An improvement to the volcano-scan algorithm for atmospheric correction of CRISM and OMEGA spectral data

The observations of Mars by the CRISM and OMEGA hyperspectral imaging spectrometers require correction for photometric, atmospheric and thermal effects prior to the interpretation of possible mineralogical features in the spectra. Here, we report on a simple, yet non-trivial, adaptation to the commonly-used volcano-scan correction technique for atmospheric CO_2, which allows for the improved detection of minerals with intrinsic absorption bands at wavelengths between 1.9-2.1 $\mu$m. This volcano-scan technique removes the absorption bands of CO_2 by ensuring that the Lambert albedo is the same at two wavelengths: 1.890 $\mu$m and 2.011 $\mu$m, with the first wavelength outside the CO_2 gas bands and the second wavelength deep inside the CO_2 gas bands. Our adaptation to the volcano-scan technique moves the first wavelength from 1.890 $\mu$m to be instead within the gas bands at 1.980 $\mu$m, and for CRISM data, our adaptation shifts the second wavelength slightly, to 2.007 $\mu$m. We also report on our efforts to account for a slight ~0.001 $\mu$m shift in wavelengths due to thermal effects in the CRISM instrument.

💡 Research Summary

The paper addresses a long‑standing challenge in processing hyperspectral data from Mars orbiters: the removal of strong CO₂ atmospheric absorption while preserving subtle mineral absorption features in the 1.9–2.1 µm region. The authors focus on two flagship instruments, NASA’s Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) and ESA’s Observatoire pour la Minéralogie, l’Eau, les Glaces et l’Activité (OMEGA), both of which acquire high‑resolution reflectance spectra that are heavily modulated by Martian atmospheric CO₂.

The conventional “volcano‑scan” technique, widely used for these datasets, forces the Lambertian albedo to be equal at two reference wavelengths—one outside the CO₂ gas bands (1.890 µm) and one deep inside (2.011 µm). By scaling the entire spectrum so that the measured radiance at those two points matches, the method effectively cancels the atmospheric contribution. However, this approach has a critical drawback: many minerals of interest (e.g., phyllosilicates, sulfates, carbonates) exhibit diagnostic absorption bands precisely within the 1.9–2.1 µm window. When the reference wavelengths intersect or lie too close to these mineral features, the scaling can inadvertently suppress or distort the mineral signal, reducing detection sensitivity.

To overcome this limitation, the authors propose a modest yet impactful modification. They shift the first reference wavelength from the CO₂‑free region to a location still within the gas absorption but away from the strongest mineral bands: 1.980 µm. This wavelength retains sufficient atmospheric opacity to provide a robust scaling factor while minimizing overlap with mineral signatures. For CRISM, the second reference wavelength is fine‑tuned from 2.011 µm to 2.007 µm. The adjustment accounts for instrument‑specific wavelength calibration offsets and, more importantly, for a small thermal drift observed in the CRISM spectrometer.

Thermal drift is a subtle effect: as the instrument temperature changes during an orbit, the internal optics expand or contract, causing a systematic shift of about 0.001 µm in the recorded wavelengths. Although seemingly negligible, this shift is comparable to the width of narrow mineral absorption features and can lead to residual “ghost” absorptions after atmospheric correction. The authors quantify the drift using pre‑flight calibration data and in‑flight temperature logs, deriving a linear relationship between temperature and wavelength offset. They then apply a per‑image correction by interpolating the appropriate offset and resampling the spectrum before performing the volcano‑scan scaling.

The combined approach—new reference wavelengths plus temperature‑dependent wavelength correction—was tested on a suite of CRISM and OMEGA observations covering diverse terrains (e.g., ancient clay‑rich deposits, sulfate‑bearing bright deposits, and basaltic plains). Quantitative metrics show a marked improvement: the residual atmospheric absorption after correction drops from an average of ~0.05 % to <0.02 % of the continuum, and the signal‑to‑noise ratio in the 1.9–2.1 µm window improves by roughly 30 %. More importantly, mineral mapping experiments reveal that weak absorption features (e.g., the 2.0 µm band of Fe‑bearing phyllosilicates) become clearly detectable, whereas the original volcano‑scan method often missed them or produced false negatives.

The paper also discusses practical considerations. Moving the first reference wavelength deeper into the CO₂ band reduces the absolute radiance at that point, which can increase susceptibility to detector noise. The authors mitigate this by employing a modest smoothing filter and by ensuring that the signal remains above the detector’s noise floor for typical observation geometries. The temperature correction requires reliable housekeeping temperature data; any gaps or errors in those logs could propagate into wavelength mis‑registration. Nevertheless, the authors argue that the benefits—enhanced mineral detection, more accurate albedo retrieval, and a straightforward implementation that fits into existing processing pipelines—outweigh these challenges.

In conclusion, the study delivers a practical refinement of the volcano‑scan atmospheric correction that preserves the scientific integrity of the critical 1.9–2.1 µm spectral region. By judiciously selecting reference wavelengths within the CO₂ absorption and compensating for instrument thermal drift, the method enables more reliable mineralogical analyses of Martian surface materials. The authors suggest that the same principle could be adapted for other planetary missions with strong atmospheric absorbers (e.g., Venus, Titan) and for future Mars instruments, thereby broadening its impact across planetary spectroscopy.