Ice Core Science Meets Computer Vision: Challenges and Perspectives

Bohleber, Pascal; Roman, Marco; Barbante, Carlo; Vascon, Sebastiano; Siddiqi, Kaleem; Pelillo, Marcello

doi:10.3389/fcomp.2021.690276

Cited by 6 publications

(3 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, more powerful techniques for image segmentation exist, for example, based on deep learning approaches (Khanal & Estrada, 2020). Like many other tools of modern computer vision, this may provide important assistance to evaluating the complex information comprised in the elemental maps in the near future (Bohleber, Roman, Barbante, et al., 2021).…”

Section: Discussionmentioning

confidence: 99%

Geochemical Characterization of Insoluble Particle Clusters in Ice Cores Using Two‐Dimensional Impurity Imaging

Bohleber

Stoll

Rittner

et al. 2023

Geochem Geophys Geosyst

Self Cite

View full text Add to dashboard Cite

Polar ice cores are an invaluable archive of past climate, thanks to recording a unique variety of proxies such as greenhouse gas concentrations and aerosol-related atmospheric impurity records over time-scales from decades up to hundreds of millennia (e.g., Fischer et al., 2021). The investigation of the deepest and highly thinned ice core layers remains a fundamental frontier in ice core research, being the key to unlock the oldest parts of the ice core record. As a new 1.5 million year-old ice record may soon be recovered from Antarctica (Brook et al., 2006;Fischer et al., 2013), it is also an increasingly pressing challenge. Given that each meter of such an "Oldest" ice core section expected around 2,500 m depth at Little Dome C may comprise more than 10,000 years (Lilien Abstract Understanding post-depositional processes altering the layer sequence in ice cores is especially needed to avoid misinterpretation of the oldest and most highly thinned layers. The record of soluble and insoluble impurities represents an important part of the paleoclimate proxies in ice cores but is known to be affected through interaction with the ice matrix, diffusion, and chemical reactions. Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) has been recognized for its micron-scale resolution and micro-destructiveness in ice core impurity analysis. Employing LA-ICP-MS for 2D chemical imaging has already revealed a close relationship between the ice grain boundary network and impurity signals with a significant soluble component, such as Na and Mg. Here we show the latest improvements in chemical imaging with LA-ICP-MS, by increasing the spatial resolution to 20 μm and extending the simultaneous analysis to also mostly insoluble impurities, such as Al and Fe. All analytes reveal signals of dispersed spots in a sample of an East Greenland ice core. Based on their average size around 50-60 times larger than an average particle and their heterogeneous elemental ratios these spots are interpreted as particle clusters. To distinguish their origin, a simple colocalization classification reveals elemental ratios consistent with marine and mineral dust aerosol. Based on already existing data from cryo-Raman spectroscopy, we discuss potential ways to integrate the two methods in a future comparison. Such a combined approach may help constraining post-depositional changes to the dust-related insoluble impurity components, such as cluster formation and chemical reactions at grain boundaries.Plain Language Summary Aerosols of marine and terrestrial origin delivered to the polar ice sheets are archived in the ice and can be studied via the analysis of ice cores. The chemical composition and size of mineral dust can deliver important information about past climatic changes and atmospheric transport. However, it has already been shown that this insoluble material is not always passively archived in the ice but can undergo changes in its chemical composition and size, for example, by forming particle aggregates. To investigat...

show abstract

Section: Discussionmentioning

confidence: 99%

Geochemical Characterization of Insoluble Particle Clusters in Ice Cores Using Two‐Dimensional Impurity Imaging

Bohleber

Stoll

Rittner

et al. 2023

Geochem Geophys Geosyst

Self Cite

View full text Add to dashboard Cite

show abstract

“…In this context, the development of automated data routines appears to be a shared frontier among both methods, cryo-Raman and LA-ICP-MS imaging. This particularly concerns taking advantage of modern techniques in computer vision and machine learning, which may yield similar revolutions in data analysis as they have already been introduced to other fields (Bohleber and others, 2021a). As both methods are micro-destructive and laser-based, a combined integration in a single setup could be envisaged for the future.…”

Section: Future Research Priorities Towards a Holistic Approach In Mi...mentioning

confidence: 99%

The new frontier of microstructural impurity research in polar ice

Stoll,

Bohleber,

Dallmayr

et al. 2023

Ann. Glaciol.

Self Cite

View full text Add to dashboard Cite

Deciphering the localisation of solid and dissolved impurities on the micron-scale in glacial ice remains a challenge, but is critical to understand the integrity of ice core records and internal deformation. Here we report on the state-of-the-art in microstructural impurity research by highlighting recent progress in bringing together cryo-Raman spectroscopy and laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). We show the potential of both methods and discuss possibilities to improve inter-method approaches aiming for a more holistic understanding of the evolution of impurity localisation throughout the ice column, including post-depositional processes. In this framework, we elaborate on future research priorities such as LA-ICP-MS imaging on firn samples and integrating a large cryo-cell with imaging capabilities.

show abstract

“…All these methods produce image data that could benefit computer vision techniques to process. [13]- [18].…”

Section: Introductionmentioning

confidence: 99%

Ice-Core Micro-CT Image Segmentation With Deep Learning and Gaussian Mixture Model

Bagherzadeh,

Freitag,

Frese

et al. 2023

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

Ice cores of polar regions (ice sheets) are one of the most prominent natural archives that can reveal essential historical information from the past environment of our planet. The ice core microstructure is a key feature in determining the principal properties of ice such as pore close-off, albedo, and melt events. Micro-CT scans can provide valuable information about the microstructure of materials, although achieving a highquality automated segmentation of porous materials, especially with phase/density changes is still a challenge. This work proposes a new method for improving the segmentation of porous microstructures where a weak segmentation (Gaussian Mixture Model) on high-resolution (30 µm) data is used as ground truth to train a deep learning model (U-net) for segmentation of low-resolution (60 µm) data. This approach has reached high segmentation accuracy in terms of quantitative metrics having the F1-score of 92.5% and an intersection over the union of 91%, with a considerable improvement compared to thresholding and unsupervised methods. Also, the segmentation results of U-net are closer to the real weight, density, and specific surface area of the specimen.

show abstract

Ice Core Science Meets Computer Vision: Challenges and Perspectives

Cited by 6 publications

References 52 publications

Geochemical Characterization of Insoluble Particle Clusters in Ice Cores Using Two‐Dimensional Impurity Imaging

Geochemical Characterization of Insoluble Particle Clusters in Ice Cores Using Two‐Dimensional Impurity Imaging

The new frontier of microstructural impurity research in polar ice

Ice-Core Micro-CT Image Segmentation With Deep Learning and Gaussian Mixture Model

Contact Info

Product

Resources

About