Review of manuscript entitled “What does the impurity variability at the microscale represent in ice cores? Insights from a conceptual approach” submitted to EGUsphere (TC) by Piers Larkman.

General comments:

This manuscript investigates the impact of measurement scale by comparing high-resolution 2D data from LA-ICP-MS with smoothed and converted 1D profiles aligned with the CFA measurement scale. High-resolution measurements are crucial for extracting climate signals from extremely thin layers of deep ice. Results from LA-ICP-MS analysis revealed that impurities, particularly sodium, are preferentially concentrated at grain boundaries rather than within the grain interiors. This indicates that impurity distribution becomes more heterogeneous as grain size increases. Consequently, signal deviation in 1D profiles (quantified as mean absolute deviation, MAD) increases with larger grain samples.

For most glaciologists, the CFA technique offers sufficiently high resolution; however, the resolution of tens of microns provided by LA-ICP-MS is an impressive advancement. Identifying the precise location of impurities is essential for understanding microstructures and deformation mechanisms.

The manuscript is well-written, well-organized, and could make a valuable contribution to the chemical analysis of ice cores and the oldest ice core project. Therefore, I recommend that the manuscript be accepted after minor revisions. I have some questions, comments, and suggestions regarding the results and their interpretation, which may require further explanation. Additionally, some figures need to be redesigned and modified to ensure they are easily understood by the readers.

Specific comments:  

The present manuscript focuses on sodium, which tends to accumulate at grain boundaries. In contrast, certain ions, such as chloride, dissolve and substitute for H₂O molecules within the grain interior. In these cases, I suspect that the impact of grain size on measurement resolution (or the difference between 1D profiles and CFA) may not be as pronounced compared to impurities that accumulate at grain boundaries.

In Figure 4, impurities appear to be distributed within the grain interiors in both EDC and RECAP LGP samples. (In Figure 3, sodium distribution does not seem to extend into the grain interiors.) While this may not be the primary focus of the current paper, the differences in impurity distribution between glacial and interglacial ice samples are intriguing.

The authors concluded that signal deviation increases with larger grain sizes. I wonder whether this deviation is influenced solely by grain size. It seems likely that impurity type (whether substituted within the grain interior or concentrated at grain boundaries) and climate period (glacial or interglacial ice) could also play a role in signal deviation. If the authors have insights on these factors, I suggest them to share their perspectives, as it would be beneficial for readers.

The discussion in Section 4.4 Potential extensions is particularly significant. As the authors mentioned, one of the key future goals of this study is to accurately extract climate signals from deep, thin ice. The deepest ice layers are subject to dynamic recrystallization due to the high-temperature environment, leading to a grain volume distribution that deviates from a gamma distribution. Moreover, migration recrystallization introduces small grains and creates complex grain boundaries, resulting in non-isometric grain shapes. Such complexities in grain boundaries and size distribution are observed not only in deep ice but also in ice samples deformed under high temperature and stress, such as those from the EastGRIP ice core. Replicating these intricate microstructures within an ice matrix model presents significant challenges.

Although it may be considered future work, the manuscript’s discussion and practical implications could be greatly enhanced by including results and assessments from ice samples with complex microstructures, such as those containing small grains and complicated grain boundaries formed by migration recrystallization. In my view, even without 3D ice matrix modeling, simply comparing LA-ICP-MS data with smoothed and combined 1D profiles (like Figure 5), and grain size (distribution) offers substantial value.

L7 in the abstract: What does “high-frequency component of signals” mean?

L15 in the abstract: This approach guides collecting layer-representative signals from LA-ICP-MS line profiles and may help to bridge the scale gap between LA-ICP-MS data and data collected from meltwater analysis.

As mentioned, this approach could assess the scale gap. On one hand, how does this approach specifically bridge the scale gap? I think this is the most interesting for general glaciologist.

L79 (Table 1) In the EDC ice samples, mean grain size in LGP is larger than that in Holocene? In Figure 3, grains in EDC LGP ice sample look small.

L63: What does “The computational representation uses the arguably most simple manifestation of a climate signal, a constant signal” mean? How does the climate signal mode affect the results and discussion of the present study such as experimental measurements and ice structure generation? Please provide brief explanation for general cryosphere’s readers.

L103: I understand that the modelling of 3D ice structure is useful. However, I didn’t see how 3D ice structure helped the verification of results and discussion. Figures 4 to 8 indicate 1D or 2D results. 2D ice matrix model is not sufficient?

L160: A comparison of the optical images and intensity maps in Fig. 3 shows sodium is concentrated preferentially at the grain boundaries compared with grain interiors for all measured samples.

Even shallow EDC sample (Holocene ice), sodium is concentrated grain boundaries. Does this mean that impurities are already concentrated at the grain boundaries during deposition?

L188: These modelled signals show the same general features as experimentally measured signals, with large spikes in intensity where profiles intersect grain boundaries.

It is difficult to determine whether experimental and modelled profiles signals have similar behaviors. Replication of experimental results by means of a model is, in my view, important in the present study. Please provide a comparative figure between the experimental and modelled results.

Figure and table:

Table 1: Please provide explanation for “Profile lateral separation (mm)”.

Figure 3: Redesign is required.

Three images (optical image grain boundary segmentation, and chemical map) are not identical at RECAP ice samples. For example, in Holocene sample, grains in the middle and right images are elongated vertically. In LGP sample, grains in the left image are elongated vertically. Please modify.

The optical image of EDC Holocene sample is low resolution, it is difficult to distinguish grain boundaries.

Size of EDC LGP samples is too small. Image sizes should be the same for all samples.

Figure 6: (a and c) It is difficult to see grain boundaries, please make contrast clearer.

(b and d) What do the color difference of grains mean?

Figure 7: Why do the smoothed profiles (d) and (e) appear to be different? (It doesn’t even look similar to panels a to c)

This is also the case in Figures S6 and S9.

Figure 8: Bottom two panels are labeled with (a) and (b). Please modify. Additionally, the spot size in panel b and caption is shown as 280 um, but the main text explain 260 um. Please correct value.

This manuscript investigates the variability of impurity signals at the microscale by comparing high-resolution 2D impurity signals to modeled 1D CFA measurements. The authors seek to improve understanding of a significant issue within LA-ICP-MS analyses, namely that localized impurities, cause significant variation in 1D signals due to their uneven distribution across the ice matrix.

The model addresses a key challenge in interpreting impurity signals and offers a method to quantify the impact of impurity localization on 2D LA-ICP-MS signals. The manuscript is well written and offers guidance on the number of profiles and level of smoothing required to generate representative signatures. This information is critical for designing LA-ICP-MS experiments. However, while the manuscript provides a step forward in the experimental design, I have concerns regarding the level to which the modelled data accurately represents the climate signal and its applications.

The authors describe in the introduction that LA-ICP-MS analyses are necessary to reconstruct climate records in deep ice and use this as a primary reason for this study. However, the authors use the most basic structure of ice in this model. This is understandable given the continued questions around methodologies for LA-ICP-MS, however, as the work currently stands there are limited implications for deep ice studies.

Additionally, while there is an attempt to quantify the CFA results using modeled results, no actual experimental data is provided to ground truth the model's ability to reconstruct CFA results. Additionally, as no concentrations are provided and no calibration was conducted, it is difficult to see whether the modeled data that this project hinges on are realistic or comparable. As a result I recommend the manuscript be reconsidered after major revisions. 

Specific comments:

L188: The statement “These modelled signals show the same general features as experimentally measured signals, with large spikes in intensity where profiles intersect grain boundaries.” is not proven as the experimental signals are not shown for comparison.

Figure 6: greater contrast is needed to see grain boundaries in a) and c). It is also very difficult to see the blue and red lines. Please make these thicker or choose different colors.

Figure 7: There is no explanation I can find for why 7d and 7e show opposite profiles. Please provide more information. This is particularly important for the author’s claim that this model can be used to compare between LA-ICP-MS and CFA results.

Line 222: I’m unclear why no calibration is used here. Particularly for comparing LA-ICP-MS to CFA signals or comparing between analyses, and ground truthing the model, this is important to understand how well the parameterized model works.  

Line 254: “In this context, the framework presented here can allow improved comparison between the outputs of different experimental setups and can form an essential foundation for inter-technique comparisons, first and foremost with CFA.” This has not been proven. The authors themselves mention that day-to-day comparisons are not comparable, and as no calibration or concentration data is provided this remains conceptual.

Line 264: “Furthermore, the simulation of a CFA signal allows a direct comparison of LA-ICP-MS and CFA signals which is only possible as this is a 3D model” This has not been shown in this paper as no comparison to experimental CFA data is provided to show these are comparable.

Line 264: What does “This facilitates a direct comparison that is not currently possible for physical ice samples as the outer portion of ice measured using CFA is not measured to avoid contamination (Dallmayr et al., 2016).” mean? I’m unclear how direct comparison is only possible with a 3D model here and why contamination control procedures impede this.