO CMEs
Environment Center and archived data are available at the National Geophysical Data Center (NGDC) of NOAA. [14] In the present study, we have used the hourly averaged integrated flux of 10 MeV particles. As shown in Figure 2, we define the SEP flux enhancement, Dlog(F10), as the logarithm of the ratio between the maximum hourly averaged flux Fpost during the 12 hours following CME onset and the minimum hourly averaged flux Fpre during the 6 hours preceding CME onset, D logðF Þ ¼ log F post logðF pre Þ ¼ log F post=F pre :
This requires the definition of pre onset and post onset values of the hourly SEP fluxes for all CMEs. When data gaps occur, such as in Figure 2, the found values may differ from the true maximum (post onset) or minimum (pre onset) values, and the effect of data gaps may be, but need not be, to decrease the magnitude of the observed SEP flux enhancement.
2.3. Geomagnetic Activity and Magnetic Storms [15] The hourly index Dst measures the depression of the ground level horizontal geomagnetic field at low to middle geomagnetic latitudes [Mayaud, 1980]. It is widely used as an indicator of the strength of magnetic storms, and in many studies [e.g., Gonzalez et al., 1994] the term ‘‘strong storm’’ have been used for periods with enhanced geomagnetic activity when Dst falls below 100 nT. We here follow this practice. [16] The Dst data are available at the World Data Center for Geomagnetism in Kyoto. Up to December 2002, we have used the final index, whereas after this date we have used the provisional Dst index.
3. CME Speed, SEP Flux, and Longitude [17] In Figures 3a and 3b, we show the estimated CME speed and the heliographic longitude for the 137 frontside halo CMEs in our database. Figure 3a shows 81 full halo CMEs, and Figure 3b shows 56 partial halo CMEs. The color of the symbols indicates whether the CME is followed by a strong magnetic storm, that is, whether we find values of the Dst index below 100 nT in the time interval from 18 to 72 hours after CME onset. [18] As expected, full halo CMEs are much more likely to be followed by a magnetic storm than partial halo CMEs. Still, the observation of a full halo CME is by itself a relatively inefficient predictor of strong magnetic storms. Only around 40% (32 out of 81) of all front-sided full halo CMEs are followed by strong magnetic storms, while less than 15% (8 out of 57) of partial halo CMEs show such an association. [19] In Figure 3a, we can discern the expected tendency that high-speed CMEs are more likely to be followed by a strong storm than CMEs with lower speeds. However, the intermixing of strong storm and no strong storm symbols in Figure 3a is relatively large, showing that the discrim-
S06006
ination between highly geoeffective and less geoeffective CMEs provided by the CME speed is not very efficient. [20] In Figures 3c and 3d we now switch from CME speed to SEP flux enhancement, Dlog(F10). As expected from the short discussion in section 1, the observation of a large SEP flux enhancement close to the onset of a full halo CME makes it more likely that the CMEs is followed by a strong magnetic storm. The degree of intermixing between strong storm and no strong storm symbols appears to be relatively small, the visual impression being that it is smaller than for CME speed. Compared to CME speed, the SEP flux enhancement may in fact provide a better discrimination between the CMEs followed by strong storms and those only followed by weaker geomagnetic activity. [21] For the partial halo CMEs, we do not find any evidence for a relation, neither with VCME nor with Dlog(F10). We note that some of the partial halo CMEs that are followed by a strong storm actually occur in close connection with a full halo CME. Disregarding these cases would further underline the fact that very few of the partial halo CMEs are followed by a strong magnetic storm. In the remainder of this paper, we therefore restrict the study to the 81 CMEs classified as ‘‘full halos.’’
4. Probabilities for Strong Magnetic Storms [22] Around 40% of the frontside CMEs classified as ‘‘full halos’’ are followed by strong storm conditions within 18 to 72 hours. From the discussion in section 1 and the scatterplots in Figure 3, we expect this fraction to be considerably larger for high-speed CMEs, and for CMEs associated with large SEP flux enhancements, than for all CMEs. In fact, the degree of CME -- strong storm association vary systematically with the speed of the CME and the magnitude of the SEP flux enhancement. This variation can be described in terms of the conditional probabilities P½Dst 100 nTjVCME V*
ð1Þ
P½Dst 100 nTjD logðF10 Þ D* ;
ð2Þ
and
where V* and D* are thresholds above which CME speeds are deemed ‘‘high’’ and SEP flux enhancements are deemed ‘‘large.’’ Figure 4a shows the conditional probabilities, based on the 81 full halo CMEs in our database, as functions of these thresholds. Note that in Figure 4a, the curves for VCME and Dlog(F10) have separate abscissae. [23] In Figure 4b, we show an alternative plot of the same data where the conditional probabilities for VCME and Dlog(F10) are plotted on a common abscissa. The full halo CMEs have been ranked from smallest to largest VCME and from smallest to largest Dlog(F10), respectively, and the conditional probabilities are plotted as functions
5 of 10
S06006
GLEISNER AND WATERMANN: SEPs AND GEOEFFECTIVENESS OF HALO CMEs
S06006
Figure 3. CME speed, SEP flux enhancement, and heliographic longitude for 137 frontside halo CMEs: (left) 81 full halo CMEs and (right) 56 partial halo CMEs. Red diamonds show CMEs followed by a strong magnetic storm, that is, CMEs that are followed by Dst below 100 nT somewhere between 18 to 72 hours after CME onset. Less intermixing of red and blue diamonds indicates a better ability to discriminate CMEs followed by a strong storm from CMEs not followed by a strong storm. For full halo CMEs (left plots), the SEP flux enhancements appear to be equally good, or better than, the CME speeds in providing such discrimination. of the rank order (see Figure 1 for the relation between physical variable and rank order). In this plot we can compare the two indicators of CME geoeffectiveness directly over a wide range of threshold levels, and also compare both with an ideal indicator providing perfect
discrimination and with an indicator not providing any discrimination at all. [24] An ideal indicator would be an observable for which we could determine a threshold above which all full halo CMEs were followed by a strong storm, and below which no CMEs were followed by a strong storm. 6 of 10
S06006
GLEISNER AND WATERMANN: SEPs AND GEOEFFECTIVENESS OF HALO CMEs
S06006
Figure 4. Fraction of frontside full halo CMEs that are followed by a strong magnetic storm within 18 to 72 hours after CME onset under the condition that VCME or Dlog(F10) exceed a certain threshold. The two plots show the same data, but have different scaling on the x-axes. (a) Thresholds quantified by the level of VCME and Dlog(F10), respectively. (b) Thresholds quantified by their rank order amongst the 81 full halo CMEs in the database. Figure 4b also shows the region bounded by an ideal indicator of CME geoeffectiveness providing perfect discrimination and a useless indicator not providing any discrimination at all. Simple indicators of CME geoeffectiveness, such as VCME or Dlog(F10), could not be expected to come close to this ideal; the dependence of magnetic storms on the magnetic field configuration of ICMEs, and the diversity of magnetic field configurations for similar magnitudes of the studied indicators, precludes this. However, the intercomparison of potential indicators, and their performances relative to an ideal indicator, nevertheless gives some insight into the relevance of the information provided by the studied indicators. [25] From the plots in Figure 4 we note that (1) the CME -- strong storm association appears to be weaker using V CME as an indicator than using Dlog(F 10 ), (2) Dlog(F10) is closer to an ideal indicator than VCME, and (3) the Dlog(F10) curve is closer to a monotonic increase than the VCME curve. This latter result can also be seen as a better separation of the strong storm and no strong storm symbols in the Dlog(F10) scatterplots than in the VCME scatterplots: less intermixing of the symbols in Figure 3c than in Figure 3a.
5. False Alarms and Missed Predictions [26] In geomagnetic forecasting, the number of false alarms and the number of missed predictions are more relevant measures than the conditional probabilities described in section 4. By applying a large enough threshold on CME speed or SEP flux enhancement we can decrease the number of false alarms at the cost of an increased number of missed predictions. Here we define the false
alarm rate as the number of false alarms in relation to the number of all alarms issued on the basis of the 81 full halo CMEs. The miss rate is here defined as the number of full halo CMEs for which no alarms were issued but that nevertheless were followed by a strong storm, in relation to all full halo CMEs followed by a strong storm. This definition of miss rate does not properly account for the case when two CMEs, of which only one gives an alarm, occur in close connection and are followed by a strong storm. To account also for this case we do not regard the weaker CME as a miss, under the condition that the two CMEs occur within a time frame of less than 24 hours. Our definition of miss rate also does not account for strong magnetic storms that are not associated with any full halo CME listed in our database. As a result, we tend to get a too low estimate of the true number of missed predictions. However, the definition of miss rate used here serves our more limited purposes of this study, namely to make a statistical comparison of CME speed and SEP flux as indicators of CME geoeffectiveness. The false alarm rates, as here defined, should give a realistic estimate of the results obtained in an operational setting. [27] In Figures 5a and 5b, we show how the rates of false alarms and missed predictions depend on the thresholds chosen for the two alternative indicators: CME speed (Figure 5a) and SEP flux enhancement (Figure 5b). In Figures 5c and 5d, we plot these data on common abscissae in order to better compare the two indicators: false alarm rates (Figure 5c) and miss rates (Figure 5d). The results
7 of 10
S06006
GLEISNER AND WATERMANN: SEPs AND GEOEFFECTIVENESS OF HALO CMEs
S06006
Figure 5. Rates of false alarms and missed predictions as a function of the threshold for the discriminating variable: (a) CME speed VCME and (b) SEP flux enhancement Dlog(F10). The bottom plots show the same data as in Figures 5a and 5b, but (c) false alarm and (d) missed-prediction rates are here plotted on common abscissae for the two indicator variables. confirm the findings presented in section 4; the SEP flux enhancement close to CME onset appears to be more efficient as an indicator of halo CME geoeffectiveness than the speed of the CME. For any threshold of CME speed, there is a range of SEP flux enhancement thresholds that at the same time gives both lower rates of false alarms and lower rates of missed predictions. [28] Selecting the appropriate threshold is a matter of making a reasonable compromise between the number of false alarms and the number of missed predictions. On the basis of what we would have selected for the ideal indicator in Figure 5, we arrive at the threshold D* 0.5 for the
SEP flux enhancement and V* 1100 km/s for the CME speed. This would give a false alarm rate of around 30% for predictions based on Dlog(F10) and 45% for predictions based on VCME, as compared to 60% if we disregard this additional information on CME geoeffectiveness. The corresponding missed-prediction rates are 23% and 40% when using SEP flux and CME speed, respectively. Other combinations of false alarm rates and missed-prediction rates are equally valid. The crucial point is that there is no choice of thresholds for which CME speed gives both a lower false alarm rate and a lower miss rate than SEP flux,
8 of 10
S06006
GLEISNER AND WATERMANN: SEPs AND GEOEFFECTIVENESS OF HALO CMEs
while there is a wide range of thresholds for which SEP flux is better than CME speed in this respect.
6. Discussion and Conclusions [29] This study shows that during the four years from January 2001 to December 2004 we would have done better, in terms of both false alarm and missed-prediction rates, using SEP fluxes rather than CME speeds to indicate whether an observed full halo CME is followed by a strong magnetic storm. The reason is that the variable Dlog(F10), defined in section 2.2, orders the full halo CMEs after their strong storm associations better than the variable VCME. This property of the SEP flux enhancement does not depend on the exact definition of the indicator variable. pre Any monotonic function of the ratio Fpost 10 /F10 , as defined in section 2.2, would give the same result. [30] These conclusions are based on 81 full halo CMEs for which alerts were issued by the NRL. Assuming that we can generalize from this somewhat limited sample, we have to ask why SEP flux enhancement would be more efficient than CME speed as an indicator of CME geoeffectiveness. Are there certain characteristics of the SEP fluxes that provide more or better information on the properties of potentially geoeffective solar wind disturbances than the CME speeds do? Or is it the current estimates of CME speed that are insufficient as descriptors of the intrinsic properties of CMEs and CME ejecta? [31] The estimation of CME speeds from white light coronagraph images is based on the observed speeds of CME structures as projected onto the plane of the sky. The speeds of a limited number of CME structures are measured, and from these measurements a single speed is selected, or somehow derived, as being representative for the CME in the vicinity of the Sun. This speed has been shown to be correlated with the CME travel speed, that is, the average CME speed out to the Earth’s orbit at a distance of 1 AU from the Sun. In this process there are several error sources and a certain degree of subjectivity, and the resulting uncertainties are likely to be substantial. SEP fluxes, on the other hand, are obtained by an objective measurement from an instrument in space. Unlike CME speed, the major uncertainties are not related to the definition of the physical quantity or the measurement of it, but rather to incomplete understanding of the causeand-effect relations between the observed CMEs, the ambient medium, and the resulting SEP flux characteristics observed at 1 AU. [32] An alternative, and perhaps more interesting, explanation is that the SEP flux enhancements observed near the Earth could provide information on the properties of CME ejecta that make them geoeffective. Recent research has shown that the relations between CMEs and the production of SEPs are complex. Nevertheless, there is a fundamental consensus that gradual SEP events result from acceleration of particles at shocks, where the strength and spatial structure of the shock determine the characteristics of the observed SEP fluxes. At the same
S06006
time, shock compression of magnetic fields is an important source of the large-magnitude IMFs that is a key factor in the generation of magnetic storms. The picture is made more complex from recent studies pointing out that preceding CMEs, preexisting energetic particles, and the density structure of the extended corona all have an influence on the production of SEPs. A better understanding of the interrelations amongst these factors, and a more complete mapping of their relations with geomagnetic storms of different strengths, might allow us to use nearEarth observation of SEPs to improve on medium-range geomagnetic forecasting. [33] Acknowledgments. Parts of this work were done within the ESA Space Weather Applications Pilot Project, under ESTEC contract 16983/03/NL/LvH. The LASCO team at the Naval Research Laboratory are gratefully acknowledged for their efforts at prompt identification and characterization of halo CMEs. The CME data used in this study were obtained from the NRL Web site http://lascowww.nrl.navy.mil/halocme.html. SEP data were obtained from NOAA/NGDC, and Dst index data were obtained from the World Data Center for Geomagnetism in Kyoto.
References Brueckner, G. E., et al. (1995), The large angle spectroscopic coronagraph (LASCO), Sol. Phys., 162, 357 -- 402. Cane, H. V., N. R. Sheeley, and R. A. Howard (1987), Energetic interplanetary shocks, radio emission, and coronal mass ejections, J. Geophys. Res., 92, 9869 -- 9874. Cane, H. V., I. G. Richardson, and O. C. St. Cyr (1998), The interplanetary events of January -- May, 1997 as inferred from energetic particle data, and their relationships with solar events, Geophys. Res. Lett., 25, 2517 -- 2520. Cane, H. V., I. G. Richardson, and O. C. St. Cyr (2000), Coronal mass ejections, interplanetary ejecta and geomagnetic storms, Geophys. Res. Lett., 27, 3591 -- 3594. Dmitriev, A. V., and M. F. Bakhareva (2000), Solar cosmic ray measurements as a tool for the magnetic storm commencement forecast, Adv. Space Res., 26, 229 -- 232. Gleisner, H., and J. Watermann (2005), Concepts of medium-range (1 to 4 days) geomagnetic forecasting, Adv. Space Res., in press, doi:10.1016/j.asr.2005.10.010. Gonzalez, W. D., J. A. Joselyn, Y. Kamide, H. W. Kroehl, G. Rostoker, B. T. Tsurutani, and V. M. Vasyliunas (1994), What is a geomagnetic storm?, J. Geophys. Res., 99, 5771 -- 5792. Gonzalez, W. D., A. Dal Lago, A. L. Clua de Gonzalez, L. E. A. Vieira, and B. T. Tsurutani (2004), Prediction of peak-Dst from halo CME/ magnetic cloud-speed observations, J. Atmos. Sol. Terr. Phys., 66, 161 -- 165. Gopalswamy, N., A. Lara, R. P. Lepping, M. I. Kaiser, D. Berdichevsky, and O. C. St. Cyr (2000), Interplanetary acceleration of coronal mass ejections, Geophys. Res. Lett., 27, 145 -- 148. Gopalswamy, N., S. Yashiro, S. Krucker, G. Stenborg, and R. A. Howard (2004), Intensity variation of large solar energetic particle events associated with coronal mass ejections, J. Geophys. Res., 109, A12105, doi:10.1029/2004JA010602. Gosling, J. T. (1993), The solar flare myth, J. Geophys. Res., 98, 18,937 -18,949. Gosling, J. T., D. J. McComas, J. L. Phillips, and S. J. Bame (1991), Geomagnetic activity associated with Earth passage of interplanetary shock disturbances and coronal mass ejections, J. Geophys. Res., 96, 7831 -- 7839.
9 of 10
S06006
GLEISNER AND WATERMANN: SEPs AND GEOEFFECTIVENESS OF HALO CMEs
Kahler, S. W. (2001), The correlation between solar energetic particle peak intensities and speeds of coronal mass ejections: Effects of ambient particle intensities and energy spectra, J. Geophys. Res., 106, 20,947 -- 20,956. Kahler, S. W., and A. Vourlidas (2005), Fast coronal mass ejection environments and the production of solar energetic particle events, J. Geophys. Res., 110, A12S01, doi:10.1029/2005JA011073. Kahler, S. W., E. Hildner, and M. A. I. van Hollebeke (1978), Prompt solar proton events and coronal mass ejections, Sol. Phys., 57, 429 -443. Kim, R.-S., K.-S. Cho, Y.-J. Moon, Y.-H. Kim, Y. Yi, M. Dryer, S. Bong, and Y.-D. Park (2005), Forecast evaluation of the coronal mass ejection (CME) geoeffectiveness using halo CMEs from 1997 to 2003, J. Geophys. Res., 110, A11104, doi:10.1029/2005JA011218. Lindsay, G. M., J. G. Luhmann, C. T. Russell, and J. T. Gosling (1999), Relationships between coronal mass ejection speeds from coronagraph images and interplanetary characteristics of associated interplanetary coronal mass ejections, J. Geophys. Res., 104, 12,515 -12,524. Luhmann, J. (1997), CMEs and space weather, in Coronal Mass Ejections, Geophys. Monogr. Ser., vol. 99, edited by N. Crooker, J. Joselyn, and J. Fehnman, pp. 291 -- 299, AGU, Washington, D. C. Malandraki, O. E., D. Lario, L. J. Lanzerotti, E. T. Sarris, A. Geranios, and G. Tsiropoula (2005), October/November 2003 interplanetary coronal mass ejections: ACE/EPAM solar energetic particle observations, J. Geophys. Res., 110, A09S06, doi:10.1029/2004JA010926. Mayaud, P. N. (Ed.) (1980), Derivation, Meaning, and Use of Geomagnetic Indices, Geophys. Monogr. Ser., vol. 22, AGU, Washington, D. C. Reames, D. V. (1999), Particle acceleration at the Sun and in the heliosphere, Space Sci. Rev., 90, 413 -- 491. Richardson, I. G. (1997), Using energetic particles to probe the magnetic topology of ejecta, in Coronal Mass Ejections, Geophys. Monogr.
S06006
Ser., vol. 99, edited by N. Crooker, J. Joselyn, and J. Fehnman, pp. 189 -- 198, AGU, Washington, D. C. Smith, Z., W. Murtagh, and C. Smithtro (2004), Relationship between solar wind low-energy energetic ion enhancements and large geomagnetic storms, J. Geophys. Res., 109, A01110, doi:10.1029/ 2003JA010044. Srivastava, N., and P. Venkatakrishnan (2002), Relationship between CME speed and geomagnetic storm intensity, Geophys. Res. Lett., 29(9), 1287, doi:10.1029/2001GL013597. St. Cyr, O. C. (2005), The last word: The definition of halo coronal mass ejections, Eos Trans. AGU, 86(30), 281. St. Cyr, O. C., et al. (2000), Properties of coronal mass ejections: SOHO LASCO observations from January 1996 to June 1998, J. Geophys. Res., 105, 18,169 -- 18,186. Valtonen, E., T. Laitinen, and K. Huttunen-Heikinmaa (2005), Energetic particle signatures of geoeffective coronal mass ejections, Adv. Space Res., 36, 2295 -- 2302. Webb, D. F., E. W. Cliver, N. U. Crooker, O. C. St Cyr, and B. J. Thompson (2000), Relationship of halo coronal mass ejections, magnetic clouds, and magnetic storms, J. Geophys. Res., 105, 7491 -- 7508. Webb, D. F., N. U. Crooker, S. P. Plunkett, and O. C. St Cyr (2001), The solar sources of geoeffective structures, in Space Weather, Geophys. Monogr. Ser., vol. 125, edited by P. Song, H. Singer, and G. Siscoe, pp. 123 -- 141, AGU, Washington, D. C. Yurchyshyn, V., H. Wang, and V. Abramenko (2004), Correlation between speeds of coronal mass ejections and the intensity of geomagnetic storms, Space Weather, 2, S02001, doi:10.1029/ 2003SW000020.
H. Gleisner and J. Watermann, Geomagnetism and Space Physics Program, Danish Meteorological Institute, Lyngbyvej 100, Copenhagen DK-2100, Denmark. (
[email protected])
10 of 10
