Improved Detection of Martian Carbonates from CRISM

Mapping the global distribution of carbonate minerals on the Martian surface is a prerequisite for understanding the aqueous and atmospheric evolution of Mars, as these minerals represent a principal sink for atmospheric CO₂ and direct tracers of past liquid water activity. To date, the orbital data from CRISM offers our best opportunity for detecting the diagnostic vibrational signatures of these minerals; however, studies have focused almost entirely on the weak short-wave infrared (SWIR) overtone and combination bands near 2.3 and 2.5 µm [1,2]. These features can be spectrally degenerate with co-occurring phyllosilicates and are not always strong enough to reveal smaller or less exposed deposits, leaving a significant fraction of the carbonates undetected.

The strongest spectroscopic signatures of carbonates lie in the mid-wave infrared (MWIR): the fundamental absorptions at 3.4 and 3.9 µm [2]. These features remain largely unexploited in CRISM data, due to the limited spectral range of the instrument, which does not fully cover the 3.9 µm band, and to the superposition of reflected solar radiance and planetary thermal emission in the 3–4 µm window, which renders band depth estimation challenging.

To overcome this issue, we developed a dual-band detection framework based on two independent pipelines processing the SWIR and MWIR parts of the spectrum. While the SWIR pipeline works similarly to conventional detection methods, refining the 2.5 µm overtone absorption, the MWIR pipeline has to deal with the presence of thermal emission. Each pipeline produces a spatially filtered detection map with fully propagated, SNR-anchored uncertainties.

In particular, the MWIR pipeline operates on the atmospherically corrected datacube and treats the observed signal as the sum of a reflected solar continuum, a surface thermal contribution, and channel noise. The solar continuum is estimated using an iteratively reweighted least-squares (IRLS) fit to a band-masked window between 1.8–2.6 µm, which serves as a robust estimator with a well-defined per-pixel noise level, avoiding the limitations that affect classical shoulder-based approaches. Using these residuals, the thermal contribution is estimated by a constrained Planck greybody fit in the 3–4 µm window, with temperature values restricted to a realistic Martian daytime range. Instrumental artifacts are subsequently identified and filtered both at the channel level and in map space before proceeding with the final detection step.

The SWIR pipeline uses a pre-processed atmospherically corrected and denoised datacube, where thermal emission is absent. The 2.5 µm overtone is isolated with an iterative continuum fit and sub-pixel band-centre localisation, while band absorption is quantified via a Gaussian-kernel-weighted band-depth integration. SNR, band-centre location, and amplitude of the detected absorption serve as criteria for distinguishing carbonates from overlapping phyllosilicates and sulphates [3].

As a first demonstration, the framework is applied to a single CRISM observation, FRT000186FA at Nili Fossae, a scene chosen for its close proximity to Jezero Crater, where carbonate presence has been independently established both from orbit [4] and in situ by the Perseverance rover [5,6]. Both pipelines detect a spatially coherent area of elevated absorption,with detection maps consistent in spatial extent and structure. 

Figure 1: Band depth integral at 2.5 µm (SWIR pipeline) for CRISM observation FRT000186FA (Nili Fossae). 

Figure 2: Band depth integral at 3.9 µm (MWIR pipeline, thermal-corrected) for CRISM observation FRT000186FA (Nili Fossae). 

Nili Fossae serves here as a validation benchmark, with the framework designed for systematic application across a geologically diverse sample of CRISM observations, moving toward a more complete carbonate inventory of Mars. 

This work is funded by the Italian Space Agency (ASI) [ASI-INAF n.23-3-HH.0]

[1] Harner, P.L., & Gilmore, M.S. (2015). Visible–near infrared spectra of hydrous carbonates, with implications for the detection of carbonates in hyperspectral data of Mars. Icarus 250, 204-214. http://dx.doi.org/10.1016/j.icarus.2014.11.037 

[2] Bishop, J.L., et al. (2021). Spectral properties of anhydrous carbonates and nitrates. Earth and Space Science, 8, e2021EA001844. https://doi.org/10.1029/2021EA001844 

[3] Plebani, E.,  et al. (2022). A machine learning toolkit for CRISM image analysis. Icarus 376, 114849. https://doi.org/10.1016/j.icarus.2021.114849 

[4] Ehlmann, B.L., et al. (2008). Orbital Identification of Carbonate-Bearing Rocks on Mars. Science 322, 1828-1832. https://doi.org/10.1126/science.1164759 

 

[5] Clavé, E., et al. (2023). Carbonate detection with SuperCam in igneous rocks on the floor of Jezero Crater, Mars. JGR Planets 128, e2022JE007463. https://doi.org/10.1029/2022JE007463

[6] Clavé, E., et al. (2026). In Situ Carbonation of Sedimentary and Igneous Rocks of Ultramafic Composition in Jezero Crater, Mars. JGR Planets 131, e2025JE009107. https://doi.org/10.1029/2025JE009107