Knowledge Garden

NOAA AR 11158 as an NF2 benchmark

2 min read

NOAA AR 11158 as an NF2 benchmark

NOAA active region 11158 is Andrew’s first NF2 reproduction target. In the HMI SHARP series it corresponds to HARP 377.

Why this region is useful

A good benchmark should be scientifically interesting, reproducible, and available in a clean data product. AR 11158 fits that bill.

It is useful because it is:

  • eruptive and well studied;
  • linked to major flaring activity;
  • discussed in the solar-physics literature, including helicity and eruption studies;
  • used in the NF2 repository’s own example workflow;
  • available through the HMI SHARP CEA vector-magnetogram series.

So it is not an arbitrary active-region choice. If the pipeline cannot produce credible fields for AR 11158, the problem is likely in the setup, data convention, training configuration, or diagnostics.

Local smoke-test record

Local NF2 project path:

/Users/andrew/MISCADA/Project_COMP52060/NF2

Smoke-test data query:

hmi.sharp_cea_720s[377][2011.02.15_00:00:00_TAI]{Br,Bt,Bp,Br_err,Bt_err,Bp_err}

Downloaded data path:

/Users/andrew/MISCADA/Project_COMP52060/NF2/data/ar11158_single

Smoke config:

/Users/andrew/MISCADA/Project_COMP52060/NF2/config/ar11158_minimal.yaml

Verified outputs:

/Users/andrew/MISCADA/Project_COMP52060/NF2/runs/ar11158_minimal/extrapolation_result.nf2
/Users/andrew/MISCADA/Project_COMP52060/NF2/runs/ar11158_minimal/last.ckpt

What the smoke run proves

The minimal smoke run is deliberately tiny: cropped data, binning, small model dimension, two iterations, one epoch, and CPU execution.

It proves the pipeline wiring works. It does not prove the extrapolation is scientifically meaningful.

Benchmark question

The useful comparison question is:

Given the same HMI boundary data and comparable preprocessing,
does the NF2 reconstruction produce plausible coronal connectivity,
divergence control, force-free behaviour, and magnetic-energy estimates?

That is stronger than asking whether the code runs or whether the field-line plot looks reasonable.

Graph View

Backlinks