cross-sensor-cal

Stage 04 Spectral Library

DO NOT EDIT OUTSIDE MARKERS The spectral library stores every spectrum with its full provenance. Each entry includes:

These fields live alongside the reflectance values in a row-oriented table or NetCDF group. Use them to filter spectra by site, compare sensors, or align data to the wavelength grid.

Quality controls

You can clean spectra before analysis:

  1. Outlier filtering – drop samples that exceed three standard deviations from the mean reflectance at any wavelength.
  2. Smoothing – apply a Savitzky–Golay or moving-average filter to reduce instrument noise while preserving absorption features.
  3. Signal-to-noise ratio (SNR) – flag spectra with low SNR and exclude them from downstream modeling.

Versioning and provenance

Each library release increments a semantic version. A manifest.json file lists source datasets, processing code commits, and software versions so you can reproduce the library or audit its origin.

Last updated: 2025-08-18