Powerpoint template for scientific posters (Swarthmore College)

1
Can wavelet analysis provide an automated
technique to create the Dst index? Agnieszka
Jach1, Piotr Kokoszka1, Lie Zhu2, Jan
Sojka2 1Mathematics and Statistics Department,
2Center for Atmospheric and Space Sciences, Utah
State University, Logan, UT

• Wavelet-based procedure
• Filter out Periodic component from details Dj
  associated with Sq system, i.e., details at
  levels j 8,9,10, corresponding to 8h, 12h and
  24h frequencies, respectively
• Eliminate Long-term component captured by SJ by
  replacing SJ with the average of X
• Remove Noise via thresholding of the wavelet
  coefficients Wj at low levels (j 1,,7),
  i.e., those associated with high frequencies
• Adjust for the magnetic latitude, and center to
  remove B
• Average four (or three) stations to get
  wavelet-based index
• Results Comparison of Wavelet Index of Storm
  Activity (WISA) with Kyoto Dst

Introduction The primary geomagnetic storm
indicator is the Dst index. This index has a well
established recipe by which ground-based
observations are processed to produce this single
planetary index. Typically four ground stations
are used in this process. Assuming the
calibration, data gaps, and measurement
uncertainties are known, the major step in
creating the Dst is the subtraction of the daily
Sq variation. Determining a quiet day Sq
variation, which has seasonal dependence as well
as shorter term ones, is the art in this
analysis procedure. Schemes for selecting this
can be as quantitative as averaging the three
quietest days in the month, i.e., a new Sq
variation is inferred each month. The question
addressed by this study is whether it is feasible
for an automated technique to remove the Sq
without operator intervention? Our study is based
on wavelet analysis. An algorithm was developed
for four ground stations for two months of
measurements (March and April 2001). The results
of applying the procedure to a year of data and
then comparing these with the Kyoto Dst are
presented. Additional comparisons with the Kyoto
Dst are presented in which the automated analysis
uses only three of the four magnetometer data
sets.

Fig.4. WISA and Dst on March 31 and July 4 from
Fig. 1. 95 confidence limits for WISA, were
calculated from the increments of WISA. 99 of
observations of Dst are within these limits,
meaning that the two indices do not differ
significantly compared to the hour-to-hour
changes.
Fig.5. Histograms for the differences between
WISA and Dst based on four triplets of Dst
stations. The excluded station is indicated by a
blank. There is no distinction between disturbed
and quiet periods. Bars span 95 of the data,
i.e., data within 2 standard deviations of the
average.
Fig.1. Minute values of WISA and hourly values
of Kyoto Dst for the year 2001. By definition,
the average of WISA is zero as a result of adding
four zero-mean magnetograms. Two days disturbed,
Mar 31, and quiet, July 4, are zoomed in Fig. 4.

• Dst index
• H component at 4 low-latitude stations, hourly
• Quietest days for each month at each station are
  used to estimate secular and diurnal variations

Conclusions Using wavelet-based procedure, we are
able to create the storm activity index that is
within 10 nT of Kyoto Dst for 2001 without any
human input. This procedure also allows us to
estimate Sq system by filtering out portions of
the wavelet details that contain three different
harmonics associated with that system. The
procedure is automated and can be applied to
two-months worth of data or more. By definition,
the average of WISA is zero and thus, at this
point, we are unable to obtain the reference
level. Future work, that is currently underway,
is to investigate the removal of magnetic
signatures arriving from non-Dst current sources,
such as the Birkeland Currents associated with
the auroral electrojets. References Sugiura, M.
1964. Hourly values of equatorial Dst for the
IGY. Ann. Int. Geophysical Year 35945.
Percival, D. B., Walden, A. T. 2000. Wavelet
Methods for Time Series Analysis. Cambridge
University Press, Cambridge. Acknowledgments Fun
ding for the WAMI (Wavelet Analysis of
Magnetosphere-Ionosphere) project is provided by
an NSF grant DMS-0413653. Data is provided by the
global network of magnetic observatories,
INTERMAGNET. For further information Please
contact ajach_at_cc.usu.edu. PDF and PS versions of
the manuscript on which the poster is based are
available upon request. Poster is available at
http//cc.usu.edu/ajach.
Fig.2. Relative comparison of WISA with Dst.
Sixty 1-min observations of WISA were averaged
and compared with one 1-h value of Dst. Dst was
adjusted by its average, -18 nT, for the year
2001. Differences were divided into two groups,
for disturbed periods and quiet periods.
Objectives Eliminate human input when creating
the storm activity index automate the
procedure.
Model for a magnetogram X X1,,XT (H
comp.) Physical representation B
Internal magnetic field (constant) A Disturbance
due to storm activity P Periodic (diurnal)
component L Long-term (seasonal or longer)
component N Noise component t time in minutes,
t 1,,T 525600 1 year j level j,
corresponds to time scale 2j minutes, j 1,,J
19 Wj are associated with Fourier frequencies
in the interval 2-(j1), 2-j
Wavelet representation Wj Wavelet
coefficients high freq. VJ Scaling
coefficients low freq. Wj Matrix defining
wavelet transform VJ Matrix defining wavelet
transform Dj, Wavelet details SJ Wavelet
smooths
Fig.3. Histograms for the disturbed and quiet
differences from Fig.2. Bars span 95 of the
data, i.e., data within 2 standard deviations of
the average. Both histograms look very normal
with standard deviations 4 nT, even though the
indices may be based on different data.