Pipeline
QA panels and metrics¶
Each completed flight line writes visual and machine-readable QA artifacts so correction, harmonization, and export decisions remain inspectable after the run.
QA PNG
Fast visual triage for reflectance, masks, brightness shifts, and sensor diagnostics.
QA PDF
A longer-form audit record when multi-page reporting is enabled for the workflow.
QA JSON
Machine-readable metrics and metadata for automation, validation, and reproducible review.
Visual QA
What the QA PNG is for¶
The PNG is the fastest way to spot unusual behavior in a flight line without loading the raster products manually. Typical panels summarize:
- reflectance distributions for representative wavelengths
- spatial mask coverage and invalid-pixel structure
- before and after brightness behavior
- BRDF correction diagnostics
- wavelength and target-sensor alignment checks
Use the PNG for quick triage before deciding whether deeper inspection is necessary.
Audit outputs
What the PDF and JSON capture¶
QA PDF
The PDF expands the visual report into a flight-line-level audit artifact with extended diagnostics, tables, optional per-band summaries, and correction details.
Think of it as the human-readable record you keep with a processing run.
QA JSON
The JSON carries the same run into a machine-readable form for dashboards, automation, and regression checks.
It is the better interface when you need to compare runs systematically.
Common QA JSON categories include:
Reflectance and masks
Bandwise min, max, median, saturated or masked fractions, and cloud, snow, shadow, water, or invalid summaries.
Correction diagnostics
BRDF coefficients, reconstruction error summaries, and brightness differences across correction stages.
Sensor harmonization
Brightness coefficients, spectral alignment checks, and regression-based metrics when that path is active.
Geometry context
Solar and view angles plus terrain summaries that help explain unusual correction behavior.
Interpretation
What to watch for¶
Reflectance above expected bounds
Values well above 1 can point to correction instability, geometry problems, or DEM issues that deserve review.
Large brightness shifts
Substantial step changes between raw, corrected, and harmonized products usually warrant a closer look at the relevant stage outputs.
High invalid or shadow fractions
These can dominate downstream summaries, especially for difficult illumination conditions or partial-scene coverage.
Noisy coefficients
Unstable BRDF or harmonization coefficients often indicate low-SNR regions or other conditions that reduce correction reliability.
QA outputs are part of the workflow contract, not optional polish. They are meant to preserve transparency when a downstream analysis needs to explain how a product was generated.
The focused stage reports and combined cross-stage synthesis are documented in Stage-by-stage scientific QA. They add explicit PASS, WARN, FAIL, and NOT EVALUATED checks without replacing the existing flightline PNG, JSON, or PDF.
Where to go next