Relationship between solar wind low-energy energetic ion enhancements and large geomagnetic storms

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109,, doi: /2003ja010044, 2004 Relationship between solar wind low-energy energetic ion enhancements and large geomagnetic storms Z. Smith and W. Murtagh Space Environment Center, National Oceanic and Atmospheric Administration, Boulder, Colorado, USA C. Smithtro 1 Center for Atmospheric and Space Sciences, Utah State University, Logan, Utah, USA Received 14 May 2003; revised 3 October 2003; accepted 27 October 2003; published 20 January [1] It is well established that energetic ion enhancements (an energetic ion enhancement will hereafter be referred to as an EIE) are partly due to acceleration by interplanetary shocks as the shocks propagate towards Earth and that arrivals of these shocks at Earth are well associated with geomagnetic storms. The observation of EIEs at satellites located at L1 is a potential tool for predicting the arrival of interplanetary shocks hours before they arrive at L1. Following an earlier study using WIND satellite data [Smith and Zwickl, 1999], we evaluate the potential of EIEs for forecasting geomagnetic storms during the period of February 1998 through December Since there are many more EIEs than large geomagnetic storms, additional associations that might improve the identification of precursors of large storms were investigated. These included probable solar sources, accompanying interplanetary shocks, and the shocks interplanetary drivers. Solar images and data from the Advanced Composition Explorer (ACE) and WIND satellites were used. The Potsdam Kp was used to specify geomagnetic storm severity. Almost all large geomagnetic storms (Kp 7) in this time period were preceded by EIEs that were associated with shocks driven by transient interplanetary disturbances. The converse is not true. Although most transient-associated EIEs were followed by some geomagnetic activity, there is a large span in the response. Most (95%) of EIEs with maximum flux pfu were followed by activity with Kp > 4 and 80% by storms with Kp 5. For a threshold of pfu, 67% of the large storms would be identified, 89% of the EIEs were followed by storms with Kp 5, and 53% by large storms (Kp 7). Using the additional information on the interplanetary drivers results in an increase in the correctly predicted events from 53% to 61%. For a threshold of pfu, all (100%) EIEs were followed by storms with Kp 5, 76% of the events were correctly predicted but 16 of the 30 large storms were missed. Most of the EIEs were followed by significant periods of southward Bz and in close to 70% of the EIEs, the Bz following the EIE was first northward before turning southward. We also investigated the relationship of the EIEs to halo or partial-halo coronal mass ejections (CMEs), and found that EIE events associated with halo CMEs are more likely to be followed by a large geomagnetic storm, but a lack of halo or partial-halo CME association does not preclude the occurrence of a large geomagnetic storm. INDEX TERMS: 2114 Interplanetary Physics: Energetic particles, heliospheric (7514); 2788 Magnetospheric Physics: Storms and substorms; 2139 Interplanetary Physics: Interplanetary shocks; 2102 Interplanetary Physics: Corotating streams; KEYWORDS: solar wind energetic ions, geomagnetic storms, predictions, space weather, interplanetary shocks Citation: Smith, Z. K., W. Murtagh, and C. Smithtro (2004), Relationship between solar wind low-energy energetic ion enhancements and large geomagnetic storms, J. Geophys. Res., 109,, doi: /2003ja Also at Air Force Institute of Technology, Wright-Patterson Air Force Base, Ohio, USA. Copyright 2004 by the American Geophysical Union /04/2003JA Introduction [2] The gradual EIEs observed in the solar wind have been well studied [Cane et al., 1988; Lario et al., 1998; Reames et al., 1996; Reames, 1999, and references therein]. In the lower-energy range (approximately kev), these EIEs result primarily from particle acceleration at discontinuities in the interplanetary (IP) medium, primarily in shocks. Both forward (F ) and reverse (R) interplanetary 1of9

2 shocks have been observed to accelerate particles but not always equally [Mason et al., 1999; Reames, 1999]. The shocks are driven either by transient solar ejections or by corotating plasma structures, as described by Gosling [1996]. We use the term transient rather than ICME (interplanetary CME), to focus attention to the distinction in time duration of the probable solar source. A transient may drive an F shock that may be followed by an R shock close behind the central part of the F shock. These transient-caused R shocks are observed less often (perhaps because they tend to be weaker and to have a much narrower longitudinal extent), and they have not been found to be as significant a source of accelerated ions as the preceding F shocks. [3] In corotating structures, however, the R shocks are often observed. These structures are formed from the highspeed solar wind that flows from coronal holes (a highspeed stream will hereafter be referred to as a HSS). As a HSS propagates outward, it interacts with the lower speed solar wind and the leading edge of the HSS steepens to form a corotating interaction region (hereafter CIR) with a fully developed forward-reverse shock structure. CIRs usually develop fully beyond 1 AU. Therefore CIRs as a source of energetic particles have been well studied primarily beyond 1 AU, using data from Ulysses [cf. Mason and Sanderson, 1999]. Nonetheless, CIRs are also found inside 1 AU [Schwenn, 1990; Balogh and Richardson, 1999], particularly at and near solar cycle minimum. In the time period of our study, we found twelve HSS/CIRs associated with the EIEs. In six cases with developed R shocks, significantly higher fluxes of EI were accelerated by the R shocks than by the F shock. This agrees with the findings of Intriligator and Siscoe [1994] for CIRs beyond about 4 AU. [4] Whether or not a shock accelerates particles, and how efficiently, is doubtless a function of the particles preacceleration energy relative to the strength and topology of the accelerating source. The theory of this particle acceleration has received much attention. Proposed processes envision acceleration to a critical energy after which the particles escape [cf. Intriligator et al., 1995; Classen et al., 1998; Li et al., 2003, and references therein]; they also consider the propagation paths of the escaped particles as they travel along field lines to the observing spacecraft [cf. Heras et al., 1995; Lario et al., 1998; Zank et al., 2000]. [5] The signature of a shock accelerated EIE observed at a spacecraft is known to depend on the location of the observing spacecraft relative to the energizing source. This is because the location of the point on the shock front that connects the spacecraft to the shock moves in a predictable way across the shock front as the shock moves outwards from the Sun and as the interplanetary magnetic field lines sweep past the particle detector. This point has been referred to as the COB (Connection to Observer) point [Lario et al., 1998]. For a shock initiated from the western hemisphere of the Sun, relative to the particle detector, this connection starts on the western flank of the shock, moves eastward through the central part of the shock (where the particle acceleration is expected to be strongest), and then progresses to the eastern flank before the shock either envelops, or passes beyond, the detector. If a shock is initiated from the central or eastern part of the Sun, the magnetic connection between the shock front and the detector occurs later and starts at a point closer to the center of the shock front before moving eastward. Therefore a detector will typically experience low-energy energetic ion profiles that depend on whether the accelerating shock was initiated from the west, central, or eastern directions. This effect was first interpreted by Heras et al. [1995], and further studied by Lario et al. [1998] and Reames al. [1999], using a numerical MHD model of the shock, showing the field lines and the flow of energized particles diffusing along field lines to variously placed observers. [6] It is also well established that shocks and EIEs correlate well with geomagnetic storms and that these storms in turn are strongly dependent on the magnitude and duration of the southward component of the IP magnetic field, Bz [cf. Tsurutani et al., 1988; Gonzalez et al., 1999; Russell and McPherron, 1973]. The strengths of geomagnetic storms are also known to be well correlated with increased solar wind dynamic pressure (and therefore velocity). However, storm-warning times obtained from IP plasma and magnetic field measurements made by spacecraft at L1 or just outside the Earth s bow shock are usually on the order of tens of minutes. The corresponding warning times provided by EIEs are on the order of hours to days. [7] The solar sources of IP disturbances have also been studied, using both disk features (to give locations in longitude as discussed above) and observations of CMEs. We searched for disk features indicating eruptive activity in H a, EUV, and X-ray data within a time period of about 1 5 days preceding the observations of each shock/eie. [8] In early CME studies, Sheeley et al. [1985] found that IP shocks and CMEs were very well correlated (only 2% of the IP shocks observed by Helios 1 during 1979 through 1982 clearly lacked CMEs). However, CMEs occur much more frequently than EIEs, IP shocks, or geomagnetic storms, and they are best observed for events directed ninety degrees from the Earth-Sun line. Therefore more recent studies focused on the wide-angle CMEs with an obvious front-side source as those more likely to be directed towards Earth. These wider CMEs are called halo or partial-halo (hereafter referred to as H/PH) CMEs. In a recent study restricted to H/PH CMEs, Cane et al. [2000] reviewed all H/PH CMEs observed by LASCO during and estimated that half of all front side H/PH CMEs intersected Earth. [9] In this paper, we assess whether EIE observations can be used to predict large geomagnetic storms, extending the work of Smith and Zwickl [1999], who used WIND data during the time period 1996 through The present study uses data primarily from ACE and geomagnetic storm intensity Kp. The concept of a threshold particle flux (the flux level to be exceeded before expecting strong geomagnetic activity) is introduced. We also review the solar sources which have the potential to greatly increase the prediction lead-time. [10] The data set is described in section 2, and the results are presented in section 3. The primary result is the relationship between the maximum flux of an EIE ( f max ) and the occurrence of a geomagnetic storm of magnitude of Kp 7. The associations of this primary relationship to other IP features are investigated in sections These features include: (1) the observation of an IP shock, (2) the southward component of the magnetic field, Bz, and (3) the drivers of these IP structures. The solar sources are discussed in section 3.6. The warning or delay times (times 2of9

3 from the particle flux crossing the threshold to the start time of the maximum Kp) are presented in section 3.7. The summary is presented in section Data [11] The ACE/Electron, Proton and Alpha Monitor (EPAM) kev energy channel particle data from February 1998 through December 2000 were searched for intervals containing EIEs. WIND data were used when ACE data were missing. Our data source was the Space Environment Center s archive of near-real-time data: the first EIE was observed at ACE in March This energy range was selected because it is above the normal solar wind energy range, but it is at the low end for energetic ions. We expect particles in this energy range to be mainly accelerated in the IP medium. Particles in the higher energy channels may come directly from the Sun without subsequent acceleration. Furthermore, both weak and strong shocks are expected to influence particle fluxes in our chosen range. [12] An EIE event is defined here as a gradual rise (excluding spikes- signatures of particles originating from the Earth s bow shock) in the flux to a level that exceeds 10 4 pfu (particle flux units, or particles cm 2 s 1 sterad 1 MeV 1 ). The rise portion of the enhancement to maximum flux is important because the rise gives warning lead-time; also, the maximum particle flux in an EIE turns out to be related to the intensity (Kp) of the subsequent geomagnetic storm. [13] Figure 1 shows two EIEs in ACE/EPAM data. The upper panel is typical of an EIE in a relatively uniform solar wind. As a shock approaches ACE, fluxes in all the energy channels rise together. The lower panel shows an atypical EIE with large fluctuations in the fluxes, making prediction of the arrival of an IP shock difficult. This is a unique event in our data set, but it illustrates dramatically the oscillations that can occur on a much smaller scale. [14] For each EIE, we reviewed the solar wind ACE (or WIND) plasma and field data together with the energetic ion data and the subsequent Kp index. We used 5-min averages for the energetic ion data, 1-min averages for the plasma and field, and 3-hour intervals for Kp. To take into account both magnitude and duration in estimating minimum Bz, 1 hour averaged data (in GSM coordinates) were used. An example of the data prepared for each EIE is shown in Figure 2. The features explained below, and their associations, are indicated in Figure 2. Table 1 lists only the larger particle events (those with f max pfu) and all intense geomagnetic events (Kp 7) regardless of f max EIEs [15] For each EIE, we list the date, time, and value of the peak flux ( f max ), and rise-time, T r, to the peak. T r is defined as the number of hours after the energetic ion flux crossed the threshold of 10 4 pfu before f max is reached. In the cases where more than one peak is seen (because of the local fluctuations in ion flux), f max is chosen as that closest to the shock arrival Shocks [16] The existence and arrival times of shocks were identified in ACE or WIND plasma and field data (denoted Figure 1. Two examples of an energetic ion enhancement (EIE): (a) a typical case and (b) a case difficult to forecast because of the oscillations in the particle flux. in Table 1 by A or W, respectively). WIND data were used when ACE data were missing or unclear. Fast reverse shocks, slow forward shocks and small shocks were included; they do not appear in Table 1 only because they are associated with weaker events (lower f max and Kp). If no shock was found, the entry is left blank Storm Sudden Commencements (SSC) [17] The day and time of each associated storm sudden commencement (SSC) is listed in Table 1. These were obtained from National Geophysical Data Center (NGDC) and Potsdam web pages (available at Kp max [18] The maximum Kp associated with each EIE was also obtained from the Potsdam and NGDC web pages. The association period was up to two days after the EIE maximum, as long as the IP structure appeared to be that in which the EIE occurred. For each event, the Kp max value, the day and time of the start of the three-hour interval of Kp max are listed. If there are more than one intervals of equal Kp max, the start time of the first interval was listed IP Drivers [19] Examination of the solar wind data gives information on the nature of the solar cause of the IP structure, whether due to an individual energetic event on the Sun (a transient) or HSS. For transients, we use the term driver rather than ICME, to focus on that part of the ICME that provides the energy to sustain the observed shock. Analogously, the high-speed plasma of a HSS can be thought of as 3of9

4 Figure 2. Example of an EIE and the accompanying solar wind parameters. the driver of a CIR. Because the transition from HSS to CIR usually occurs beyond 1 AU and because in our data set we found that the stronger EIE-producing structures were the not yet fully developed CIRs, we use HSS to refer to this category. We classified each event according to the IP structure into the following three categories: Category (1) when there appeared to be an individual energetic event on the Sun (a transient); Category (2) when a HSS appeared to be responsible; and Category (3) when we were uncertain about the driver s nature. We identified drivers for all of the events in Table 1: the weaker events, not listed, often had Category 3 drivers. Only three of the IP drivers associated with the strong events of Table 1 are HSSs; the majority of drivers are transients. [20] Other features considered for the potential precursors of large storms included magnetic structures, large abrupt density enhancements and current sheet/sector crossings. However, as discussed by Wimmer-Schweingruber et al. [1999], these features are typically components of the HSS/ CIR structures. HSS/CIRs are readily identified in the solar wind plasma and, in our study period, were never followed by strong geomagnetic activity (Kp > 6). This will be discussed in sections 3.3 and Bz [21] The value of minimum Bz (Bz min = largest value of Bz-south, measured in GSM coordinates) associated with each EIE was obtained from the one hour averaged Bz data, in order to take into account both duration and magnitude of Bz. For Category 1 (transient) events, Bz min occurs in the time between the shock arrival and the end of the interval of Kp max. For HSSs however, Kp max is often found within the CIR structure as it passes over Earth, after the leading F shock, and preceding the EIE maximum and R shock. Therefore our analysis of EIEs associations with Bz min is presented in section 3.5 only for transient events (Category 1) and ignores the three HSS-driven events in Table 1. [22] We also investigated whether, in the interplanetary structure following the EIE and shock, the polarity of the magnetic field remained northward or southward or if it reversed at least once. If the polarity reversed at least once, then the leading polarity was used. One minute averaged data were used to study these polarities Solar Source [23] We attempted to identify the probable solar sources of each transient event, when possible, in both coronagraphic and other solar observations. For coronal mass ejections (CMEs), halo or partial halo (H/PH) information used was obtained from the SOHO/LASCO experiment team preliminary list of H/PH CMEs (defined by Plunkett [2001]) on Solar sources of EIEs were sought in a time interval from approximately one to five days before the arrival of the EIE flux maximum at 1 AU. Reports of H/PH CMEs in this time-interval are noted in Table 1, together with the inferred latitude and central meridian distance of its chromospheric origin. When SOHO/LASCO data were either unavailable, or no H/PH was observed, or no disk-association was given in the LASCO reports, then sources were sought in the NOAA Space Environment Center (SEC) records of optical, image, X-ray-flare, and radio data. An event where no association could be made with confidence is given as na (no association) in the event column and blanks in the date and location columns. Since the H/PH manifestation is seen at high altitudes in the solar atmosphere and other visible manifestations may be seen lower (earlier), multiple, different onset times can be reported; however, the study of the detailed timing is not within the scope of this paper. [24] The full data set contains 115 events. Since each event is unique and the flux often fluctuates around the threshold, our selection was biased towards being inclusive rather than exclusive, particularly in the case of overlapping events (discussed in section 3.4). 3. Results 3.1. Relationship Between f max and Kp [25] In our study period, we identified 115 EIE events with f max pfu; for 38 of these events, f max reaches at least pfu. The data set contains 30 large storms (Kp 7, classified as strong, severe, or extreme in the NOAA Space Weather scales). In Table 1, we show the 38 events with f max pfu plus the 10 events with lower f max that are followed by large storms. All of the large storms in the time period of this study were accompanied by EIEs with f max 10 4 pfu. However, for one event, the rise in EI flux did not follow the patterns described here for an EIE (too abrupt a jump in flux to maximum). Therefore all but one of the large storms could have been predicted by this relationship. [26] The relationship between Kp and f max for the complete data set is displayed in Figure 3. This plot shows that in general, the geomagnetic storms with higher Kp follow stronger EIEs (larger f max ). All but three of the 30 large geomagnetic storms with Kp 7 had f max pfu. 4of9

5 Table 1. Events List Number a b EIE Shock c SSC d Kp (max) e IP f Bz.m g Solar Source h Y M D UT Tr *f max D UT s/c sh D UT D UT max Dr nt Event D UT Location Halo i na nh A F X S17E42 H A F M S17E18 H A F X S15W15 H A F na W60 PH A F na nd A F X1/3B N31E09 nd A F M6/3B N20E07 nd A F M3/2N N23 W58 nd A F EIT/DF N19E10 H A F M8/2B N18W21 H na nd A F M3/SF S22W23 nd A F EIT W PH A F M N18E24 H A F DF N29W13 PH A f na N20W72 H A F na nh W F C N21E05 PH A F na H A F na PH A F na nd A F na nh A F C7/DF S17W40 H A F C9/2F N18W58 H na nh C7/1N S15W08 nh A F X2/3B N21E15 H A F C6/1N N22W18 nh A F na PH A F X1/2N N18E27 H B F X2/2B N17E27 nh B F X5/3B N17E00 H A F M2/1N S11E36 nd A F EIT N11W09 H A F DSF N26W37 PH A F M1/2N S17W09 H A F C N14E01 H A F EIT/C S07E05 H A F EIT S07E05 H A F C6/1F N01W18 H A F na nd A F EIT S1 E39 H A F na nh B F M7/EIT N10W73 H A F na nd A F na N21E06 H A F X N21W30 H a First column is the event number in the original set of 115 events. Here we list only the events with f max pfu, orkp 7. The bold event numbers indicate the large storms (Kp 7) with f max < pfu. b EIE is the Energetic Ion Enhancement, the time of the maximum flux in the ACE/EPAM 47 65keV energy channel; Y, M, D, UT are the year, month, day, and time of the maximum of the EIE event. Tr is the rise-time (in hours), from the time the flux crosses the threshold of pfu to the maximum; *fmax is the maximum flux of the EIE event, in pfu c D and UT are the day and time at which the interplanetary shock was observed at the L1 (blank if no shock). The notation s/c is the spacecraft used; A and W denote the ACE/Wind for the spacecraft, B denotes that this time is obtained from the ACE/MAG instrument when the SWEPAM data were not available, sh is shock type: F is fast forward shock. d D and UT: day and time of the Storm Sudden Commencement, if one reported, otherwise blank. e D, UT: day and time of the start of the 3 hour interval,and the maximum Kp value observed. f IP dr is Interplanetary driver: these are classified as Transient (1) and HSS (2). g Bz.m is minimum value of Bz in GSM coordinates, from the 1-hour averages of the ACE/MAG data. h Event classification: from LASCO reports, when available, otherwise from SEC records. Here na = no association given because either none was reported, or there were several candidates. D, UT, and location are day, time, and location of the solar event (if one listed, otherwise blank). i H = halo, PH = partial-halo, nh = no halo nor partial-halo observed, nd = no LASCO observations available. Using this low threshold results in both a high (90%) hit rate and high (66%) false alarm rate. If the threshold were set to f max 10 5 pfu, then six large storms would have been missed (80% hit rate) but the corresponding false alarm rate is lower (52%). Therefore we seek other parameters that can help reduce the false alarm rate. The events in Figure 3 are sorted on interplanetary driver (whether a HSS or transient). This will be discussed in section 3.3. [27] This data set was compiled based on energetic ion data. We note that all large storms that occurred during the 5of9

6 Figure 3. The f max versus Kp, showing the category of the driver of the IP disturbance. 35-month study period are included in this data set because they followed EIEs. Moreover, all but 17 of the 115 EIEs with f max pfu in this period had associated geomagnetic enhancements of Kp 4. This is of particular interest for predictive purposes because, if the flux reaches higher values, then the probability of a large storm ensuing increases (e.g., for f max pfu, all EIEs were followed by storms with Kp 5 and 67% by large storms). [28] This study focuses on large storms. Having established the primary relationship between f max and Kp, we now examine other relationships that might improve the identification of precursors of the large storms in order to reduce the false alarm rate. The two most useful characteristics that can provide longer lead times are the IP driver (whether HSS or transient) and whether individual, or overlapping events Relationship Between f max, Shocks, and Storm Sudden Commencements [29] This study confirms the well-known association between EIEs and shocks. In the time interval studied, over 90% of EIEs with f max > pfu were accompanied by shocks at the spacecraft and SSCs at Earth. Only four of the 30 large (Kp 7) geomagnetic storms lack associated shocks and five lack SSCs (in one case because it immediately followed a smaller storm). The stronger EIEs that were not associated with clear shocks were usually accompanied by solar wind discontinuities with large density enhancements (>30 cm 3 ) and two were followed by SSCs. For the 36 events with 10 4 < f max <510 4 pfu, the association of EIEs with IP shocks is approximately 55%. Therefore although the EIEs that precede large geomagnetic storms usually have shocks at or near their maxima, the absence of a shock observation does not preclude the occurrence of a large geomagnetic storm. [30] It is important to keep in mind that the accelerating source (IP shock) of an EIE is usually at some distance from the particle sensor and may not ever pass over the observing spacecraft. The same is true of the drivers of the IP shocks; that is, if the driver does not actually encounter the spacecraft, then the driver s nature may not be readily identified. [31] The shocks are mainly F shocks but, of the 92 shocks identified, six were R shocks. Two of these six were part of a F-R shock pair; such an F-R shock pair is the signature of a HSS having developed into a CIR. We confirm that if present, the reverse shock of a HSS dominates over the F shock as a source of energetic ions, as noted in previous studies [cf. Intriligator et al., 1995; Decker et al., 1999]. [32] Detailed MHD properties of shocks that accelerate particles have been extensively studied for transients and for CIRs [cf. Classen et al., 1998; Li et al., 2003]. Although knowledge of these detailed shock properties is not available to a forecaster in real time, they may be fundamental to the understanding of the physical processes upon which this EIE-forecasting technique is based Interplanetary Drivers [33] Whether or not an EIE is likely to be followed by a large geomagnetic storm appears to be determined by the nature of the driver of its source, whether HSS or transient. This is seen in Figure 3, where the relationship of Kp to f max for the complete data set is shown, sorted on driver category. All but one of the Kp 7 events are clearly associated with transients. Note that, however large f max,the EIEs due to HSSs in this time interval were all followed by storms with Kp 6. Therefore if the IP structure is identified as a HSS, a Kp 6 may be expected, even if the EIE is F shock-driven. HSSs have the advantage of being identified in the IP data before or at the time of the EIE maximum. However, this may be a function of solar cycle as this data set covers the rise and maximum of the solar cycle, when transients dominate. [34] The category of events for which the plasma and field data showed no clear drivers and usually no (or weak) shocks, was found to contain mainly events with low f max and low Kp (Kp 4). Therefore this investigation of the IP drivers focuses on the larger events Overlapping Events [35] In most EIEs, the EI flux rises above threshold to a maximum that is accompanied by one shock and one SSC. However, there are times when multiple EIEs occur during a period of elevated EI flux. In these events, referred to as overlapping events, the EI flux does not return back to below the threshold between successive EIE maxima but continues to increase or decreases slowly. Each EIE in an overlapping series is listed as a separate event in Table 1. Usually a smaller EIE, sometimes accompanied by a separate shock, SSC, and even separate Kp maximum, occurs during the rise (or decline) of a larger EIE. In the case where the smaller EIE follows a larger one, the EI flux profile decreases slowly and has the appearance of a plateau. Note, however, that a plateau, or broad maximum, is also the characteristic signature of single event originating from the center of the solar disk. These events can usually be distinguished by simultaneously reviewing the solar wind plasma and field data to determine whether or not compound (or multiple) events are seen there. If compound or multiple events are seen in the plasma and field data, then overlapping events are to be expected. [36] Overlapping events are included here because they form a subset of EIEs and because the expected Kp of the ensuing storms differ from the case of only one EIE. They originate from solar events that are closely spaced in time so the resulting IP disturbances may interact en route to 1 AU. There were nine sets of overlapping EIEs in this data set. During the interval June 1999, there were four EIE 6of9

7 events, three of which had separate shocks and SSCs. In general, the following EIEs (the second, third etc. of such a series) tend to be followed by storms with Kp values that are the same or higher than those of the preceding event (and the same or higher than would be expected from an individual EIE) even if the f max is significantly lower Role of Bz [37] The severity of geomagnetic storms is known to correlate well with both the magnitude and the duration of the southward component of Bz (negative Bz). Therefore the distribution of the minimum Bz (Bz min = largest value of hourly averaged Bz-south) within transient events was considered to check whether these values could assist in identifying events that are followed by large storms. We obtained the value of Bz min encountered between the EIE maximum (or shock) and the subsequent interval of maximum Kp. As noted in section 2, only transient events are considered here. [38] This distribution is shown in Figure 4, sorted by f max, for the 55 transient-associated EIEs with f max 10 5 pfu and the large storms with f max <10 5 pfu. A number of points overlap, so the Kp values are displaced to show the three f max categories. This plot shows the well-known relationship between Bz min and Kp. All but one of the large geomagnetic storms have at least one 1-hour interval with averaged Bz 5 nt. The exception, when viewed in the 1 min resolution data, shows an interval of more than 1/2 hour with Bz 10 nt and Bz reaches a minimum value of close to 20 nt. These results therefore differ from those of Tsurutani et al. [1988] in that a 3 hour interval of Bz < 10 nt is not required for a large storm. However, those authors examined only the large storms as measured in Dst, in a different time period ( ) and we use simple hourly averages. [39] Figure 4 shows that for this time period, EIE events with at least one 1-hour average of Bz 10 nt are likely to be followed by a storm (100% in this data set were followed by storms with Kp 5, and 77% with Kp 7). Events with Bz min < 20 nt were all followed by storms with Kp 7. Events with less strong southward fields (Bz min > 10 nt) are also likely to be followed by geomagnetic activity but with a wider range of Kp (the majority fall between 4 and 8). This plot also suggests that there exists, for each level of Bz south, a minimum Kp level that has a high probability of being reached or exceeded for transient events. [40] The majority (4 of the 6) of large storms missed by using a threshold of f max 10 5 pfu, were accompanied by Bz 10 nt for at least 1 hour. Of the subset of all EIE events with both f max 10 5 pfu and hour-averaged Bz 10 nt, only 27% were followed by activity with Kp 6. This number would be reduced to 17% by removing the two events associated with HSSs. [41] Figure 4 also shows that almost all the EIEs were followed by periods of southward Bz, in the solar wind structure related to the EIE. Therefore the polarities were investigated: whether the 1 hr averaged Bz was initially southward at the shock/eie and turned northward, changed from north to south, or did not change direction. The magnetic field was usually (following more than 75% of the EIEs) found to contain a mix of polarities, with southward Bz for at least part of the time. There were, Figure 4. Bz min versus Kp, sorted on f max. however, significantly more structures led by northward fields. The largest storms (those with Kp 8) mainly followed structures where the field was initially northward. The larger number of structures led by northward fields in this time period may be due to a solar cycle dependence, such as that reported from observations of solar wind interaction regions [Rosenberg and Coleman, 1980], from solar flare observations [Garcia, 1990]; or from a change in the solar wind QI index described in the study of magnetic clouds by Osherovich et al. [1999]. It may also be related to the north south asymmetries described by Bothmer and Schwenn [1998] who found a solar cycle dependence in their study of the structure of magnetic clouds. Our studies however use different definitions. The study of Bothmer and Schwenn considers the rotations only of the cloud itself, whereas we focus on the magnetic field changes in the entire solar wind structure from the shock/eie through the driver Solar Sources [42] Identification of the solar origins of the IP disturbances that cause EIEs would provide significantly longer warning times of EIE occurrences. Furthermore, knowing the longitude of the solar source is helpful in interpreting the signatures of the EIEs caused by transient events. As transients were the main cause of large geomagnetic storms in this time period, only transient-associated EIEs (including overlapping events) with f max pfu are considered in this Section, along with the six events with lower f max but Kp 7. As previously mentioned, western sources, where the observer is magnetically well connected to the IP shock soon after its formation, are expected to give longer EIE rise times. Eastern sources do not become magnetically connected until the shock is close to the observer; consequently, the EIE rise is expected to be much steeper and shorter. [43] Associations with solar sources were made for 42 events (32 of which appear in Table 1) and, as expected, more events originated from central locations (±20 of central meridian) than west of W20 and the fewest from sites east of E40. Although the expected relationship between rise times and heliolongitude is generally seen in this data set, there is substantial scatter in the distribution. This is due to several factors, including our use of a threshold flux level, overlapping events, the effect of inhomogeneous background solar wind through which each IP disturbance travels [Heinemann, 2002], the properties of the IP shock, and the difficulty inherent in associating a 7of9

8 Table 2. CME EIE Associations With Large Storms a Number events with large storms Number of events without large storms Percent of the 25 large storms preceded by H, PH, nh: Percent of the 11 storms with Kp 8 preceded by H, PH, nh: H PH nh % 12% 20% 82% 9% 9% a Large storm is one with Kp 7 ( strong-extreme in the NOAA Space Weather scales). CME data was available for 45 of the EIE events, 25 of these were associated with large (Kp 7) storms. H = halo, PH = partialhalo, nh = neither H nor PH. single solar event with each EIE and IP shock. The use of a threshold flux level to determine lead time means that only a part of the actual rise time is used, and this discriminates against the weaker (lower f max ) events. Overlapping events (defined in section 3.4) can produce an EIE signature that resembles one long rise, emulating a western origin. The varying background level of the energetic ion flux in the kev energy channel affects the rise time because the shock-accelerated particles have to rise above background before being considered. The detailed properties of the section of the IP shock that is magnetically connected to the observer influence the injection of energetic ions [Lario et al., 1998]. For these reasons, substantial scatter is expected in the distribution of rise-times versus heliolongitude. [44] We also investigated the associations of EIEs with halo/partial halo (H/PH) CMEs. During this study period, SOHO/LASCO data were available for 45 of the EIE events with both f max pfu and transient drivers and for the six additional EIEs with lower f max that were followed by large (Kp 7) storms. Twenty-five of these 45 EIEs were associated with large storms. These 45 EIEs were divided into categories according to whether they were followed by large storms and whether they were preceded by halo or partial halo CMEs, or neither, labeled nh. The statistics are shown in Table 2. The first two rows give the number of EIE events associated with large storms in each of the H, PH, nh categories. The next two rows give the percentage of the total number of storms that fall in each of the CME categories. This table shows that for this time period, more than half (57%) of the EIEs preceded by halo CMEs were also followed by large storms. However, the association is not good for partial halo CMEs. From the point of view of the large storms, 68% were preceded by both an EIE and halo CME. For storms with Kp 8, the percentage rises to 82%. [45] Using EIEs alone, the corresponding figures for Kp 7 storms is 80% (24 of the 30 large storms were preceded by EIEs with f max pfu). Of the six large storms that were missed by setting the EIE f max threshold at f max 10 5 pfu), two were preceded by Hs, one by a PH and three had no H/PH associations (nh). [46] These statistics, based on a relatively small sample of H/PH CMEs that occur in association with the larger EIEs, show similar associations with geomagnetic storms to those found by Cane et al. [2000]. However, this data set indicates that halo CMEs are more important than partial halo CMEs. In general, EIE events associated with halo CMEs are more likely to be followed by a large geomagnetic storm than those with no halo CME associations, but a lack of an associated H/PH CME does not preclude the occurrence of a large geomagnetic storm Delay Times [47] The delay time, T d, is defined as the time interval between the energetic ion flux crossing the threshold value of 10 4 pfu and the start time of the 3-hour interval of maximum Kp (Kp max ). The delay times are analyzed only for the EIE events with associated transient drivers, including overlapping EIEs. [48] The 84 EIEs associated with clear transient drivers were sorted according to T d and f max. The resulting distributions are shown in Figure 5. The bottom (black) histogram shows the distribution of T d for all events with f max 10 6 pfu. Similarly, the other f max ranges are stacked on each other, so that the outer envelope of the histogram gives the distribution for all 84 events. [49] Delay times range from 0 to 48 hours. The average T d is about 15 hours. These 84 events include all but one of the large (Kp 7) storms. All of the 10 events for which T d < 0.5 hour are followed by geomagnetic activity with Kp < 4. Figure 5 shows that about half of the events (36 of 84) fall in the f max range of pfu. Although the weaker EIE events tend to have shorter delay times, events with f max pfu are more evenly distributed about the average. 4. Summary [50] We examined all EIEs, defined as a gradual rise in flux in the ACE/EPAM kev energy channel to flux levels 10 4 pfu, from February 1998 through December 2000, in order to investigate their relationship to large geomagnetic storms (storms with Kp 7, strong to severe in the NOAA Space Weather Scales). In general, the larger the maximum flux ( f max ), the greater the likelihood that a large geomagnetic storm will follow. All of the 30 large geomagnetic storms observed in this time period were associated with ion flux maxima f max >10 4 pfu; however, one EIE did not follow the pattern described here that gives warning of an ensuing storm. Almost all (90%) large geomagnetic storms (Kp 7) were preceded by EIEs with maximum fluxes pfu: however, 66% of these EIE events were followed by weaker geomagnetic activity. For EIE events with maximum fluxes exceeding pfu, 67% Figure 5. Delay times for several f max thresholds for transient events. 8of9

9 of the large storms would be identified, 89% of the EIEs were followed by storms with Kp 5, and 47% by levels of activity with Kp <7. [51] In an effort to identify other solar wind signatures that would help select the EIEs likely to be followed by large storms, a number of interplanetary (IP) features were examined heuristically. The nature of the IP driver, whether transient or high speed stream (HSS), was found to be the most useful characteristic. Using this additional information on the interplanetary drivers lowered the false alarm rates for these two thresholds from 66% and 47%, to 53% and 39%, respectively. The other useful characteristic of the EIEs is whether is whether or not it is part of an overlapping series. [52] It is of interest to note that when EIEs were associated with high-speed streams with a developed F-R shock structure, it was the R shock that accelerated the energetic ions. In this time period, the geomagnetic activity associated with HSS/CIRs was always of levels Kp max 6. [53] The larger EIEs correlate very well with IP shocks and SSCs and are almost always accompanied (or driven) by IP structures with periods of negative Bz). However, except for very strong southward Bz values (Bz min 20 nt), which were all followed by large storms, a low Bz was not found to be a guarantee of an ensuing large storm, nor can a moderate negative Bz predict with certainty that a large storm will not follow. Note that these findings are based only on 1-hour averaged data. [54] Most of the IP structures associated with EIEs contain intervals with both north and south components of the magnetic field but, in this time period, there were significantly more structures led by northward fields than were led by southward fields. These findings could well be dependent on phase and nature of the solar cycle. [55] The associations of EIEs with halo/partial-halo CMEs were also investigated. This study suggests that the halo CME association is better than that of a partial-halo and, in general, EIEs associated with halo CMEs are more likely to be followed by a large geomagnetic storm than those with no halo CME associations. However, lack of an associated halo CME does not preclude the occurrence of a large geomagnetic storm. [56] Note that this study period covers the rise and maximum of the solar cycle, when transients dominate. Whether all the results will hold during the decay phase and minimum remains to be checked. [57] Acknowledgments. We thank H. A. Garcia, R. Zwickl, E. Hildner, T. Onsager, and H. Singer for their assistance in reviewing this paper and useful discussions. We thank the ACE and WIND instrument teams for making this data available to the community; in particular D. J. McComas and C.W. Smith for the ACE plasma and field data, R. Gold and S. M. Krimigis for the ACE energetic particle data, and K. W. Ogilvie, R. P. Lepping, and R.P. Lin for the WIND plasma, field, and energetic particle data. We also thank the GeoForschungsZentrum, Potsdam, and the National Geophysical Data Center for the geomagnetic indices. This work was partially funded by the NASA Living With a Star (LWS) Targeted Research and Technology program through NOAA Work Order W-10,118. [58] Arthur Richmond thanks Gary P. Zank and another reviewer for their assistance in evaluating this paper. References Balogh, A., and I. G. Richardson (1999), CIRs in the inner heliosphere: A summary from Helios, Space Sci. Rev., 89, Bothmer, V., and R. Schwenn (1998), The structure and origin of magnetic clouds in the solar wind, Ann. Geophys., 16, Cane, H., D. Reames, and T. T. von Rosenvinge (1988), The role of interplanetary shocks in the longitudinal distribution of solar energetic particles, J. Geophys. Res., 93, Cane, H. V., I. G. Richardson, and O. C. St. Cyr (2000), Coronal mass ejections, interplanetary ejecta and geomagnetic storms, Geophys. Res. Lett., 27, Classen, H. T., G. Mann, and E. Keppler (1998), Particle acceleration efficiency and MHD characteristics of CIR-related shocks, Astron. Astrophys., 335, Decker, R. B., M. I. Desai, and G. M. Simnett (1999), Origins, injection, and acceleration of CIR particles: Spectra: Temporal and radial evolution, Space Sci. Rev., 89, Garcia, H. A. (1990), Evidence for solar cycle evolution of NS flare asymmetry during solar cycles 20 and 21, Solar Phys., 127, Gonzalez, W. D., B. T. Tsurutani, and A. Clua de Gonzalez (1999), Interplanetary origin of geomagnetic storms, Space Sci. Rev., 88, Gosling, J. T. (1996), Corotating and transient solar wind flows in three dimensions, Annu. Rev. Astron. Astrophys., 34, Heinemann, M. (2002), Effects of solar wind inhomogeneities on transit times of interplanetary shock waves, J. Atmos. Sol. Terr. Phys., 64, Heras, A. M., B. Sanahuja, D. Lario, Z. K. Smith, T. R. Detman, and M. Dryer (1995), Three low energy particle events: Modeling the influence of the parent interplanetary shock, Astrophys. J., 445, Intriligator, D. S., and G. L. Siscoe (1994), Stream interfaces and energetic ions closer than expected: Analyses of Pioneers 10 and 11 observations, J. Geophys. Res., 21, Intriligator, D. S., G. L. Siscoe, G. Wibberenz, H. Kunow, and J. T. Gosling (1995), Stream interfaces and energetic ions II: ULYSSES test of Pioneer results, J. Geophys. Res., 22, Kahler, S. W. (1987), Coronal mass ejections, Rev. Geophys., 25, Lario, D., B. Sanahuja, and A. M. Heras (1998), Energetic particle events; Efficiency of interplanetary shocks as 50 kev < E < 100 MeV proton accelerators, Astrophys. J., 509, Li, G., G. P. Zank, and W. K. Rice (2003), Energetic particle acceleration and transport at coronal mass ejection-driven shocks, J. Geophys. Res., 108(A2), 1082, doi: /2002ja Mason, G. M., and T. R. Sanderson (1999), CIR associated energetic particles in the inner and middle heliosphere, Space Sci. Rev., 89, Osherovich, V., J. Fainberg, and R. G. Stone (1999), Solar wind quasiinvariant as a new index of solar activity, Geophys. Res. Lett., 26, Plunkett, S. (2001), Halo CMEs in the rise phase of solar cycle 23, Eos. Trans. AGU, 82(47), Fall Meet. Suppl., Abstract SH12B Reames, D. V. (1999), Particle acceleration at the Sun and in the heliosphere, Space Sci. Rev., 90, Reames, D., L. M. Barbier, and C. K. Ng (1996), The spatial distribution of particles accelerated by coronal mass ejection-driven shocks, Astrophys. J., 466, Rosenberg, L. R., and P. J. Coleman Jr. (1980), Solar cycle dependent north-south field configurations observed in solar wind interacting regions, J. Geophys. Res., 85, Russell, C. T., and R. L. McPherron (1973), Semiannual variations in geomagnetic activity, J. Geophys. Res., 78, 92. Schwenn, R. (1990), Large scale structure of the interplanetary medium, in Physics and Chemistry of Space, vol. 20, edited by R. Schwenn and E. Marsch, pp , Springer-Verlag, New York. Sheeley, N. R., Jr., R. A. Howard, M. J. Koomen, D. J. Michels, R. Schwenn, K. H. Mulhauser, and H. Rosenbauer (1985), Coronal Mass Ejections and Interplanetary shocks, J. Geophys. Res., 90, Smith, Z., and R. Zwickl (1999), Forecasting geomagnetic storms using energetic particle enhancemements, in Solar Wind 9, edited by S. R. Habbal et al., AIP Conf. Proc., 471, pp Tsurutani, B. T., W. D. Gonzalez, F. Tang, S. I. Akasofu, and E. J. Smith (1988), Origins of interplanetary southward magnetic fields responsible for major magnetic storms near solar maximum ( ), J. Geophys. Res., 93, Wimmer-Schweingruber, R. F., R. J. Forsyth, and N. U. Crooker (1999), CIR morphology, turbulence, discontinuities and energetic particles: Morphological structures, Space Sci. Rev., 89, Zank, G. P., W. K. M. Rice, and C. C. Wu (2000), Particle acceleration and coronal mass ejection driven shocks: A theoretical model, J. Geophys. Res., 105, 25,079 25,095. W. Murtagh and Z. Smith, Space Environment Center, National Oceanic and Atmospheric Administration, 325 Broadway, Boulder, CO 80305, USA. (william.murtagh@noaa.gov; zdenka.smith@noaa.gov) C. Smithtro, Center for Atmospheric and Space Sciences, Utah State University, Logan, UT , USA. (smithtro@cc.usu.edu) 9of9