Module vignette 5: review QA outputs¶
Notebook: View the QA notebook in the repository. GitHub displays the cells; clone or download the file to run them.
Use this module to inspect a completed or partially completed flightline without recomputing its scientific products. QA PNG files support quick visual triage; QA JSON files support reproducible checks and summaries.
Render QA from existing outputs¶
spectralbridge-qa \
--base-folder spectralbridge_output \
--quick \
--save-json
The quick path uses deterministic subsampling. Use the installed command's
--help output before changing sampling or RGB options.
Review the result¶
For each flightline, inspect:
- the QA PNG for spatial structure, masks, distributions, and correction or sensor anomalies;
- the QA JSON for machine-readable metrics, run context, and flagged issues;
- the merged Parquet for the values behind any anomaly.
QA is evidence for review, not an automatic declaration that a scientific product is valid. Investigate unexpected bounds, large brightness shifts, high invalid fractions, or unstable coefficients in context.
Render from Python¶
from pathlib import Path
from spectralbridge.qa_plots import render_flightline_panel
flight_dir = Path("spectralbridge_output") / (
"NEON_D13_NIWO_DP1_L020-1_20230815_directional_reflectance"
)
png_path, metrics = render_flightline_panel(
flightline_dir,
quick=True,
save_json=True,
)
print(png_path)
print(metrics.get("issues", []))
Continue¶
- Carry on from a partial run
- Technical details: QA panels and metrics and validation metrics