the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Landcover succession for recently drained lakes in permafrost on the Yamal peninsula, Western Siberia
Abstract. Drained Lake Basins (DLBs) are dominant features in lowland permafrost landscapes of the Arctic. Here we present a novel approach describing and quantifying the succession progression of recently drained basins using a landcover unit retrieval scheme developed specifically for the Arctic tundra biome. The added value compared to commonly used Normalized Difference Vegetation Index (NDVI) trend analyses is demonstrated. Landcover units were linked to DLB ages (years passed since a drainage event occurred). The data were divided into bioclimatic subzones and the landcover units grouped according to their characteristics, first related to vegetation and second to wetness gradients (dry, moist and wet). A regression analyses of NDVI values and fraction of each landcover unit group provided the justification for the utility of the units in our research. The regression results showed the highest correlation with NDVI values for the wetness group ‘Moist’ and the vegetation group ‘Shrub Tundra’ (R2 = 0.458 and R2 = 0.444). There was no correlation (R2 = 0.066) found between NDVI and the fraction of group ‘Wet’ . This highlights the importance of an alternative to NDVI such as the use of landcover units to describe wetland area changes. Finally, our results showed different trajectories in the succession of landcover units in recently DLBs with respect to different bioclimatic subzones. Remaining water in the basin after a lake drainage event was highest for the most southern subzone (median 6.28 %). The open water fraction dropped below one percent for all subzones after five to ten years since drainage. The results of this study contribute to an improved understanding of DLB landcover change in permafrost environments and to a better knowledge base of these unique and critically important landforms.
- Preprint
(12465 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on egusphere-2024-699', Anonymous Referee #1, 23 May 2024
reply
This study uses a land cover classification developed from Sentinel data to look at the vegetation of 51 drained thaw lake basins on the Yamal peninsula, and how land cover changes over the first years (up to 10) after drainage. All in all, the paper presents the methods and results adequately, but does not, either the Introduction or Discussion, convince the reader that this is important, interesting work. For example, the Discussion starts with the sentence, “We showed that landcover changes in DLBs can not be detected with an analysis focusing on changes in the NDVI only.” Your most important result is not this negative finding, but rather that the landcover classes show the changing succession of land cover and consistent patterns in proportion of moisture groups during the years after lake drainage.
The introduction does not convey the importance of drained thaw lake basins on the Yamal peninsula. There is mention of general permafrost thaw leading to CO2 and methane emissions, but nothing directly related to thaw lake basins and their drainage. How do the permafrost characteristics and associated greenhouse gas emissions change with lake drainage? How are these lakes important to wildlife or reindeer herders when they are full and when they are drained? Some of the information from the first paragraph of the study area section could be moved into the introduction, particularly lines 78-81.
Section 4.1 – include, at least in the Appendix, a table which lists of all the DLBs included in the analysis, their drainage dates, area, and bioclimate subzone. It’s not clear if all 51 lakes identified are include in the study, or if older ones may have been excluded (see comment below re. line 249-250).
The inclusion of the peripheral area around the DLBs in the NDVI analysis is helpful. It shows that the method is capturing the relevant changes.
In Section 4.3, Table 3 presents data that is shown in Figure 5. I suggest moving the table to the Appendix, as I see no benefit in making readers look through both the table and the figure.
In Section 4.4, you report ground data for moisture class grouping (B), and describe cover for broad species groups. You do not compare your ground data with land cover units or land cover groups (A). You should report this and discuss these results, as you do the moisture class grouping. Also, refer to Figure A5 in this section (lines 311?).
Discussion – If the fact that there is more water retained in lakes in southern subzones is important enough to mention in the abstract (lines 12-13), then it should be discussed in the Discussion section.
Line 6 – change “A regression analyses” to “Regression analyses”
Lines 21-22 – rewrite sentence. Not clear what you are trying to say here.
Line 24 – change “cycles span over thousands of years. Those” to “cycles can span thousands of years. These”
Line 25 – introduce the acronym DLB
Lines 44-45 – change “identified the largest changes within the first five years after drainage for the NDVI.” to “identified the largest changes in NDVI within the first five years after drainage.”
Lines 63-67 – Most of this paragraph belongs in the methods. Maybe keep the second sentence, in the same paragraph as Lines 61-62.
Line 73 – change “Siberian Plain, the highest” to “Siberian Plain. The highest”
Line 74 – change “is increasing” to “increases”
Line 85-86 – move “the Vaskiny Dachi research station is located” to the beginning of the sentence
Figure 1 – in inset, remove red dots within the Yamal box.
Line 124 – change “is derived” to “was derived”
Line 125 – change “is generated” to “was generated”
Line 128 – change “is taken” to “was taken”
Line 131 – change “achieve” to “archive”
Line 131 – change “extents” to “extends”
Line 133 – change “is used for our study. Data are available” to “was used for our study. Data were available”
Line 142 – change “corresponding plant” to “corresponding vascular plant”
Line 143 – change ”50 species” to “50 vascular species”
Line 146-154 – change “mean July temperatures” to “historic mean July temperatures”. Your figure A1 shows that these temperatures are now higher
Line 149 – change “and woody plants increase in stature to hemiprostrate dwarf shrubs (<15 cm tall),” to “and hemiprostrate dwarf shrubs (<15 cm tall) are present,”
Line 159 – change “In a second step cloud free” to “In a second step, cloud free”
Line 161 – change “Landcover was derived using Sentinel-1 and Sentinel-2 data “ to “Landcover was modified from Bartsch et al. (2023a)” Or something similar indicating that the landcover built on this previous study.
Line 187- change “temperature below freezing point was used only” to “only data with temperatures below the freezing point were used.”
Line 231 – change “has been” to “was”
Line 242 – change “extend” to “extent”
Line 244 – change “expands over” to “covers”
Line 248-249 – how could a DLB age be -5? Does that mean it refilled 5 years ago? Please make this clear in the text.
Lines 249-250 – please explain further why you excluded the DLBs that drained earlier than 2012. “age gaps and inconsistent time series respectively” is not clear. With respect to what?
“Further analyses were therefore limited to a basin age of 10 years” does this mean that you excluded DLBs that drained more than 10 years ago or that you included only the first 10 years for lakes that had drained more than 10 years ago?
Lines 256-257 – Please explain either here or in the discussion why the non-drained lakes in subzones B&C had such high NDVI values. Are they generally shallower with emergent vegetation?
Lines 264-265 – this small percent change is not worth mentioning, since it is based on just one lake. The general patterns of decreasing water and wetland types, and increasing barren and dry types is the main result.
Lines 266-271 – Way too much detail for such a small sample size. Please report more general patterns, such as that #15 dry to moist tundra, partially barren was an important component of all these DLB, at all ages, but not in Subzone B.
Lines 272-276 – de-emphasize changes in years 9 and 10, when you only have 3 lakes.
Figure 3 – The brown color for #19 partially barren is different in the top two vs. bottom two bar charts. They should be changed to match. I prefer the top, which has a stronger contrast between #8 and #19 (and for Figure 4 also). Since you include color boxes for the subzones, in the caption you should refer readers to Figure 1 for Subzone mapping.
Line 283 – change “‘Water’” to “#1 ‘Water’”
Line 284 – change “‘Wetlands’” to “#2 ‘Wetlands’” – be consistent and always include the unit # (on Line 286 as wel)l. It is difficult for the reader to have to constantly refer back to Table 1.
Line 288 – remove “CAL unit” for simplicity and consistency.
Figure 4 caption – include the information that this lake is located at site 1 in Figure 1.
Line 289 – I would add that this progression for the first 6 years at one lake matches well with the patterns recorded for the set of Subzone E DLBs.
Line 295 – “higher proportion” – higher than what? Ahah – than the average overall cover in that subzone. Add reference to Figure 4.
Lines 297-301 (including Table 4) – move to Supplement – not particularly meaningful. Decrease in water cover expected, and shown by subzone in figures 3 and 5
Line 303 – change “ ‘Wet’ CALU group (grouping B).” to “ ‘Wet’ group (grouping B, Table 1).”
Line 307-308 – explain this more fully: “a wide window on wetness information reaching from ‘Moist’ to ‘Wet’.”
Lines 302-308 – You compare the ground data with wetness classes, so that we have some idea that the wetness classes matched the ground data. You give us no information on how the ground data match with the land cover units or the plant physiognomy (Group A) classes.
Lines 310-311 – need to introduce Figure 6
Line 315 – deemphasize results based on n=1. You can say that the fraction of the dry group increased over time, staying above 50% after 4 years.
Line 328 – change “analyse” to “analysis”
Line 328 – “341 different data points were available representing a basin at a certain age” Please clarify. Do you mean that you averaged the NDVI values for each DLB, to get one average value for each lake for each time period?
Line 332 – change “correlation between the NDVI and the ‘Shrub Tundra’” to “correlation between the NDVI and the percent cover of ‘Shrub Tundra’”
Line 333 – “it is rather low” – what does “it” refer to – NDVI or correlation (which is negative) or the strength of the correlation (R value)
Line 334 – and the “Grassland” group
Line 344 – delete “CALU”
Line 345 – delete “out”
Line 394 – change “showed Ermokhina et al. (2023)” to “Ermokhina et al. (2023) showed”
Line 409 – change “question, when” to “question of when”
Line 410 – delete comma
Line 417 – change “knowledge for processes” to “knowledge of processes”
Line 421 – change “It could be shown” to “We showed”
Line 425 – change – “ecotopes what is reflected” to “ecotopes, which is reflected”
Figure A2 caption - Describe which dataset the red line defining the DLB came from
Table A1 – include the corresponding imagery information for each sample point – landcover unit and group (A and B) of the nearest pixel (or group of pixels)
Figure A4 caption – change “conducted” to “collected”
Citation: https://doi.org/10.5194/egusphere-2024-699-RC1 -
AC1: 'Reply on RC1', Clemens von Baeckmann, 29 May 2024
reply
Thank you very much for your comprehensive and constructive comments. Below, you will find our responses. We have numbered the points and are focusing on those that highlight potential issues and contain specific questions. However, we will take all comments into account during revisions.
1) This study uses a land cover classification developed from Sentinel data to look at the vegetation of 51 drained thaw lake basins on the Yamal peninsula, and how land cover changes over the first years (up to 10) after drainage. All in all, the paper presents the methods and results adequately, but does not, either the Introduction or Discussion, convince the reader that this is important, interesting work. For example, the Discussion starts with the sentence, “We showed that landcover changes in DLBs can not be detected with an analysis focusing on changes in the NDVI only.” Your most important result is not this negative finding, but rather that the landcover classes show the changing succession of land cover and consistent patterns in proportion of moisture groups during the years after lake drainage.
Reply: Thank you for your valuable comment. Given that our study is exclusively focused on the Yamal Peninsula, we aimed to present the results with most caution. However, we suggest the following rephrasing:
- Old: We showed that landcover changes in DLBs can not be detected with an analysis focusing on changes in the NDVI only. Additional information on succession stages and landscape change trajectories could be derived using the landcover units.
- New: We showed the changing succession of land cover and consistent patterns in proportion of moisture groups during the years after lake drainage. In addition, our results convincingly show the capability of this landcover mapping approach to capture relevant post-drainage landcover change processes on a basin scale. Furthermore, we demonstrated that this change in the landcover cannot be detected with an analysis focusing solely on changes in NDVI.
2) The introduction does not convey the importance of drained thaw lake basins on the Yamal peninsula. There is mention of general permafrost thaw leading to CO2 and methane emissions, but nothing directly related to thaw lake basins and their drainage. How do the permafrost characteristics and associated greenhouse gas emissions change with lake drainage? How are these lakes important to wildlife or reindeer herders when they are full and when they are drained? Some of the information from the first paragraph of the study area section could be moved into the introduction, particularly lines 78-81.
Reply: We suggest adding the following publications to our introduction and rephrasing it to include these phrases.
- Publications:
- Laptander, Roza / Horstkotte, Tim / Habeck, Joachim Otto / Rasmus, Sirpa / Komu, Teresa / Matthes, Heidrun / Tømmervik, Hans / Istomin, Kirill / Eronen, Jussi T. / Forbes, Bruce C. Critical seasonal conditions in the reindeer-herding year: A synopsis of factors and events in Fennoscandia and northwestern Russia 2024-03 Polar Science, Vol. 39 Elsevier BV p. 101016, https://doi.org/10.1016/j.polar.2023.101016
- Treat, Claire C. / Virkkala, Anna‐Maria / Burke, Eleanor / Bruhwiler, Lori / Chatterjee, Abhishek / Fisher, Joshua B. / Hashemi, Josh / Parmentier, Frans‐Jan W. / Rogers, Brendan M. / Westermann, Sebastian / Watts, Jennifer D. / Blanc‐Betes, Elena / Fuchs, Matthias / Kruse, Stefan / Malhotra, Avni / Miner, Kimberley / Strauss, Jens / Armstrong, Amanda / Epstein, Howard E. / Gay, Bradley / Goeckede, Mathias / Kalhori, Aram / Kou, Dan / Miller, Charles E. / Natali, Susan M. / Oh, Youmi / Shakil, Sarah / Sonnentag, Oliver / Varner, Ruth K. / Zolkos, Scott / Schuur, Edward A. G. / Hugelius, Gustaf Permafrost Carbon: Progress on Understanding Stocks and Fluxes Across Northern Terrestrial Ecosystems 2024-02 Journal of Geophysical Research: Biogeosciences, Vol. 129, No. 3 American Geophysical Union (AGU), https://doi.org/10.1029/2023JG00763
- New: (lines 78-81 moved to line 39) …et al., 2021). Disappearing lakes were previously reported specifically for the southern part of the Yamal Peninsula (Smith et al., 2005; Nitze, 2018). Central Yamal is known for rising temperatures and changes associated with unusually warm summers (thaw slumps, active layer deepening etc.; e.g. Babkina et al., 2019; Bartsch et al., 2019a). The whole region has been shown to be a hot spot of thaw lake change (Nitze, 2018).
- New: (at line 27) …flora and fauna. The lakes are crucial for wildlife and reindeer herders both when full and drained (Kumpula et al., 2011; Laptander et al., 2024). When full, they provide essential water sources and support diverse ecosystems (Laptander et al., 2024). However, late freezing of lakes can delay reindeer migration, causing herders to move reindeer to winter pastures later than in previous decades (Kumpula et al., 2012; Laptander et al., 2024). Early snowmelt and unsafe ice cover can further complicate migration and increase the risk of avalanches and floods. When drained, the lakes create new grazing areas with abundant and diverse vegetation, attracting reindeer in the summer. This vegetation succession offers rich pastures but may pose health risks to reindeer due to unfamiliar plant species (Laptander et al., 2024).
- New: (at line 43) …et al., 2022). Permafrost thaw often increases soil moisture and lake extent, generally enhancing CH4 emissions (Schuur et al., 2022; Treat et al. 2024). In contrast, lake drainage significantly alters permafrost and greenhouse gas emissions by reducing soil moisture and changing hydrology, leading to decreased CH4 emissions (Treat et al. 2024). However, predicting CO2 and CH4 fluxes is complex due to spatial variability in vegetation, soil carbon stocks, and geomorphology across the permafrost domain. Additionally, distinguishing between wetlands and lakes in remote sensing data remains challenging, risking double counting of emission sources (Treat et al. 2024).
3) Section 4.1 – include, at least in the Appendix, a table which lists of all the DLBs included in the analysis, their drainage dates, area, and bioclimate subzone. It’s not clear if all 51 lakes identified are included in the study, or if older ones may have been excluded (see comment below re. line 249-250). The inclusion of the peripheral area around the DLBs in the NDVI analysis is helpful. It shows that the method is capturing the relevant changes.
Reply: We plan to publish the full dataset once paper is accepted. We propose to include the following information for each DLB: ID, lat, lon, area, drainage year, subzone.
4) In Section 4.4, you report ground data for moisture class grouping (B), and describe cover for broad species groups. You do not compare your ground data with land cover units or land cover groups (A). You should report this and discuss these results, as you do the moisture class grouping. Also, refer to Figure A5 in this section (lines 311?).
Reply: Given that the landcover product isn't designed for the purpose of distinguishing between different species, we found it beneficial to compare them based on their wetness grouping. Generally, increasing the amount of in situ data would enhance the precision of these comparisons. Additionally, the sample plots are in a line with varying intervals, rather than being equally distributed across the basin as it would be useful for such analysis. However, we suggest including in the Table A1 the corresponding Landcover Unit and adding this paragraph into the Discussion part:
- (lines 359-361) …local species abundance (Lantz, 2017; Loiko et al., 2020). The landcover product is not designed to differentiate between various plant species. The grouping into the wetness gradients ‘Dry’, ‘Moist’, and ‘Wet’ is supported by field vegetation data, despite limitations. Enhanced precision in these comparisons could be achieved through increased in situ data availability. Additionally, distributing sample plots evenly across the basin, rather than along a line with varying intervals, would improve the analysis. However, the wetness separation is further justified by indicator plants (see Figure A4 and Table A1). For example…
5) Discussion – If the fact that there is more water retained in lakes in southern subzones is important enough to mention in the abstract (lines 12-13), then it should be discussed in the Discussion section.
Reply: When extending the drainage data beyond the 10 years of drainage data (which we shortened due to the limited number of available basins), it appears that water levels for individual basins can fluctuate. Consequently, the remaining water decreases over the course of those 10 years, but various factors may influence further changes. Due to the lack of comprehensive data, we are careful in interpreting water fractions only. We suggest adding the following sentence into the Discussion:
- (line 384) …over time after drainage. Our results show that more water retained in lakes after a drainage event in southern subzones for the first 10 years after drainage, however this can be biased since additional factors may also contribute to further changes like floodplain related flooding or basin depth. Due to the lack of comprehensive data, we approach the interpretation of water fractions cautiously. Previously published work…
6) Lines 21-22 – rewrite sentence. Not clear what you are trying to say here.
Reply: We suggest this change:
- Old: The forming of characteristic landforms can be described as a process of the disturbance of the thermal equilibrium of the ground (Jones et al., 2011).
- New: Temperature changes are responsible for disrupting the balance of heat in the ground and influence the formation of those characteristic landforms (Jones et al., 2011).
7) Line 248-249 – how could a DLB age be -5? Does that mean it refilled 5 years ago? Please make this clear in the text.
Reply: Our initial annual Landcover database covers basin ages from 5 years before the drainage event to 24 years after the event. In this database, the year of the drainage event is represented as year 0.
- Old: Basin ages in our database ranged from -5 to 24 years, where year 0 represents the year of the event of the drainage.
- New: Our annual Landcover database included Basin ages from 5 years before the drainage event to 24 years after the drainage event took place. The year 0 represents the year of the event of the drainage.
8) Lines 249-250 – please explain further why you excluded the DLBs that drained earlier than 2012. “age gaps and inconsistent time series respectively” is not clear. With respect to what? “Further analyses were therefore limited to a basin age of 10 years” does this mean that you excluded DLBs that drained more than 10 years ago or that you included only the first 10 years for lakes that had drained more than 10 years ago?
Reply: The issue arises when we have older basins and aim to apply a fully consistent space-for-time approach. To avoid the gaps in the space-for-time sequence, we've set a limit on basin age to 10 years. This ensures that we maintain a consistent approach and have a relatively adequate number of basins covering various drainage dates. Because we separated between the Bioclimate Subzones, we're unable to allocate full 10 years to each Subzone. Consequently, in Figure 5, we increased the number of available basins by combining the different ages into two groups (1 - 5 and 6 - 10 years). We suggest the following rephrasing:
- Old: Splitting the data into the bioclimate subzones resulted in age gaps and inconsistent time series respectively. Further analyses were therefore limited to a basin age of 10 years. Age groups (1 - 5 and 6 - 10 years) were built to increase the number of recently DLBs.
- New: The separation of the data into the Bioclimate Subzones resulted in data gaps for the more northern Subzones and no full data coverage of 10 years could be achieved. Further analyses (see Figure 5) were therefore carried out introducing two age groups (1 - 5 and 6 - 10 years).
9) Lines 256-257 – Please explain either here or in the discussion why the non-drained lakes in subzones B&C had such high NDVI values. Are they generally shallower with emergent vegetation?
Reply: Thank you for bringing up this interesting aspect. We hadn’t considered the depth of the lakes yet. We suggest including the following sentence:
- (At line 257) The NDVI values in the Arctic region increase from colder to warmer bioclimate subzones, which is also shown in Figure 2.
- Raynolds, M. K. / Walker, D. A. / Maier, H. A. NDVI patterns and phytomass distribution in the circumpolar Arctic 2006-06 Remote Sensing of Environment, Vol. 102, No. 3–4 Elsevier BV p. 271-281. https://doi.org/10.1016/j.rse.2006.02.016
10) Line 307-308 – explain this more fully: “a wide window on wetness information reaching from ‘Moist’ to ‘Wet’.”
Reply: We suggest changing the sentence (line 307):
- Old: The coverage of Carex and Poaceae reached from 0% to over 90%, covering a wide window on wetness information reaching from ‘Moist’ to ‘Wet’.
- New: A broad range of fractional cover, spanning from 0% to over 90%, is observed for Poaceae in the wet group and for Carex in the moist group.
11) Lines 302-308 – You compare the ground data with wetness classes, so that we have some idea that the wetness classes matched the ground data. You give us no information on how the ground data match with the land cover units or the plant physiognomy (Group A) classes.
Reply: see Comment 4). We suggest including the corresponding Landcover Unit in Table A1. Due to the limited number of field data points available, we chose to compare them using a smaller set of distinct categories. From the 19 different Landcover Units, Group B includes only four (water, wet, moist, dry), whereas Group A includes seven categories (water, wetland, grassland, lichen and moss, shrub tundra, forest, and barren).
12) Line 328 – “341 different data points were available representing a basin at a certain age” Please clarify. Do you mean that you averaged the NDVI values for each DLB, to get one average value for each lake for each time period?
Reply: Yes, we averaged the NDVI values for each DLB for each age. We suggest following rephrasing:
- Old: 341 different data points were available representing a basin at a certain age, used for the regression analyse.
- New: 341 different data points were available for the regression analyses, with each data point representing a specific basin at a certain age. These data points include the mean NDVI value for that basin at the corresponding age, along with the fraction of the Landcover Group (e.g., Barren). One basin can have multiple data points if there are data available for different ages of that basin, which may vary for other basins.
13) Line 333 – “it is rather low” – what does “it” refer to – NDVI or correlation (which is negative) or the strength of the correlation (R value)
Reply: thanks for spotting the mistake, we suggest removing the minus (–) in the R square values in the Figures, and following rephrasing:
- Old: Negative correlation was detected for the ‘Water’ unit and the ‘Barren’ group whereas it is rather low in the latter case.
- New: Negative correlation was detected for the ‘Water’ unit and the ‘Barren’ group whereas the R value is rather low in the latter case.
Citation: https://doi.org/10.5194/egusphere-2024-699-AC1
-
AC1: 'Reply on RC1', Clemens von Baeckmann, 29 May 2024
reply
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
289 | 54 | 26 | 369 | 17 | 15 |
- HTML: 289
- PDF: 54
- XML: 26
- Total: 369
- BibTeX: 17
- EndNote: 15
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1