A Test of Real-Time Prediction of Magnetic Cloud Topology and Geomagnetic Storm Occurrence from Solar Signatures

A. H. McAllister¹, S. F. Martin², N. U. Crooker³,
R. P. Lepping⁴, and R. J. Fitzenreiter⁴

¹ Helio Research, Boulder, Colorado. Also at High Altitude Observatory, Boulder, Colorado.
² Helio Research, La Crescenta, California.
³Center for Space Physics, Boston University, Boston, Massachusetts.
⁴Goddard Space Flight Center, Greenbelt, Maryland.

Abstract

Introduction

Coronal Mass Ejections (CMEs) are large eruptions of coronal magnetic fields and plasma that are directly observable in white light coronagraphs [ Hundhausen, 1997]. CMEs may form into flux ropes as they develop [e.g., Low and Smith , 1993; Chen , 1996] and/or as they leave the Sun [ Gosling , 1990]. Roughly 30% of CMEs identified in the solar wind show the large-scale field rotations characteristic of flux ropes [ Gosling , 1990]. Interplanetary CMEs with clear field rotations coincident with low temperature and strong magnetic field are called ``magnetic clouds" [ Burlaga , 1991]. Many more cases, without clear field rotations, also can be fit with oblique passages through a flux rope model [ Marubashi , 1997].

Earth-directed CMEs are not yet directly visible against the solar disk, but over the last four decades efforts have been made to identify their signatures as the solar sources of geomagnetic storms. Strong correlations have been found with strong active region flares [ Dodson and Hedeman , 1964], Ha filament eruptions [ Joselyn and McIntosh , 1981], and the less impulsive, less energetic flares called ``long duration X-ray events" (LDEs) [ Sheeley et al. , 1975; Webb et al. , 1976]. With the advent of the high-energy coronal images from the Yohkoh Soft X-ray Telescope (SXT) came the ability to observe much fainter LDEs, seen as coronal X-ray arcades [e.g., McAllister et al. , 1996a, b]. The SXT images, which, owing also to their regular cadence, include a large number of events, have allowed the retrospective identification of the solar signatures of many specific storms [ Lemen et al. , 1996; McAllister et al. , 1996b; Weiss et al. , 1996; McAllister and Crooker , 1997].

With the launch of the Solar and Heliospheric Observatory (SOHO) sattelite, SXT has been joined by Extreme Ultraviolet Imaging Telescope (EIT), which can also observe coronal arcade formation and, often, the eruption of prominences, as well, and by Large Angle Spectroscopic Coronagraph (LASCO), with wide-field coronagraphs that can detect many Earth-directed or ``halo" CMEs, so named because of the hazy ring that encircles the occulting disk as their angular extent increases away from the Sun [ Howard et al. , 1982; Brueckner et al. , 1998]. While there are still some cases with almost no solar signature below the outer white light corona [e.g., Webb et al. , 1998], the ability to recognize signatures associated with CME launches on the disk has increased greatly.

At the same time, the ability to predict topological properties of magnetic clouds from solar signatures has also increased greatly. Marubashi [1986, 1997] first demonstrated that the elevation angle (angle of the filament axis relative to the ecliptic) of a flux rope axis fit to a cloud correlates well with the tilt of the associated erupting filament (relative to the solar equator). Rust [1994] and Bothmer and Schwenn [1994] demonstrated that the chirality of a cloud correlates with the solar hemisphere of its origin. By identifying the cloud flux rope as the filament itself, Rust [1994] attributed the correlation to the fact that filament chirality patterns on the Sun are statistically sorted by hemisphere [ Martin et al. , 1994] and that these patterns imply that filaments are flux ropes with left-handed chirality in the Northern Hemisphere and right-handed in the Southern Hemisphere. Later Martin and McAllister [1997, 1999] argued that if filaments are flux ropes at all, their chirality is opposite to that assumed by Rust [1994], that magnetic clouds are composed of the much larger-scale coronal arcade fields, and that the skew (angle of coronal loops viewed from above relative to the underlying neutral line) pattern of the coronal fields, which they found to be tightly correlated with the chirality of the underlying filaments, unambiguously determines cloud chirality. In spite of these opposing interpretations of filament skew patterns, which hinge on whether the barbs (lateral extensions of filaments or prominence feet) extending from filaments are concave upward or downward, either the filament chirality or the arcade skew pattern can be used as a predictive tool for cloud chirality. While the hemisphere of origin predicts chirality in a statistical sense, observations of the filament or posteruptive arcade can provide more precise predictions covering the occasional cases where the hemispheric pattern does not apply [ McAllister et al. , 1998].

In addition to chirality and axis elevation, one can predict the field direction along a magnetic cloud axis from observations of arcade skew in the context of the large-scale solar magnetic field configuration. As pointed out by Bothmer and Schwenn [1994, 1998], a cloud with a given chirality can have either northward or southward leading fields depending upon the direction of the field along its axis. Bothmer and Rust [1997] found a solar cycle variation in the leading fields of clouds reflecting the solar cycle variation of the direction of the solar dipole field, and this correlation, combined with the hemispherical chirality correlation, in itself can serve as a statistical predictive tool for cloud axis polarity. On a case-by-case basis, however, one can improve the statistical prediction by examining the posteruptive arcade fields in X-ray data in the context of the larger-scale structure of the solar magnetic field to infer cloud axis polarity.

The predictive tools for magnetic cloud topology have been further developed in statistical studies [e.g., Zhao and Hoeksema , 1997; Mulligan et al. , 1998] and applied in retrospective case studies [e.g., Crooker et al. , 1998]. In this paper, we report on the first application of these tools to predict magnetic cloud topology and a geomagnetic storm in real time from solar observations. The mixed and somewhat unexpected results provide insights for future empirical models. We present the solar observations in section 2, the prediction based upon them in section 3, the solar wind observations and the geomagnetic response in section 4, and a comparison with the predictions in section 5. The main conclusions are summarized in Section 6.

The Solar Observations

Fig. 1 A 14.6 arcmin square section of the September 27, 1997, Kitt Peak synoptic magnetogram, taken at 1529 UT. This has been matched (including rotation) to the SXT image of 2226 UT (see Figure 4). The black is negative flux, and the white is positive flux. The amplitude has been cropped at ± 200 Guass. The limb is visible at the upper right corner, and the bar in the lower left corner is 1 arcmin long. The thin vertical lines indicate central meridian.

On September 27, 1997 an extended region of strong plage, evolving from an aging active region, was crossing central meridian at roughly 30° N(Figure 1). It extended diagonally from 15°N 20°W to 40°N 15°E, along a neutral line tilted at roughly 45° to the equator. The positive (white) and negative (black) flux had been conspicuously sheared along the polarity inversion line (neutral line), as is typical for isolated active regions under differential rotation and meridianal flow [van Ballegooijen et al., 1998]. The neutral line was within a filament channel defined in Ha by fibrils whose long axes are nearly parallel to the filament [Martin et al., 1992]. The channel was partially filled by several thin filaments (Figure 2).

The largest of the filaments was to the east. Its western end was low-lying, threading into the central region of strong field, and rooting in positive plage. The eastern end rose to a height of 50,000km or more, being visible for roughly 62", but rooting another 15" to the east, in negative plage. Although still a thin filament, it had developed barbs and could be identified as a dextral filament [Martin et al. 1994], the dominant pattern in the northern hemisphere. Filament chirality can also be determined from the direction of the axial magnetic field or from the orientation of fibrils relative to plagettes (Ha features [Martin et al., 1992]). As mentioned above, the filament terminates in positive polarity plage at its western end and negative at the eastern end. Viewed from the side with dominantly positive photospheric polarity (i.e., from the north) the axial field points to the right, indicating a dextral filament. The fibrils from the plagette above the middle of the filament (Figure 2) emanate to the east, consistent with the direction of the field along the spine. This means they emanate to the right when viewed from the positive polarity side of the filament, making the filament dextral by this definition, as well.

Fig. 2 Helio Research Ha images showing the filament eruption (left side) and subsequent flare (bright patches, center and right side). The frames were taken at 1831 UT, 2213 UT, and 2313 UT on September 27, 1997, and 1704 UT on September 28. The final frame shows the filament reforming. The frames have been rotated by 30o so that solar north is at the top, matching the orientation of all the other figures. Note that the elevation of the eastern end of the filament changes its apparent orientation, owing to projection. The white bar in each frame is 1 arc min long.

Having been under observation throughout the day with the Helio Research solar telescope, the filament erupted between 1900 and 2400 UT. Eruption was followed by a small arcade LDE flare that was still visible more than 12 hours later. The eruption was partially observed at Helio Research in Ha and from the L1 point in front of the Earth by the SOHO/EIT instrument at 195 Å (Figure 3). At this wavelength, EIT was able to record the filament in both absorption and later in emission and also the cooling coronal arcade that formed below it. Coupled with Yohkoh SXT coverage, this event allows detailed analysis of the relationship of the filament eruption to other coronal changes seen in EUV and SXR. In brief, the eruption can be analyzed in three phases: a slow rise phase (1923-2206 UT, Figure 3 top), a rapid rise phase (2206-2310 UT, Figure 3 middle), and the final eruption phase (2310-2400 UT, Figure 3 bottom). This process was accompanied (after ~2230 UT) by brightenings of nearby loops in the surrounding corona, including some axial loops close to the filament [Khan et al., 1998]. As the filament was still rising, a coronal arcade formed under the lower western end (stronger fields), appearing in SXR after 2130 UT (Figure 4) and in EUV after 2230 UT (Figure 3 middle, bottom). This LDE flare was barely above the GOES background (B0 (10^-7 W/m²) rising to B2.5) and might have been missed without the coronal images [McAllister et al. 1996b; Webb et al., 1998], although it might have been inferred from the associated Ha flare (see Figure 2).

Fig. 3 Three frames from the EIT 195 Å data, taken at (top) 2045, (middle) 2229, and (bottom) 2329 UT on September 27, 1997. They cover roughly 16 x 10 arc mins. A 1 arc min scale is shown in each frame. In Figure 3(top) the dark filament (arrow) can be seen rising above the filament channel, east of the bright central region. In Figure 3(middle) the filament has risen further and is beginning to fade (right arrow). There is a new bright loop (left arrow) behind the dark filament and footpoint brightening near its western end (to the right of the arrows). In Figure 3(bottom) the filament is whipping westward and upward (right arrow), bright loops curve up from its eastern end (left arrow), and the bright arcade loops are well formed under the western end (straight below the right arrow).

As has been observed in many Yohkoh events [McAllister et al., 1992; Moore et al., 1997], the coronal structures near the stronger fields are more compact and low-lying at one end (the western), while the weaker fields at the other end (eastern) balloon upwards (Figure 4). At the western end, the spine of the filament approached the surface asymptotically while, at the eastern end, it was rooted nearly vertically. The overall profile is that of a half pear, with the steep vertical foot not strongly visible in Ha. The eruption took place as the high eastern end expanded further, ballooning upward, seeming to draw the visible filament after it, like an elastic rope draping over a rising balloon. Eventually the low-lying western end appears to have snapped, so that the filament whipped upward, pivoting about the vertical eastern footpoint. This type of asymmetrical eruption is common [cf. Hanaoka et al., 1994; McAllister et al., 1996b].

Fig. 4 A composite AlMg SXT image from 2229 UT on September 27, 1997 (the flaring region has been filled in from a short exposure image take 2 min later). The bright early flare loops are crossing central meridian (vertical line segments), at roughly 33o N. The NW limb is high-lighted in the upper right corner. The image is 14.6 arc min square (the bar in the lower left is 1 arc min long). The arcade loops (both flare and ballooning) have a left skew of roughly 45o.

The Helio Research observations of the filament eruption and portions of the subsequent flare, coupled with first look images from SXT and EIT, strongly indicated that an Earth-directed CME had been launched late on September 27, 1997. This was confirmed by the report of a halo CME from the SOHO/LASCO coronagraphs, which appeared in the C2 field of view at 0127 UT on September 28th (S. Plunkett, personal communication, 1997). The halo appeared initially in the north and east, consistent with a filament eruption and CME north and east of disk center. Although there were five other CMEs observed by LASCO from 0525 UT on September 27 to 1531 UT on September 28, all of these appear to be at or behind the limb, making them unlikely to have been directed toward Earth (C. St. Cyr, personal communication, 1999). Only two are in the Northern Hemisphere, one behind the NE limb and one at the W limb. The latter, appearing in LASCO C2 at 1235 UT on September 27, coincides with the activation of the filament but not with its eruption. While it may be causally related to the eruption of the halo CME and the filament, being at the limb it is not a likely source of the magnetic cloud observed at Earth (section 4).

Magnetic Cloud Predictions.

On the basis of the solar observations described above, we were able to predict not only that Earth was likely to encounter a magnetic cloud but also some properties of that cloud. These predictions were formulated and published by e-mail on September 28, 1997, and on the ISTP web site on September 29, 1997, available at http://www-spof.gsfc.nasa.gov/istp, and are quoted below. Additional information drawn from the fuller data sets later available, and a few relevant references have been added in square brackets. These additions were not part of the original predictions.

"The location of the filament indicates that the associated CME will very likely intersect Earth, sometime on September 30, assuming roughly average transit time. [A 3 day transit actually requires a slightly above average, but not unusual speed of ~550 km/s.] The season, being just after the fall equinox, is most favorable for geomagnetic activity associated with northern hemisphere [solar] events [Crooker and Cliver, 1994; McAllister and Crooker, 1997]."

"The global magnetic field rooted near the north polar hole, must be southward [positive. This contributed a southward dipole field high over the eruption site]. The extended plage, has a positive leading region and a negative trailing region (new cycle [23]) with the leader having the same polarity as the pole. [It lies along a mid-latitude diagonal polarity inversion line merging into the western end of a section of the polar crown neutral line (see also section 4 below). Marubashi [1997] has shown that the orientation of the axis of a magnetic cloud is generally preserved from that of a filament erupting behind a CME. In this case] we predict a roughly east-west axis for the CME (or possible magnetic cloud) with the axial field pointing to the east. The NE-SW tilt (if it is maintained into the interplanetary medium) will lend a slight northern component, near the core, but the leading outer layers should have a southward direction."

"The relative location of the [positive and negative] plages, suggests a left-skewed connection in the corona, as expected for the northern hemisphere. Yohkoh and EIT images do show left-skewed arcade [loops at various heights] over the filament matching that expected from the filament chirality [dextral] based on work by Martin and McAllister [1998]. [The left-skew of the coronal fields is seen in both pre-eruption and post-eruption loops. Following the models of flux rope formation] we can predict that the associated CME will form a left-handed flux rope [Gosling, Birn, and Hesse, 1995; Martin and McAllister, 1997]. This matches the dominant northern hemisphere pattern found in cloud studies by Bothmer and Schwenn [1994] and Bothmer and Rust [1997]."

We note that at the time of these predictions only some preliminary coronal images were available. They were enough to confirm the general left-skewed nature of the overlying corona and to suggest that a bright, posteruptive coronal arcade may have appeared. We also suggested the possibility that although this event was well on the disk, signatures might be seen by the LASCO and HAO Mauna Loa coronagraphs. LASCO observations of a CME appearing in the C2 (middle) coronagraph at 0149 UT on September 28, 1997, were announced the next day (by G. Brueckner on the ISTP web page).

Solar Wind Disturbance and Geomagnetic Response

Fig. 5 Solar Wind data for October 1, 1997, from Wind. Magnetic field (magnitude B and the longitude and latitude angles phi and theta) and plasma parameters (thermal velocity, density n, and speed) are shown.

Several days later, interplanetary signatures of the CME associated with the disappearing filament on September 27 were observed near Earth. On October 1-2, 1997, both Wind and IMP 8 encountered a magnetic cloud preceded by a shock-like discontinuity. At onset, IMP 8 was in the solar wind on the dusk side of the magnetosphere near the dawn-dusk terminator, heading toward the bow shock, and Wind was in the dawn magnetosheath, also near the dawn-dusk terminator but heading away from the bow shock.

Figures 5 and 6 show the magnetic field and plasma data from Wind and IMP 8, respectively. Wind passage across the bow shock into the solar wind was marked by a sharp drop in field magnitude $B$ and a decrease in fluctuations of the plasma parameters around 1130 UT on October 1, with two brief bow shock encounters at ~1600 UT. IMP 8 passage in the opposite direction, across the bow shock into the magnetosheath, was marked by a sharp rise in $B$ around 0230 UT on October 2. Despite the complications introduced by these bow shock crossings, the signatures of the magnetic cloud are clear.

The disturbance began with a shock-like discontinuity that passed both spacecraft at 0100 UT. The signature is clearest in the data from IMP 8, since, unlike Wind, it was located in the solar wind at that time. Figure 6 shows a rise in speed, density n, and thermal speed, characteristic of a shock. We refrain from claiming that the discontinuity was an actual interplanetary shock, however, because there was only a modest rise in $B$ and then an unusual return to ambient values a few minutes later. Shortly afterward (~33 min later), there was a marked rise in $B$, but it is clearly not associated with this event. Similar discontinuity signatures were apparent in the Wind data in Figure 5 except for a longer delay between the rises in speed and $B$ and a decrease rather than increase in thermal speed. The latter may have been due to a change in spacecraft position in the magnetosheath flow pattern in response to increasing dynamic pressure.

Fig. 6 Solar Wind data for October 1, 1997, from IMP 8. Magnetic field (magnitude B and the longitude and latitude angles phi and theta) and plasma parameters (thermal velocity, density n, and speed) are shown. The data gaps represent time spent in the magnetosheath.

The clearest signatures of the magnetic cloud itself are found in the data from Wind since, at this later time, unlike IMP 8, Wind was completely immersed in the solar wind. The shaded cloud interval in Figure 5 coincides with the characteristic low thermal speed and magnetic field rotation [e.g., Burlaga, 1991]. The field rotation in this case occurs primarily in magnetic longitude phi, from roughly the lower (45°) to the upper (225°) limit of away polarity along the Parker spiral (where phi = 0° points directly toward the Sun). The IMP 8 field data in Figure 6, although interrupted by the bow shock crossing, show a similar pattern.

Another possible interpretation is that the southward field lay in an extended sheath region and that the body of the transient did not reach the spacecraft until the field turned northward. This interpretation is suggested by the interval of steady low temperature characteristic of clouds [e.g., Burlaga, 1991] that begins with the northward turning. On the other hand, this interpretation places the shock perhaps too far from the leading edge of the body that created it.

The cloud interval selection is strongly confirmed by two additional findings. First, counterstreaming electrons in the heat flux energy range were present throughout the interval. Counterstreaming is usually taken as the signature of CME field lines rooted at the Sun at both ends [e.g., Gosling, 1990]. This case, however, was complicated by the fact that Wind was located in Earth's foreshock region, where Parker spiral fields connect to the bow shock. Electrons energized there travel upstream along the field and can mimic counterstreaming in closed CME structures. To test for this possibility, the depth of the spacecraft into the foreshock region, measured from the magnetic tangent, was calculated using measured field directions and a model of bow shock shape and position adjusted according to the measured dynamic pressure [Fairfield, 1971; Roelof and Sibeck, 1993]. Within an accuracy equivalent to adjusting the model parameters by 5% or the field direction by 5°, Wind was in the foreshock region for a total of only two of the 30 hours in the cloud interval: 1700-1800, 1845-1915, and 1945-2015 UT on October 1, with an additional brief period at 2115 and some possible brief periods during the 0000-0500 interval on October 2. Except for these minor periods of foreshock ambiguity, we conclude that the cloud coincided with counterstreaming signaling a closed CME structure.

The second finding confirming our cloud interval selection is that the field data provide a very good fit to a force-free flux rope model [e.g., Lepping et al., 1990]. Additionally, the model yields the following parameters, some of which can be compared with predicted cloud properties: helicity, left-handed; axis longitude (GSE, phi), 144° axis elevation, 41° distance from axis, 39% of radius. These properties are illustrated in Figure 8. Although the flux rope that fits the data has southward fields on its leading edge, as shown in the view looking toward the Sun in Figure 8b (see caption), these do not intersect the calculated trajectory of the spacecraft, which, owing to the rope's canted orientation, passes through the southeastern quadrant of its cylindrical cross-section. Figure 8c shows that the ecliptic plane projection of the cloud axis was nearly aligned with the Parker spiral (phi = 138°, for the average cloud speed of 450 km/s, compared to the observed phi = 144°). This orientation implies passage through the leading leg of a flux rope loop, the apex of which lay beyond 1 AU, as illustrated. The loop is attached to the Sun at both ends, consistent with the presence of counterstreaming electrons throughout the cloud, as discussed above.

The solar wind disturbance created by the September 27 CME is an excellent example of an event in which the geoeffective southward magnetic field was encountered not within the cloud itself but in its sheath [cf. Tsurutani et al., 1988]. Figure 5 shows that the magnetic latitude theta (GSE) turned southward about halfway through the sheath, near 0730 UT on October 1, and remained southward until the beginning of the cloud at 1700 UT. The southward field was responsible for a magnetic storm with a classical Dst signature, shown in Figure 7. It began with a sudden commencement at the shock-like discontinuity at 0100 UT, where Dst rose to positive values. After returning to pre sudden commencement values, Dst decreased sharply in response to the southward turning of the solar wind field, around 0730 UT, marking the beginning of the main phase decline to -108 nT near 1500 UT. The subsequent smooth recovery began as the southward field weakened and then turned northward with cloud onset at 1700 UT. No sustained activity followed the main phase because there was neither high-speed flow nor polarity favorable for the Russell-McPherron effect after cloud passage [cf. McAllister and Crooker, 1997]. However, as pointed out by Crooker and Cliver [1994], for CMEs that carry sector boundaries, if the trailing polarity is unfavorable, then the leading polarity in the sheath must be favorable and lead to enhanced peak storm strength [ Crooker et al., 1992; Crooker and Cliver, 1993]. How this occurs in the case presented here is described below.

Fig. 7 Dst for September 30 through October 6, 1997. There is an initial positive excursion followed by the steep drop and rapid recovery on October 1 of the storm. Note the lack of significant extended activity indicating the absence of a trailing high-speed stream.

The source of the southward field in the sheath may have been solar rather than interplanetary. More likely sources are a preceding CME [e.g., Burlaga et al., 1987; Zhao, 1992], as observed by LASCO (section 2), or the magnetic arcade fields high in the corona [cf. Tsurutani et al., 1998] (see also section 6). Although magnetically open, as indicated by no counterstreaming electrons, the arcade fields nevertheless could have maintained their southward orientation if they became open through reconnection in one leg, near the Sun [cf. Crooker et al., 1998]. Normally sheath fields are assumed to be part of the ambient medium, consisting of Parker spiral fields with random, possibly Alfvenic, fluctuations supplying southward components. These can become strong owing to compression. In addition, draping of the ambient field around the body of the CME can bend the field southward [e.g., McComas et al., 1989]. Further, if the Parker spiral field has favorable polarity, the Russell-McPherron effect can increase the southward components in Earth's tilted dipole coordinate system (GSM) by as much as 40% [ Russell and McPherron, 1973; Crooker et al., 1992]. These concepts apply to a certain extent in the present case. Figure 5 shows strong fluctuations in the sheath, and the away polarity is favorable for the Russell-McPherron effect in this fall season. The persistence of the southward field after 0730 UT on October 1 for nearly 10 hours, however, does not appear to be a random fluctuation. In addition, draping of the ambient field around the body of the CME [e.g., McComas et al., 1989] and deflection by a leading shock [e.g., Wu and Dryer, 1989] can bend the field southward. The Russell-McPherron effect increases the geoeffectiveness of this persistent southward field by 33%: its average strength in GSM coordinates, which automatically incorporate the effect, is 8.8 nT, compared to 6.6 nT in GSE coordinates.

Comparison with the Prediction

Figure 8 illustrates that our prediction was successful on several counts. The predicted left-handed helicity of the cloud, based on the skew of the overlying arcade, and the axial fields pointing away from the Sun in the leading leg, based on the polarity pattern surrounding the filament, were confirmed. The NE-SW alignment of the filament channel in Figure 8a corresponds well with the observed 41^o elevation of the cloud axis above the ecliptic plane in Figure \ref{cartoon}b. (The apparent tilt of the filament (~63^o) is higher than the tilt of the channel (~45^o) owing to projection effects.) We also note that the intersection of Earth with a point behind and below the apex of the cloud is consistent with the Northern Hemisphere location of the solar event and the somewhat eastward propagation of the CME indicated by the LASCO observations.

Fig. 8 Schematic diagrams showing relationships between solar and magnetic cloud topology. (a) At the Sun, the skewed arcade overlying the filament becomes the magnetic cloud, and the overlying arcade high in the corona reflecting the Sun's dipolar field becomes the southward field in the sheath of the cloud. The tilt angle of the filament is based on the filament channel at its base. (b) In a view from beyond Earth looking toward the Sun, the cloud axis is nearly aligned with the filament axis in Figure 7a, and the coiled field on the leading edge of the cloud has a southward component which Earth fails to intersect. (Note; a true projection onto the plane perpendicular to the Sun-Earth axis would show a projected rather than actual axis elevation angle, which would be 10o higher.) (c) In an ecliptic plane projection of the flux rope loop, Earth passes through its leading leg, with the apex well beyond 1 AU and the cloud axis aligned with the Parker spiral.

The predicted leading southward magnetic field in the flux rope most likely was also born out (Figure 8b). This is supported by the results of the flux rope model fit to the data. The southward field was not encountered, however, owing to the tilt of the rope and the spacecraft trajectory through it, as discussed in Section 4. Thus the prediction of a geomagnetic storm based on southward field in the leading edge of the cloud failed. That a storm occurred, however, owing to southward field in the sheath of the cloud, may not have been entirely fortuitous. Both the southward field in the leading edge of the cloud that was not encountered and the field high in the corona that may have been responsible for the southward field in the sheath reflect the large-scale dipole field of the Sun upon which statistical predictions are based [ Bothmer and Rust, 1997; Mulligan et al., 1998]. Nevertheless, it is clear from this case that successfully predicting the topology of a magnetic cloud does not necessarily mean successful prediction of storm occurrence.

Although predicting the time of arrival of the magnetic cloud and, consequently, the time of storm occurrence was part of this study only to the extent of what might be expected based on average transit times, we note here two points regarding timing. The first point concerns the above-mentioned fact that the southward field responsible for the storm preceded the cloud. Had the speed of the cloud been known, for example, through radio techniques [e.g., {\it Reiner et al.}, 1998],\nocite{Reietal:98} the main phase of the storm would have begun about nine hours earlier than predicted. The second point concerns the deduced geometry of the flux rope loop illustrated in Figure 8c. The transit time of the leading edge of the cloud was 91 hours (taking 2200 UT on September 27 as the launch time), which translates to an average speed of 455 km/s. The observed speed at the leading edge was slightly faster, ~475 km/s. For passage through the leading leg of a flux rope loop whose apex was already beyond 1 AU, these speeds imply that the cloud gradually accelerated from a slow start and that its apex attained a considerably higher speed. How much higher would determine how much beyond 1 AU the apex was located at the time of encounter, a distance which is probably exaggerated in Figure 8c. Nevertheless, the implication seems qualitatively reasonable. In addition, the long delay of 16 hours between passage of the shock-like discontinuity and the leading edge of the cloud is consistent with the proposed geometry, since the delay increases with longitudinal distance from the apex. From this postevent analysis, it is clear that predicting time of arrival depends not only on knowledge of cloud speed but on cloud geometry, as well. In the future, time of arrival may prove to be the most difficult parameter to predict successfully.

Summary

For the first time, we have applied in real time the most recently developed techniques for predicting the topology of a magnetic cloud from solar signatures [ Martin and McAllister, 1997, 1999]. Beginning with a ground-based observation of an erupting filament in Ha light, we went beyond the technique of Rust [1994] for predicting chirality based on hemisphere of origin by examining the skew of the arcade overlying the erupting filament in X-ray data to check whether the pattern matched what was expected from that hemisphere. In addition, we went beyond the technique of Marubashi [1986, 1997] for predicting the elevation angle of the cloud axis based on the filament orientation by using its surrounding magnetic field structure to predict its polarity, as well. The resulting prediction provided a good match to the cloud parameters from a flux rope model fit to data obtained three days later at 1 AU.

The successful prediction of cloud topology contrasts with the prediction of how the cloud would produce a geomagnetic storm. Although a storm ensued, it was not caused by southward field in the cloud itself, as predicted, but rather by southward field in its sheath. The spacecraft trajectory through the model flux rope fit to the cloud data precluded encounter with the southward field in the rope owing to geometrical factors. Of interest for future predictions, however, is the possibility that the southward field in the sheath can be derived from southward arcade fields high in the corona, which can easily be observed and folded into a prediction.

A crucial factor missing from the prediction was the time of arrival of the magnetic cloud. In a postevent analysis of cloud geometry, we demonstrated that uncertainty in Earth's trajectory through a cloud compounds any uncertainty in the cloud speed, which may make successful predictions of storm onset the most challenging aspect of future space weather forecasting.

Acknowledgements

We thank the Yohkoh SXT, SOHO EIT, Wind and IMP-8 teams for sharing their data. The research at Helio Research was supported by NASA grant NAG5-4641. The research at Boston University was supported by NASA under grant NAG5-7049 and by NSF under grant ATM98-05064. Many thanks to Chris St. Cyr for checking the LASCO and EIT data for other CMEs on September 27 and 28. We are grateful to Adam Szabo for his assistance in magnetic field data preparation and magnetic cloud model fitting. Thanks Xue Pu Zhao and David F. Webb for their assistance in evaluating the JGR paper.

References

Bothmer, V., and D. M. Rust, The field configuration of magnetic clouds and the solar cycle, in Coronal Mass Ejections: Causes and Consequences, edited by N. Crooker, J. Joselyn, and J. Feynman, pp. 139-146, Washington, D.C. AGU, 1997.

Bothmer, V., and R. Schwenn, Eruptive prominences as sources of magnetic clouds in the solar wind, Space Science Reviews, 70, 215-220, 1994.

Bothmer, V., and R. Schwenn, The structure and origin of magnetic clouds in the solar wind, Ann. Geophys., 16, 1-24, 1998.

Brueckner, G. E., et al., Geomagnetic storms caused by coronal mass ejections (CMEs): March 1996 through June 1997, GRL, 25, 3019-3022, 1998.

Burlaga, L. F., Magnetic clouds, in Physics of the Inner Heliosphere II, edited by R. Schwenn, and E. Marsch, pp. 1-22. Springer-Verlag, New York, 1991.

Burlaga, L. F., K. W. Behannon, and L. W. Klein, Compound streams, magnetic clouds, and major geomagnetic storms, JGR, 92, 5725-5734, 1987.

Chen, J., Theory of prominence eruption and propegation: Interplanetary consequences, JGR, 101, 27,499-27,519, 1996.

Crooker, N. U., and E. W. Cliver, Reply to the comment by W. D. Gonzalez, A. L. Clua de Gonzalez, and B. T. Tsurutani on "The semiannual variation of great geomagnetic storms and the postshock Russell-McPherron effect preceeding coronal mass ejecta", by N. U. Crooker, E. W. Cliver, and B. T. Tsurutani, GRL, 20, 1661-1662, 1993.

Crooker, N. U., and E. W. Cliver, Postmodern view of M-regions, JGR, 99, 23,383-23,390, 1994.

Crooker, N. U., E. W. Cliver, and B. T. Tsurutani, The semiannual variation of great geomagnetic storms and the postshock Russell-McPherron effect preceeding coronal mass ejecta, GRL, 19, 429-432, 1992.

Crooker, N. U., et al., Sector boundary transformation by an open magnetic cloud, JGR, 103, 26,859-26,868, 1998.

Dodson, H. W., and E. R. Hedeman, Problems of differentiation of flares with respect to geophysical effects, Planet. Space Sci., 12, 393-418, 1964.

Fairfield, D. H., Average and unusual locations of the Earth's magnetopause and bow shock, JGR, 76, 6700-6716, 1971.

Gosling, J. T., Coronal mass ejections and magnetic flux ropes in interplanetary space, in Physics of Magnetic Flux Ropes, Geophys. Monogr. Ser., vol. 58, edited by C. T. Russell, E. R. Priest, and L. C. Lee, pp. 343-364, AGU, Washington, D.C., 1990.

Hanaoka, Y., et al., Simultaneous observations of a prominence eruption followed by a coronal arcade formation in radio, soft x-ray and Ha, PASJ, 46, 205-216, 1994. Howard, R. A., D. J. Michels, N. R. Sheeley, and M. J. Koomen, The observation of a coronal transient directed at the Earth, ApJ Let., 263, L101-L104, 1982.

Hundhausen, A. J., Coronal mass ejections, in Cosmic Winds and The Heliosphere, edited by J. R. Jokipii, C. P. Sonett, and M. S. Giampapa, pp. 256-296, Univ. of Ariz. Press, Tucson, 1997.

Joselyn, J. A., and P. S. McIntosh, Disappearing solar filaments: A useful predictor of geomagentic activity, JGR, 86, 4555-4564, 1981.

Khan, J. I., Y. Uchida, A. McAllister, Z. Mouradian, I. Soru-Escaut, and E. Hiei, Yohkoh soft x-ray observations related to the prominence eruption, flare and coronal mass ejections of 1992, May 7, Astron. Astrophys., 336, 753-768, 1998.

Lemen, J. R., L. W. Acton, D. Alexander, A. B. Galvin, K. L. Harvey, J. T. Hoeksema, X. Zhao, and H. S. Hudson, Solar identification of solar-wind disturbances observed at Ulysses, in Solar Wind Eight, edited by D.~Winterhalter, et al., pp. 92-95, AIP Conf. Proc., 382, 1996.

Lepping, R. P., J. A. Jones, and L. F. Burlaga, Magnetic field structure of interplanetary magnetic clouds at 1 AU, JGR, 95, 11,957-11,965, 1990.

Low, B. C., and D. F. Smith, The free energies of partially-open coronal magnetic fields, ApJ, 410, 413-425, 1993.

Martin, S., and A. H. McAllister, Predicting the sign of helicity in erupting filaments and coronal mass ejections, in Coronal Mass Ejections: Causes and Consequences, edited by N. Crooker, J. Joselyn, and J. Feynman, pp. 127-138, Washington, D.C. AGU, 1997.

Martin, S. F., and A. H. McAllister, The skew of x-ray coronal loops overlying Ha filaments, Ap J, in press, 2001.

Martin, S. F., W. H. Marquette, and R. Bilimoria, The solar cycle pattern in the direction of the magnetic field along the long axes of polar filaments, in The Solar Cycle, edited by K. L. Harvey, p. 53. Astron. Soc. Pacific Conference Series 27, 1992.

Martin, S. F., R. Bilimoria, and P. W. Tracadas, Magnetic field configurations basic to filament channels and filaments, in Solar Surface Magnetism, edited by R. Rutten, and C. J. Schrijver, pp. 303-338. Kluwer Academic Pub., 1994.

Marubashi, K., Structure of the interplanetary magnetic clouds and their solar origins, Adv. Space Res., 6, 335-338, 1986.

Marubashi, K., Interplanetary flux ropes and solar filaments, in Coronal Mass Ejections: Causes and Consequences, edited by N. Crooker, J. Joselyn, and J. Feynman, pp. 147-156, Washington, D.C. AGU, 1997.

McAllister, A. H., and N. U. Crooker, Coronal mass ejections, corotating interaction regions, and geomagnetic storms, in Coronal Mass Ejections: Causes and Consequences, edited by N. Crooker, J. Joselyn, and J. Feynman, pp. 279-289, Washington, D.C. AGU, 1997.

McAllister, A., Y. Uchida, S. Tsuneta, K. Strong, L. Acton, E. Hiei, M. Brunner, T. Watanabe, and K. Shibata, The structure of the coronal soft x-ray source associated with the dark filament disappearence of September 28, 1991 using the Yohkoh soft x-ray telescope, PASJ, 44, L205-L210, 1992.

McAllister, A. H., M. Dryer, P. McIntosh, H. Singer, and L. A. Weiss, A large polar crown coronal mass ejection and a "problem" geomagnetic storm: April 14-23, 1994, JGR, 101, 13,497-13,515, 1996a.

McAllister, A. H., H. Kurokawa, K. Shibata, and N. Nitta, A filament eruption and accompanying coronal field changes on november 5, 1992, Solar Physics, 169, 123-149, 1996b.

McComas, D. J., J. T. Gosling, S. J. Bame, E. J. Smith, and H. V. Cane, A test of magnetic field draping induced bz perturbations ahead of fast coronal mass ejecta, JGR, 94, 1265-1471, 1989.

Moore, R. L., B. Schmieder, D. H. Hathaway, and T. D. Tarbell, 3-D magnetic field configuration late in a large two-ribbon flare, Solar Phys., 176, 153-169, 1997.

Mulligan, T., C. T. Russell, and J. G. Luhmann, Solar cycle evolution of the structure of magnetic clouds in the inner heliosphere, GRL, 25, 2959-2963, 1998.

Reiner, M. J., M. L. Kaiser, J. Fainberg, J.-L. Bougeret, and R. G. Stone, On the origin of radio emissions associated with the January 6-11, 1997 CME, GRL, 25, 2493--2496, 1998.

Roelof, E. C., and D. G. Sibeck, Magnetopause shape as a bivariate function of interplanetary magnetic field Bz and solar wind dynamic pressure, JGR, 98, 21,421-21,450, 1993.

Russell, C. T., and R. L. McPherron, Semiannual variation of geomagnetic activity, JGR, 78, 92-108, 1973.

Rust, D. M., Spawning and sheeding helical magnetic fields in the solar atmosphere, GRL, 21, 241-244, 1994.

Sheeley, N. R., et al., Coronal changes associated with a disappearing filament, Solar Phys., 45, 377-392, 1975.

Tsurutani, B. T., W. D. Gonzalez, F. Tang, S. I. Akasofu, and E. J. Smith, Origin of interplanetary southward magnetic fields responsible for major magnetic storms near solar maximum (1978-1979), JGR, 93, 8519-8531, 1988.

Tsurutani, B. T., et al., The January 10, 1997 auroral hot spot, horseshoe aurora and first substorm: A CME loop? GRL, 25, 3047-3050, 1998.

van Ballegooijen, A., N. P. Cartledge, and E. R. Priest, Magnetic flux transport and the formation of filament channels on the sun, Ap J, 501, 866-881, 1998.

Webb, D. F., A. S. Krieger, and D. M. Rust, Coronal x-ray enhancements associated with Ha filament disappearances, Solar Phys., 48, 159-, 1976.

Webb, D. F., E. W. Cliver, N. Gopalswamy, H. S. Hudson, and C. St. Cyr., The solar origin of the January 6, 1997 CME and subsequent geomagnetic cloud and storm, GRL, 24, 2469-2472, 1998.

Weiss, L. A., J. T. Gosling, A. H. McAllister, D. J. McComas, A. J. Hundhausen, J. T. Burkepile, J. L. Phillips, K. T. Strong, and R. Forsyth, A comparison of Yohkoh soft x-ray coronal events with observations of coronal mass ejections at Ulysses, Astron. Astrophys., 316, 384-395, 1996.

Wu, C. C., and M. Dryer, Predicting the initial IMF B_z polarity's change at 1 AU caused by shocks that precede coronal mass ejections, GRL, 23, 1709-1712, 1996.

Zhao, X.-P., Interaction of fast steady flow with slow transient flow: A new cause of shock pairs and interplanetary B_z event, JGR, 97, 15,051-15,055, 1992.

Zhao, X.-P., and J. T. Hoeksema, Is the geoeffectiveness of the 6 January 1997 CME predictable from solar observations?, GRL, 24, 2965-2968, 1997.

Last updated February 27, 2001. AHM.