Small Events Detection

1
Small Events Detection

• Fludra1, D. Haigh1,2, D. Bewsher1, V.
  Graffagnino1, P.R. Young1
• 1 Rutherford Appleton Laboratory, UK
• 2 Birmingham University, UK

2
eSDO Virtual Observatory Access to SDO
Data UKs Particle Physics and Astronomy
Research Council (PPARC) has funded the eSDO
project to prepare algorithms for the analysis of
SDO data and make them available to the solar
community using the virtual observatory (poster
26 - Culhane et al. More details in a talk by
Elizabeth Auden in session S6 on Friday).
Small Events Detection will be one of ten
eSDO algorithm packages.
3
What are Small Events? - Size
Aschwanden et al. 2000, ApJ, 535, 1047
Length 4 30
Area 4 500 arcsec2
Any event smaller than 20 - 30, down to AIA
pixel size.
4
What are Small Events? - Energy Range
Benz Krucker, 2002, ApJ,
Aschwanden et al. 2000, ApJ, 535, 1047
Nanoflares 1023 1026 erg Microflares 1026
1029 erg
5
Why study Small Events? Is the quiet Sun corona
heated by nanoflares? If ?gt2, small-scale
events provide the dominant contribution to the
heating of the corona. Some recent analyses give
?lt2, suggesting nanoflares cannot heat the quiet
sun corona.
6
Past research on small event detection Krucker
and Benz, 1998, ApJ, 501, L213 Berghmans, Clette
and Moses, 1998, AA, 336, 1039 Benz and Krucker,
1999, AA, 341, 286 Parnell and Jupp, 2000, ApJ,
529, 554 Aschwanden et al. (I) 2000, ApJ, 535,
1027 Aschwanden et al. (II) 2000, ApJ, 535,
1047 Harra, Gallagher and Phillips, 2000, AA,
362, 371 Aschwanden and Parnell, 2002, ApJ, 572,
1048 Benz and Krucker, 2002, ApJ, 568,
413 Bewsher, Parnell and Harrison, 2002, Sol.
Phys., 206, 21 marks algorithms selected for
testing. This talk presents first results from
Bewsher et al. method.
7
CDS Quiet Sun Brightenings
CDS NIS in O V 630 A line
Sit stare time series

? 5 arc min ?
Movie duration 2 hours

Time

• Advantage of using O V observations for testing
  SED algorithms
• 15 s time cadence and long series
• high variability in the transition region

Quiet Sun areas show thousands of short-lived
intensity enhancements

8
CDS time series in one of 71 pixels, with 2500
exposures and 15 s cadence. Note that previous
TRACE EIT nanoflare studies dealt with 13-25
frames with 80-120 s cadence.
9
Peak detection
10
Algorithm from Bewsher et al. 2000, Sol.
Phys. I. Peak detection Stage 1. Identify
maximum peak () in time series Stage 2. Identify
minimum troughs (?) on either side of
peak. Intensity jump between peak and troughs is
greater than ne e, so peak is kept. Stage 3.
Identify largest peaks () on either side of
original peak, and minimum troughs (?) on either
side. If intensity jumps are larger than ne e, so
those peaks are kept. Stage 4. Identify next
largest peaks ()on either side of () peaks and
their minimum troughs (?). If none of the
intensity jumps from peaks to troughs are greater
than ne e, peaks are neglected and time series is
not investigated further. II. Pixel
Grouping Adjacent pixels that peak within /- n?t
(n0, 1, 2,) are grouped together to form an
event. All 8 neighbour pixels are examined around
each pixel already classified as part of the
event. The lightcurves of the pixels are
integrated over all the pixels in an event to
produce an event' lightcurve.
11
III. Check! A check is made to see if each
event' still meets the same criteria as for
individual pixels, i.e. its peaks and troughs are
identified and compared in the same way as
discussed in Section I. If a peak is identified
in the summed event' intensity, and this peak is
still greater than above the troughs, then the
event is counted. IV. Picking Troughs The
nearest minimum (NM) trough on either side of the
peak is identified. Look at all other troughs
between the NM trough and the peak and
calculate increase in intensity between troughs
and small peaks in between. If change in
intensity is less than 2e, then the NM trough is
correct. If the change in intensity is greater
than 2e, then the nearer trough is correct.
Finally, the intensity difference between the
peak and the corrected trough is checked to
ensure that they meet the criteria for an
event, i.e. that the intensity increase is
still greater than ne e.
12
Example results from CDS sit stare time
series 4x240 slit 71 spatial pixels
(3.4) 15 s cadence 10 hour run 2500
exposures Threshold high, large peaks only Min
event size 1 pixel time coincidence of /- 3
steps Result 52 events
Time
distance
13

Example 2
Time
Threshold 3x lower, 255 events
distance
14
Example of individual events in the CDS sit
stare time series
15
Example of individual events in the CDS sit
stare time series
16
Example of detected events in CDS rastered
images 60 rasters, 10 min cadence. Time
coincidence requirement ?t0
? 240 ?
? 240 ?
17
Example of a detected event in the CDS rastered
image
18
(No Transcript)
19
DISCUSSION (1) How many nanoflares can we
expect Krucker Benz, 1998, ApJ Let. 7x7
field of view, 42 minutes. Energy range 8x1024
2.8x1026 erg gt3 sigma 11,150 events. gt
5sigma 2600 events Aschwanden et al. 2000
173 A, gt3 sigma 3130 events 195 A gt3 sigma
900 events Parnell and Jupp, 2000 (TRACE 7x7
fov) gt3sigma 11,700 pixels 4500 events in 80
minutes Our estimate CDS 1,500 events/hour
in a 4x4 fov. Expect 50,000 100,000
events/hour on the entire disk
20
DISCUSSION (2) Should we catalogue all small
events? Down to what energy and size? Should we
catalogue events in each AIA channel separately,
or attempt to identify the same events in
different channels? Data products What event
parameters should be stored for each AIA channel
start/peak/end time, coordinates of all pixels,
area, duration, peak intensity, total intensity
How can individual event parameters be used?
Time evolution of individual events heating
phase, decay phase Are statistical frequency
distributions of these parameters sufficient?
21
DISCUSSION (3) Analyse full disk or a central
part of disk? (can events near the limb merge and
spoil statistics?) Analyse 24-hours/day or
shorter samples (1-hour/day, 1-hour/week
etc.)? Frequency distributions of event
energies Conversion of signal (DN/s) to event
energy, through Te and emission measure (and
Ne). What statistical significance level should
we adopt for event detection? 3s ? Cosmic ray
removal, solar rotation correction, other issues?
22
DISCUSSION (4) What constitutes an event
every act of energy release even if occurring
within a bigger event? Example