Fix low-risk QC/post-processing bugs (no catalog outputs change)

[mdenolle]

Summary

Fixes four low-risk bugs found in a QC / post-processing audit. None of these change any published catalog output — they fix a display bug, two crash bugs, and a divide-by-zero in an unused-by-the-filter metric.

The amplitude-plot fix resolves the undergrad reports about ARIDs 999395, 999382, 999384, 178428: the amplitude values were always correct, but the full-window QC plot misplaced the pick marker and amplitude window, so the true peak appeared to fall outside the purple band.

Changes

File	Fix
4_relocation/quality_control/examine_..._w_amp.ipynb	Both plot functions: replaced the fake np.linspace(-window_before, window_after, npts) time axis with tr.times(reftime=time_pick) (honors tr.stats.starttime); added sdata.merge(method=1) so gappy multi-segment channels are plotted/selected as one record instead of only tr_select[0]. Display-only.
4_relocation/event_waveform_processing.py	Guard print(picks_df.loc[mask]) with if measurements: — mask was undefined for the first event with no data → NameError.
3_post_processing/get_waveform_amplitude.py	Same crash guard.
utils/qc_utils.py (calc_snr, 3 sites)	Guard 20*log10(signal/noise) against zero amplitudes → NaN instead of inf/-inf. Not used by the QC pass/fail filter.

Why these are safe

• Amplitude plot: cosmetic; the stored Amplitude values (from calculate_amplitudes.py) are unchanged and were verified correct.
• The two batch scripts are superseded by calculate_amplitudes.py; the fix only prevents a crash.
• calc_snr's SNR column is not consumed by the rms<2.5 & p_picks>4 & s_picks>4 filter.

Verification

• All three .py files pass python -m py_compile.
• Notebook re-parses; embedded outputs untouched (diff confined to the two function cells).

Deliberately NOT included (change outputs / need a decision)

• QC pick-count threshold: qc_metrics_..._cc.ipynb uses p_picks > 4 & s_picks > 4 (≥5), but the output file is named ..._p_4_s_4_... (≥4). Reconcile intent before changing the catalog.
• Morton nearest-neighbor match has no acceptance threshold; concat_anss does no dedup / magnitude-type reconciliation.
• Aligning the qc_metrics pick-assignment inputs to the _cc catalog version.
• Removing the 3_post_processing ↔ 4_relocation duplicate files; rewriting hardcoded /wd1//home/hbito paths.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR addresses several QC/post-processing robustness and display issues: it prevents two batch scripts from crashing when no measurements are produced, corrects QC notebook time-axis alignment for amplitude windows/pick markers, and avoids divide-by-zero/-inf SNR values by returning NaN.

Changes:

• QC amplitude plotting: use true trace-relative sample times (tr.times(reftime=time_pick)) and merge gappy channels before selecting/plotting.
• Crash guards: only print picks_df.loc[mask] when measurements exist (prevents NameError).
• SNR computation: return NaN instead of inf/-inf when signal/noise amplitudes are zero.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
utils/qc_utils.py	Guards SNR computation to avoid divide-by-zero and log-of-zero producing inf/-inf.
4_relocation/quality_control/examine_Cascadia_updated_catalog_picks_assignment_ver_3_w_amp.ipynb	Fixes QC plot time axis to honor tr.stats.starttime, and merges gappy channels so selection/plotting uses the full record.
4_relocation/event_waveform_processing.py	Prevents NameError by guarding debug printing on measurements.
3_post_processing/get_waveform_amplitude.py	Same crash guard as relocation script.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.