Monitoring the Transformation of Arctic Landscapes: Automated Shoreline Change Detection of Lakes Using Very High Resolution Imagery
Round 1
Reviewer 1 Report
Comments to the Authors
Kaiser et al. have developed an automated workflow to facilitate detection of lake shoreline changes by analysing multi-temporal high resolution imagery for two areas in Alaska. The results of this study facilitate development of improved tools that can be used to assess Arctic landscape change and to further understanding of the causes of these changes and their impacts. Permafrost scientists, hydrologists and landscape ecologists will find this study interesting and the tools applicable to their image analysis. The MS is generally well written and provides a clear rationale for the study and a good description of methodology and its testing and validation. The limitations and errors associated with the automated workflow are also adequately assessed and described. With minor revisions the MS should be acceptable for publication. A few comments are offered below for consideration by the authors in preparing a revised MS.
The title and in the text refer to shoreline change detection of thermokarst lakes. However, the criteria used in the workflow to determine whether the water body is a thermokarst lake is not provided. Not all lakes in the permafrost regions are initiated through thermokarst processes, although subsequent changes to shorelines may be related to thermokarst processes. Oxbow lakes are mentioned in the MS and these features are not specific to permafrost regions. They result from fluvial erosion of meanders and become isolated from the river when the meander is cut off. They are therefore initiated through non thermokarst processes and may be subject to permafrost related process after their formation. In other regions, lakes may be structurally controlled or the product of glacial/quaternary history (see for example Wolfe et al. 2020). It might be better to refer to just refer to “lake shoreline change detection” rather than specifically shoreline change detection of thermokarst lakes.
The direction of shoreline movement is discussed on page 9 and is predominantly southward (including SW and SE) and towards the east. There is however no discussion regarding the reason for this preferential orientation. I realize that the paper focusses mainly on development of the automated workflow but some comment on potential reasons would be useful. I note that there is discussion regarding reasons for changing water levels and shoreline position so further discussion of direction of shoreline movement would seem to fit with this. Does prevailing wind have an influence on direction of movement – see for example Cote and Burn (2002).
“Shoreline movement” can occur in a direction that results in a larger lake or a smaller lake. The authors discuss changing water levels and do mention lower levels in drier periods and potentially smaller lakes. It is not clear whether the authors are referring to any movement in shoreline position when they provide summary values in the text and tables or specifically movement that results in enlargement of the lake (i.e. are + and – movement considered). It might also be useful to consider providing the change in lake area (% change) as this might give the reader a better idea of the changes in small vs larger lakes.
Multi-panel figures are provided and are referred to as left, right, top, bottom. This makes reference to figures in the text awkward. The panels in each figure should be labelled A, B etc. as this would make it easier and clear to refer to them in the text.
Additional specific comments are provided below keyed to line number
L 1 – Consider revising. Thermokarst is a product of permafrost thaw, with the presence of ponded water influencing the ground thermal regime which can further contribute to thaw.
L22-23 – Consider referring to “ice-rich permafrost”. You might also consider adding that poor drainage is also important in pond/lake formation.
L30 – You could give some idea of the size here.
L 33-34 – revision suggested “… using analysis of multi-temporal imagery at different…”
L36 – the resolution of this imagery could be given
L 38 – revision suggested “….very high resolution (better than xx m) airborne…”
L 82 – You could have a map of AK in figure 2 to provide the reader a better idea of where these areas are.
L 96-97 – Suggested revision “Both sites are underlain by ice-rich permafrost with ice wedge polygons”
L98-102 – Be clear here that the multi-temporal imagery is only available for the one site.
L 156 – What is the rational for removing water bodies of this size – wouldn’t this mean that you would be missing smaller ponds and the initiation of a pond?
L190 – 197 – Is this the size in 2006? Shouldn’t the first category be 0.1 to 0.36 ha since water bodies smaller than 0.1 ha have been filtered out?
Table 2 – see previous comment regarding the smallest lake category. Number of detected lakes are the ones that appear on both images – have I got this right? (maybe this could be clearer in the caption). Can you also provide the number of lakes detected on all 4 images?
L250-280 – Are you referring to the right figures in this section? Figure 5 is mentioned before figure 4. You may need to renumber figures.
L250-253 – Awkward sentences. See earlier comment regarding labelling of panels in figures as A, B etc. Also consider integrating the figure reference into your text rather than a sentence that describes the figure and repeats information that should be in the figure caption.
L259-261 – Reasons for this?
L262 – Suggested revision “The median rate of shoreline movement for ….
L264 – suggested revision “Lakes larger than 23 ha…”
L262 – 267 – Can you give some idea of variability, for example the range in the rate for each size category along with the median?
L270 – suggested revision “For 13 lakes (32%) the shoreline moved towards SE ….” (the shorelines moved rather than the lake)
L291 – 297 – This is an important point – could you show this in a figure? For the other errors mentioned in the list below, you could also show examples in a figure.
L354 – High movement rate resulting in lake enlargement?
L351-353 – Were these other studies in similar terrain with similar climate, permafrost conditions etc? Were these movement rates also determined from high res imagery or through other methods including field investigations?
L354 – 367 – How important is the timing of the imagery. The dates of the imagery range from early July to mid August especially since there are seasonal differences as you mention here. Would it be desirable for example to compare images towards the end of the thaw season for example, to get an idea in overall shoreline change. Consideration of precipitation and temperature conditions during the time periods considered would seem to be important to include in the analysis and interpretation.
References cited in comments:
Cote MM, Burn CR (2002) The oriented lakes of Tuktoyaktuk peninsula, western arctic coast, Canada: a GIS-based analysis. Permafrost and Periglacial Processes 13:61-70
Wolfe, SA et al. (2020) Oriented-lake development in the context of late Quaternary landscape evolution, McKinley Bay Coastal Plain, western Arctic Canada. Quaternary Science Reviews 242: 106414 https://doi.org/10.1016/j.quascirev.2020.106414
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The topic of this paper is interesting and important in the context of the increased vulnerability of the cryosphere due to climate changes. This study presents an automated workflow that allows the calculation of the thermokarst lake shoreline. In my opinion, the paper is suitable for publication in Remote Sensing, but minor issues might be addressed.
Despite this is concise methodological research, I suggest including an insert map with the localisation of the study area in Fig. 2.
It is not clear if this study is conducted in a region where available thermokarst lakes inventory obtained from high-resolution imagery exist. Muster et al (2017) did an inventory of this kind and I was wondering if there are comparable lakes data in this region (would be nice to compare your findings with their results). Maybe add a phrase if necessary on this issue.
It is not clear if this methodology presented here is better than other solutions presented in the Introduction (lines 51-64) or other available tools, such as Digital Shoreline Analysis System (Himmelstoss et al., 2018). Maybe, within Discussions, you may refer to this.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
