NF2 FITS loader component conventions

NF2’s default FITS loader expects CEA vector component paths keyed like this:

fits_path:
  Br: path/to/Br.fits
  Bt: path/to/Bt.fits
  Bp: path/to/Bp.fits
error_path:
  Br_err: path/to/Br_err.fits
  Bt_err: path/to/Bt_err.fits
  Bp_err: path/to/Bp_err.fits

The relevant local source file is:

/Users/andrew/MISCADA/Project_COMP52060/NF2/nf2/loader/fits.py

Magnetic-field mapping

Inside FITSDataset, the loader reads Bp, Bt, and Br, applies the same crop/binning to each map, and then stacks:

b = np.stack([p_map.data, -t_map.data, r_map.data]).transpose()

So the boundary field used by NF2 is:

NF2 Bx = Bp
NF2 By = -Bt
NF2 Bz = Br

This is the mapping to cite in the report. Do not describe it vaguely as “CEA to Cartesian”; the sign convention is the bit that can bite.

Error mapping

The error maps are stacked as:

b_err = np.stack([p_error_map.data, t_error_map.data, r_error_map.data]).transpose()

Therefore:

NF2 Bx_err = Bp_err
NF2 By_err = Bt_err
NF2 Bz_err = Br_err

There is no sign flip on Bt_err because the error is an uncertainty magnitude.

Why this note matters

Coordinate mistakes are the classic silent failure in magnetogram pipelines. A wrong sign can still produce a smooth field-line plot, but the inferred handedness, currents, and topology can be wrong.

For AR 11158 and other SHARP CEA runs, this mapping is the local source of truth.