Validation Metrics¶
This page documents the validation metrics generated by cross-sensor-cal and recorded primarily in the QA JSON file. These metrics characterize reflectance quality, BRDF stability, harmonization accuracy, and mask distributions.
Reflectance metrics¶
- min/max/median per band
- out-of-range counts (>1.5 or <0)
- percent masked
- per-band histograms (in QA PNG/PDF)
High out-of-range values may indicate issues in BRDF correction or DEM artifacts.
Mask metrics¶
- cloud fraction
- cloud-shadow fraction
- snow fraction
- water fraction
- invalid/no-data fraction
These metrics help identify flights with adverse conditions or sensor anomalies.
BRDF metrics¶
- per-band BRDF model coefficients
- reconstruction RMSE
- flags for unstable or ill-conditioned fits
These diagnostics help identify pixels or bands where geometric normalization may be unreliable.
Convolution / harmonization metrics¶
- brightness adjustment coefficients
- per-band regression statistics (if applicable)
- spectral alignment differences
- RMSE for harmonized vs. reference spectra
These values help assess the quality of sensor harmonization.
Geometry and ancillary metrics¶
- solar zenith / azimuth
- view zenith / azimuth
- DEM slope/aspect summaries
These values allow interpretation of scene illumination conditions.
Using metrics for quality control¶
Typical red flags:
- Large brightness shifts (>0.1 reflectance)
- High invalid fraction (>20%)
- Poor BRDF fits (RMSE > threshold for the site/sensor)
- Unusual wavelength alignment differences
These indicators suggest reviewing the flight line before using it in ecological analyses.