Forecasting Coronal Mass Ejections from Magnetograms ...lasp.colorado.edu/sdo/meetings/session_4_5_6/... · In the red intervals the gradient of the line-of-sight field is ≥ 50

D. A. Falconer (UAH/MSFC/NSSTC), R. L. Moore, G. A. Gary (NASA/MSFC/NSSTC)

S. Balasubramanian (UAH/ NSSTC)

Forecasting Coronal Mass Ejections fromMagnetograms: Active Region

Nonpotentiality, Complexity, and Size

Forecasting Coronal Mass Ejections from MagnetogramsD. A. Falconer (UAH/MSFC/NSSTC), R. L. Moore, G. A. Gary (NASA/MSFC/NSSTC) S.

Balasubramanian (UAH/ NSSTC)

Objective

Assess measures of active-region nonpotentiality, complexity, and size as predictors ofCME productivity.

Published

From a sample of 17 bipolar active regions, we showed that measures of active-regionnonpotentiality from vector magnetograms are correlated with active-region CME productivity(Falconer, Moore, & Gary 2002, ApJ, 569, 1016). One measure of nonpotentiality can bemeasured using only a conventional line-of-sight magnetogram (Falconer, Moore, & Gary, JGRSpace Physics, 108(A10), 1,380).

New

1. We have expanded the bipolar sample to 36 vector magnetograms of 31 active regions (Falconer,Moore, & Gary 2004 submitted).

2. We have begun to expand our sample to multi-bipolar active regions (12 at present).

3. We have found a measure of active region complexity that is promising as a predictor of CMEs.

Active-Region CME ProductivityAn active region’s CME productivity during a window of time. Primary window used is ±2 days,but also have evaluated windows of ±1, ±4, 0-1, 0-2, and 0-4 days.

Total magnetic flux

Best constant alpha

Current magnetic twist

Length of strong-gradient Neutral Line

Length of strong-gradient main Neutral Line

Length of strong-shear Neutral Line

Length of strong-shear main Neutral Line

Net Current

Name

Size/Flux=∫BZda, |BZ|>100 G_

Normalized Nonpotentialityminimizes=∫(Bt sin(φo- φf)

2_BC

Normalized Nonpotentiality=_IN/__IN

Non-normalized NonpotentialityFig 3LSG

Non-normalized NonpotentialityFig 1LSGM

Non-normalized NonpotentialityFig 3LSS

Non-normalized NonpotentialityFig 1LSSM

Non-normalized Nonpotentiality(=∫BT•dl)IN

Type of MeasureDefinitionSymbol

Active-Region Magnetic Measures

1. LSGM, measurable from line-of-sight magnetograms (e.g. MDI and Kitt Peak), is correlated with ourother measures of nonpotentiality and is well correlated with active-region CME productivity.

2. Non-normalized measures of nonpotentiality: IN, LSSM and LSGM are strongly correlated (ConfidenceLevel ~99.96%) with each other.

3. Normalized measures of nonpotentiality: αI and αBC are well correlated (Confidence Level ~ 99.5%).

4. Total magnetic flux has a statistically-significant correlation with the non-normalized measures(Confidence Level ~97%) but not with normalized measures (Confidence Level ~55%) (Table 3, Figure2).

5. Normalized and non-normalized measures have statistically significant CME prediction ability(Confidence level 97-99.7%, success rates 75-80%).

6. Total magnetic flux does not have a statistically-significant correlation with CME productivity(Confidence Level ~65%), its success rate though is ~60% which is consistent with Canfield, Hudson,& McKenzie 1999 correlation between sunspot area and eruptive flares.

7. Due to ease of measuring, LSGM and LSSM are the two most promising measures for CME forecasting.

Finding from Multi-Bipolar Study

1. Generalized LSG and LSS have similar predictive ability as their main neutral line counterparts (LSGMand LSSM).

Findings from Bipolar Study

_= ∫ BZda (BZ>100G) LSSM: length of the red sections.

LSGM: length of the red sections. IN =∫Bt⋅dl, of red contours

Figure 1 The Magnetic Measures. Upper Left: Part of a MSFC vector magnetogram showing anonpotential bipolar active region, with ±25G, 100G, 250G, and 500G contours of line-of-sight magnetic field(solid contours: positive). The field of view is 115x115 arcsec. The potential (arrows) and observed (dashes)transverse fields are shown with only 1 in 49 pixels plotted. The shortest arrows (or dashes) are 150G. UpperRight: The main neutral line on which the observed transverse field is greater than 150G is the thick red andgreen curve, with the color indicating whether the shear angle is larger (red) or smaller (green) than 45o. LSS isthe length of all neutral line segments with strong shear. For a purely bipolar active region such as this LSS isequal to LSSM. Lower-Left: The main neutral line on which the potential transverse field is greater than 150Gis the thick red and green curve, with the color indicating whether the gradient of the line-of-sight field isstronger (red) or weaker (green) than 50G/Mm. LSG is the length of all neutral line segments with stronggradient. For this active region LSG is equal to LSGM. Lower-Right The 180o ambiguity of the observedtransverse field has been resolved, with the arrows representing the observed field. αIN= µIN/Φ, where the redcontours are the integration contours used to obtain IN from Ampere's law. αBC is the alpha that minimizesthe quantity ∫(Btsin(_o-_f))

2da.

Figure 2 Correlation of magnetic measures with CME productivity and with each other. Each panel shows onemagnetic measure plotted against another for 36 magnetograms marked with either an X or an O. The X’s representactive regions that produced CMEs, within 0-2 days of the day of the magnetogram; the circles represent activeregions that did not. The X’s tend to be above (or right) of the threshold lines, and the O’s tend to be below (or left).An exception to this rule is in the rightmost panel, where the O’s and X’s have no strong correlation with the fluxthreshold (horizontal line). The confidence level of each correlation is the percentage shown and the correspondingsuccess rate is the percentage in the parenthesis below the confidence level.

Bipolar Active Regions, 36 Magnetogram Sample

LSGM (Mm) IN (A) LSSM (Mm)_ I

N (1

0-8 m

-1)

_ BC

(10-8

m-1

)

_(M

x)

LSS

LSG

Generalized Neutral Line Length Measures for Multibipolar Active Regions

Figure 3. LSS and LSG in a CME-productive multi-bipolar active region. Upper left panel: MSFC vectormagnetogram of AR 9628 on 25 September 2001, showing the line-of-sight field (contours, 25, 250, 500 G, solid forpositive polarity, dotted for negative polarity), the observed transverse field (green dashes), and the potentialtransverse field (red arrows). The longest dashes and arrows are for transverse field strengths ≥ 500 G and theshortest are for 150 G. To avoid overlap, the dashes/arrows are centered 7 pixels apart in x and y; each dash/arrowshows the transverse field found in its centered pixel. Upper right panel: The strong-observed-field (observedtransverse field ≥ 150 G) segments of the neutral lines displayed in the gray-scaled line-of-sight component of thevector magnetogram. In the red intervals the shear angle is ≥ 45°; in the green intervals the shear angle is < 45°. LSSis the total length of all of the red intervals. Lower left panel: The strong-potential-field (potential transverse field ≥150 G) segments of the neutral lines displayed in the contoured line-of-sight component of the vector magnetogram.In the red intervals the gradient of the line-of-sight field is ≥ 50 G/Mm; in the green intervals the gradient is < 50G/Mm. LSG is the total length of all of the red intervals. Lower right panel: LASCO C2 running-difference image ofa CME that exploded from this active region two days after the day of the magnetogram. The black and white“confetti” noise in this image is from high-energy particles produced by this CME and is a striking signature of thehazardous space weather produced by halo and west-limb fast CMEs.

Figure 4 Our generalized nonpotentiality measures as predictors of CME productivity in the 0-2 day window. Eachpanel is similar to those in Figure 2, with the main-neutral line version of the measure plotted against the generalizedversion of the measure. The generalized measures using all neutral lines are as reliable as the main neutral linemeasures for CME prediction.

Generalized Measures versus Main Neutral Line Measures

LSGM (Mm)LSSM (Mm)L

SG (M

m)

LSS

(Mm

)

Figure 5 Generalized neutral line length measures applied to the bipolar, multibipolar and combined samples.Each panel is similar to those in Figure 2. For each sample, the generalized measures have ~ 75% success rate inpredicting whether an active region will or will not be CME productive in the 0-2 day window. Note how theconfidence level varies in the three cases even though the success rate is the same. This is due to the differentsample sizes (36, 12, and 48 respectively).

LSS (Mm)

LSG

(Mm

)

LSG

(Mm

)

LSG

(Mm

)

LSS (Mm) LSS (Mm)

Bipolar Sample Multibipolar Sample Combined Sample

Figure 6: Comparison of complexitymeasures and preliminary results ofthe measures as predictors of CMEproductivity. The two magnetogramsindicate the range of complexity of activeregions classed as primary bipolar in ouroriginal 17 magnetogram sample. Davidand Ron independently ranked themagnetic complexity of each line-of-sight magnetogram by visual inspection(middle, left). The number given (0.82)is the Spearman rank coefficient,indicating a strong positive correlation.We have also explored three quantitativemethods of ranking active-regioncomplexity: 1) Percentage of the strongfield neutral-line length that is not fromthe main neutral line. 2) The number ofSunspots. 3) Percentage of flux that isnot part of the main bipole . Thecorrelations of the ranking by thesemethods with David’s subjective rankingare shown. The % magnetic flux not partof the main bipole has the strongestcorrelation. The lower middle tableevaluates the correlation of David’ssubjective ranking with whether theactive region was classified as a gammaactive region. The lower-right tablegives the success rate of each method inpredicting future CME productivity (0-2days).

33Not Gamma

83Gamma

≥7th<7thActive Region

Visual ComplexityRank

Confidence Level= 66%

Success Rate=65%

52%Gamma Classification

82%% of Flux not in MainBipole

70%Number of Sunspots

70%% of SecondaryNeutral Line Length

64%Visual Ranking

Success Rate inPredicting

CMEs

Predictor

Complexity as a Predictor of CMEs

Table 3 Correlation Confidence Levels and Agreement Numbers forPairs of Magnetic Measures.

LSSM LSGM αIN αBC Φ IN 99.999% (32) 99.96% (30) 99.999% (32) 87% (24) 97% (26) LSSM - 99.999% (32) 99.5% (28) 97% (26) 99.5% (28) LSGM - - 99.5% (28) 87% (24) 97% (26) αIN - - - 99.5% (28) 65% (22) αBC - - - - 26% (18)

Notes:1.The confidence level of the correlation is given first and the agreement number isgiven in parentheses for each pair of measurements.2. The confidence level is determined from the Fisher Test. It is 100% x (one minusthe chance that the population of the 2x2 contingency table could occur by randomchance).3. Agreement number: number of times (out of 36) the two measures agree, i.e., areeither both at or above or both below threshold.

Table 4 Correlation of Measures with CME Productivity ±2 days ±1 days ±4 days 0-2 days 0-1 days 0-4 days LSSM 99.5% (28) 98% (27) 99.9% (28) 97% (27) 92% (26) 95% (26) LSGM 99.96% (30) 99.8% (29) 99.9% (28) 99.7% (29) 99% (28) 99.3%(28) IN 99.5% (28) 98% (27) 99.9% (28) 97% (27) 92% (26) 95% (26) αIN 99.5% (28) 98% (27) 99% (26) 97% (27) 92% (26) 82% (24) αBC 99.96% (30) 98% (27) 99.9% (28) 97% (27) 92% (26) 82% (24) _ 65% (22) 35% (21) 80% (22) 57% (23) 34% (22) 82% (24)

Note: Number in parentheses is the number of correct predictions out of 36.

Table 2 Example 2x2Contingency Table

Confidence level 99.999%Agreement number 32

NoteTh. Stands for threshold

Statistical Results for 36 Bipolar Vector Magnetograms of 31 BipolarActive Regions

361224Total

=24222< Th.

=12102≥ Th.

Total≥ Th.< Th.IN

LSSM