IceSpec: An ice-core hyperspectral imaging framework. PAGES Magazine. 2025. Kurbatov A.V., Breton D.J., Hargreaves G., LaBombard C., Nunn R., Rumsey R. and Zhizhin M.
https://pastglobalchanges.org/publications/pages-magazines/pages-magazine/138632
We developed a novel hyperspectral line-scan imaging system for ice cores. The system nondestructively captures light-scattering features via dark-field illumination. Its calibrated visible and near-infrared spectral data enhance analysis and archiving, advancing research on stratigraphic impurities, ice dynamics and paleoclimate.
https://doi.org/10.22498/pages.33.2.50
“IceSpec: An Ice-Core Hyperspectral Imaging Framework.” 2025. Past Global Changes Magazine 33 (2): 50–51. https://doi.org/10.22498/pages.33.2.50.
We co-authored it with D. Breton and Roisin Rumsey
Background
Ice cores are premier paleoclimate archives due to their visible summer and winter layers, which enable precise depth-age timescales (Alley et al. 1997). Snowfall traps impurities and air, preserving a record of past environmental conditions. Reconstructing these conditions requires analyzing chemical constituents, particle content, trapped air bubbles, and the size and orientation of impurities and ice crystals within stratigraphic layers. However, snow-to-ice transformation, driven by pressure and temperature gradients, compacts layers, diffuses impurity signals, and thins or deforms the polycrystalline structure. These processes complicate stratigraphic interpretation, particularly in older ice samples, motivating multidisciplinary studies to better understand these processes for improved reconstructions of ice-core paleoarchives.
Visual inspection has been foundational for studying ice cores since the 1950s (Gow et al. 1968; Langway 2008). Early research relied on sketches, notes, photography and video, and documented stratigraphy semi-quantitatively, mostly as depth of counted layers or intervals of interpreted stratigraphic anomalies (Alley et al. 1997). High-resolution digital imaging, introduced through line-scan systems (McGwire et al. 2008; Svensson et al. 2005; Takata et al. 2004), drastically improved quantification of uncertainties in annual-layer detection (Winstrup et al. 2012), and directly provided image maps of stratigraphic disturbances, while supporting long-term data archives. Illumination techniques also advanced notably with the inclusion of dark-field illumination. This method enhances image contrast by scattering indirect light off low concentration inclusions in the ice matrix, with gray-value intensity correlating to mineral-dust loading in monochrome images (Faria et al. 2018; Morcillo et al. 2020). Digital images also streamline the development of various algorithms to quantify air-bubble count, size, orientation and geometry. Recent innovations incorporate hyperspectral cameras into line-scan systems, operating in visible and near-infrared (VNIR) (Garzonio et al. 2018) and near-infrared (NIR) ranges (McDowell et al. 2024). The hyperspectral method enables the identification of specific impurities in ice samples by comparing their unique spectral signatures to a comprehensive library of known spectral profiles (Kokaly et al. 2017). The proliferation of artificial intelligence (AI)-based, open-source tools, and the rapid growth of information technology lowers barriers to developing data-unmixing algorithms and user-friendly software tools for data classification. Ultimately, hyperspectral imaging has the potential to revolutionize the study of ice cores, enabling the capture of high-resolution spectral data that can reveal ultra-high-resolution details of ice-core records.