the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Sep 2022
Contemporary (2016–2020) land cover across West Antarctica and the McMurdo Dry Valleys
Abstract. Continental-scale land cover information is essential to furthering our understanding of the terrestrial environment, atmosphere and climate change. Several global land cover products have been released in recent years but they typically do not include Antarctica. The lack of land cover data in Antarctica is concerning because mountain glaciers and icecaps there have been losing mass at a rate well above the global average, leading to expansion of proglacial regions. Proglacial regions comprise transient land cover types with high rates of geomorphological activity that delivers sediment into the Southern Ocean and supports its rich biodiversity. With Antarctic mountain glaciers and icecaps projected to lose more mass in the coming decades, and active layer soils expected to increase in thickness, it is timely to establish a baseline land cover dataset for Antarctica with which future classifications can be compared. Here, we use Landsat-8 Operational Land Imager (OLI) images to classify six proglacial regions of Antarctica at 30 m resolution, with an overall accuracy of 77.0 % for proglacial land classes. We conducted this classification using an unsupervised K-means clustering approach, which circumvented the need for training data and was highly effective at picking up key land classes, such as vegetation, water, and different sedimentary surfaces. We have highlighted the spatial pattern in land cover and emphasise a need for more and higher quality field data. The land cover maps produced from this paper are available at: Stringer, C. (2022). Contemporary (2016–2020) land cover classification across West Antarctica and the McMurdo Dry Valleys (Version 1.0) [Data set]. NERC EDS UK Polar Data Centre. https://doi.org/10.5285/5A5EE38C-E296-48A2-85D2-E29DB66E5E24.
- Preprint
(2198 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on essd-2022-250', Anonymous Referee #1, 25 Oct 2022
This manuscript developed a contemporary land cover map of West Antarctica and the McMurdo Dry Valleys from the combined Landsat-8 images during 2016-2020, using the unsupervised K-means method. Six sites were chosen for their classification and analysis. I have the following concerns:
- Since their dataset only cover six sites and only one-year map (i.e., contemporary) was provided, which greatly reduces the application value of their dataset. I don't think this dataset will be widely used.
- Although the authors state that they used the Landsat data rather than other high-resolution data (e.g., Sentinel-2) for classification due to their consideration of its long-term series and ensuring future robust and seamless comparisons, I don't think this is a good reason, as only one-year land use map was provided in this manuscript and the land use cover in this area has not changed much over the past few decades. In addition, considering that the spatial resolution of Sentinel-2 is significantly higher than that of Landsat-8, classification using Sentinel-2 data is a better choice.
- I didn’t see innovation in their methods. In addition, the accuracy of K-mean classification largely depends on the choice of K value and the selected features. However, I didn’t see any accuracy comparisons for different K values and classification features in the manuscript.
- Where are fig. 7 and fig. 8?
Based on the above comments, I don't think the current manuscript is enough to be published in ESSD. Therefore, at this stage I will reject the manuscript, I hope my decision will not disappoint the authors.
Citation: https://doi.org/10.5194/essd-2022-250-RC1 -
AC1: 'Reply on RC1', Christopher Stringer, 02 Nov 2022
Thank you very much for taking the time to review our manuscript. We have provided some individual responses to the issues you have highlighted in our manuscript. We can see that you have provided the recommendation that our manuscript be rejected, but we hope that the responses below will encourage you and the editor to allow us the chance to revise our manuscript.
Comment 1: "Since their dataset only cover six sites and only one-year map (i.e., contemporary) was provided, which greatly reduces the application value of their dataset. I don't think this dataset will be widely used."
Response 1: These six sites represent the main proglacial regions in Antarctica, including the two largest sites of the Dry Valleys and the Ulu Peninsula of James Ross Island. These are regions that are currently without a unified land cover classification map. Knowledge of contemporary land cover is needed to further understand the ecological and sedimentological make-up of Antarctica. In addition, land cover maps have applications for climatology, since land cover affects albedo and surface roughness. As an indication of their potential value, we would point out that since being published as a preprint; this manuscript has had a “Research Interest Score” higher than 92 % of the articles published on Research Gate in 2022.
The regions not covered by this land classification are excluded because they are extensively covered by ice or cloud cover, or are nunataks that are largely unimportant in the context of a changing Antarctica. We would be happy to further signpost BAS datasets that provide polygons of ice cover and exposed bedrock/land in our manuscript; we could even include these into our dataset to produce a more amalgamated map of Antarctica. While we appreciate a time series would be interesting, this contemporary land classification still provides a much-needed data set and represents a first step towards realising that ultimate, and more comprehensive, analysis.
Comment 2: "Although the authors state that they used the Landsat data rather than other high-resolution data (e.g., Sentinel-2) for classification due to their consideration of its long-term series and ensuring future robust and seamless comparisons, I don't think this is a good reason, as only one-year land use map was provided in this manuscript and the land use cover in this area has not changed much over the past few decades. In addition, considering that the spatial resolution of Sentinel-2 is significantly higher than that of Landsat-8, classification using Sentinel-2 data is a better choice."
Response 2: Whether the land cover of these regions has changed in the past or not, or indeed will change in the future, is very much unknown. Given the rapid changes in temperature and in the cryosphere observed in this region, we might suggest that change would be expected. Landsat, therefore, remains the most appropriate dataset to use, allowing for a seamless comparison using both past and future imagery. We are happy to clarify this further within the manuscript.
Comment 3: I didn’t see innovation in their methods. In addition, the accuracy of K-mean classification largely depends on the choice of K value and the selected features. However, I didn’t see any accuracy comparisons for different K values and classification features in the manuscript.
Response 3: The number of clusters was chosen based on trial and error and we found 40 clusters for the first-order classification, and 75 for the second-order, adequately allowed us to identify key land cover features (including streams, lakes, talus slopes, exposed bedrock and vegetation). Given this justification for choosing the number of clusters, and given the detailed accuracy assessment we conducted of our final classification, we are unsure what further analysis of the number of K clusters used would add to understanding the produced land classification. Ultimately, the key with this methodology, in our opinion, is the expert interpretation of what the pixels within each cluster represent; we have shown with our accuracy assessment that this has been done robustly, and we have high confidence in the quality of the output. Given the opportunity, we would be happy to add a few sentences to clarify that the final product is dependent on the number of clusters, and how they are interpreted, and that different numbers may be more appropriate at other sites (and will affect the final accuracy).
Comment 4: “Where are fig. 7 and fig. 8?”
Response 4: We would like to apologise for the references to fig. 7 (line 306) and fig. 8 (line 312). These were included in error. These should refer to figures 5 and 6 respectively. This can be easily amended.
Yours sincerely,
Christopher Stringer (on behalf of all co-authors)
Citation: https://doi.org/10.5194/essd-2022-250-AC1
-
RC2: 'Comment on essd-2022-250', TC Chakraborty, 25 Oct 2022
Summary: In the study titled “Contemporary (2016–2020) land cover across West Antarctica and the McMurdo Dry Valleys”, the authors use k-means unsupervised clustering to classify 6 proglacial regions in Antarctica using Landsat images. While I appreciate the motivation behind the study, the analysis done is simplistic and does not address the motivation sufficiently. As such, I would suggest rejection at this stage with the potential for resubmission after significant revision.
Major Comments:
- The authors argue that since the Antarctic is changing faster than other land surfaces, there is a need for land cover datasets specific to Antarctica. So, they choose 6 regions and use Landsat image between 2016-2020 to do this classification. The bands chosen are such that the classification could be expanded to include earlier years in the future. However, they only provide a static map using these limited years of observations. The study would be much more complete if they provided the land cover maps from 1976 to 2020. Otherwise, the work seems preliminary to the point that there would be no use of this dataset till the annually varying dataset is released.
- The changes in land cover in Antarctica has significant seasonal components that is impacted by ENSO oscillations and other factors. As such, I am unsure how useful an annual land cover map is for analysis of ice cover loss.
- The 6 regions represent a tiny fraction of Antarctica; and as such, the study does not really address the main motivation of the work, which is the need for a continental scale land cover dataset for Antarctica.
- K-means is a really simple method and the field has advanced in terms of classification methods. More importantly, since they already use finer resolution labelled data for validation, it would be much more useful to use supervised learning, which generally performs better than unsupervised methods.
- Why were the number of clusters chosen through trial and error instead of using commonly used elbow methods?
- There are missing figures.
Citation: https://doi.org/10.5194/essd-2022-250-RC2 -
AC2: 'Reply on RC2', Christopher Stringer, 02 Nov 2022
Thank you very much for taking the time to review our manuscript and for suggesting that resubmission may be possible. We have provided some individual responses to the issues you have raised and note that many of these are similar to those provided by Reviewer 1, so there may be some overlap in the two responses.
Comment 1: The authors argue that since the Antarctic is changing faster than other land surfaces, there is a need for land cover datasets specific to Antarctica. So, they choose 6 regions and use Landsat image between 2016-2020 to do this classification. The bands chosen are such that the classification could be expanded to include earlier years in the future. However, they only provide a static map using these limited years of observations. The study would be much more complete if they provided the land cover maps from 1976 to 2020. Otherwise, the work seems preliminary to the point that there would be no use of this dataset till the annually varying dataset is released.
Response 1: We acknowledge that a time series of land classifications would be interesting. However, the contemporary land cover map is a useful and sought-after product in its own right; especially given recent datasets that have produced global land cover classifications have failed to include Antarctica. This dataset will be widely relevant for ecologists, sedimentologists and climatologists, to name a few (see also response to reviewer 1).
Comment 2: The changes in land cover in Antarctica has significant seasonal components that is impacted by ENSO oscillations and other factors. As such, I am unsure how useful an annual land cover map is for analysis of ice cover loss.
Response 2: We agree that determining ice cover loss from such land cover maps would be problematic, but this was not the intention of our data set. A time series of land cover maps would be interesting to judge quantify changes in vegetation, lake formation and hydrology, rather than ice cover loss.
Comment 3: The 6 regions represent a tiny fraction of Antarctica; and as such, the study does not really address the main motivation of the work, which is the need for a continental scale land cover dataset for Antarctica.
Response 3: We acknowledge that these regions are only a small fraction of the total continent. However, they include the two largest proglacial areas of Antarctica (Dry Valleys and Ulu Peninsula of James Ross Island). There are very few existing datasets that provide land cover data for Antarctica, and existing datasets tend to focus on even smaller sub-regions. Existing global land cover inventories typically do not include Antarctica; therefore, these data are truly novel. The regions excluded from this analysis are nunataks or ice, with a small number of proglacial regions excluded due to cloud cover in imagery. We would be happy to make use of existing ice and bare rock datasets to produce a more complete map of Antarctica.
Comment 4: K-means is a really simple method and the field has advanced in terms of classification methods. More importantly, since they already use finer resolution labelled data for validation, it would be much more useful to use supervised learning, which generally performs better than unsupervised methods.
Response 4: We appreciate that K-means is a simple methodology, and we did explore the use of other methods (supervised learning techniques). However, we decided K-means was the most appropriate methodology and believe our outputs to be robust. Despite its simplicity, K-means is a tried and trusted methodology that is easy to accuracy assess, and an appropriate methodology given the lack of consistent and independent training data across Antarctica. We decided that the best approach was to use expert judgement to interpret clusters, rather than to train a supervised learning method with unreliable training data.
Comment 5: Why were the number of clusters chosen through trial and error instead of using commonly used elbow methods?
Response 5: Our approach was to use K-means as a way of producing clusters from which we could interpret land cover. We found trial and error was suitable for determining this, as the breaks were clear enough to decide manually. The final product has still been accuracy assessed and shows a good accuracy level.
Comment 6: There are missing figures.
Response 6: I would like to apologise for the references to fig. 7 (line 306) and fig. 8 (line 312). These were included in error. These should refer to figures 5 and 6 respectively; all figures are present in the manuscript. This can be easily amended.
Yours sincerely,
Christopher Stringer (on behalf of all co-authors).
Citation: https://doi.org/10.5194/essd-2022-250-AC2
Status: closed
-
RC1: 'Comment on essd-2022-250', Anonymous Referee #1, 25 Oct 2022
This manuscript developed a contemporary land cover map of West Antarctica and the McMurdo Dry Valleys from the combined Landsat-8 images during 2016-2020, using the unsupervised K-means method. Six sites were chosen for their classification and analysis. I have the following concerns:
- Since their dataset only cover six sites and only one-year map (i.e., contemporary) was provided, which greatly reduces the application value of their dataset. I don't think this dataset will be widely used.
- Although the authors state that they used the Landsat data rather than other high-resolution data (e.g., Sentinel-2) for classification due to their consideration of its long-term series and ensuring future robust and seamless comparisons, I don't think this is a good reason, as only one-year land use map was provided in this manuscript and the land use cover in this area has not changed much over the past few decades. In addition, considering that the spatial resolution of Sentinel-2 is significantly higher than that of Landsat-8, classification using Sentinel-2 data is a better choice.
- I didn’t see innovation in their methods. In addition, the accuracy of K-mean classification largely depends on the choice of K value and the selected features. However, I didn’t see any accuracy comparisons for different K values and classification features in the manuscript.
- Where are fig. 7 and fig. 8?
Based on the above comments, I don't think the current manuscript is enough to be published in ESSD. Therefore, at this stage I will reject the manuscript, I hope my decision will not disappoint the authors.
Citation: https://doi.org/10.5194/essd-2022-250-RC1 -
AC1: 'Reply on RC1', Christopher Stringer, 02 Nov 2022
Thank you very much for taking the time to review our manuscript. We have provided some individual responses to the issues you have highlighted in our manuscript. We can see that you have provided the recommendation that our manuscript be rejected, but we hope that the responses below will encourage you and the editor to allow us the chance to revise our manuscript.
Comment 1: "Since their dataset only cover six sites and only one-year map (i.e., contemporary) was provided, which greatly reduces the application value of their dataset. I don't think this dataset will be widely used."
Response 1: These six sites represent the main proglacial regions in Antarctica, including the two largest sites of the Dry Valleys and the Ulu Peninsula of James Ross Island. These are regions that are currently without a unified land cover classification map. Knowledge of contemporary land cover is needed to further understand the ecological and sedimentological make-up of Antarctica. In addition, land cover maps have applications for climatology, since land cover affects albedo and surface roughness. As an indication of their potential value, we would point out that since being published as a preprint; this manuscript has had a “Research Interest Score” higher than 92 % of the articles published on Research Gate in 2022.
The regions not covered by this land classification are excluded because they are extensively covered by ice or cloud cover, or are nunataks that are largely unimportant in the context of a changing Antarctica. We would be happy to further signpost BAS datasets that provide polygons of ice cover and exposed bedrock/land in our manuscript; we could even include these into our dataset to produce a more amalgamated map of Antarctica. While we appreciate a time series would be interesting, this contemporary land classification still provides a much-needed data set and represents a first step towards realising that ultimate, and more comprehensive, analysis.
Comment 2: "Although the authors state that they used the Landsat data rather than other high-resolution data (e.g., Sentinel-2) for classification due to their consideration of its long-term series and ensuring future robust and seamless comparisons, I don't think this is a good reason, as only one-year land use map was provided in this manuscript and the land use cover in this area has not changed much over the past few decades. In addition, considering that the spatial resolution of Sentinel-2 is significantly higher than that of Landsat-8, classification using Sentinel-2 data is a better choice."
Response 2: Whether the land cover of these regions has changed in the past or not, or indeed will change in the future, is very much unknown. Given the rapid changes in temperature and in the cryosphere observed in this region, we might suggest that change would be expected. Landsat, therefore, remains the most appropriate dataset to use, allowing for a seamless comparison using both past and future imagery. We are happy to clarify this further within the manuscript.
Comment 3: I didn’t see innovation in their methods. In addition, the accuracy of K-mean classification largely depends on the choice of K value and the selected features. However, I didn’t see any accuracy comparisons for different K values and classification features in the manuscript.
Response 3: The number of clusters was chosen based on trial and error and we found 40 clusters for the first-order classification, and 75 for the second-order, adequately allowed us to identify key land cover features (including streams, lakes, talus slopes, exposed bedrock and vegetation). Given this justification for choosing the number of clusters, and given the detailed accuracy assessment we conducted of our final classification, we are unsure what further analysis of the number of K clusters used would add to understanding the produced land classification. Ultimately, the key with this methodology, in our opinion, is the expert interpretation of what the pixels within each cluster represent; we have shown with our accuracy assessment that this has been done robustly, and we have high confidence in the quality of the output. Given the opportunity, we would be happy to add a few sentences to clarify that the final product is dependent on the number of clusters, and how they are interpreted, and that different numbers may be more appropriate at other sites (and will affect the final accuracy).
Comment 4: “Where are fig. 7 and fig. 8?”
Response 4: We would like to apologise for the references to fig. 7 (line 306) and fig. 8 (line 312). These were included in error. These should refer to figures 5 and 6 respectively. This can be easily amended.
Yours sincerely,
Christopher Stringer (on behalf of all co-authors)
Citation: https://doi.org/10.5194/essd-2022-250-AC1
-
RC2: 'Comment on essd-2022-250', TC Chakraborty, 25 Oct 2022
Summary: In the study titled “Contemporary (2016–2020) land cover across West Antarctica and the McMurdo Dry Valleys”, the authors use k-means unsupervised clustering to classify 6 proglacial regions in Antarctica using Landsat images. While I appreciate the motivation behind the study, the analysis done is simplistic and does not address the motivation sufficiently. As such, I would suggest rejection at this stage with the potential for resubmission after significant revision.
Major Comments:
- The authors argue that since the Antarctic is changing faster than other land surfaces, there is a need for land cover datasets specific to Antarctica. So, they choose 6 regions and use Landsat image between 2016-2020 to do this classification. The bands chosen are such that the classification could be expanded to include earlier years in the future. However, they only provide a static map using these limited years of observations. The study would be much more complete if they provided the land cover maps from 1976 to 2020. Otherwise, the work seems preliminary to the point that there would be no use of this dataset till the annually varying dataset is released.
- The changes in land cover in Antarctica has significant seasonal components that is impacted by ENSO oscillations and other factors. As such, I am unsure how useful an annual land cover map is for analysis of ice cover loss.
- The 6 regions represent a tiny fraction of Antarctica; and as such, the study does not really address the main motivation of the work, which is the need for a continental scale land cover dataset for Antarctica.
- K-means is a really simple method and the field has advanced in terms of classification methods. More importantly, since they already use finer resolution labelled data for validation, it would be much more useful to use supervised learning, which generally performs better than unsupervised methods.
- Why were the number of clusters chosen through trial and error instead of using commonly used elbow methods?
- There are missing figures.
Citation: https://doi.org/10.5194/essd-2022-250-RC2 -
AC2: 'Reply on RC2', Christopher Stringer, 02 Nov 2022
Thank you very much for taking the time to review our manuscript and for suggesting that resubmission may be possible. We have provided some individual responses to the issues you have raised and note that many of these are similar to those provided by Reviewer 1, so there may be some overlap in the two responses.
Comment 1: The authors argue that since the Antarctic is changing faster than other land surfaces, there is a need for land cover datasets specific to Antarctica. So, they choose 6 regions and use Landsat image between 2016-2020 to do this classification. The bands chosen are such that the classification could be expanded to include earlier years in the future. However, they only provide a static map using these limited years of observations. The study would be much more complete if they provided the land cover maps from 1976 to 2020. Otherwise, the work seems preliminary to the point that there would be no use of this dataset till the annually varying dataset is released.
Response 1: We acknowledge that a time series of land classifications would be interesting. However, the contemporary land cover map is a useful and sought-after product in its own right; especially given recent datasets that have produced global land cover classifications have failed to include Antarctica. This dataset will be widely relevant for ecologists, sedimentologists and climatologists, to name a few (see also response to reviewer 1).
Comment 2: The changes in land cover in Antarctica has significant seasonal components that is impacted by ENSO oscillations and other factors. As such, I am unsure how useful an annual land cover map is for analysis of ice cover loss.
Response 2: We agree that determining ice cover loss from such land cover maps would be problematic, but this was not the intention of our data set. A time series of land cover maps would be interesting to judge quantify changes in vegetation, lake formation and hydrology, rather than ice cover loss.
Comment 3: The 6 regions represent a tiny fraction of Antarctica; and as such, the study does not really address the main motivation of the work, which is the need for a continental scale land cover dataset for Antarctica.
Response 3: We acknowledge that these regions are only a small fraction of the total continent. However, they include the two largest proglacial areas of Antarctica (Dry Valleys and Ulu Peninsula of James Ross Island). There are very few existing datasets that provide land cover data for Antarctica, and existing datasets tend to focus on even smaller sub-regions. Existing global land cover inventories typically do not include Antarctica; therefore, these data are truly novel. The regions excluded from this analysis are nunataks or ice, with a small number of proglacial regions excluded due to cloud cover in imagery. We would be happy to make use of existing ice and bare rock datasets to produce a more complete map of Antarctica.
Comment 4: K-means is a really simple method and the field has advanced in terms of classification methods. More importantly, since they already use finer resolution labelled data for validation, it would be much more useful to use supervised learning, which generally performs better than unsupervised methods.
Response 4: We appreciate that K-means is a simple methodology, and we did explore the use of other methods (supervised learning techniques). However, we decided K-means was the most appropriate methodology and believe our outputs to be robust. Despite its simplicity, K-means is a tried and trusted methodology that is easy to accuracy assess, and an appropriate methodology given the lack of consistent and independent training data across Antarctica. We decided that the best approach was to use expert judgement to interpret clusters, rather than to train a supervised learning method with unreliable training data.
Comment 5: Why were the number of clusters chosen through trial and error instead of using commonly used elbow methods?
Response 5: Our approach was to use K-means as a way of producing clusters from which we could interpret land cover. We found trial and error was suitable for determining this, as the breaks were clear enough to decide manually. The final product has still been accuracy assessed and shows a good accuracy level.
Comment 6: There are missing figures.
Response 6: I would like to apologise for the references to fig. 7 (line 306) and fig. 8 (line 312). These were included in error. These should refer to figures 5 and 6 respectively; all figures are present in the manuscript. This can be easily amended.
Yours sincerely,
Christopher Stringer (on behalf of all co-authors).
Citation: https://doi.org/10.5194/essd-2022-250-AC2
Data sets
Contemporary (2016 - 2020) land cover classification across West Antarctica and the McMurdo Dry Valleys (Version 1.0) [Data set] Stringer, C. https://doi.org/10.5285/5A5EE38C-E296-48A2-85D2-E29DB66E5E24
Model code and software
Contemporary (2016–2020) land cover across West Antarctica and the McMurdo Dry Valleys [Code] (Version 1) Stringer, C. https://doi.org/10.5281/zenodo.6720051
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 595 | 221 | 41 | 857 | 30 | 30 |
- HTML: 595
- PDF: 221
- XML: 41
- Total: 857
- BibTeX: 30
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1