the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 01 Jul 2025
Automated Detection and Chemical Characterization of Anomalous Grains in Scientific Ocean Drilling Legacy Cores
Abstract. The dynamic behavior of glaciers under varying climatic conditions plays a crucial role in Earth's climate systems, necessitating reliable records of glacial extent throughout the Cenozoic era. Ice-rafted debris (IRD) provides insights into past iceberg activity, glacial erosion rates, sediment transport, and meltwater delivery. Polar IRD reconstructions can enhance our understanding of ice sheet responses to climate shifts, however, traditional methods for detecting IRD are often destructive and labor-intensive. This study proposes a new methodological framework for detecting IRD in marine sediments as a proxy for reconstructing polar paleoclimate variability and glacial dynamics. Our new multi-proxy approach utilizes micro-X-ray fluorescence (μXRF) 2D imaging and computed tomography (CT) 3D scanning on five Deep Sea Drilling Project (DSDP) Site 266 archive half-sections. Our findings reveal that non-destructive scanning techniques have significant potential to effectively identify and quantify IRD, with μXRF providing high-resolution geochemical mapping and CT imaging enabling three-dimensional visualization of sediment structures. Machine learning classification of CT data using the Waikato Environment for Knowledge Analysis (WEKA) in the Fiji software package is a powerful tool for fast detection of anomalous components in IODP cores. However, we established that the finding of such coupled machine learning-CT approaches must be ground-truthed to ensure accuracy: our application of automated (k-means cluster analyses) chemical fingerprinting based on the μXRF data revealed many of the grains identified in the CT classification were false positives due to contamination from drill pipe flakes. We confirmed this contamination using scanning electron microscopy energy dispersive spectroscopy (SEM-EDS) on targeted samples. Fundamentally, our multi-proxy approach underscores that this approach alone cannot provide the geochemical fingerprinting necessary for unequivocal IRD identification. We therefore recommend a combined approach applying μXRF for geochemical fingerprinting and CT for structural detection to ensure a comprehensive, non-destructive method for IRD verification and preliminary provenance analysis. Ultimately, our work underscores the importance of applying stringent ground truthing to machine learning and high-resolution image datasets to ensure the data quality necessary for robust palaeoclimate reconstructions.
- Preprint
(3325 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-2760', Anonymous Referee #1, 29 Jul 2025
Auer and colleagues present new methods for non-destructive study of clasts in marine drill cores, motivated by the need to study ice-rafted debris without excessive sampling of the limited amount of material. This work builds on many years of trying to determine the best approaches to do this and explores new methods focused on CT scanning and uXRF. It is a super interesting approach and I enjoyed reading about it—although I was not super convinced it would be the best approach for IRD quantification in these specific cores. That said, the science is sound, it is an interesting method, and promotes the use of legacy cores. I think it will be of interest to the wider community as they evaluate the best approaches for IRD quantification. I think it could be published with limited revision.
As a commentary—not a criticism of this work—in my own research, I have become more and more convinced that there is no better way to quantify IRD than breaking out the sieves and actually looking at what is there. For the record, this is coming from someone who initially really pushed for automated approaches. In my recent experience, we’ve found that other particles (like iron sulfides) can confuse many different detection algorithms and are responsible for the differences between x-ray/CT counts and physical sieved wt% data. In the author’s work, metal flakes pose a similar problem. However, the datasets we’ve been developing from sieved samples have ended up quite different from x-ray derived methods in some intervals (supervised and unsupervised), with the sieved data seeming to make a lot more sense when compared to other proxy data. I was left wondering after Section 4.3 if the authors actually thought that these approaches were useful or if they would still require significant work looking at sieved fractions in order to ground truth their interpretations.
As a final note, I wonder if the authors might comment a bit more about the metal flakes and if this is an issue that is likely limited to early DSDP rotary cored un-lithified sediments and to what extent the issue might persist in the more common legacy APC cored sediments. From my experience, I had thought that for APC cores, these flakes are generally limited to the near the core liner and generally in greater quantities near the top of the cores—and thus a bit easier to work around. I know that there has been a bit of documentation about this from the paleomagnetic community, with one of the primary reasons for the development of the ‘u’-channel method to sample the more pristine centers of the cores to get away from the metal flakes that were sometimes found near the core liner and messed up the half-core pass-through measurements (see the first u-channel study: Mead et al., 1986; https://doi.org/10.1029/PA001i003p00273). There is also documentation on rust contamination from Carl Richter’s 2007 “Handbook for Shipboard Paleomagnetists.” As the author’s conclusions do not mention this complication, despite being a major caveat, I wonder if they think that their approach here might be better suited for legacy APC cores than legacy rotary cores (for soft mud), like reported here?
Minor comments:
Line 84: You do such a good job reviewing the literature, I am surprised you don’t list some examples of CT scans being used to quantify IRD at the end of this sentence, including many studies that have previously developed automated detection routines from CT data.
Line 181-182: Figure call out to Fig. 2, but it doesn’t seem to align with the statement which seems to be suggesting that you are going to show a picture of the 19 grains and 1 fish tooth. Please clarify this statement. Or is it meant to call out to Figure 8?
Figure 4 caption: Should cycle by circle?
Figure 7: spelling? Shon -> shown
Line 304-307: This would be true IF the object filled an entire voxel. Based on the resolution of the CT scanner, it is unlikely that the flake would be large enough and the voxels would be an integration of the high density steel and matrix sediment (thus, more likely in the range of a lithic clast). Unless the flakes were sufficiently thick?
Citation: https://doi.org/10.5194/egusphere-2025-2760-RC1 -
RC2: 'Comment on egusphere-2025-2760', Anonymous Referee #2, 29 Jul 2025
To whom this may concern,
The manuscript by Auer et al. complements a growing number of studies that seek to advance the potential of high-resolution scanning methods to automate counting and classification of particles with paleoenvironmental relevance. But while there are interesting aspects to this work, notably the use of XRF systems that scan the entire surface of a core (rather than just a line), but also the application of plug-ins that allow machine learning classification of particles, I believe that there several major issues in the design of this study require reframing and major revision:
Ground-truthing: Throughout the manuscript, the authors highlight the need for ground truthing of (semi)automatic approaches like those presented here. Yet it appears that few aspects of their study have been ground truthed. For example, just one sample was IRD-counted using traditional (wet-sieving) approaches, out of three core sections. Also, no information is provided on the provided on the bedrock provenance of IRD in the analyzed sediments, making it hard to contextualize the XRF and SEM data. In this respect, I also noted that the former uses Fe and Mn as diagnostic indicators, while the former reveals a more diverse composition where both these elements do not always dominate.
Classification: while I am not an expert in the applied classification schemes, the fact that the metal flakes deriving from drill pipes were not clearly distinguished based on their distinct geochemistry (measured with XRF) and density (measured with CT), does raise questions about the way they have been applied here.
Resolution: While the authors certainly do not hide this fact, the major offset in resolution between their XRF and CT data makes it hard (impossible) to inter-compare both datasets. Indeed, while they argue that their material did not contain IRD that could be easily identified by eye to test the potential of their scanning techniques, the used scanner can only resolve particles larger than 1000 micrometers (medium sand: visible). This is twice as high as the resolution of their XRF scans, although it is casually mentioned that these could have been generated at a 100-micrometer resolution.
Novelty: the above issues with ground-truthing and resolution restrict the added value of this work, compared to existing studies that have, for example, automatically counted IRD particles in the fine sand fraction, while also ground-truthing these profiles against traditionally measured datasets, and harnessing the 3-D potential of CT data. What does not help either is that quite a few of these studies are not mentioned or cited by the authors (also see my in-text comments). In addition, the authors sometimes revert to the use of rather hyperbolic statements that feel a bit out of place, considering the above.
In addition to these main/major concerns, I have provided detailed in-text comments in the manuscript, and hope that these will prove useful to improve this manuscript.
Data sets
Supplementary Materials and Supplementary Text for the Submission "Automated Detection and Chemical Characterization of Anomalous Grains in Scientific Ocean Drilling Legacy Cores" G. Auer et al. https://doi.org/10.5281/zenodo.15543202
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 406 | 28 | 8 | 442 | 32 | 30 |
- HTML: 406
- PDF: 28
- XML: 8
- Total: 442
- BibTeX: 32
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1