The disturbances in the geomagnetic fields are caused by fluctuations in the solar wind impinging on the Earth. The disturbances may be limited to the high-latitude polar regions, unless the interplanetary magnetic field (IMF) carried by the solar wind has long periods (several hours or more) of southward component (Bz < 0) with large magnitudes (greater than 10-15 nT). The occurrence of such a period stresses the magnetosphere continuously, causing the magnetic field disturbances to reach the equatorial region. The degree of the equatorial magnetic field deviation, the measure of the magnitude of geomagnetic storms, is usually given by the Dst index. This is the hourly average of the deviations of the H (horizontal) component of the magnetic field measured by several ground stations in mid- to low-latitudes. Dst = 0 means no deviation from the quiet condition, and Dst < -100 nT means large storms. During the March 1989 storm that caused the province-wide blackout in Quebec, Canada, the Dst index reached approximately -600 nT.
There is another type of storms. These recur with periods of approximately 27 days, the solar rotation period, and are associated with the high-speed solar wind originating in ``coronal holes'' at the Sun. Such storms tend to be moderate. Severe storms tend to be nonrecurrent and are difficult to predict.
The new prediction method takes advantage of the distinct physical attributes of geoeffective solar structures in the following way. (1) They are magnetically well organized, so that the leading edge of a structure has clear relations to the solar wind that has yet to arrive at the observing platform. (2) The long durations of the solar wind drivers of large storms can provide forecasting time, in the range of several hours to more than 20 hours, depending, among other factors, on the solar cycle.
Theoretical model of a magnetic cloud.
A solar flux rope propagating toward the Earth (the green asterisk on the right) from the Sun (on the left). (Figure from Chen et al. [1995]. See Chen [1996] for the theory.) The dimensions of the model flux rope are shown to scale in relation to the Sun-Earth distance (1 AU). The color scheme shows the average magnetic field strength, ranging from strong (red, of the order of a gauss) to weak (blue, of the order of 10 nT). The helical curve illustrates a characteristic magnetic field line. Magnetic clouds may indeed be structurally simple as depicted here. Recent observations indicate that magnetic field lines of magnetic clouds do remain connected to the Sun and that the field lines toward the outer edge of a flux rope are more twisted [Larson et al., 1997]. This property is implied by the model structure and the magnetic field described below.
Click [here] to view the time evolution of the magnetic cloud.
Magnetic field profile associated with the above model magnetic cloud.
As the model magnetic cloud moves past the Earth, an observer (e.g., a satellite) sees a magnetic field varying in time. In this example, the northward component (Bz > 0) arrives at the Earth first [adapted from Chen (1996)]. The continued expansion in the minor radial direction is taken into account. Although no ambient magnetic field is included outside the cloud edges (the vertical dashed lines), the solar wind typically has fluctuating field of roughly 5 nT in magnitude. A nonzero By outside the flux rope will cause theta (panel c) to decrease.
The By and Bz components and theta are shown. The above theoretical model closely resembles the observed magnetic cloud. The By component, which is in the east-west direction, peaks where Bz = 0. The sign difference in By between the model and the observed cloud is insignificant. The solar wind speed V and density are also shown. Two major gaps in the data are indicated by vertical dashed lines.
Because the Earth passage of magnetic clouds takes 10-20 hours to a few
days, the method allows prediction of IMF forward in time. As a
result, the method can yield advance forecasting time of several to
more than 10 hours. This is far in excess of the warning time
achieved by the current neural network techniques which attempt to
predict the response of the magnetosphere to the measured solar wind
data. (See Chen et al. [1996, 1997]
for more detail.)
1. Burlaga, L. F., Magnetic clouds and force-free fields with constant
alpha, J. Geophys. Res., 93, 7217, 1988.
2. Chen, J., and D. A. Garren, Interplanetary magnetic clouds: Topology
and driving mechanism, Geophys. Res. Lett., 20, 2319, 1993.
3. Chen, J., P. J. Cargill, and S. P. Slinker, Plasma Physics of
Solar-Terrestrial Coupling, 1995 NRL Review,
NRL/PU/5230--95-274 May 1995.
4. Chen, J., Theory of prominence eruption and propagation: Interplanetary
consequences, J. Geophys. Res., 101, 27,499, 1996.
5. Chen, J., P. J. Cargill, and P. J. Palmadesso, Real-time identification
and prediction of geoeffective solar wind structures, Geophys. Res.
Lett., 23, 625, 1996.
6. Chen, J., P. J. Cargill, and P. J. Palmadesso, Predicting solar wind
structures and their geoeffectiveness, J. Geophys. Res., 102,
14701, 1997.
7. Joselyn, J. A., Geomagnetic activity forecasting: The state of the art,
Rev. Geophys., 33, 383, 1995.
8. Larson, D. E., et al., Tracing the topology of the October 18-20, 1995,
magnetic cloud with 0.1-100 keV electrons, Geophys. Res. Lett.,
24, 1911, 1997.
A Feature-Based Method of Predicting Geoeffective Solar Wind and
Geomagnetic Storms
This technique takes advantage of the fact that the solar wind (SW)
features that cause large storms, long uninterrupted durations of
strong southward or northward magnetic field, make such geoeffective
structures clearly distinguishable from the nongeoeffective (``background'')
solar wind. They can be described as magnetic flux ropes propagating
past the Earth. If the magnetic field of an interplanetary flux rope
is measured, the field vector shows a characteristic rotation which is
much slower than the time scale of IMF fluctuations in the background
solar wind. The slow rotation of the magnetic field can be recognized
and used to infer the magnetic field profile of the solar wind that
has yet to arrive at the Earth. This means that the leading edge
of the SW structure has a structural relationship to the trailing edge.
The basic technique falls under the
broad category of artificial intelligence. Another
artificial intelligence approach uses neural networks, and a number of
neural network space weather forecasting techniques are under
development elsewhere. The basic difference is that our technique seeks
to estimate the magnetic field profile (Bz) forward in time, in the
solar wind that has yet to arrive at the observing platform. The
response of the magnetosphere to the predicted Bz profile is then
inferred. In contrast,
the neural network approach attempts to predict the response of the
magnetosphere to the solar wind that has been observed.
Test Criteria and Dst Threshold
The test is being conducted to examine the degree of success one can
expect from the method in its present form with
respect to the following specific objectives.
The first is to estimate the eventual
duration tau and maximum value of Bz field (denoted Bzm) of
each solar wind event being encountered. The second objective is to
estimate the Bz(t) profile of the solar wind stream
that has yet to come. The third objective is to determine whether
the event is geoeffective or not geoeffective according to the estimated
Bzm' and tau'. The threshold for geoeffectiveness is chosen to be
Dst < -80 nT for two hours or longer. Our overall objective is to accurately
identify and predict the solar wind events that cause large
geomagnetic storms rather than the detailed response of the magnetosphere.
We judge the success of the technique as follows.
If a solar wind event leads to P1 > 0.5 for 2 hours or longer, we regard
the outcome as a positive prediction of the occurrence of Dst < -80 nT.
We then compare this outcome with the actual or provisional Dst value,
whichever may be available. If Dst does fall and remain below this
threshold for two hours or longer,
the prediction is judged to be correct. If P1 does not exceed 0.5 for
two hours or longer but Dst does fall below the -80 nT threshold,
then the result is
judged to be a false negative (a miss). If P1 produces a positive
prediction for storm but Dst does not fall below the -80 nT threshold,
the outcome is judged to be a false positive (a false alarm).
The association between the solar wind events and
the Dst values can be subjective, but we generally do not encounter obvious
ambiguity.
References