Executive Summary
This paper establishes the rigorous mathematical foundation for orbital continuum removal and non-negative least squares library unmixing against a public laboratory spectral library. Blind cross-validation against ground-truth field spectroscopy is reported in the paper, without reliance on black-box heuristics.
Key Methodological Findings
Automated hull continuum removal preserves the diagnostic doublet separation across the shortwave infrared window.
Alunite (2.165 µm) and well-crystallised kaolinite (2.208 µm) are cleanly separated under open archive data with zero false-positive bleed into adjacent phyllosilicate domains.
Leave-one-region-out cross-validation across independent geological terranes supports the score the platform reports.
Physics & Inversion Methodology
Radiance calibrated at sensor Level 1C is atmospheric-corrected to surface reflectance using coupled radiative transfer code. Continuous absorption envelopes are continuum-removed with dynamic tie-points before multi-endmember linear unmixing with sparsity constraints.
Diagnostic Spectral Windows
| Band Regime | Wavelength Range | Physical / Chemical Feature |
|---|---|---|
| SWIR-1 | 1.40 – 1.80 µm | O-H and H2O fundamental vibration overtone features |
| SWIR-2 | 2.00 – 2.45 µm | Al-OH, Fe-OH, Mg-OH, and CO3 diagnostic lattice combination doublets |
| VNIR | 0.40 – 1.00 µm | Fe3+ electronic crystal field transitions (jarosite, hematite, goethite) |
Citation & Digital Object Identifier (DOI)
@techreport{axine_the_spectral_advantage_in_mineral_exploration_2026,
author = {Axine Labs Research},
title = {The Spectral Advantage in Mineral Exploration},
institution = {Axine Labs Technical Whitepaper Series},
number = {1},
year = {2026},
doi = {pending (Zenodo, 2026)},
url = {https://axinelabs.com/papers/the-spectral-advantage-in-mineral-exploration}
}