HMI & Photospheric Flows 1. Review of methods to determine surface plasma flow; 2. Comparisons between methods; 3. Data requirements; 4. Necessary computational resources; 5. Possible improvements to methods.
HMI & Photospheric Flows
1. Review of methods to determine surface plasma flow;
2. Comparisons between methods; 3. Data requirements;4. Necessary computational
resources; 5. Possible improvements to
methods.
General Approach• From 2D data arrays, f1(x1,x2) & f2(x1,x2), find vector
flow v(x1,x2) consistent with:
1. Observed evolution, f(x1,x2) = f2(x1,x2) – f1(x1,x2)
2. Other possible assumptions: – Magnetic induction eqn., Bn/t = t(vnBt-vtBn)– Continuity equation, f/t + t(vtf) = 0 – Doppler velocities – more later
• v(x1,x2) might have 2 or 3 components
General Approach, cont’d:
• Ideally, with finite difference equations, cadence should beat “Courant cadence” tC = x/vmax
• analog of numerical Courant condition: – time step limited by propagation speed of information
x pixel size; vmax expected max. flow speed
• “low cadence” is t > tC
t tC very rare in solar physics!
(usually, t >> tC)
• Pixels = .5” ~ 363 km, resolution ~ 1.5” ~ 1100 km• Photospheric csound (kT/m)1/2 9 km/s • Courant Cadence:
tHMI (363 km)/(9 km/s) 40 sec.
• LOS Mag. Field Cadence, tLOS ~ 60 sec. • Vector Mag. Field: tVEC ~ 600 sec.
• Typical v ~ 2 km/s, and resolution ~ 1100 km, so tPRACTICAL ~ 550 sec.
HMI Capabilities
Current Methods
1. Local Correlation Tracking (LCT)
2. “Inductive” Methods (ILCT, MEF, …)
3. Feature Tracking (FT)
1. Local Correlation Tracking (LCT)
• Take subregions, pixels wide, of f1 & f2, find, e.g.,– shift x that minimizes difference f ; or– shift x of peak in (Fourier) correlation func’n
• Sub-pixel shifts found by interpolation – SLOW!• Most algorithms solve advection equation,
f/t + (vtt) f = 0• Can be used on intensity images, LOS, & vector
magnetograms from HMI. • Cadence must be slow enough that fnoise < fadvection
• Workable with very low cadence data: t 100tC
27 Jan 2005 HMI Planning Meeting 7
LCT applied to magnetograms: Démoulin & Berger’s (2003) analysis of flux transport velocity
Motion of flux across photosphere, uf, is a combination of horizontal & vertical flows acting on non-vertical fields.
nn
ttf v
B
Bvu 0)B(
B
nftn
tu
LCT, cont’d
Hence, flows uLCT from LCT on magnetograms:1. are not generally identical to plasma velocity v 2. solve advection equation, not continuity equation
1. Given vector B, can assume uf = uLCT, and thereby find v from uLCT algebraically (ADC).
2. Q: How good does LCT do? A: Pretty good!
27 Jan 2005 HMI Planning Meeting 9
27 Jan 2005 HMI Planning Meeting 10
27 Jan 2005 HMI Planning Meeting 11
A Comparable Data Set:Flare Genesis Experiment
• Balloon-borne (Antarctic) observations of NOAA 8844, 25 Jan 2000
• 54 vector magnetograms, ~2.3/5.3 min. per
• hi-res: .18” pixels (130.5 km), ~520 x 520 pix
• LCT differenced over ti +/-10, for t ~85 min.
• Doppler maps, too! (No info. on method.)
• Tracking of white light images underway
27 Jan 2005 HMI Planning Meeting 12
FGE Movie
27 Jan 2005 HMI Planning Meeting 13
FGE: White Light vs. Mag
27 Jan 2005 HMI Planning Meeting 14
FGE: Larger
27 Jan 2005 HMI Planning Meeting 15
• Near future: Improvement in sub-pixel interpolation – added speed.
• Future: Convert to FORTRAN; parallelize.
• Compute on tiles, not on each pixel.
Modifications to LCT
2. Inductive Methods
• Use finite diff. approx. to magnetic induction equation’s normal comp. as add’l constraint.
• Purely inductive methods need t tC
• Methods currently available: ILCT, MEF, Kusano et al. (2002), MSR (Georgoulis et al., 2005, in prep.)
• All methods return (vx, vy, vz) at photosphere, where (vB) = 0; parallel flow unconstrained by ind’n eqn.
• Post-processing with Doppler data can give v || B– NB: NOT Doppler from Stokes I (Chae et al., 2004)
Inductively Derived Flows are Consistent with Induction Eqn’s Normal Component!
zyxzyxz B
t
B)( v)( vvBB
Directly measuredDirectly measured Derived InductivelyDerived Inductively
)v()v(
)v()v(
yzyzyxyxy
xzxzxyxyx
Bz
Bxt
B
Bz
Byt
B
vBvB
vBvB
Directly measuredDirectly measured Derived by Derived by new method?new method?
What about other components?
Derived Derived InductivelyInductively
From NLFFFFrom NLFFFExtrapolation?Extrapolation?
at photosphere, z = 0 above photosphere, z > 0
27 Jan 2005 HMI Planning Meeting 19
A) ILCT: Modify LCT solution to match induction equation
• Solve for , with 2D divergence and 2D curl (n-comp), and the approximation that uf=uLCT:
.ˆBvB n tttntnfn vBuLet
2tn
t
B
2B tLCTn u
NB: if only BLOS is known, we can still solve for , !
B) Minimum Energy Fit (MEF)
• Also uses induction equation’s normal component to derive flow, with additional assumption that integral of squared velocity is minimized.
• Applicable to vector magnetograms.
• More from D. Longcope, shortly!
Other Inductive Methods
• Kusano et al. (2002): get v from LCT flow, derive additional flow for consistency with induction equation.
• Georgoulis (2005, in prep): Use (i) “minium structure” & (ii) “coplanarity” assumptions, with (iii) induction equation to derive (iv) velocity perpendicular to magnetic field. (System overconstrained.)
27 Jan 2005 HMI Planning Meeting 22
Prelim. Comparison of Inductive Methods
• Used MHD simulations of Magara (2001)
• Given B(x,y,z=0,t), “practioners” computed v(x,y,z=0,t), and were then told actual v.
27 Jan 2005 HMI Planning Meeting 23
Some Prelim Comparisons
27 Jan 2005 HMI Planning Meeting 24
Some Prelim Comparisons
27 Jan 2005 HMI Planning Meeting 25
Some Prelim Comparisons
27 Jan 2005 HMI Planning Meeting 26
Some Prelim Comparisons
3. Feature Tracking• Useful with WL images & magnetograms.
• Algorithms:– White Light: L. Strous– Active region fields: B. Welsch, G. Barnes – Quiet Sun fields: C. DeForest, M. Hagenaar, C.
Parnell, B. Welsch
• Does not return v(x,y); rather, gives velocity of “patches” of photosphere.
• Easily incorporated in pipeline.
27 Jan 2005 HMI Planning Meeting 28
Feature Tracking in AR 8038
27 Jan 2005 HMI Planning Meeting 29
Conclusions• Planned data cadences are compatible with
existing velocity inversion algorithms.• LCT can be used to derive flows in HMI’s
intensity, LOS, and vector field maps.• ILCT, MEF suitable for determining three-
component photospheric magnetic flows. • Doppler data from Stokes’ profiles (zero crossing
of V, or central minima of Q,U) desirable.• Significant improvement in computational
performance of LCT algorithms is needed for real-time analysis.