Quality Standards¶

speclib maintains a tiered quality system to ensure data integrity while welcoming community contributions.

Quality Tiers¶

Data Requirements¶

All spectral data in speclib must meet these minimum standards regardless of tier:

Calibration¶

• Calibrated reflectance only — all spectra must be calibrated against a known reference standard
• Raw DN values, uncalibrated measurements, and transmission spectra are not accepted
• Contributors must document their calibration method (e.g., Spectralon white reference panel, vicarious calibration, atmospheric correction)
• Reflectance values must be on a 0.0-1.0 scale (fractional, not percentage)

Wavelength¶

• Wavelengths must be in micrometers (um)
• Wavelength arrays must be sorted in ascending order
• Non-uniform spacing is acceptable (store as measured, do not resample)

Metadata¶

Every spectrum must carry:

• Material name and category classification
• Source library identifier and original record ID
• Measurement type (laboratory, field, airborne, spaceborne)
• License for data use
• Instrument used for measurement

Spectral Integrity¶

• Spectral gaps are stored as NaN, never interpolated
• Out-of-range reflectance values (outside 0.0-1.0) are flagged with a warning, never silently clipped
• Delete values from upstream sources (e.g., USGS -1.23e+34) are converted to NaN

Contribution Review Process¶

Community contributions go through a two-stage review:

Stage 1: Automated Validation (CI)¶

The CI pipeline automatically checks:

• contribution.yaml has all required fields
• calibrated: true is set
• calibration_method is non-empty
• Contributor affiliation and email are provided
• CSV data parses correctly
• Wavelengths are sorted ascending
• Reflectance values are in [0.0, 1.0]
• Wavelength and reflectance arrays are the same length

Stage 2: Maintainer Review¶

A project maintainer manually verifies:

• Contributor credentials — institutional affiliation is legitimate
• Instrument and calibration — measurement setup is appropriate for the claimed data
• Data quality — spectra look physically reasonable (no obvious artifacts, saturation, or noise issues)
• Metadata completeness — description, locality, and citation are adequate
• License compatibility — data license is compatible with the project (GPL-3.0 for code, CC-BY-4.0 recommended for data)

Promotion¶

After merging, community contributions remain in the COMMUNITY tier. A maintainer can promote contributions to CURATED status after additional verification (e.g., comparing against known reference spectra, confirming measurement conditions).

Instruments¶

speclib accepts data from any calibrated spectrometer or imaging spectrometer. Common instruments include:

Drone-mounted spectrometers and hyperspectral cameras are welcome as long as the data has been radiometrically calibrated and atmospherically corrected.