5 th SECCHI Consortium Meeting, May 5 2007 Tomographic Reconstruction of CMEs from White Light...

Post on 21-Dec-2015

213 views 0 download

Tags:

transcript

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Tomographic Reconstruction of CMEs from White Light

Coronograph Data

FITS ingest, Visualization, and PIXON Current State

Alex Antunes, J.W. Cook, J. Newmark (NRL)

A. Yahil (Stony Brook University)

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Abstract

We discuss our 3D tomographic reconstruction approach with PIXON for SECCHI and LASCO data: setting up the geometry using datafile FITS headers, calculation of statistical noise, and input/output for the PIXON tool. In previous meetings we have discussed our PIXON 3D reconstruction tool and shown early results from synthetic modeled coronal structures. 3D tomographic reconstruction works best with multiple overlapping yet distinct viewpoints, while early in the STEREO mission, we are still at small angles of separation. We illustrate reconstruction geometry and set-up for SECCHI Cor2 data. We also discuss issues in incorporating HI data in reconstructions.

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Intent: Determine underlying ne density

of a CME

Means:

● FITS ingest

● Solver (Pixon or conjugate gradient or FM)

● Visualizing Datacubes

Sample: Chen flux rope with noise

Discussion and 'To Do'.

3D Reconstruction Update

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

FITS Ingest

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Fit to the noise, then stopDN to Photons:

dataphotons = dn2photons * (dataDN – biasmean)sigmaphotons= √dataphotons

sigmaDN = sigmaphotons/dn2photons

Fractional Method:Noise = dataL1.0 * sigmaDN/dataL0.5 in DN

secchi_prep Method:Noise = secchi_prep[sigmaDN]

Result= the same

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Solving (and LOS) Problems

Original Data, 45º apart

Rendered Solution

3-axis projection of solution datacube

Inverse modeling

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

What is Pixon?(We interrupt for some definitions)

● "Data": 2-dimensional images, from a spacecraft or created via a

rendering of a model.

● "Image": a 3-dimensional data cube containing electron density

measurements. An image is rendered to produce data.

● Pixon: software using the PIXON algorithm for reconstruction.

"Classic" uses a cartesian grid, "Tetrahedral" uses an arbitrary grid.

● Raytrace WhiteLight, a renderer with front-end GUI.

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

And Visualize!

(a busy IDL desktop)

● Row of images tv_multi,array_of_data

● Datacube axes threeview, imgcube

● Interactive 3-D GUI render_rot_gui,imgcube

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Sample 1: Chen Fluxrope Model

Model at 0º, 45º, 90º

Same, with noise added

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Fluxrope Reconstruction

Noisy model at 0º, 45º, 90º

Reconstruction at 0º, 45º, 90º

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Fluxrope Solution Densitycube

view down 'x' view down 'y' view down 'z'

Linear plot

Log plot

(Fit for 408 minutes, not to completion)

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Computational LimitsPrimary limit is memory, second is run-time.

Theory:●Typical 32-bit architecture can address 2GB of memory,

for 8 byte N3 arrays → N≤812.

Practice:●32-bit IDL can rarely allocate multiple large-N arrays: with 2GB RAM, IDL managed only (4) 5123 float arrays●Pixon uses N3 * 1.2×10-7 GB. N=512 → 16GB swap●RAM max space must be unfragmented, contiguous memory

Today:● Pixon N=256 runs (barely) with 2GB of physical RAM● Plan is to test N=512 on a 64-bit RAM=16MB system

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

To Do

1) Complete testing, GUI

2) Reconstruction of A/B Cor2 + LASCO C2 CMEs

3) Reconstruction of A/B/LASCO events, multiple instruments

4) Reconstruction of an A/B HI CME (or comet?)

5) Higher resolution reconstructions (current limit, N=256)

6) Mix of inverse and forward methods

7) Commit software to SolarSoft archive

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data

Contact Info

Alex (Sandy) Antunes

alexander.antunes@nrl.navy.mil

http://ares.nrl.navy.mil /~antunes

5th SECCHI Consortium Meeting, May 5 2007Tomographic Reconstruction of CMEs from White Light Coronagraph Data