Interannual Variability of Sea Ice Dirtiness in the East Siberian Sea Based on Satellite Data

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript provides a timely and meaningful contribution by examining the interannual variability of sea ice dirtiness in the East Siberian Sea using MODIS satellite imagery from 2000 to 2025. The topic is highly relevant to Arctic environmental studies, particularly regarding sea ice dynamics and remote sensing applications. The manual interpretation approach, while labor-intensive, generates a valuable and reliable dataset that may serve as a useful reference for the development of future automated detection methods. The analysis also offers important physical insights by relating the occurrence of dirty ice to wind-wave activity and ice formation processes. Nevertheless, the study remains predominantly descriptive and depends heavily on subjective visual interpretation. The methodological details are limited, raising concerns about reproducibility and the accuracy of the results. Essential aspects, such as the justification of the “three-point scale”, evaluation of internal consistency, and lack of quantitative validation, require further clarification. Additionally, the manuscript’s presentation could be improved, as some sections lack clarity or contain redundancy, and several figures and tables are insufficiently described. In summary, the study holds promise but requires substantial revision to strengthen its methodological rigor, analytical depth, and overall clarity before it can be considered for publication.

Major Comments:

1. The core methodology, manual classification of sea-ice “dirtiness” from satellite imagery, requires substantially more detail to ensure reproducibility. It is unclear who performed the visual interpretations and how consistency and potential biases were controlled: if a single analyst performed all classifications, report an intra-observer repeatability assessment; if multiple analysts were involved, quantify inter-observer agreement. The phrase “visual and analytical analysis of images on average for a month” is ambiguous and should be clarified.
2. The authors should provide a step-by-step workflow that includes image selection criteria and preprocessing, the number of images interpreted per year and per region, how cloud cover and shadows were handled, and the quality-control procedures used. A summary table or figure showing the number and spatial-temporal distribution of analyzed images is recommended.

 

3. The manuscript must also state whether any automated tools or objective indices (e.g., reflectance thresholds, spectral indices, or segmentation algorithms), software (with versions), or parameter settings were used to assist identification, or whether all classifications were entirely manual. Finally, including an explicit algorithmic description or a workflow diagram would substantially improve methodological transparency and reproducibility.
4. The manuscript repeatedly refers to a “3-point scale”,however, Figure 3 displays four categories (0–3), resulting in an inconsistency. If Level 0 is included, the system should be defined as a four-class classification scheme in accordance with standard academic conventions. The rationale behind this classification also requires clearer explanation. Specifically, was the scale directly adopted from the cited Russian nomenclature, or was it modified to align with the objectives of this study? If the former, the authors should justify the chosen thresholds (i.e., 1/3 and 2/3 of the ice area) and discuss their applicability for interpretation based on MODIS satellite imagery.
5. Moreover, the authors should clarify how sea-ice regions were delineated and how the proportion of dirty ice within each region was quantified. For instance, at the 250 m spatial resolution of MODIS data, what is the minimum area of contamination that can be reliably identified through visual interpretation? Providing these methodological details would greatly enhance the clarity and reproducibility of the study.
6. The study currently lacks a quantitative evaluation of classification accuracy. For example, although Figure 2 presents expedition and Landsat images for illustration, no quantitative comparison (e.g., percentageor accuracy metrics) is provided. It is encouraged to validate a subset of the MODIS-based classifications against independent observations. Possible approaches include: (a) comparing MODIS-derived results with higher-resolution Landsat series imagery, as suggested in Figure 2, to assess spatial consistency; (b) conducting cross-validation using available field data or regional observations; or (c) reporting the frequency of ambiguous classifications between “dirty” and “clean”
7. I have some concerns regarding the sea ice area results presented in this study. First, in Figure 3, the color used to represent clean sea ice is identical to that of the background, making it difficult to visually distinguish or accurately quantify the clean sea ice area. Furthermore, the proportion of clean sea ice appears to be unrealistically low, and this result should be carefully verified for physical plausibility. It is suggestedto re-examine the classification and visualization scheme to ensure that the representation and derived proportions are both accurate and interpretable.
8. Figure 4 suggests an apparent upward trend in the extent of dirty ice; however, a regression analysis should be performed to quantify this trend and evaluate its statistical significance. Additionally, it would be valuable to present the temporal variation of clean sea ice over the past 26 years, together with the overall trend of total sea ice extent in the East Siberian Sea.Based on the current results, the proportion of clean sea ice appears to be very small, while the total sea ice extent in the East Siberian Sea seems to show an overall increasing trend. This observation appears inconsistent with the manuscript’s statement that “the obtained results on the interannual variability of ice dirtiness are consistent with studies indicating intensified dynamic processes in the Arctic due to climate change and a decrease in sea ice extent.”

 

9. The visible-range approach used in this study may lead to potential misclassification of ice dirtiness, as other surface features—such as melt ponds, algal growth, or ice shadows—can exhibit similar spectral characteristics. The authors should clarify how sediment-laden ice was distinguished from dark meltwater pools or algal blooms on the ice surface. In addition, based on Figure 2, it remains unclear how thin sea ice was differentiated from dirty ice, since both can display comparable reflectance properties in the visible spectrum. Providing explicit classification criteria or auxiliary spectral information would help substantiate the accuracy of the visual interpretation.
10. The manuscript would benefit from comprehensive language editing to enhance clarity and readability.

 

Minor Comments:

1. Abstract. The abstract should explicitly summarize the key research findings of this study, rather than relying on vague or general statements.
2. The manuscript frequently refers to factors such as the marine environment and coastal river systems. It is recommended to include an overview map of the study area to clearly illustrate its characteristics, boundaries, and spatial extent.
3. Line 78-80. The term “the Laptev East Siberian Seas” is non-standard.
4. Figure 1 is overly simplistic and lacks essential map elements, such as a scale bar and latitude/longitude markers.
5. 5. Line 132 and 150.Sections 3.2 and 3.3 should be corrected to 2.2 and 2.3.
6. 6. Line 163-165. The sentence “The input wind data for the wave model forcing are U and V components of wind …” appears repeatedly on line 268 and is highly similar to other instances. This redundancy should be addressed, and the text should be rewritten for clarity and conciseness.
7. 7. Figure 3. Itshould be redrawn to clearly distinguish Level 0 sea ice from the background.
8. 8. Figure 4. It is recommended to include data on clean sea ice and total sea ice in Figure 4 to enable a more comprehensive comparison. Additionally, the figure would benefit fromstandardized formatting: the x-axis should display variable names and units directly, rather than only in the figure title. For the linear trend of dirty ice, a regression analysis with statistical significance testing should be performed, and the results should be clearly annotated on the figure.
9. 9. Table 1. There appear to be data inconsistencies: the year 2017 is listed in both the “medium” and “large” columns with differing values. The authors should clarify and resolve this discrepancy.
10. 10. Figure 5. The legend for “depth” does not indicate the measurement unit.
11. 11. Figure 6. It is recommended to revise the title“Dates of fast ice formation during the ice season 2023/2024 (a)”for clarity.

Author Response

Comments 1: This manuscript provides a timely and meaningful contribution by examining the interannual variability of sea ice dirtiness in the East Siberian Sea using MODIS satellite imagery from 2000 to 2025. The topic is highly relevant to Arctic environmental studies, particularly regarding sea ice dynamics and remote sensing applications. The manual interpretation approach, while labor-intensive, generates a valuable and reliable dataset that may serve as a useful reference for the development of future automated detection methods. The analysis also offers important physical insights by relating the occurrence of dirty ice to wind-wave activity and ice formation processes. Nevertheless, the study remains predominantly descriptive and depends heavily on subjective visual interpretation. The methodological details are limited, raising concerns about reproducibility and the accuracy of the results. Essential aspects, such as the justification of the “three-point scale”, evaluation of internal consistency, and lack of quantitative validation, require further clarification. Additionally, the manuscript’s presentation could be improved, as some sections lack clarity or contain redundancy, and several figures and tables are insufficiently described. In summary, the study holds promise but requires substantial revision to strengthen its methodological rigor, analytical depth, and overall clarity before it can be considered for publication.

Response 1: The authors are very grateful to the Reviewer for such an attentive attitude to our article, a detailed analysis of errors and valuable suggestions for improving the article.

We agree that the manual interpretation approach generates a valuable and reliable dataset that may serve as a useful reference for the development of future automated detection methods. Authors have extensive experience in special shipboard ice observations from icebreakers, experience in studying the physical properties of ice during winter expeditions at the Cape Baranova polar station in the Severnaya Zemlya archipelago and in the Khatanga Bay, as well as many years of experience in interpretation of satellite images to draw sea ice charts, which are regularly used by navigators in the Northern Sea Route. Since using satellite ice data in practice during shipboard expeditions, we noticed that so far no automatic method provides us with accurate data, and all satellite ice data requires validation. Obtaining information about some characteristics of the ice cover in real field conditions is still based only on visual observation methods and has no automatic analogues. That is why studying natural processes requires not only computational and modeling work, but also descriptive work describing real natural processes that ice researchers have observed in the field.

Indeed, we have not explained the method of interpretation of dirty ice sufficiently, and we have responded to all your comments and suggestions below, as well as made appropriate edits to the text of the article.

Major Comments:

Comments 2: The core methodology, manual classification of sea-ice “dirtiness” from satellite imagery, requires substantially more detail to ensure reproducibility. It is unclear who performed the visual interpretations and how consistency and potential biases were controlled: if a single analyst performed all classifications, report an intra-observer repeatability assessment; if multiple analysts were involved, quantify inter-observer agreement.

Response 2: We have completely rewritten and significantly expanded the Materials and Methods section, where we explained in detail the method of interpretation of dirty ice. However, even with the most detailed explanation, many nuances cannot be explained, as they depend on the field experience of an expert in studying sea ice, experience in interpretation of satellite data, and knowledge of the features of ice formation in the region under study. The work of any expert who compiles ice maps is based on this principle.

Comments 3: The phrase “visual and analytical analysis of images on average for a month” is ambiguous and should be clarified.

Response 3: In the introduction, it was replaced with the phrase " This mapping approach includes visual interpretation of images for the period from snow melt on ice surface until the landfast ice decay, which is resulted in one map of dirty ice for each year." Later, this point was described in more detail in the Materials and Methods. (Line 95-97)

 

Comments 4: The authors should provide a step-by-step workflow that includes image selection criteria and preprocessing, the number of images interpreted per year and per region, how cloud cover and shadows were handled, and the quality-control procedures used. A summary table or figure showing the number and spatial-temporal distribution of analyzed images is recommended.

Response 4: The methodology is now explained in sufficient detail in the Materials and Methods. See the point 2.2.- the second stage of the workflow. We have shown how many images are used on the example of 2021. We believe that the summary table for all years will increase the volume of the article without providing the information necessary to understand the research.  (Line 220-245)

 

Comments 5: The manuscript must also state whether any automated tools or objective indices (e.g., reflectance thresholds, spectral indices, or segmentation algorithms), software (with versions), or parameter settings were used to assist identification, or whether all classifications were entirely manual. Finally, including an explicit algorithmic description or a workflow diagram would substantially improve methodological transparency and reproducibility.

Response 5: The information has been added to the Materials and Methods section (Line 152-245).

 

Comments 6: The manuscript repeatedly refers to a “3-point scale”,however, Figure 3 displays four categories (0–3), resulting in an inconsistency. If Level 0 is included, the system should be defined as a four-class classification scheme in accordance with standard academic conventions. The rationale behind this classification also requires clearer explanation. Specifically, was the scale directly adopted from the cited Russian nomenclature, or was it modified to align with the objectives of this study? If the former, the authors should justify the chosen thresholds (i.e., 1/3 and 2/3 of the ice area) and discuss their applicability for interpretation based on MODIS satellite imagery.

Response 6: We are thankful for the correction and have checked the terminology in accordance with international rules. The scale was taken from the Russian nomenclature and we referred to it (Line 133-139). In the Materials and Methods, we have explained in detail why the scale is also applicable for satellite images in the visible range. We carry out a visual assessment of dirtiness, rather than direct measurements of the area of dirtiness (Line 140-146).

Comments 7: Moreover, the authors should clarify how sea-ice regions were delineated and how the proportion of dirty ice within each region was quantified. For instance, at the 250 m spatial resolution of MODIS data, what is the minimum area of contamination that can be reliably identified through visual interpretation? Providing these methodological details would greatly enhance the clarity and reproducibility of the study.

Response 7: We have explained the interpretation process in as much detail as possible in Materials and Methods (Line 220-245 and Figure 5). The objective of this study was to compile generalized rather than detailed maps of ice dirtiness in the East Siberian Sea. This is necessary not to directly calculate the area of dirty ice (this cannot be done visually), but to assess the intensity of processes resulting to dirty ice. This helps to understand which processes cause dirtiness of huge areas of sea ice in some years, and why they do not in other years. It also helps to understand why there is the same dirtiness in a particular area every year.

Comments 8: The study currently lacks a quantitative evaluation of classification accuracy. For example, although Figure 2 presents expedition and Landsat images for illustration, no quantitative comparison (e.g., percentageor accuracy metrics) is provided. It is encouraged to validate a subset of the MODIS-based classifications against independent observations. Possible approaches include: (a) comparing MODIS-derived results with higher-resolution Landsat series imagery, as suggested in Figure 2, to assess spatial consistency; (b) conducting cross-validation using available field data or regional observations; or (c) reporting the frequency of ambiguous classifications between “dirty” and “clean”

Response 8: It is impossible to compare the values of the in-situ ice area with the areas on satellite images, since field observations are always point-based while satellite observations are area-based. We also cannot compare the exact values of the area of contaminated ice in the Landsat and MODIS images, since Landsat observations would still be subjective, and spatial coverage of Landsat images is not enough for full comparison. Even if we had full coverage, we would need some automatic method for retrieval of ice dirtiness from Landsat and MODIS data in order to conduct an accurate comparison.  The problem is that there is no such method developed. Anyway, for developing such method, we first need to visually determine how the dirty ice is displayed in the satellite images. This was the main task of our work. When developing a visual method of interpreting dirty ice from satellite images, we used field data from expeditions (photo, video, data of ice drilling), according to which the expert checked whether the ice was really dirty in the area where he identifies dirty ice.

We explained this in Materials and Method.

Comments 9: I have some concerns regarding the sea ice area results presented in this study. First, in Figure 3, the color used to represent clean sea ice is identical to that of the background, making it difficult to visually distinguish or accurately quantify the clean sea ice area. Furthermore, the proportion of clean sea ice appears to be unrealistically low, and this result should be carefully verified for physical plausibility. It is suggestedto re-examine the classification and visualization scheme to ensure that the representation and derived proportions are both accurate and interpretable.

Response 9: The Figure covers the water area of the East Siberian Sea, where the land is highlighted in yellow, – and dirty ice is shown in dark gray and black colors. The other area is located in the north of the East Siberian Sea where the ice is clean as long as it is formed at great depths. Accordingly, the white area in the Figure represents compact clean ice. In June, the entire water area still remains ice-covered, and only polynyas and leads may appear in some areas, which has little effect on the total area of contaminated ice.

We do not agree that the proportion of clean ice is unrealistically low. In the article, we explained in detail that the East Siberian Sea is the shallowest among other shelf seas in the Arctic, and dirty ice exactly forms in shallow waters (Figure 2 Line 307-317).

We have discussed about the accuracy of interpretation above. This work, which uses visual interpretation, cannot be evaluated as a basis for accurate determination of areas. Visual interpretation gives us a basis for studying the nature of dirty ice formation in a particular area and helps to roughly estimate spatial distribution of contaminated ice in the Arctic shelf seas.

Comments 10: Figure 4 suggests an apparent upward trend in the extent of dirty ice; however, a regression analysis should be performed to quantify this trend and evaluate its statistical significance. Additionally, it would be valuable to present the temporal variation of clean sea ice over the past 26 years, together with the overall trend of total sea ice extent in the East Siberian Sea.Based on the current results, the proportion of clean sea ice appears to be very small, while the total sea ice extent in the East Siberian Sea seems to show an overall increasing trend. This observation appears inconsistent with the manuscript’s statement that “the obtained results on the interannual variability of ice dirtiness are consistent with studies indicating intensified dynamic processes in the Arctic due to climate change and a decrease in sea ice extent.”

Response 10: We study the dirtiness of the ice that is formed during the entire winter season. In June and July, we see the dirt  on the ice because sub-surface dirty layers become open in those areas where was ice formation in the autumn. Throughout the winter season and until the end of June, the entire East Siberian Sea is covered with compact ice. Therefore, the area of the ice is taken as 100% of the Sea area.

When talking about interannual changes in the ice area, the researchers usually mean the September values, when the ice area is minimal. This is not relevant to our study. That is why we do not consider the overall changes in the ice area as long as these changes are determined for the period of minimal ice coverage. We study the dirtiness that is accumulated during the period of 100% ice coverage (winter season). That is why we say that area of dirty ice increases with stronger dynamic processes. Stronger dynamic processes, in turn, lead to a decrease in the entire ice area.

Comments 11: The visible-range approach used in this study may lead to potential misclassification of ice dirtiness, as other surface features—such as melt ponds, algal growth, or ice shadows—can exhibit similar spectral characteristics. The authors should clarify how sediment-laden ice was distinguished from dark meltwater pools or algal blooms on the ice surface. In addition, based on Figure 2, it remains unclear how thin sea ice was differentiated from dirty ice, since both can display comparable reflectance properties in the visible spectrum. Providing explicit classification criteria or auxiliary spectral information would help substantiate the accuracy of the visual interpretation.

Response 11: Figure 4 (the new Figure that we added to the article) shows that melt ponds are clearly distinguished from dirty ice, if clean ice underlays the melt water.

In our study, we do not separate the types of dirtiness, as these cannot be separated in satellite images. We consider dirty ice in general. However, according to field studies in the shallow waters of the Siberian shelf, in particular at Cape Baranova and in Khatanga Bay, and according to ship expeditions, predominant type of dirtiness is  of terrigenous origin..

We study the ice in June, when melt processes begin in the East Siberian Sea, and at the same time the  thickness of ice remains maximum. According to data  obtained from stationary polar stations of the Russian Hydrometeorological Service network, the maximum ice thickness in the East Siberian Sea ranges from 1.5 to 2 meters. We provided this data in our previous article, which may be found at the link: https://link.springer.com/article/10.1134/S0001433822120039 (Figures 5 and 6)

Comments 12: The manuscript would benefit from comprehensive language editing to enhance clarity and readability.

Response 12: We reworked the text with the help of a professional translator.

Minor Comments:

Comments 13: Abstract. The abstract should explicitly summarize the key research findings of this study, rather than relying on vague or general statements.

Response 13: We have updated the abstract (Line 21-29).

Comments 14: The manuscript frequently refers to factors such as the marine environment and coastal river systems. It is recommended to include an overview map of the study area to clearly illustrate its characteristics, boundaries, and spatial extent.

Response 14: We have added Figure 2 into the article, which shows the borders of the region under study as well as the main geographical names.

Comments 15: Line 78-80. The term “the Laptev East Siberian Seas” is non-standard.

Response 15: We corrected it, meaning Laptev and East Siberian Seas. (Line 88)

Comments 16: Figure 1 is overly simplistic and lacks essential map elements, such as a scale bar and latitude/longitude markers.

Response 16: The Figure 1 has been corrected.

 

Comments 17: Line 132 and 150.Sections 3.2 and 3.3 should be corrected to 2.2 and 2.3.

Response 17: It has been corrected.

Comments 18: Line 163-165. The sentence “The input wind data for the wave model forcing are U and V components of wind …” appears repeatedly on line 268 and is highly similar to other instances. This redundancy should be addressed, and the text should be rewritten for clarity and conciseness.

Response 18: It has been corrected.

 

Comments 19: Figure 3. Itshould be redrawn to clearly distinguish Level 0 sea ice from the background.

Response 19: We mentioned above that white area is clean ice, since in June the entire East Siberian Sea is covered with ice. We have made the outlines of each chart brighter.

Comments 20: Figure 4. It is recommended to include data on clean sea ice and total sea ice in Figure 4 to enable a more comprehensive comparison. Additionally, the figure would benefit fromstandardized formatting: the x-axis should display variable names and units directly, rather than only in the figure title. For the linear trend of dirty ice, a regression analysis with statistical significance testing should be performed, and the results should be clearly annotated on the figure.

Response 20: We have added axis names and regression analysis. It makes no sense to add data on clean ice, since in the winter-spring period the entire sea is covered with ice. It means that, for example, if the area of dirty ice is 60%, then the area of clean ice is 40%.

Comments 21: Table 1. There appear to be data inconsistencies: the year 2017 is listed in both the “medium” and “large” columns with differing values. The authors should clarify and resolve this discrepancy.

Response 21: We corrected the mistake: not 2017, but 2018 in the third column

Comments 22: Figure 5. The legend for “depth” does not indicate the measurement unit.

Response 22: We removed Figure 7, added depths in Figure 2, and indicated the units of measurement.

Comments 23: Figure 6. It is recommended to revise the title“Dates of fast ice formation during the ice season 2023/2024 (a)”for clarity.

Response 23: The title was corrected.

With great gratitude for the valuable comments from the authors

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript presents a valuable and comprehensive study on the interannual variability of dirty sea ice in the East Siberian Sea using satellite imagery from 2000 to 2025. The authors have manually interpreted MODIS visible-range images to produce maps of ice dirtiness on a 3-point scale, a method that aligns with historical aerial reconnaissance techniques. The topic is relevant, especially in the context of changing Arctic conditions, and the dataset spanning 25 years is a significant contribution.

The study is well-structured, and the methodology is clearly described. The inclusion of wave modeling and historical context adds depth to the analysis. However, several major issues need to be addressed before the manuscript can be considered for publication. I have following suggestions:

 

• While the manual interpretation method is well-explained, the process of aggregating monthly data into a single annual map needs more detail. How were inconsistencies in cloud cover or temporal gaps handled? A flowchart or step-by-step summary would improve reproducibility.
• The authors mention that automatic methods are error-prone, but no quantitative validation of the manual method is provided. How was inter-observer variability controlled? Were field data or high-resolution images (e.g., Landsat) used to validate the dirtiness classifications?
• The increasing trend in dirty ice area is highlighted, but no statistical tests (e.g., Mann-Kendall, p-values) are presented to support this claim. The use of standard deviation for grouping years is acceptable, but trend analysis should be statistically robust.
• The study rightly identifies the limitations of optical imagery (cloud cover, daylight dependence). The discussion would be significantly strengthened by addressing emerging remote sensing techniques that can overcome these limitations. Specifically, Global Navigation Satellite System-Reflectometry (GNSS-R) shows great promise for sea ice sensing. Research from groups like those at Memorial University  has demonstrated that L-band signals from GNSS-R are sensitive to sea ice properties, including the presence of sea ice, sea ice concentration, and sea ice thickness. L-band signals can penetrate clouds and are independent of solar illumination, offering a viable solution for continuous monitoring in the Arctic. A discussion on how future studies could integrate GNSS-R data to complement and validate optical-based dirtiness maps would be a valuable addition to the manuscript.
• The manuscript would benefit from thorough proofreading to correct minor grammatical errors and improve fluency.
• Some acronyms (e.g., IBCAO, OLI) are defined late or not at all. Please ensure all acronyms are defined at first use.

Author Response

Comments 1: This manuscript presents a valuable and comprehensive study on the interannual variability of dirty sea ice in the East Siberian Sea using satellite imagery from 2000 to 2025. The authors have manually interpreted MODIS visible-range images to produce maps of ice dirtiness on a 3-point scale, a method that aligns with historical aerial reconnaissance techniques. The topic is relevant, especially in the context of changing Arctic conditions, and the dataset spanning 25 years is a significant contribution.

Response 1: We are very thankful to the Reviewer for the acceptance of our work and for the suggested edits. We have revised the article in accordance with the comments and highlighted all corrections in red. Almost all Figures have been reworked.

Comments 2: The study is well-structured, and the methodology is clearly described. The inclusion of wave modeling and historical context adds depth to the analysis. However, several major issues need to be addressed before the manuscript can be considered for publication. I have following suggestions:

While the manual interpretation method is well-explained, the process of aggregating monthly data into a single annual map needs more detail. How were inconsistencies in cloud cover or temporal gaps handled? A flowchart or step-by-step summary would improve reproducibility.

Response 2: We have completely reworked and expanded the Materials and Methods section, where detailed description of the interpretation method has been added. However, even with the most detailed explanation, many nuances cannot be explained, as they depend on the field experience of an ice expert, his/her experience in satellite data interpretation as well as knowledge of the features of ice formation in the region of interest. The work of any expert producing ice maps is based on this experience and knowledge. Еhe process of aggregating monthly data into a single annual map is shown now on Figure 5 (Line 220-245)

Comments 3: The authors mention that automatic methods are error-prone, but no quantitative validation of the manual method is provided. How was inter-observer variability controlled? Were field data or high-resolution images (e.g., Landsat) used to validate the dirtiness classifications?

Response 3: It is impossible to compare the values of the in-situ ice area with the areas on satellite images, since field observations are always point-based while satellite observations are area-based. We also cannot compare the exact values of the area of contaminated ice in the Landsat and MODIS images, since Landsat observations would still be subjective, and spatial coverage of Landsat images is not enough for full comparison. Even if we had full coverage, we would need some automatic method for retrieval of ice dirtiness from Landsat and MODIS data in order to conduct an accurate comparison.  The problem is that there is no such method developed. Anyway, for developing such method, we first need to visually determine how the dirty ice looks in satellite images. This was the main task of our work. When developing the visual method of dirty ice detection in satellite images, we used the archive of field data from expeditions (photo, video, data of ice drilling). Using field data the expert could understand whether the ice is really dirty in those areas where he/she identified dirty ice in satellite images.

We explained this in Materials and Method.

Comments 4: The increasing trend in dirty ice area is highlighted, but no statistical tests (e.g., Mann-Kendall, p-values) are presented to support this claim. The use of standard deviation for grouping years is acceptable, but trend analysis should be statistically robust.

Response 4: For statistical analysis, a series of data containing only 26 values is not significant, and in-depth statistical analysis is possible on a longer series of values. We have added regression analysis, and replaced “trend” to “tendency” (Figure 7)

Comments 5: The study rightly identifies the limitations of optical imagery (cloud cover, daylight dependence). The discussion would be significantly strengthened by addressing emerging remote sensing techniques that can overcome these limitations. Specifically, Global Navigation Satellite System-Reflectometry (GNSS-R) shows great promise for sea ice sensing. Research from groups like those at Memorial University  has demonstrated that L-band signals from GNSS-R are sensitive to sea ice properties, including the presence of sea ice, sea ice concentration, and sea ice thickness. L-band signals can penetrate clouds and are independent of solar illumination, offering a viable solution for continuous monitoring in the Arctic. A discussion on how future studies could integrate GNSS-R data to complement and validate optical-based dirtiness maps would be a valuable addition to the manuscript.

Response 5: We are thankful to the Reviewer for the idea. This is a topic for a large independent study, and it’s really difficult to add this to the current article. We deal with microwave radiometry of sea ice, including the L-band data, in our parallel research. We have obtained interesting results, which are presented in papers:

1) Alekseeva T. A., Sokolova J. V., Afanasyeva E. V., Tikhonov V. V., Raev M. D., Sharkov E. A., Kovalev S. M., Smolyanitskiy V. M. The Contribution of Sea-Ice Contamination to Inaccuracies in Sea-Ice Concentration Retrieval from Satellite Microwave Radiometry Data during the Ice-Melt Period // Izvestiya, Atmospheric and Oceanic Physics. 2022. Vol. 58. No. 12. P. 1470–1484. DOI: 10.1134/S0001433822120039

2) Tikhonov V. V., Khvostov I.V., Alekseeva T. A., Romanov A.N., Afanasyeva E. V., Sokolova J. V., Sharkov E.A., Boyarskii D. A., Komarova N. Yu. Analysis of the Winter Hydrological Regime of the Yenisei, Pechora, and Khatanga Estuaries Using SMOS Data // Izvestiya, Atmospheric and Oceanic Physics. 2022. Vol. 58. No. 12. P. 1519–1531. DOI: 10.1134/S0001433822120234

3) Tikhonov V.V., Khvostov I.V., Romanov A.N., Sharkov E.A. A Model of Microwave Emission from Mouth Regions of Arctic Rivers Providing for Radiometer Pixel Land Contamination // Izvestiya, Atmospheric and Oceanic Physics. 2024. Vol. 60. No. 9. P. 1020–1030. DOI: 10.1134/S0001433824700981

4) Tikhonov V.V., Katamadze D.R., Alekseeva T.A. et al. Analysis of ice concentration in the Kara Sea based on SMOS MIRAS data using machine learning methods // Sovremennye problemy distantsionnogo zondirovaniya Zemli iz kosmosa.  2024. Vol. 21. No. 6. P. 344–355 (in Russian), DOI: 10.21046/2070-7401-2024-21-6-344-355.

Comments 6: The manuscript would benefit from thorough proofreading to correct minor grammatical errors and improve fluency.

Response 6: We have revised and corrected the text.

Comments 7: Some acronyms (e.g., IBCAO, OLI) are defined late or not at all. Please ensure all acronyms are defined at first use.

Response 7: We have checked and corrected all the abbreviations in the text.

With great gratitude for the valuable comments from the authors

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

Your manuscript titled “Interannual variability of sea ice dirtiness in the East Siberian Sea based on satellite data” presents an interesting dataset and a visual interpretation approach for mapping dirty ice in the East Siberian Sea over a 25-year period. The topic is both relevant and timely, with clear implications for climate research, Arctic navigation, and remote sensing applications.

However, the manuscript needs substantial revision to address methodological weaknesses, the absence of validation, and to improve clarity and critical discussion.

 

1) Validation

The maps are based entirely on expert visual interpretation of MODIS imagery, with no quantitative validation using in situ observations, automated methods, or comparison with other datasets. This lack of validation reduces confidence in the results. I suggest adding at least a limited validation or a dedicated discussion of the approach’s main limitations.

2) Methodological uncertainty

The three-point scale for classifying ice dirtiness is inherently subjective and no uncertainty analysis is provided. The paper should discuss potential observer bias, reproducibility of the manual classification, and how inter-operator differences might influence the results.

3) Limited scope

While the study focuses specifically on the East Siberian Sea, some of the conclusions appear to be extended to the entire Arctic. It would be useful to clarify the regional context, discuss how generalisable the findings are, and highlight local factors (e.g. sediment sources, ice drift patterns) that may limit broader applicability.

4) Language and presentation

Some parts of the text contain awkward phrasing and literal translations that affect readability. A careful language revision would help improve academic fluency and make the manuscript clearer for international readers.

5) Discussion

The discussion section is mainly descriptive and would benefit from a more analytical and critical tone. It would strengthen the paper to compare your results with previous studies, address the implications for Arctic processes, and clearly acknowledge the limitations of the proposed method.

Additional suggestions:

-Improve figure captions so that each figure clearly explains what is shown and how it supports the analysis.

-Include a table summarizing the classification criteria with examples for each dirtiness level.

-Clarify the temporal resolution of the dataset and explain why June–July imagery was selected.

-Ensure consistent terminology throughout (e.g., “dirty ice” vs. “ice dirtiness”).

-Add a brief description of the preprocessing steps applied to the MODIS and Landsat data.

 

This is a promising study, but it would greatly benefit from methodological refinement and a deeper discussion of its broader implications.

Comments on the Quality of English Language

Several sections contain awkward phrasing and literal translations that hinder readability. The manuscript would benefit from thorough language editing to improve academic fluency and clarity.

Author Response

Comments 1: Your manuscript titled “Interannual variability of sea ice dirtiness in the East Siberian Sea based on satellite data” presents an interesting dataset and a visual interpretation approach for mapping dirty ice in the East Siberian Sea over a 25-year period. The topic is both relevant and timely, with clear implications for climate research, Arctic navigation, and remote sensing applications.

Response 1: We are thankful to the Reviewer for the acceptance of our work and for the suggested edits. We have revised the article in accordance with the comments and highlighted all corrections in red. Almost all Figures have been reworked.

Comments 2: However, the manuscript needs substantial revision to address methodological weaknesses, the absence of validation, and to improve clarity and critical discussion.

Validation

The maps are based entirely on expert visual interpretation of MODIS imagery, with no quantitative validation using in situ observations, automated methods, or comparison with other datasets. This lack of validation reduces confidence in the results. I suggest adding at least a limited validation or a dedicated discussion of the approach’s main limitations.

Response 2: It is impossible to compare the values of the in-situ ice area with the areas on satellite images, since field observations are always point-based while satellite observations are area-based. As for comparison with other datasets, theoretically it could be done with the use of Landsat data (or some other high-resolution satellite data). However, we cannot compare the exact values of the area of contaminated ice in the Landsat and MODIS images, since Landsat observations would still be subjective, and spatial coverage of Landsat images is not enough for full comparison. Even if we had full coverage, we would need some automatic method for retrieval of ice dirtiness from Landsat and MODIS data in order to conduct an accurate comparison.  The problem is that there is no such method developed. Anyway, for developing such method, we first need to visually determine how the dirty ice looks in satellite images. This was the main task of our work. When developing the visual method of dirty ice detection in satellite images, we used the archive of field data from expeditions (photo, video, data of ice drilling). Using field data the expert could understand whether the ice is really dirty in those areas where he/she identified dirty ice in satellite images.

We explained this in Materials and Method.

Comments 3: Methodological uncertainty

The three-point scale for classifying ice dirtiness is inherently subjective and no uncertainty analysis is provided. The paper should discuss potential observer bias, reproducibility of the manual classification, and how inter-operator differences might influence the results.

Response 3: We have explained in detail the manual classification process in Materials and Methods.

Comments 4: Limited scope

While the study focuses specifically on the East Siberian Sea, some of the conclusions appear to be extended to the entire Arctic. It would be useful to clarify the regional context, discuss how generalisable the findings are, and highlight local factors (e.g. sediment sources, ice drift patterns) that may limit broader applicability.

Response 4: This study is focused on the East Siberian Sea only. In the future, we plan expand our research on other Arctic seas. After that we will be able to give a preliminary answer, if the East Siberian Sea results can be extrapolated to the entire Arctic.

Comments 5: Language and presentation

Some parts of the text contain awkward phrasing and literal translations that affect readability. A careful language revision would help improve academic fluency and make the manuscript clearer for international readers.

Response 5: We revised the text and corrected the errors.

Comments 6: Discussion

The discussion section is mainly descriptive and would benefit from a more analytical and critical tone. It would strengthen the paper to compare your results with previous studies, address the implications for Arctic processes, and clearly acknowledge the limitations of the proposed method.

Response 6: We have described the limitations of our method in the Materials and Methods section. We have not met such an analysis of the spatial distribution of ice dirtiness in any other works to compare with. Consideration of the consequences for Arctic processes requires further research. 

Additional suggestions:

Comments 7: Improve figure captions so that each figure clearly explains what is shown and how it supports the analysis.

Response 7: We've reworked all the Figures and adjusted the captions.

Comments 8: Include a table summarizing the classification criteria with examples for each dirtiness level.

Response 8: We have added Table 1 and selected better examples for each dirtiness level.

Comments 9: Clarify the temporal resolution of the dataset and explain why June–July imagery was selected.

Response 9: We explained this issue in the Materials and Methods section. (Line 220-230)

Comments 10: Ensure consistent terminology throughout (e.g., “dirty ice” vs. “ice dirtiness”).

Response 10: We corrected it where necessary. It should be noted that «dirty ice» is a description of ice state while «dirtiness» is an ice characteristics measured by a 4-point scale.

Comments 11: Add a brief description of the preprocessing steps applied to the MODIS and Landsat data.

Response 11: We added this to the Materials and Methods section (Line 220-222).

Comments 12: This is a promising study, but it would greatly benefit from methodological refinement and a deeper discussion of its broader implications.

Response 12: We are thankful to the Reviewer for valuing of our work. We have clarified the interpretation method as much as possible. Now we continue to work with additional data.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The paper called “Interannual variability of sea ice dirtiness in the East Siberian Sea based on satellite data” has conducted a lot of work, but has a small amount of innovation. There are still some issues that need to be revised.

 

Abstract Section

L16-18 In here, “There are also automatic methods for determining dirty ice from satellite data”. Why choose manual mode instead of automatic drawing of dirty ice map?

L19-21 why not write about interannual variations?

 

L22 The “spatiotemporal variation” may be the Keywords.

 

Introduction Section

Adding some literature on the impact of dirty ice on albedo, ice strength, and ice concentration identification.

L79-80 To compare with Figure 3, place names and latitude and longitude coordinates should be added to Figure 1. In addition, it is also drawn in three color charts like Figure 3.

L201 In Table 1, there were two occurrences in 2017, which belong to the medium and high area categories.

L216 Box the research area in Figure 5 and label the depth units.

 

Material and Methods Section

L113 what date is a certain period?

L151-153 L165 How is the spatial resolution of the WAVEWATVH Ⅲ model matched with the spatial resolution of wind speed data?

L159 Formula 1 symbols should use standardized subscripts, e.g. C_x instead of C_x.

 

Results Section

L177 Suggest labeling some place names in Figure 3 and making the latitude and longitude clearer. Is there a certain relationship between the distance between the dirty ice area and the shore? It is suggested to increase the discussion.

L186 L393 The vertical axis in Figure 4 and 10 is missing the axis name and unit.

L250-253 Only the wave height for 2023 has been calculated here. The annual wave height should be calculated, showing a relatively large one in 2023. It is suggested to increase the discussion.

L357 Label the place name in Figure 8. Suggest drawing dirty ice areas at different times in Figure 8.

L373 Box the research area in Figure 9.

 

Reference Section

Check all symbols after “doi”, they should be dots instead of colons.

 

Author Response

Comments 1: The paper called “Interannual variability of sea ice dirtiness in the East Siberian Sea based on satellite data” has conducted a lot of work, but has a small amount of innovation. There are still some issues that need to be revised.

Response 1: The authors are very thankful to the Reviewer for such an attentive attitude, a detailed analysis, and valuable suggestions for improving the article. We have substantially revised the article in accordance to the Reviewer’s comments. All corrections are highlighted in red. Almost all Figures were reworked.

Comments 2: Abstract Section

L16-18 In here, “There are also automatic methods for determining dirty ice from satellite data”. Why choose manual mode instead of automatic drawing of dirty ice map?

Response 2: Ice specialists from national ice services throughout the world recognize visual interpretation of satellite images as the most reliable method for determining the characteristics of sea ice (https://o365coloradoedu.sharepoint.com/sites/NSIDC-MT-IICWG/Shared%20Documents/Forms/AllItems.aspx?id=%2Fsites%2FNSIDC%2DMT%2DIICWG%2FShared%20Documents%2FGeneral%2Fother%2Dinfo%2FSummary%5FReport%2DIce%5FService%5FSurvey%5Fon%5FAutomated%5FProducts%2Epdf&parent=%2Fsites%2FNSIDC%2DMT%2DIICWG%2FShared%20Documents%2FGeneral%2Fother%2Dinfo&p=true&ga=1, https://nsidc.org/sites/default/files/wmo_574.pdf ).

Of course, automated methods for determining ice characteristics are being implemented into the processes of preparing ice maps; however, these methods are still of an auxiliary nature. It should also be noted that the developed automatic methods are mostly aimed to determine sea ice concentration and stage of development (sea ice age). Today, very few research groups deal with the issue of automatic determination of ice dirtiness. Automatic methods were offered in [21-24], but their results, first of all, require validation. Besides, these methods are not publicly available.

Comments 3: L19-21 why not write about interannual variations?

Response 3: Abstract was revised and supplemented with information about interannual variations.

 L22 The “spatiotemporal variation” may be the Keywords.

Response 3: Thanks for the suggestion, we added it.

Comments 4: Introduction Section

Adding some literature on the impact of dirty ice on albedo, ice strength, and ice concentration identification.

Response 4: In this case, we do not provide any exact research results. We believe that the list of the literature should not be expanded. The effect of dirtiness on albedo is a basic statement that follows from the general physics course. The albedo of clean white ice and the albedo of ice with dirty inclusions are naturally different. As well as the strength of clean ice and the strength of ice with dirty inclusions is different. Moreover, our study does not focus on investigating these issues. This paragraph refers to our previous work, where this issue was discussed in more detail and several links to other studies were given.

Comments 5: L79-80 To compare with Figure 3, place names and latitude and longitude coordinates should be added to Figure 1. In addition, it is also drawn in three color charts like Figure 3.

Response 5: Figure 1 has been corrected. Figure 1 was taken from other research and we referred to it, therefore we can not change the color. But we added three colored chart near the original legend.

Comments 6: L201 In Table 1, there were two occurrences in 2017, which belong to the medium and high area categories.

Response 6: We corrected this error. It should be 2018 in the third column, not 2017.

 

Comments 7: L216 Box the research area in Figure 5 and label the depth units.

Response 7: We made an additional Figure (now it is Figure 2) with the main toponyms and the research polygon. The units of depth were added. Accordingly, Figure 5 was removed.

Comments 8: Material and Methods Section

L113 what date is a certain period?

Response 8: We removed the word “certain” and further explained the methodology in detail in Materials and Methods (Line 95-97).

Comments 9: L151-153 L165 How is the spatial resolution of the WAVEWATVH Ⅲ model matched with the spatial resolution of wind speed data?

Response 9: Indeed, the spatial resolution of the wind model in the nested grid with a 9’ x 3’ resolution does not match the 0.25-degree (15-minute) wind resolution. The model resolution was chosen to adequately represent the specific features of ice formation and the fetch conditions limited by the ice edge in the East Siberian Sea.

Comments 10: L159 Formula 1 symbols should use standardized subscripts, e.g. Cx instead of C_x.

Response 10: The symbols in the formula were adjusted

Comments 11: Results Section

L177 Suggest labeling some place names in Figure 3 and making the latitude and longitude clearer. Is there a certain relationship between the distance between the dirty ice area and the shore? It is suggested to increase the discussion.

Response 11: We have corrected Figure 3. We will definitely expand the discussion in our subsequent work, as we plan to add an analysis of the dependence between spatial distribution of dirty ice and factors affecting dirtiness.

Comments 12: L186 L393 The vertical axis in Figure 4 and 10 is missing the axis name and unit.

Response 12: The figures have been corrected in accordance with the comment.

Comments 13: L250-253 Only the wave height for 2023 has been calculated here. The annual wave height should be calculated, showing a relatively large one in 2023. It is suggested to increase the discussion.

Response 13: The article is already overloaded with a description of the methodology and features of dirty ice distribution in the East Siberian Sea. The relationship between wave height and dirtiness, as well as other factors, will be highlighted in a separate study using additional field data. So far, we have only shown the most illustrative example of future research.

Comments 14: L357 Label the place name in Figure 8. Suggest drawing dirty ice areas at different times in Figure 8.

Response 14: We have added toponyms and areas of dirty ice.

Comments 15: L373 Box the research area in Figure 9.

Response 15: We believe that neither the presented schematic map nor the context of its presentation need to draw a research polygon. The map is given for understanding the general circulation in the Arctic Ocean as well as possible location of contaminated ice areas identified by Gordienko and Laktionov in the 20th century. In the article, we indicated the research area several times in previous Figures. It includes the entire East Siberian Sea, whose location may be seen on this map.

Comments 16: Reference Section

Check all symbols after “doi”, they should be dots instead of colons.

Response 16: All DOIs have been corrected

With great gratitude for the valuable comments from the authors.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors' response, which declines to incorporate a discussion on GNSS-R, is unsatisfactory and misses a critical opportunity to strengthen the manuscript. The suggestion was not to undertake a new "large independent study" but to provide a forward-looking discussion that places their optical findings within the broader context of modern remote sensing capabilities.

The limitations of optical imagery (cloud cover, daylight dependence) are central to this work, and failing to address how these limitations can be overcome by complementary technologies like GNSS-R constitutes a significant oversight. The authors' mention of parallel work on L-band microwave radiometry makes their refusal to include this discussion even more perplexing. If they possess relevant, published findings, it is imperative that they cite them here to provide a more comprehensive perspective.

Therefore, I must insist that the authors revise the manuscript to include a concise discussion in the relevant section (e.g., the Discussion or Future Work). This addition should:

1. Explicitly state how the limitations of optical data highlighted in their study can be mitigated by L-band signals, which are unaffected by clouds and daylight.

2. Briefly introduce GNSS-R as a promising technique for continuous Arctic monitoring and validation of optical-based maps.

3. Cite their own relevant L-band microwave radiometry publications to substantiate this point and demonstrate the feasibility of such an integrated approach.

This revision is not optional; it is essential to elevate the impact and scholarly rigor of the paper. The manuscript cannot be recommended for publication until this necessary context is provided.

Author Response

We sincerely thank the Reviewer for his/her contribution to making our article better and more meaningful. Following the Reviewer’s recommendations, we have added the discussion about microwave remote sensing methods to the Data and Methods (Discussion) section. These methods seem to be promising due to their independence from weather and daylight conditions; however, they cannot be applied for identifying dirty ice areas because of similarities between the complex permittivity of ice contaminants (sediments, terrigenous deposits) and metamorphosed snow or wet snow/ice cover. The explanation can be found in the text added into the Data and Methods section (new text is colored by blue, Line 129-143).

We are very thankful to the Reviewer for such an attentive attitude towards our work.

Best regards,

Authors

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors,
You have followed the reviewers' suggestions excellently, improving both methodological clarity and scientific depth. I consider the new version of the paper to be more robust and publishable.

Author Response

We are very thankful to the Reviewer for such estimation of our work.

Best regards,

Authors

Reviewer 4 Report

Comments and Suggestions for Authors

The new version has been revised according to the comment. I agree to publish.

Author Response

We are very thankful to the Reviewer for such estimation of our work.

Best regards,

Authors

Round 3

Reviewer 2 Report

Comments and Suggestions for Authors

I have no more comments.