Download S4_Aschwanden

Feature & Pattern Recognition
in EUV Corona :
Prospects for AIA
Markus Aschwanden
(Lockheed Martin Solar Astrophysics Laboratory)
AIA/HMI Science Teams Meeting, Monterey, Feb 13-17, 2006
Session S4: Feature Recognition: Needs and Techniques
Content of talk :
(1) Scientific Motivation for reconstructing 3D structures:
- Hydrodynamics of elementary coronal loops
- Hydrodynamics and evolution of flare loops
- Hydrodynamics and evolution of filaments
- Tracing the real coronal magnetic fields
(2) Methods and Problems of extracting 3D structures:
- Fingerprinting (automated detection) of curvi-linear structures
- Background subtraction, disentangling, confusion problems
- Disentangling of multi-temperature, multi-thread structures
- 3D reconstruction of 2D curvi-linear features
- Stereoscopic analysis with 2 spacecraft (STEREO 2006)
(3) Conclusions
(1a) Hydrodynamic modeling of elementary and composite loops
Problems:
-Isolated loops don’t exist
-Every background consists of loops itself
-Disentangling of nested loop strands often impossible
due to lack of 3D information and insufficient resolution
-Background is often ill-defined because it requires
modeling of background loops ad infinitum
(1b) Hydrodynamics and evolution of flare loops
Aschwanden (2002)
- Spatio-temporal tracing of magnetic reconnection
- Hydrodynamics, heating, cooling of postflare loops
- Footpoint (double) ribbon separation and X-point height h(s)
- Shear vs. height relation of reconnecting field lines
(1c) Hydrodynamics and evolution of filaments
Envold (2001)
Aulanier & Schmieder (2002)
-Geometry and multi-threat structure of filaments
(helicity, chirality, handedness  conservation, fluxropes)
-Spatio-temporal evolution and hydrodynamic balance
-Stability conditions for quiescent filaments
-Hydrodynamic instability and magnetic instability
of erupting filaments leading to flares and CMEs
(1d) Tracing the real coronal magnetic field
Aschwanden et al. (1999)
Wiegelmann & Neukirch (2002)
-Tests of (theoretical) potential field, linear force-free,
and nonlinear force-free magnetic field extrapolation by
comparison with observed EUV loops (projected in 2D)
-3D reconstruction of EUV loop coordinates with “dynamic
solar-rotation stereoscopy” or “two-spacecraft observations”
(1e) Measuring the twist of magnetic field lines
Aschwanden (2004)
-Measuring the number of turns in twisted loops
-Testing the kink-instability criterion for stable/erupting loops
-Monitoring the evolution of magnetic relaxation (untwisting)
between preflare and postflare loops
(1e) Measuring the twist of magnetic field lines
Aschwanden (2004)
-Measuring number of turns in (twisted) sigmoids
before and after eruption
-Test of kink-instability criterion as trigger of flares/CMEs
(1f) Measuring the twist of erupting fluxropes
Gary & Moore (2004)
-Measuring number of turns in erupting fluxropes
-Test of kink-instability criterion as trigger of flares/CMEs
2) Pattern and Feature Recognition:
Methods and Problems
(2a) Fingerprinting (automated detection) of curvi-linear structures
Louis Strous (2002) http:/www.lmsal.com/~aschwand/stereo/2000easton/cdaw.html
-Strous detects curvi-linear segments from brightness gradients
in 3x3 neighborhood areas
-Problems: incompleteness of coronal loops
no discrimination between noisy pixels and loops
combination of curvi-linear segments to full loops
(2a) Fingerprinting (automated detection) of curvi-linear structures
Lee, Newman & Gary
improve detection of
coronal loops with
“Oriented connectivity
Method” (OCM):
-median filtering
-contrast enhancement
-unsharp mask
-detection threshold
-directional connectivity
Lee, Newman, & Gary (2004), 17th Internat. Conf. On Pattern Recognition, Cambridge UK, 23-26 Aug 2004
Lee, Newman, & Gary (2004),  see poster by Lee et al. at this workshop
Simulation results:
-OCM renders most
of the loop structures
Remaining problems:
-crossing loops
-misconnections
-ambiguous connections
-faint loops
-crowded regions
Lee, Newman, & Gary (2004)
Lee, Newman, & Gary 2006, “Dynamic Aperture-based
Solar Loop Segmentation” (in prep.)
Loop detection in
triple-filter TRACE data
(171 A, 195 A, 284 A)
1998-Jun-12 1205:20 UT
-Manual tracing (10 pts)
-spline interpolation
x(s),y(s)
-1D stretching with
bilinear interpolation
-multiple strands visible
-spatial offsets of loop
centroids in 3 filters
-background loops
-background moss
Loops
Widths
Loop/Backgr. Instrument Ref.
1
10
30
1
41
~12 Mm
?
CDS
?
170%150% EIT
7.10.8 Mm 30%20% EIT
~5.8 Mm
76%34% TRACE/CDS
3.71.5 Mm
?
TRACE
234
1.40.2 Mm
8%3%
TRACE
Schmelz et al. (2001)
Schmelz et al. (2003)
Aschwanden et al. (1999)
DelZanna & Mason (2003)
Aschwanden et al. (2000)
(no highpass filter)
Aschwanden & Nightingale
2005 (with highpass filter)
(2d) 3D reconstruction of loop structures
Full testing of theoretical
magnetic field
extrapolation models
with EUV-traced loops
requires 3D
reconstruction of
loop coordinates
[x(s), y(s), z(s)]

(1) Solar-rotation
dynamic stereoscopy
(2) Two-spacecraft
stereoscopy
Matching/Fitting
of EUV tracings
and extrapolated
field lines allows
to constrain free
parameters:
Alpha of nonlinear
force-free field
model.
Wiegelmann & Neukirch (2002)
(2e) Stereoscopic analysis with 2 STEREO spacecraft
-Identification of a common feature from two views is difficult
for nested structures (loop arcades, active region loops)
-Stereoscopic 3D-reconstruction is least ambiguous for small
stereo-angles, but 3D accuracy is best for large stereo-angles:
optimum at angles of ~10-30 deg.
Conclusions
(1) Stereoscopic reconstruction of 3D structures
can be used to detect and quantify coronal loops,
filaments, and postflare loops.
Conclusions (cont.)
(2) 3D-reconstruction of coronal loop structures is
most useful to test theoretical models of magnetic
field extrapolations.
Conclusions (cont.)
(3) Hydrostatic/hydrodynamic modeling of coronal loops
requires careful disentangling of neighbored loops,
background modeling, multi-component modeling,
and multi-filter temperature modeling. Accurate
modeling requires the identification of elementary loops.
Conclusions (cont.)
(4) The latest TRACE study has shown the existence
of elementary loop strands with isothermal crosssections, at FWHM widths of <1500 km. TRACE
has a pixel size of 0.5” and a point-spread function
of 1.25” (900 km) and is able to resolve some of them,
while AIA (0.6” pixels, PSF~1.5”=1100 km) will
marginally resolve the largest ones. Multi-filter analysis
is a necessity to discriminate elementary from
composite loops.