Permafrost Stability Mapping on the Tibetan Plateau by Integrating Time-series InSAR and Random Forest Method

Zhao, Fumeng; Gong, Wenping; Ren, Tianhe; Chen, Jun; Tang, Huiming; Li, Tianzheng

doi:https://doi.org/10.5194/tc-2022-9

Preprints

https://doi.org/10.5194/tc-2022-9

Preprints

24 Jan 2022

| 24 Jan 2022

Status: this preprint was under review for the journal TC but the revision was not accepted.

Permafrost Stability Mapping on the Tibetan Plateau by Integrating Time-series InSAR and Random Forest Method

Fumeng Zhao, Wenping Gong, Tianhe Ren, Jun Chen, Huiming Tang, and Tianzheng Li

Abstract. Ground deformation is an important index for evaluating the stability and degradation of the permafrost. Due to limited accessibility, in-situ measurement of the ground deformation of permafrost area on the Tibetan Plateau is a challenge. Thus, the technique of time-series Interferometric Synthetic Aperture Radar (InSAR) is often adopted for measuring the ground deformation of the permafrost area, the effectiveness of which is however degraded in the areas with geometric distortions in Synthetic Aperture Radar (SAR) images. In this study, a method that integrates InSAR and random forest method is proposed for an improved permafrost stability mapping on the Tibetan Plateau; and, to demonstrate the application of the proposed method, the permafrost stability mapping in a small area located in the central region of the Tibetan Plateau is studied. First, the ground deformation in the concerned area is studied with InSAR, in which 67 Sentinel-1 scenes taken in the period from 2014 to 2020 are collected and analyzed. Second, the relationship between the environmental factors (i.e., topography, land cover, land surface temperature, and distance-to-road) and the permafrost stability is mapped with the random forest method, based on the high-quality data extracted from initial InSAR analysis. Third, the permafrost stability in the areas where the visibility of SAR images is poor or the InSAR analysis results are not available is mapped with the trained random forest model. Comparative analyses demonstrate that the integration of InSAR and random forest method yields a more effective permafrost stability mapping, compared to the sole application of InSAR analysis.

Received: 12 Jan 2022 – Discussion started: 24 Jan 2022

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.

Download & links

Fumeng Zhao, Wenping Gong, Tianhe Ren, Jun Chen, Huiming Tang, and Tianzheng Li

Status: closed

RC1:
'Comment on tc-2022-9', Anonymous Referee #1, 17 Feb 2022

General Comments:

The authors propose a combination of InSAR time-series data from one viewing direction with one ML approach (Random Forest Method) to map the permafrost deformation in the Tibetan Plateau, emphasizing the area where radar visibility problems take place.

As InSAR delivers ground displacements along the slant-looking direction, visibility problems such as layover and shadow often arise in mountainous areas; it occurs in the descending track in this study. In such a case, InSAR users will usually take advantage of other data imaged from another direction that is the ascending track in this study. However, instead of using the ascending InSAR data, the authors employ the ML method to infer a permafrost stability map even at unmeasured areas. In other words, it appears as if the authors derived some signals from virtually nothing. In terms of the overall design of research work, I am not willing to recommend the authors' approach to my friends.

Furthermore, there are a couple of serious issues in the authors' interpretation of InSAR data and time-series analysis. They do not mention anything about the corrections of the tropospheric errors/artifacts. Even if they employ the time-series analysis approach with plenty of SAR images, it is impossible to ignore the tropospheric errors particularly when they use the entire image frame; the larger the imaged area, the larger the tropospheric errors. Although the authors attribute the apparent seasonal signals to the subsidence and uplift due to the freeze-thaw cycle of the active layer, we should first eliminate or minimize the tropospheric errors. Secondly, while it is related to the previous comment, Figure 14 derived from ascending track clearly indicates that the "deformation rates" are closely correlated with the local topography. Those are called topography (elevation)-correlated noise, which is again caused by tropospheric delays. They can be corrected, by fitting with the DEM. Thirdly, while InSAR tells us the surface displacements relative to non-deformed point(s), it is not clear where the reference pixels are located; the reference pixels should be stable not only in one InSAR image but also over the entire observation period. I, therefore, recommend reanalyzing the InSAR time-series data based on the ascending track, considering the points above. It is not clear why they must use the Random Forest Method; permafrost stability and landslide susceptibility follow totally different physical mechanisms. The authors should show both descending and ascending data over flat areas as verification of deformation signals as they are mostly vertical.

Specific and technical comments:

L30: As the focus is now on Tibet, those papers outside Tibet and/or Global should be removed.

L48: Delete "that"

L51: Unclear sentence

L58: The two references are not related to permafrost.

L109: Delete "in which"

L135: Replace "spatial" with "temporal"

L176: "relatively flat and homogeneous" conflicts with L115, "mountainous terrain"

L276: Is 0.8 true? There is a big deviation near the end.

Figure 6: When are the periods in the four years? Show month and date.

L319: Disagree with "in general agreement"

L333: Michaelides et al (2019) examined the post-fire area, where there occurred a big change in surface vegetation. But the authors are now examining unburned areas. If we follow the suggestion by Michaelides et al (2019), we expect significant deformation signals over "Bare lands" as there would be no insulation effects. The authors should check if there exist such signals.

Figure 12: Leveling route by Wu et al (2018) should be clarified, whereas only one leveling data was shown.

Citation: https://doi.org/10.5194/tc-2022-9-RC1
- AC1: 'Reply on RC1', Fumeng Zhao, 15 Jun 2022
  
  Thank you for your time to carefully review our manuscript and provide many high-quality comments, we sincerely appreciate the comments that have helped sharpen this paper. The detailed responses to the comments are presented in the supplement file.
  
  Citation: https://doi.org/10.5194/tc-2022-9-AC1
RC2:
'Comment on tc-2022-9', Anonymous Referee #2, 28 Mar 2022
The paper aims to develop a method for permafrost stability mapping on the Tibetan Plateau, which combines integrates InSAR and random forest. The work is an innovative and very worthwhile attempt, and it has a good guiding for disaster research in some regions with a complex geological environment like the Qinghai Tibet Plateau particularly. However, some minor issues still need to be improved. The specific comments are given as follows.

There are two spelling mistakes in line 48 and 305 that “too that many” and a sudden “s”.

Line 177: please explain why do you use vertical ground deformation, but not LOS ground deformation, i.e. what are the advantages over here by doing so?

It is mentioned in line 265 that permafrost instability mainly distributed in the valley areas with low altitude. However, in your Fig. 4(a), there are many areas with high deformation that distribute in high altitude mountainous areas. Please explain!

It is mentioned in line 266 that the ground deformation mainly took place in the west-facing slopes. In theory, it is right due to the “descending” approach of the satellite. However, in Figure 4b and 4c it seems that there are more points on the east-facing slopes. Why?

In line 370, the threshold values are set as ±0.15 mm/year and -40 mm/year. Please state or provide a scientific basis of setting up such values.

In line 397, the ROC curve is used to evaluate the accuracy of the model, but where is the ROC figure?
Citation: https://doi.org/10.5194/tc-2022-9-RC2
- AC2: 'Reply on RC2', Fumeng Zhao, 15 Jun 2022
  
  Thank you for your time to carefully review our manuscript and provide many high-quality comments, we sincerely appreciate the comments that have helped sharpen this paper. The detailed responses to the comments are presented in the supplement file.
  
  Citation: https://doi.org/10.5194/tc-2022-9-AC2
RC3:
'Comment on tc-2022-9', Anonymous Referee #3, 05 May 2022
This study proposes a new method that integrates InSAR time series and machine learning (random forest) for mapping permafrost stability in a selected region in central Tibet. This is probably one of the first efforts of such integration tailored towards estimating stability. However, the current framework relating InSAR-estimated surface deformation to permafrost stability is both conceptually flawed and poorly explained. The quality of the manuscript in its current form is further jeopardized by vague and even sometimes careless description and presentation. I raise a few major issues as detailed below and choose not to list minor editorial comments.

1. To permafrost scientists and engineers, ‘permafrost stability’ can be expressed from various perspectives, e.g., mechanical strength, permafrost thickness, temperature, active layer thickness, ground ice content, thaw settlement, engineering, hydrology, and even carbon stock and fluxes. In the abstract and introduction, the authors put forward a link between ground deformation (or should be ground *surface* deformation) as an observable for permafrost stability and the whole study has built on this concept. Since the authors didn’t define what they mean by ‘permafrost stability’, I had to guess they were mainly concerned with thaw settlement.

2. There are a few fundamental and practical issues that need to be critically addressed if using thaw settlements to infer permafrost stability.

(a) As pointed out by the authors, thaw-season subsidence is predominately caused by thawing of the active layer, not permafrost. The authors used an annual cycle + linear trend to separate seasonal and linear deformation (eq 1). Such a simple model is a reasonable choice and was adopted in many InSAR studies on permafrost. However, it is clear from Figure 5 (esp. P2) that InSAR time series often deviate from this simple time pattern. Such deviations may cause errors in the estimated trend, esp. given that the duration of the InSAR time series is only about 6 years.

(b) Since the presented results include both seasonal and linear deformation and in many places the authors didn’t explicitly state whether the deformation is seasonal or trend, I got completely lost and wasn’t sure what kinds of deformation are presented and what were used as input into the random forest. E.g. it is unclear what kind(s) of deformation are shown in Figures 7 and 8. I can only guess from the units that they are thaw-season subsidence. But wouldn’t you mainly use the trend?

(c) Only till the result section 4.2.1, the authors stated the thresholds in deformation trends to classify stable vs unstable ground as “+/- 0.15 mm/year and -40 mm/year”. Such important criteria need to be justified and introduced in the methodology. What are the bases of these thresholds? E.g., why trends larger than 40 mm/yr mean unstable permafrost, as it may seem small to different experts. Are these trends in the vertical direction or in line-of-sight (LOS) direction? 0.15 mm/year is extremely small compared to nominal uncertainties of InSAR measurements. Without estimating the uncertainties in the measured trend, is it still meaningful to set such a small threshold?

(d) Not stated explicitly in the paper, but I suppose the authors converted LOS deformation trends to vertical by assuming the ground motion is purely vertical; and used vertical trends as input to random forest and all the InSAR results presented are in the vertical direction. Then another major flaw lies in the ignorance of lateral flow on slopes in periglacial landscapes. Depending on the geomorphic type and nature of the processes, lateral movement on certain such as landforms such as solifluction sheets, rock glaciers, and even fluvial fans can move (much) faster than 40 mm/year, yet the underlying permafrost could be stable. Whereas mass wasting associated with thermokarst processes such as active layer detachment slides and thaw slumps (Figure 16 shows one example) can show very fast movement due to degrading permafrost. Without differencing the nature of deformation on flat vs slope regions and knowing the surface geomorphology, the inferred ‘permafrost stability’ would be unreliable over slopes.

(e) Yet, one of the selling points of this work is to use random forest to fill gaps in ‘poor visibility areas’ on slopes. This machine-learning-enabled advantage cannot solve the fundamental issue raised in (d).

3. The methodologic description of how integrating InSAR with random forest to infer permafrost stability is very vague to me. I raised a few concerns related to this methodology above and would summarize the key ones below.

(a) InSAR observations: seasonal or trend or both? No uncertainties.

(b) Classification of stable vs unstable: what are the bases?

(c) What are the exact inputs and outputs of the random forest?

(d) What are the reasons for selecting the topographic and climate variable? E.g. it makes little intuitive sense to include curvature and it turns out that curvature is the least important factor.

(e) What land surface temperature did you use? Annual ground surface temperature?

The authors should pay more attention to properly citing references and following scientific rigor. Here are a few examples:

Line 40: Schaefer et al., 2015 was about retrieving active layer thickness from InSAR, not about using the thickness as an index for permafrost stability.

Line 44-45: The sentence is about sparse field-based measurements. But the two papers cited are both based on remote sensing.

Line 48-49: Widhalm et al., 2017 mainly used SAR data, didn’t involve ‘many environmental factors’.

Line 62: Schaefer et al., 2015 was not a Tibet study.

Line 140: Ran et al. (2021) were concerned with permafrost temperature and thickness, not ground deformation.

(This list can be very long)

Two relevant and important papers were published recently and should provide some guidance and inspiration.

Ran et al. "Biophysical permafrost map indicates ecosystem processes dominate permafrost stability in the Northern Hemisphere." Environmental Research Letters 16.9 (2021): 095010.

Chen et al. "Magnitudes and patterns of large-scale permafrost ground deformation revealed by Sentinel-1 InSAR on the central Qinghai-Tibet Plateau." Remote Sensing of Environment 268 (2022): 112778.

In addition to the issues raised earlier, there are a few places of superficial or incorrect expressions of permafrost concept, such as:

Line 28: shrinking of ‘permafrost boundaries’ should be ‘permafrost extent.

Line 34: thickening active layer is not the root cause of carbon release from permafrost

Line 278: ground surface, not permafrost, heaves; ground (the active layer to be exact) is freezing, not frozen in September.

Line 519: deterioration should be degradation.
Citation: https://doi.org/10.5194/tc-2022-9-RC3
- AC3: 'Reply on RC3', Fumeng Zhao, 15 Jun 2022
  
  Thank you for your time to carefully review our manuscript and provide many high-quality comments, we sincerely appreciate the comments that have helped sharpen this paper. The detailed responses to the comments are presented in the supplement file.
  
  Citation: https://doi.org/10.5194/tc-2022-9-AC3

Status: closed

RC1:
'Comment on tc-2022-9', Anonymous Referee #1, 17 Feb 2022

General Comments:

The authors propose a combination of InSAR time-series data from one viewing direction with one ML approach (Random Forest Method) to map the permafrost deformation in the Tibetan Plateau, emphasizing the area where radar visibility problems take place.

As InSAR delivers ground displacements along the slant-looking direction, visibility problems such as layover and shadow often arise in mountainous areas; it occurs in the descending track in this study. In such a case, InSAR users will usually take advantage of other data imaged from another direction that is the ascending track in this study. However, instead of using the ascending InSAR data, the authors employ the ML method to infer a permafrost stability map even at unmeasured areas. In other words, it appears as if the authors derived some signals from virtually nothing. In terms of the overall design of research work, I am not willing to recommend the authors' approach to my friends.

Furthermore, there are a couple of serious issues in the authors' interpretation of InSAR data and time-series analysis. They do not mention anything about the corrections of the tropospheric errors/artifacts. Even if they employ the time-series analysis approach with plenty of SAR images, it is impossible to ignore the tropospheric errors particularly when they use the entire image frame; the larger the imaged area, the larger the tropospheric errors. Although the authors attribute the apparent seasonal signals to the subsidence and uplift due to the freeze-thaw cycle of the active layer, we should first eliminate or minimize the tropospheric errors. Secondly, while it is related to the previous comment, Figure 14 derived from ascending track clearly indicates that the "deformation rates" are closely correlated with the local topography. Those are called topography (elevation)-correlated noise, which is again caused by tropospheric delays. They can be corrected, by fitting with the DEM. Thirdly, while InSAR tells us the surface displacements relative to non-deformed point(s), it is not clear where the reference pixels are located; the reference pixels should be stable not only in one InSAR image but also over the entire observation period. I, therefore, recommend reanalyzing the InSAR time-series data based on the ascending track, considering the points above. It is not clear why they must use the Random Forest Method; permafrost stability and landslide susceptibility follow totally different physical mechanisms. The authors should show both descending and ascending data over flat areas as verification of deformation signals as they are mostly vertical.

Specific and technical comments:

L30: As the focus is now on Tibet, those papers outside Tibet and/or Global should be removed.

L48: Delete "that"

L51: Unclear sentence

L58: The two references are not related to permafrost.

L109: Delete "in which"

L135: Replace "spatial" with "temporal"

L176: "relatively flat and homogeneous" conflicts with L115, "mountainous terrain"

L276: Is 0.8 true? There is a big deviation near the end.

Figure 6: When are the periods in the four years? Show month and date.

L319: Disagree with "in general agreement"

L333: Michaelides et al (2019) examined the post-fire area, where there occurred a big change in surface vegetation. But the authors are now examining unburned areas. If we follow the suggestion by Michaelides et al (2019), we expect significant deformation signals over "Bare lands" as there would be no insulation effects. The authors should check if there exist such signals.

Figure 12: Leveling route by Wu et al (2018) should be clarified, whereas only one leveling data was shown.

Citation: https://doi.org/10.5194/tc-2022-9-RC1
- AC1: 'Reply on RC1', Fumeng Zhao, 15 Jun 2022
  
  Thank you for your time to carefully review our manuscript and provide many high-quality comments, we sincerely appreciate the comments that have helped sharpen this paper. The detailed responses to the comments are presented in the supplement file.
  
  Citation: https://doi.org/10.5194/tc-2022-9-AC1
RC2:
'Comment on tc-2022-9', Anonymous Referee #2, 28 Mar 2022
The paper aims to develop a method for permafrost stability mapping on the Tibetan Plateau, which combines integrates InSAR and random forest. The work is an innovative and very worthwhile attempt, and it has a good guiding for disaster research in some regions with a complex geological environment like the Qinghai Tibet Plateau particularly. However, some minor issues still need to be improved. The specific comments are given as follows.

There are two spelling mistakes in line 48 and 305 that “too that many” and a sudden “s”.

Line 177: please explain why do you use vertical ground deformation, but not LOS ground deformation, i.e. what are the advantages over here by doing so?

It is mentioned in line 265 that permafrost instability mainly distributed in the valley areas with low altitude. However, in your Fig. 4(a), there are many areas with high deformation that distribute in high altitude mountainous areas. Please explain!

It is mentioned in line 266 that the ground deformation mainly took place in the west-facing slopes. In theory, it is right due to the “descending” approach of the satellite. However, in Figure 4b and 4c it seems that there are more points on the east-facing slopes. Why?

In line 370, the threshold values are set as ±0.15 mm/year and -40 mm/year. Please state or provide a scientific basis of setting up such values.

In line 397, the ROC curve is used to evaluate the accuracy of the model, but where is the ROC figure?
Citation: https://doi.org/10.5194/tc-2022-9-RC2
- AC2: 'Reply on RC2', Fumeng Zhao, 15 Jun 2022
  
  Thank you for your time to carefully review our manuscript and provide many high-quality comments, we sincerely appreciate the comments that have helped sharpen this paper. The detailed responses to the comments are presented in the supplement file.
  
  Citation: https://doi.org/10.5194/tc-2022-9-AC2
RC3:
'Comment on tc-2022-9', Anonymous Referee #3, 05 May 2022
This study proposes a new method that integrates InSAR time series and machine learning (random forest) for mapping permafrost stability in a selected region in central Tibet. This is probably one of the first efforts of such integration tailored towards estimating stability. However, the current framework relating InSAR-estimated surface deformation to permafrost stability is both conceptually flawed and poorly explained. The quality of the manuscript in its current form is further jeopardized by vague and even sometimes careless description and presentation. I raise a few major issues as detailed below and choose not to list minor editorial comments.

1. To permafrost scientists and engineers, ‘permafrost stability’ can be expressed from various perspectives, e.g., mechanical strength, permafrost thickness, temperature, active layer thickness, ground ice content, thaw settlement, engineering, hydrology, and even carbon stock and fluxes. In the abstract and introduction, the authors put forward a link between ground deformation (or should be ground *surface* deformation) as an observable for permafrost stability and the whole study has built on this concept. Since the authors didn’t define what they mean by ‘permafrost stability’, I had to guess they were mainly concerned with thaw settlement.

2. There are a few fundamental and practical issues that need to be critically addressed if using thaw settlements to infer permafrost stability.

(a) As pointed out by the authors, thaw-season subsidence is predominately caused by thawing of the active layer, not permafrost. The authors used an annual cycle + linear trend to separate seasonal and linear deformation (eq 1). Such a simple model is a reasonable choice and was adopted in many InSAR studies on permafrost. However, it is clear from Figure 5 (esp. P2) that InSAR time series often deviate from this simple time pattern. Such deviations may cause errors in the estimated trend, esp. given that the duration of the InSAR time series is only about 6 years.

(b) Since the presented results include both seasonal and linear deformation and in many places the authors didn’t explicitly state whether the deformation is seasonal or trend, I got completely lost and wasn’t sure what kinds of deformation are presented and what were used as input into the random forest. E.g. it is unclear what kind(s) of deformation are shown in Figures 7 and 8. I can only guess from the units that they are thaw-season subsidence. But wouldn’t you mainly use the trend?

(c) Only till the result section 4.2.1, the authors stated the thresholds in deformation trends to classify stable vs unstable ground as “+/- 0.15 mm/year and -40 mm/year”. Such important criteria need to be justified and introduced in the methodology. What are the bases of these thresholds? E.g., why trends larger than 40 mm/yr mean unstable permafrost, as it may seem small to different experts. Are these trends in the vertical direction or in line-of-sight (LOS) direction? 0.15 mm/year is extremely small compared to nominal uncertainties of InSAR measurements. Without estimating the uncertainties in the measured trend, is it still meaningful to set such a small threshold?

(d) Not stated explicitly in the paper, but I suppose the authors converted LOS deformation trends to vertical by assuming the ground motion is purely vertical; and used vertical trends as input to random forest and all the InSAR results presented are in the vertical direction. Then another major flaw lies in the ignorance of lateral flow on slopes in periglacial landscapes. Depending on the geomorphic type and nature of the processes, lateral movement on certain such as landforms such as solifluction sheets, rock glaciers, and even fluvial fans can move (much) faster than 40 mm/year, yet the underlying permafrost could be stable. Whereas mass wasting associated with thermokarst processes such as active layer detachment slides and thaw slumps (Figure 16 shows one example) can show very fast movement due to degrading permafrost. Without differencing the nature of deformation on flat vs slope regions and knowing the surface geomorphology, the inferred ‘permafrost stability’ would be unreliable over slopes.

(e) Yet, one of the selling points of this work is to use random forest to fill gaps in ‘poor visibility areas’ on slopes. This machine-learning-enabled advantage cannot solve the fundamental issue raised in (d).

3. The methodologic description of how integrating InSAR with random forest to infer permafrost stability is very vague to me. I raised a few concerns related to this methodology above and would summarize the key ones below.

(a) InSAR observations: seasonal or trend or both? No uncertainties.

(b) Classification of stable vs unstable: what are the bases?

(c) What are the exact inputs and outputs of the random forest?

(d) What are the reasons for selecting the topographic and climate variable? E.g. it makes little intuitive sense to include curvature and it turns out that curvature is the least important factor.

(e) What land surface temperature did you use? Annual ground surface temperature?

The authors should pay more attention to properly citing references and following scientific rigor. Here are a few examples:

Line 40: Schaefer et al., 2015 was about retrieving active layer thickness from InSAR, not about using the thickness as an index for permafrost stability.

Line 44-45: The sentence is about sparse field-based measurements. But the two papers cited are both based on remote sensing.

Line 48-49: Widhalm et al., 2017 mainly used SAR data, didn’t involve ‘many environmental factors’.

Line 62: Schaefer et al., 2015 was not a Tibet study.

Line 140: Ran et al. (2021) were concerned with permafrost temperature and thickness, not ground deformation.

(This list can be very long)

Two relevant and important papers were published recently and should provide some guidance and inspiration.

Ran et al. "Biophysical permafrost map indicates ecosystem processes dominate permafrost stability in the Northern Hemisphere." Environmental Research Letters 16.9 (2021): 095010.

Chen et al. "Magnitudes and patterns of large-scale permafrost ground deformation revealed by Sentinel-1 InSAR on the central Qinghai-Tibet Plateau." Remote Sensing of Environment 268 (2022): 112778.

In addition to the issues raised earlier, there are a few places of superficial or incorrect expressions of permafrost concept, such as:

Line 28: shrinking of ‘permafrost boundaries’ should be ‘permafrost extent.

Line 34: thickening active layer is not the root cause of carbon release from permafrost

Line 278: ground surface, not permafrost, heaves; ground (the active layer to be exact) is freezing, not frozen in September.

Line 519: deterioration should be degradation.
Citation: https://doi.org/10.5194/tc-2022-9-RC3
- AC3: 'Reply on RC3', Fumeng Zhao, 15 Jun 2022
  
  Thank you for your time to carefully review our manuscript and provide many high-quality comments, we sincerely appreciate the comments that have helped sharpen this paper. The detailed responses to the comments are presented in the supplement file.
  
  Citation: https://doi.org/10.5194/tc-2022-9-AC3

Fumeng Zhao, Wenping Gong, Tianhe Ren, Jun Chen, Huiming Tang, and Tianzheng Li

Viewed

Total article views: 1,295 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
904	331	60	1,295	62	58

HTML: 904
PDF: 331
XML: 60
Total: 1,295
BibTeX: 62
EndNote: 58

Views and downloads (calculated since 24 Jan 2022)

Month	HTML	PDF	XML	Total
Jan 2022	106	42	3	151
Feb 2022	88	27	5	120
Mar 2022	74	20	6	100
Apr 2022	43	13	0	56
May 2022	58	13	2	73
Jun 2022	50	22	8	80
Jul 2022	19	6	0	25
Aug 2022	15	7	0	22
Sep 2022	20	6	1	27
Oct 2022	20	8	1	29
Nov 2022	20	12	0	32
Dec 2022	15	4	0	19
Jan 2023	13	10	1	24
Feb 2023	21	10	0	31
Mar 2023	23	6	0	29
Apr 2023	13	6	2	21
May 2023	12	6	2	20
Jun 2023	14	3	1	18
Jul 2023	14	20	2	36
Aug 2023	11	16	1	28
Sep 2023	19	9	2	30
Oct 2023	31	4	2	37
Nov 2023	9	4	3	16
Dec 2023	19	5	1	25
Jan 2024	18	4	0	22
Feb 2024	17	9	0	26
Mar 2024	27	11	2	40
Apr 2024	21	2	5	28
May 2024	14	4	3	21
Jun 2024	11	6	2	19
Jul 2024	11	5	2	18
Aug 2024	25	2	2	29
Sep 2024	21	2	0	23
Oct 2024	8	4	1	13
Nov 2024	4	3	0	7

Cumulative views and downloads (calculated since 24 Jan 2022)

Month	HTML	PDF	XML	Total
Jan 2022	106	42	3	151
Feb 2022	88	27	5	120
Mar 2022	74	20	6	100
Apr 2022	43	13	0	56
May 2022	58	13	2	73
Jun 2022	50	22	8	80
Jul 2022	19	6	0	25
Aug 2022	15	7	0	22
Sep 2022	20	6	1	27
Oct 2022	20	8	1	29
Nov 2022	20	12	0	32
Dec 2022	15	4	0	19
Jan 2023	13	10	1	24
Feb 2023	21	10	0	31
Mar 2023	23	6	0	29
Apr 2023	13	6	2	21
May 2023	12	6	2	20
Jun 2023	14	3	1	18
Jul 2023	14	20	2	36
Aug 2023	11	16	1	28
Sep 2023	19	9	2	30
Oct 2023	31	4	2	37
Nov 2023	9	4	3	16
Dec 2023	19	5	1	25
Jan 2024	18	4	0	22
Feb 2024	17	9	0	26
Mar 2024	27	11	2	40
Apr 2024	21	2	5	28
May 2024	14	4	3	21
Jun 2024	11	6	2	19
Jul 2024	11	5	2	18
Aug 2024	25	2	2	29
Sep 2024	21	2	0	23
Oct 2024	8	4	1	13
Nov 2024	4	3	0	7

Viewed (geographical distribution)

Total article views: 1,275 (including HTML, PDF, and XML) Thereof 1,275 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 17 Nov 2024

Short summary

In this study, a new permafrost stability mapping is obtained by integrating time-series InSAR and machine learning method, this method provides another alternative for measuring permafrost degradation when the ground temperature is limited to the site-specific measurements. Also, the influences of topography and vegetation coverage on the ground deformations are studied to illustrate that the permafrost stability is high related to the environmental factors.


Total:	0
HTML:	0
PDF:	0
XML:	0