the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Oct 2025
A satellite-based ice fraction record for small water bodies of the Arctic Coastal Plain
Abstract. Ice cover of water bodies in the northern high latitudes (NHL) is highly sensitive to the changing climate, and its dynamics exert substantial impacts on the NHL ecosystems, hydrological processes, and the carbon cycle. Yet, operational quantification of ice cover dynamics for smaller water bodies (e.g., ≤ 25 km²) over vast, remote NHL regions remains limited. Here, we developed an ice fraction dataset for small water bodies (900 m² to 25 km²) across the Arctic Coastal Plain of Alaska (ACP) from 2017 through 2023, using Sentinel-1 Synthetic Aperture Radar (SAR) imagery, texture features, and Daymet air temperature data. The dataset has a spatial resolution of 1 km and a temporal resolution of approximately 6 days. Compared with the Google Dynamic World (DW) product derived from Sentinel-2 optical remote sensing, our dataset shows high consistency with DW (R = 0.91, RMSE = 0.19) while having enhanced temporal coverage due to less SAR constraints from solar illumination, cloud cover, and atmospheric conditions. Validation against in-situ observations suggests that our dataset is more capable of capturing small water body ice phenology (e.g., freeze-up and break-up dates) relative to DW, with an 11-day reduction in mean absolute error. Our ice fraction dataset reveals high spatial heterogeneity in ice conditions mainly occurring in June for small water bodies across the ACP. The ice phenology analysis over three selected subregions further shows that a warmer transition period generally leads to earlier ice break-up and later freeze-up, while the responses of ice fraction to warming climate vary among and within individual water bodies. The resulting dataset is anticipated to fill a gap in ice phenology studies for small water bodies, improve our understanding on the interactions between ice dynamics and climate change, and enhance the coupled modelling of ice and carbon processes. The S1 ice fraction dataset is publicly available at https://doi.org/10.5281/zenodo.17033546 (Lin et al., 2025)
- Preprint
(8731 KB) - Metadata XML
-
Supplement
(1723 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on essd-2025-503', Anonymous Referee #1, 19 Nov 2025
-
RC2: 'Comment on essd-2025-503', Anonymous Referee #2, 20 Nov 2025
This work introduces a new product to support ice cover research in the north. It also presents a novel method that can be used in other regions to monitor ice cover fraction. This will be highly beneficial for those working in northern areas. The writing is clear and well-structured, and the data is easily accessible and well-labelled for the most part. I have only a few minor suggestions to improve clarity for the reader or user.
‘small water bodies’, to me, refers to lakes, rather than rivers – so the validation sites being rivers seemed to come out of the blue while reading. The authors should consider clarifying in the abstract that they are referring to lakes and rivers as small water bodies.
While I understood the product had 2 bands of data, it wasn’t clear to me that each image did not cover the entire study area, so some clarification on that could be added to the text.
Overall model performance shows many pixels in the moderate-to-large error category (section starting around line 304). Given the data limitations, I agree that this is still a very useful product. The per-pixel quality product, however, could use some clarification. The text lists the values as percentages, but the product loads with a scale of 0 – 153.8. What is the link between uncertainty and the RRMSE values in the quality file? Perhaps even just in the .md file, some explanation of what exactly the RRMSE in the tif are in terms of uncertainty would be helpful.
Also, can the user be given some cautions to watch for regarding reasons for high RRMSE? e.g., some of the larger errors occur in these regions (it did not appear to me to be particularly spatially based, from a brief review of the data product), or on this size or type of water body, etc.? or is there no discernable set of reasons? With the understanding that this is a data paper and not the venue for a deep exploration of the reasons, a brief comment or two to help the user would be beneficial.
A few minor things to note:
Figure 1: The map should include a panel with an overview of the site’s location for context. Even just an outline of Alaska would help to see where it is situated.
Line 193: The authors explain that ascending and descending are processed separately, but don’t mention why. For clarity, it might be helpful to mention here for those less familiar with radar and the orbital times. (I fully agree with the methods used and the separate processing; this is just a clarification.)
Figure 2: b) Constructing the dataset using the 4 types of input data is clear. In panel C, then, the data set goes through the RF model, and it appears that two of the original datasets are then used again to generate the ice cover maps. The text makes it clear that the RF model was applied to S1 to generate the 10m map. Perhaps the authors could make panel C clearer for workflow, but this might just be a matter of my interpretation.
Line 344: “This discrepancy may result from a mismatch between the observed freeze-up phase and the phase captured by remote sensing,” can the authors clarify what they mean here?
Citation: https://doi.org/10.5194/essd-2025-503-RC2
Data sets
Data and Code for paper "A satellite-based ice fraction record for small water bodies of the Arctic Coastal Plain" Hong Lin, Jinyang Du, and John S. Kimball https://doi.org/10.5281/zenodo.17033546
Model code and software
Data and Code for paper "A satellite-based ice fraction record for small water bodies of the Arctic Coastal Plain" Hong Lin, Jinyang Du, and John S. Kimball https://doi.org/10.5281/zenodo.17033546
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 242 | 38 | 23 | 303 | 40 | 22 | 25 |
- HTML: 242
- PDF: 38
- XML: 23
- Total: 303
- Supplement: 40
- BibTeX: 22
- EndNote: 25
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript is a valuable contribution to the delineation of lake ice/water cover using SAR imagery. Although coverage is limited to the ACP, the algorithm shows promise for application to broader areas across the northern hemisphere. The data provides advantages over optical imagery as expected of active microwave. The comparison to DW is a provides suitable validation for the ice fraction product.
There are some minor comments that would be good to address. Overall, the manuscript quality is very high but there a few key points that should be addressed.
There is clear indication that this product could be extended to be an operational product. Was there a reason that other operational products were not compared? For example, the CCI lakes lake ice cover product is available at roughly a 1km resolution and covers some of the lakes in the study area. A comparison to the CCI product would be beneficial due to the similarity between methods, both use a random forest algorithm to classify ice cover.
Another question for the authors relates to the choice of texture as a variable for the classifier. The citation provided was conducted for sea ice, however, to the reviewers knowledge no formal exploration of texture has been done for lake ice. Did the authors conduct any investigation into texture values for lake ice? For example, does the texture provide any context for heterogenous surfaces during freeze-up? break-up? Was an investigation done into the temporal evolution of the texture pattern?
There is no variable importance analysis provided - was this conducted? It would be of interest to users to see how the valuable the different variables were in the random forest classifier. The classifier used both SAR parameters and temperature variables, how does the model rate these? The concern here being that the classifier is being driven by temperature rather than SAR/EO data which is the original goal.