/image

Run first-pass imaging on this calibrated MS: $ARGUMENTS

Read .claude/skills/radio-interferometry/11-imaging.md before starting. Prerequisite: CORRECTED_DATA populated on the target field(s) — run /project:calibrate (and /project:polcal for IQUV) first if not.

Workflow:

1. ms_workflow_status(ms_path, workdir) — confirm CORRECTED is populated.

2. Step 0 — confirm field and Stokes with the user per 11-imaging.md §0. Never assume mosaic vs. single pointing. Default Stokes='I' unless polcal is complete and the user wants IQUV.

3. Gather placeholders from tool outputs (no hand math):

   • ms_observation_info(ms_path) → telescope, center_freq_hz
   • ms_spectral_window_list(ms_path) → bandwidth_hz
   • ms_antenna_list(ms_path) → dish_diameter_m, n_ant
   • ms_baseline_lengths(ms_path) → max_baseline_m, expected_beam_major_arcsec
   • ms_scan_list(ms_path) → t_on_source_s for the target field

4. Derive tclean parameters per 11-imaging.md §Steps 1–7:

   • specmode (§1): 'mfs' unless user asks for a cube
   • deconvolver (§2): 'mtmfs' with nterms=2 if bw/freq > 0.2, else 'hogbom'
   • cell (§3): 1 / (max_baseline_lambda × 3), rounded to 1 sf
   • imsize (§4): primary-beam-FWHM / cell, rounded up to composite number
   • Fresnel check (§5) → wprojplanes
   • gridder (§6): 'awp2' mosaic + EVLA/ALMA, 'wproject' single-pointing + W, else 'standard'
   • threshold (§7): 3 × radiometer RMS from SEFD table; format as mJy

5. ms_tclean(ms_path, imagename, field, stokes, specmode, deconvolver, nterms, gridder, wprojplanes, cell, imsize, weighting='briggs', robust=0.5, niter=50000, threshold, savemodel='modelcolumn', workdir, execute=False) → run the generated script. Wait for completion.

6. ms_image_stats(imagename.image.pbcor, psf_path=imagename.psf) — quality metrics.

7. Apply 11-imaging.md §Step 9 gates:

   • rms_jy vs. radiometer prediction (2× tolerance)
   • dynamic_range thresholds per field type
   • beam_major_arcsec vs. expected_beam_major_arcsec (20% tolerance)
   • peak_jy sanity check

If rms_jy > 3 × radiometer estimate, escalate — the problem is calibration, not imaging. ms_residual_stats on the target's CORRECTED column before re-imaging.

Output: the derived tclean parameters, image paths, image-quality metrics with pass/fail per gate, and a recommendation (ship, re-image with different parameters, or go back to calibration).