Guidance Checklist on Expressing Uncertainty Information

The following list, based on USGS Landsat 8 OLI L1TP products, represents the primary contributors to per-pixel radiometric uncertainty which should be measured/estimated in order to develop/provide a per-pixel product uncertainty.

Primary Components: 

• SI Traceability (Calibration) Uncertainty
  • Can include SI-Traceable Source, laboratory equipment, sensor hardware, etc.
  • Provides the scene-independent baseline/floor for uncertainty
• Noise Interpolation Propagation
  • Derived from per-detector noise model and interpolation kernel
  • Significant for darker targets and low Signal-to-Noise sensors
• Interpolation Modeling Uncertainty
  • Error due to interpolation not perfectly reconstructing the underlying signal
  • Strongly scene-dependent (high-contrast edges, urban areas, high spatial frequency content)
• Coupled Geometric–Radiometric Uncertainty
  • Radiometric error due to pixel location uncertainty (i.e. radiometry of a different target being presented as another target)
  • Proportional to the product of image gradient x geometric uncertainty
• Bad Pixels / Saturation / Defective Detectors
  • Includes samples from defective detectors, saturated pixels, dropped data, etc.
  • These do not have a valid uncertainty estimate and are masked and propagated through the interpolation kernel

Additional material is available below  

The WGCV-56 presentation "Discussion: Expressing Uncertainty Information" is available here

The WGCV-56 presentation "Outcomes from 2026 Uncertainty Workshop" is available here

Links

CEOS Committee on Earth Observation Satellites: CEOS.org

CEOS WGCV Working Group on Calibration and Validation: CEOS WGCV