¹Xing Wu, ²Beatrice Baschetti, ^1,3Yang Liu, ²Cristian Carli, ^1,3Xiang Zhou, ¹Yazhou Yang, ¹Yongliao Zou
Icarus (in Press) Link to Article [https://doi.org/10.1016/j.icarus.2026.117235]
¹State Key Laboratory of Space Weather, National Space Science Center, Chinese Academy of Sciences, Beijing, 100190, China
²Italian National Institute for Astrophysics (INAF) – Institute for Space Astrophysics and Planetology (IAPS), Via del Fosso del Cavaliere, 100, Rome 00133, Italy
³College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing 100049, China
Copyright Elsevier

Quantitative spectral unmixing is essential for identifying and characterizing mineral assemblages on planetary surfaces. The Hapke and Shkuratov models are the most widely used radiative transfer models (RTMs) for interpreting reflectance spectra, yet their quantitative accuracy remains largely untested due to limited laboratory validation. In this study, both models are applied to binary and ternary laboratory powder mixtures composed of phyllosilicates, including nontronite and saponite, sulfates represented by hexahydrite, and basaltic analogs relevant to Martian surface materials. The effects of endmember variability, spectral noise, and spectral sampling interval on unmixing performance are systematically evaluated. The results show both models reproduce measured spectra and compositional trends accurately when correct endmembers are used, achieving abundance retrievals within ~10 wt% for binary mixtures and ~ 15 wt% for ternary mixtures, except in systems with strong reflectance contrasts. The Shkuratov model provides lower errors and greater stability overall, whereas the Hapke model shows slightly better tolerance to compositional mismatch between Al-rich and Al-poor nontronite. Incorporating multiple compositional variants as an endmember bundle effectively mitigates mismatch effects. Estimated grain sizes fall within realistic physical ranges but show large uncertainties for bright or spectrally neutral materials, reflecting reduced model sensitivity to grain size variations. Additionally, we also find the unmixing performance remains robust under spectral noise levels of at least 25 dB and spectral sampling intervals up to 50, consistent with the capabilities of Mars orbital instruments. These results demonstrate that radiative transfer based unmixing, particularly using the Shkuratov model, provides a reliable and physically grounded framework for quantitative mineralogical analysis of Martian hyperspectral data.