Improving Space Weather Forecasts Using Coronagraph Data

1
Improving Space Weather Forecasts Using
Coronagraph Data S.P. Plunkett1, A. Vourlidas1,
D.R. McMullin2, K. Battams3, R.C.
Colaninno4 1Naval Research Laboratory,
Washington, DC 20375, 2Praxis, Inc., Alexandria,
VA, 3Interferometrics, Inc., Chantilly, VA 20151,
4George Mason University, Fairfax, VA 5th GOES
Users Conference, New Orleans, LA Paper Number
P1.50

• Introduction
• The objectives of this study were to
• Determine whether measures of CME asymmetry in
  coronagraph images could be used to objectively
  aid forecasts of whether a CME or its associated
  shock will strike Earth, and the magnitude and
  duration of any resulting geomagnetic storm
• Determine whether estimated shock speeds derived
  from metric Type II radio burst observations, or
  CME speeds derived from coronagraph observations,
  provide a better measurement of the initial event
  speed at the Sun to be used as an input to models
  for predicting the arrival of an interplanetary
  (IP) shock or ICME at Earth.

• CME Arrival Time Model Input Comparisons
• The models that we considered were
• Shock Time of Arrival Version 2 (STOA-2, Moon et
  al 2002)
• Empirical CME Arrival Model (ECA, Gopalswamy et
  al 2000, 2001)
• Empirical Shock Arrival Model (ESA, Gopalswamy et
  al 2005).
• Initial event speed inputs for each of the models
  were obtained as follows
• CME leading edge speeds from the LASCO catalog
• CME lateral expansion speeds perpendicular to the
  direction of propagation of the leading edge
• Metric Type II shock speeds from NGDC
  (http//www.ngdc.noaa.gov/stp/SOLAR/ftpsolarradio.
  html), supplemented by estimated shock speeds
  from the Fearless Forecast archive
  (http//www.gi.alaska.edu/pipermail/gse_ff).
• Other inputs required by the STOA-2 model (flare
  onset time and location, piston driving time, and
  background solar wind speed) were obtained from
  NGDC and Fearless Forecast archives.

• CME Asymmetry Analysis
• For each event, we selected a single LASCO/C3
  image in which the CME leading edge was at an
  apparent heliocentric distance from 12 14 R?.
  We used excess mass images, in which a pre-event
  image was subtracted, and the image data were
  photometrically calibrated in units of B/B?.
• Asymmetry parameters were determined by fitting
  an ellipse to the outline of the CME in this
  image, and by applying the cone model formulation
  of Xie et al (2003). The asymmetry parameters
  were
• Ellipse eccentricity (ratio of minor to major
  axes of fitted ellipse)
• Offset of ellipse center from Sun center
• Cone apex position (solar latitude and longitude)
  and cone width
• CME mass and center of mass within the fitted
  ellipse.

Results for 24-hour hit window
Example of a user-defined ellipse fit to the
outline of a CME on February 15, 2001.

• Event Selection
• We selected for analysis events from Solar Cycle
  23 (1997 2005) that have observations and
  measured speeds of an Earth-directed CME and are
  isolated sufficiently to allow unique
  associations with geomagnetic activity.
• Selected CMEs with apparent width 120? from the
  SOHO/LASCO catalog (http//cdaw.gsfc.nasa.gov/CME_
  List)
• Restricted selection to those events with no
  other Earth-directed CME in a window of ? 2 days
  centered on the time of first observation of the
  CME
• For the analysis involving comparison of
  different inputs speeds to CME propagation
  models, we also restricted our selection to those
  CMEs that were observed within a window of ? 2
  hours of the onset time of a metric Type II radio
  burst for which the frequency drift rate could be
  converted to an estimated coronal shock speed,
  and in which the location of the associated flare
  lay within the apparent angular span of the CME
  in LASCO.
• A total of 101 events were selected for the
  analysis of CME asymmetry.
• A total of 56 events were selected for the
  comparison of CME propagation model predictions
  with different initial speed inputs.

• Contingency Table Analysis and Verification
  Statistics
• Definition of 2?2 contingency table
• Hit (H) A shock or ICME is both predicted and
  observed within a hit window of ? N hours
• Correct Null (CN) No shock or ICME was
  predicted, and none was observed within a window
  of 1 5 days following a CME
• Miss (M) A shock or ICME was detected within a
  window of 1 5 days following a CME, but no
  shock or ICME was predicted within a window of ?
  N hours of the detection
• False Alarm (FA) A shock or ICME was predicted,
  but none was observed, within a window of 1 5
  days following a CME.

• Correlation of Asymmetry Parameters with
  Geomagnetic Activity
• Form a histogram of the asymmetry parameter
  values.
• Calculate the average value of the geomagnetic
  index in each bin.
• Plot the average value of the geomagnetic index
  against the binned asymmetry index, and search
  for correlations.
• Sample results from this type of analysis are
  shown below.

Results for the STOA-2 model with different
proxies for the initial shock speed, using a ? 24
hour hit window
Left Kp and Dst indices versus ellipse offset
from Sun center. Right Percentage of events with
Kp 5 or Dst 50 nT versus ellipse offset.
Western events are more geoeffective.
Left Kp and Dst indices versus cone width.
Right Percentage of events with Kp 5 or Dst
50 nT versus cone width. Events with width gt
50? are most likely to drive storms.
Standard meteorological metrics were used to
evaluate the relative success of the different
speed inputs to each model. Note Some metrics do
not strictly apply to the ECA and ESA models,
since these models always predict that a CME at
the Sun will result in an ICME or shock at Earth.
Results for the ESA model with different proxies
for the initial shock speed, using a ? 24 hour
hit window

• IP Shock and Geomagnetic Storm Associations
• We used a two-step approach similar to Zhang et
  al (2003) to determine associations between CMEs
  and ICMEs, IP shocks or geomagnetic storms at
  Earth
• Search for the arrival of an IP shock or ICME at
  Earth within 1 5 days following an
  Earth-directed CME
• Used the ICME plasma speed (where available) and
  the CME leading edge speed to constrain possible
  solar sources of the IP event by calculating a
  range of possible transit times from the Sun to 1
  AU.
• IP shocks in the solar wind were identified from
  ACE and WIND shock lists.
• IP shocks at Earth were identified by the
  occurrence of a Sudden Commencement (SC) or
  Sudden Impulse (SI) in the geomagnetic field.
• ICMEs were identified from a list provided by
  Cane and Richardson (private communication).
• Geomagnetic storms were identified using both Kp
  and Dst indices.

• Prediction of Shock Arrival at Earth
• No single asymmetry parameter showed a strong
  correlation with the detection of a shock at
  Earth.
• We identified a composite parameter (CME momentum
  mass ? apparent speed of the leading edge
  divided by center of mass offset), that showed a
  clear correlation with shock arrival at Earth. We
  refer to this parameter as the CME mass flow,
  since it has dimensions of mass/time.

Results for the ECA model with different proxies
for the initial CME speed, using a ? 24 hour hit
window

• Summary of Results
• Forecast quality does not depend significantly on
  the data source that is used for the initial
  event speed input to the STOA-2, ESA or ECA
  models.
• All models performed worse than the expected
  performance from a random forecast in predicting
  the arrival of a shock or ICME at Earth or L1 for
  this sample of events.

References Xie, H. et al, JGR 109,
doi10.1029/2003JA010226, 2003. Gopalswamy, N. et
al, GRL 27(2), 145 148, 2000. Gopalswamy, N. et
al, JGR 106, 29,207 29,217, 2001. Moon, Y.-J.
et al, GRL 29(10), 1390, doi10.1029/2002GL014865,
2002.
Earth-directed CME mass flow and shock
associations. 93 of events with mass flux gt 1012
grams/second have associated shocks at Earth.