ration of Artificial Data in Support of SDO Nagi N. Mansour, NASA ARC Alan Wray, NASA ARC Thomas Hartlep, Stanford CTR Alexander Kosovichev, Stanford HEPL Thomas Duvall, NASA GSFC Mark Miesch, UCAR
Generation of Artificial Data in Support of SDO-HMI
Nagi N. Mansour, NASA ARC
Alan Wray, NASA ARC
Thomas Hartlep, Stanford CTR
Alexander Kosovichev, Stanford HEPL
Thomas Duvall, NASA GSFC
Mark Miesch, UCAR
Two efforts in progress:
(1) Direct simulation of wave propagation in solar interior
(2) Large-eddy simulation of the near-surface convection zone
€
∂t ′ ρ = −Φ
€
∂t Φ = −∇2 c 2 ′ ρ ( ) +1
r2∂r (r
2g ′ ρ )
€
+ f
€
−σ ′ ρ
€
−σΦ
Randomforcing
DampingLayer aboveSurface
Wave equation
Equations
Direct Simulation of Wave Propagation in the Solar Interior
• Spherical harmonics, and Basis-Spline in radial direction
• Resolution in all directions adjustable with r• Treatment of coordinate singularity by
enforcing regularity condition at the center
• Non-reflecting boundary by means of a damping layer at the top
• Temporally random forcing of each spherical harmonics mode
NumericalMethod
€
′ ρ l,m (r) = r lP(r2)
Example of B-Splines
OscillationPowerSpectraradial-symmetricSound Speed of aStandard Solar Model
with gravity term
without gravity term
StellarBox
• Rectangular geometry
• 50 50 43 Mm
• Compressible, radiation-hydro equations
• LTE radiation, 14 ray angular quadrature
• Non-ideal (tabular) EOS; tabular, binned opacity
• 4th order Padé derivatives
• 3rd (or 4th) order Runge-Kutta in time
• No-penetration, hydrostatic-pressure b.c.’s
• MPI parallelization
StellarBox MPI code
500x500x500
0
2
4
6
8
10
12
0 100 200 300 400 500 600
Number of processors
Mega-updates/sec
Scaling results on Columbia
Current status
• 500500500 on 100 processors (typically)• ~7-8 hour runs• 700-800 steps/run• Also used on brown dwarf stars