SHORT TERM PREDICTION OF SOLAR ACTIVITIES

1
Automated McIntosh-based Classification of
Sunspot Groups from Solar Images
Tufan Colak, Rami Qahwaji EIMC, University of
Bradford
SIP WORKSHOP III Dublin 08/09/06
2
Objective
3
Solar Images

• CA II K3 - Meudon

• MDI Continuum-SOHO

01 MAY 2001 1123
01 MAY 2001 0000
4
Solar Disk Detection
Group (object) Detection
Original Image
Binary Image
Mask Image
Projection
Solar Disk
5
Solar Feature Detection

• Divide the image into regions.
• Find the mean (µ) of the each region.
• Using equations (1) and (2) define a threshold
  value.
• The value of every pixel is compared against its
  local threshold value and is replaced by a white
  pixel only if its initial value is smaller than
  the intensity threshold for sunspots and larger
  than the intensity threshold for active regions.

Sunspots
Active Regions
Local Intensity Filtering
a0.9
a0.3
6
Solar Feature Detection
Local Intensity Filtering
Modified Region Growing
Firstly the adjacent seeds are combined and
secondly, the unwanted seeds are eliminated.
7
Solar Feature Detection
8
Grouping of Sunspots
90 30 0 -30 -90
90 30 0 -30 -90
90 30 0 -30 -90
Carrington Coordinate System
9
Grouping of Sunspots
90 30 0 -30 -90
90 30 0 -30 -90
90 30 0 -30 -90
0
Carrington Coordinate System
10
Grouping of Sunspots
11
Classification of Sunspot Groups
McIntosh Classification

• Modified Zurich Class
• Penumbra Largest Spot
• Distribution

12
Classification of Sunspot Groups
Umbra Penumbra
PixelltMean Umbra
13
Classification of Sunspot Groups
Polarity
(McIntosh 1990)
Calculate the number of spots. Calculate the
distance between sunspots within a group.
14
Classification of Sunspot Groups
McIntosh Classification

• Modified Zurich Class
• Penumbra Largest Spot
• Distribution

15
Classification of Sunspot Groups
Rudimentary??? Mature???
Rudimentary
Mature

• SOLUTION??
• Calculate the number of pixels, starting from
  umbra to N,S,E,W.
• Compare the values

Rudimentary
(McIntosh 1990)
16
Classification of Sunspot Groups
Symmetry N-S Diameter
- Calculate the Horizontal and Vertical
projections
- Look for abnormalities
SYMMETRIC
ASYMMETRIC
17
Classification of Sunspot Groups
McIntosh Classification

• Modified Zurich Class
• Penumbra Largest Spot
• Distribution

18
Classification of Sunspot Groups
Distribution
-Count the spots with mature penumbra

• Calculate the number of spotsgt1
• Find ratio between this and total number of
  sunspots
• Ratio gt0.50 intermediate

(McIntosh 1990)
19
Classification of Sunspot Groups
McIntosh Classification

• Modified Zurich Class
• Penumbra Largest Spot
• Distribution

20
Results
01/MAY/2001 0000
G1 Cen(x,y)447, 258 Heli(Lat,Lon)N27E8
Nss1 Len1.4687 Hsx G2 Cen(x,y)181,
288 Heli(Lat,Lon)N24E49 Nss14 Len16.3691
Fao G3 Cen(x,y)932, 331 Heli(Lat,Lon)N20W68
Nss1 Len4.55448 Cko G4 Cen(x,y)545,
419 Heli(Lat,Lon)N6W3 Nss13 Len11.3046
Eao G5 Cen(x,y)856, 571 Heli(Lat,Lon)S9W46
Nss1 Len2.28397 Hax G6 Cen(x,y)221,
572 Heli(Lat,Lon)S10E37 Nss3 Len4.55607
Bxo G7 Cen(x,y)328, 572 Heli(Lat,Lon)S10E22
Nss1 Len0.241203 Axx
01/MAY/2001 0020
11010501 0020 N28E07 A 9442 HSX
1 1 40 0105 1.6 0105 1.6 184
2LEAR 11010501 0020 N26E48 B 9445
EAO 22 12 350 0105 4.7 0105 4.7 190
2LEAR 11010501 0020 N20W69 BG 9433 DKI
9 7 390 010425.8 010425.1 175
2LEAR 11010501 0020 N07W06 B 9441
EAO 20 11 230 010430.6 010430.7 185
2LEAR 11010501 0020 S10W47 A 9436
HAX 2 2 110 010427.6 010427.5 181
2LEAR 11010501 0020 S10E36 B 9444
BXO 4 4 30 0105 3.7 0105 3.8 191 2LEAR
11010501 0700 S11E21 B 9447A CRO 3
2 10 0105 2.9 0105 2.7 155 3SVTO
National Geophysical Data Centre NGDC
21
Grouping of Sunspots
Results
Fao N24E49
Hsx N27E08
Cko N20W68
Eao N06W03
Axx S10E22
Hax S09W46
Bxo S10E37
22
Results

• Tested on 15 consecutive images and 5 separate
  images so far.
• Grouping
• Over 146 groups, 122 groups detected correctly.
• Classification (On the 122 correctly detected
  groups.)
• Modified Zurich Class 97 correct
• Penumbra Largest Spot 89 correct
• Distribution 92 correct
• Timing (on 2 GHz)
• Grouping lt 2 seconds
• Classification lt0.5 seconds

23
Conclusions Discussions

• Automated classification is possible. (75 to 79
  success rate for correct grouping so far)
• Need for more images.
• Classification doesnt work well for spots close
  to the limb (After 60 degrees).
• Grouping doesnt work well , if there are more
  than one sunspot groups within an active region.
  (Putting a distance limit can be solution)

24
Future work

• More testing.
• Implementation of machine learning in various
  stages.