Next Article in Journal
The Coastal Imaging Research Network (CIRN)
Next Article in Special Issue
Retrieving the Infected Area of Pine Wilt Disease-Disturbed Pine Forests from Medium-Resolution Satellite Images Using the Stochastic Radiative Transfer Theory
Previous Article in Journal
Spectral Comparison of UAV-Based Hyper and Multispectral Cameras for Precision Viticulture
 
 
Article
Peer-Review Record

An Effective Method for InSAR Mapping of Tropical Forest Degradation in Hilly Areas

Remote Sens. 2022, 14(3), 452; https://doi.org/10.3390/rs14030452
by Harry Carstairs 1,*, Edward T. A. Mitchard 1, Iain McNicol 1, Chiara Aquino 1, Andrew Burt 2, Médard Obiang Ebanega 3, Anaick Modinga Dikongo 4, José-Luis Bueso-Bello 5 and Mathias Disney 2
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Remote Sens. 2022, 14(3), 452; https://doi.org/10.3390/rs14030452
Submission received: 4 November 2021 / Revised: 13 January 2022 / Accepted: 13 January 2022 / Published: 18 January 2022
(This article belongs to the Special Issue Forest Disturbance Monitoring Using Satellite Remote Sensing)

Round 1

Reviewer 1 Report

Dear authors,
I read this paper with interest, and I found it rather clear, well written, and also rather fairly discussed. Indeed, this paper addresses a very relevant point through the ability of simultaneous InSAR data to provide not only forest degradation maps but also severity maps through forest AGB loss, having also in mind the current operational issue given the lack of revisit for such spaceborne configuration.
 However, the paper as it is suffers from important flaws (or open questions), limiting its interest to serve as relevant recommendations for space agencies. Most important is the conclusion about the number of look which is for me a simple artefact, optimizing on the one hand the sensitivity to AGB variations, but on the other hand the uncertainties (see theoretical errors related phase and amplitude coherence deviation vs number of looks) and I guess the false alarm detection. With this respect, a more quantitative and systematic assessment based on statistics (e.g confusion matrices or classification metrics like precision,recall, etc). Also important is to extent the comparison with existing methods based on more operational configuration (radar or optical), as performed here with Sentinel-1 except that it remains unclear if the authors have used published tools/products or if they develop it by their own (meaning of course a mitigation of methodology plus data type comparison).

Author Response

Dear reviewer 1, 

Thank you for taking the time to review our paper and for your excellent comments.

Point 1: number of looks [which is for me a simple artefact, optimizing on the one hand the sensitivity to AGB variations, but on the other hand the uncertainties (see theoretical errors related phase and amplitude coherence deviation vs number of looks) and I guess the false alarm detection. With this respect, a more quantitative and systematic assessment based on statistics (e.g confusion matrices or classification metrics like precision,recall, etc).]

You are correct that multilooking generally represents an optimisation between sensitivity and uncertainty. Our analysis, however, is always performed at a scale of 1-ha, so the question is technically about whether the averaging is performed in the complex or real domain. We have now clarified this in the abstract, methods (2.6) and conclusion, in addition to the emphasis given to this point in the discussion under the heading Importance of Few-Look Interferograms. With this in mind, it is not trivial to find that multilooking (specifically defined as spatial averaging before phase height calculation) should lead to better sensitivity to AGB loss than spatial averaging of phase height. Our results show (figure 6) that there is no longer a trade off - smaller number of looks resulted in better sensitivity and less uncertainty. We feel that given our limited ground data (4 points of change and 11 control points), it would not be meaningful to present further statistics.

Point 2: Extend comparison with existing methods [based on more operational configuration (radar or optical), as performed here with Sentinel-1 except that it remains unclear if the authors have used published tools/products or if they develop it by their own (meaning of course a mitigation of methodology plus data type comparison).]

We have clarified section 3.5 to make it clear we compare our results to a published product. This is why we only touch lightly on the methods used to map degradation using Sentinel-1. The reason for making this comparison in particular is simply that we know of no other remote sensing product that claims to quantify tropical forest degradation on a large scale, and we have also clarified this in section 3.5. As we mention in the introduction and discussion, optical data is severely limited in this region by cloud cover –we could not find even a single pair of optical images that would allow an analysis over these forest plots.

Re: Language

You checked the box “Extensive editing of English language and style required”, but we can only assume this an accident, as it contradicts your comment “I found it rather clear, well written…”. Please let us know if there are issues with the language you would like us to address.

Reviewer 2 Report

The authors have explored the use of TanDEM-X InSAR for mapping forest degradation in the tropics. The paper is well written and makes a valuable contribution to the field. I have a few minor suggestions for the authors to consider.

Line 6: worlds’ should be world’s

Line 30: Perhaps pluralize spatial patterns and temporal changes

Line 36: Remote sensing is ‘well-placed’?

Line 37: Consider rewording and remove ‘true’

Line 81: Ref [26] could go at the end of the sentence

Line 84: Consider ‘this is potentially a major issue, as much of the world’s forests are in steep terrain’

Lines 100-114: I don’t think much of this is necessary. It is enough to state the main objectives of the paper. I would particularly not mention main findings

Line 138: change ‘10s-100s’ to ‘<100m’

Line 142: Spell out full name for DLR

Line 144: full name for HH

Table 1 caption: add full name for HoA

Line 157: is time relevant?

Section 2.3. I think there is almost too much detail in the methods, especially section 2.3. For example, I don’t think you need to specify exact plot sizes when they are all more-or-less 1 ha. Also, line 187 could just say ‘DBH was measured at a height of 1.3m’

Line 237: where is the value for wood density from?

Line 257: surely if one side is parallel to north, the whole plot is

Maybe I missed something, but I’m wondering why you use SRTM for slope and not the TanDEM-X itself, given its purpose was to derive a DEM

Lines 266:277: this section is a little unclear

Line 319: is ‘typically’ located?

Line 336 or top of page 10: ‘outwith’ ?? -outside?

The sub-headings in the results section could be improved. I think 3.1 for example is about ‘Field based biomass change’ (which is almost a pre result for this paper)

Lines 419-422: This is more discussion than results I think

Figure 7: The strength of the relationship is strong, but I am wondering whether this is an artifact of the low number of field plots (4). It is possible that you just got lucky and more plots would actually reduce the correlation. The authors do touch on this, but it could be discussed more in the discussion section.

Line 448: Add ‘direction’ after pass and remove ‘allowed’

Line 454: Should ‘direction’ read ‘selection’?

Line 461: Might be useful to highlight this area on the figure 10

Page 16: Much of this seems more like discussion than results

Section 3.5: This section could be added to the methods, not just the final paragraph of the intro.

The Discussion section is a little general and not much comparison with other studies is made. Should section headings here have numbers?

Lines 524-525: I am not 100% convinced by the ‘carbon emissions’ angle (or its relevance to this work). In your study, merchantable timber was extracted, presumably for construction, so the removal of ‘carbon’ is really a transfer from a live to dead pool (i.e., it remains locked in the timber).

Author Response

Dear reviewer 2, 

Thank you for taking the time to review our paper and for your excellent comments. Please see our responses below.

Line 6: worlds’ should be world’s

Corrected

Line 30: Perhaps pluralize spatial patterns and temporal changes

Changed to “spatial and temporal patterns”

Line 36: Remote sensing is ‘well-placed’? Line 37: Consider rewording and remove ‘true’

Agreed – switched sentence order here and used “well-placed”

Line 81: Ref [26] could go at the end of the sentence

Agreed

Line 84: Consider ‘this is potentially a major issue, as much of the world’s forests are in steep terrain’

Thank you, this reads better.

Lines 100-114: I don’t think much of this is necessary. It is enough to state the main objectives of the paper. I would particularly not mention main findings

We feel it can be useful to have a “road-map” of the paper to help orientate the reader. However, if you or the editor feel strongly that it is not necessary then we can delete this paragraph.

Line 138: change ‘10s-100s’ to ‘<100m’

Changed to “< 1km” as “<100m” would not be accurate, but we agree this format looks clearer.

Line 142: Spell out full name for DLR

Added in full name

Line 144: full name for HH

Clarified this is “horizontal polarisation both in transmission and reception”

Table 1 caption: add full name for HoA

Inserted this into the caption

Line 157: is time relevant?

It could possibly be relevant if atmospheric conditions are typically different at certain times of day. However, as this is not something we investigate and there are more significant differences between the ascending and descending images, we agree this information is extraneous and have deleted this sentence.

Section 2.3. I think there is almost too much detail in the methods, especially section 2.3. For example, I don’t think you need to specify exact plot sizes when they are all more-or-less 1 ha. Also, line 187 could just say ‘DBH was measured at a height of 1.3m’

Agreed the exact plot sizes is too much detail – we have removed this and also some of the detail on DBH measurements which can be found in the reference.

Line 237: where is the value for wood density from?

This is described in the previous sub-section (see line 208).

Line 257: surely if one side is parallel to north, the whole plot is

Yes, I suppose technically two sides are parallel to north, not one. Changed this to “orientated parallel to north”.

Maybe I missed something, but I’m wondering why you use SRTM for slope and not the TanDEM-X itself, given its purpose was to derive a DEM

Yes, the TanDEM-X DEM would have been more up to date, but SRTM is more readily available without special permission required and is also posted at 30m. For our purposes, SRTM was sufficient to draw conclusions about the effect of slope.

Lines 266:277: this section is a little unclear

We have condensed this section down to the key points.

Line 319: is ‘typically’ located?

Yes not always – added ‘typically’.

Line 336 or top of page 10: ‘outwith’ ?? -outside?

Sorry, a bit of Scottish dialect! Changed to outside of.

The sub-headings in the results section could be improved. I think 3.1 for example is about ‘Field based biomass change’ (which is almost a pre result for this paper)

Agreed the headings were not particularly useful. We have made each of them more explicit.

Lines 419-422: This is more discussion than results I think

We feel it is important to show these linear trends (and briefly justify using them) here, as they are used to produce the maps at the end of the results section. The caveats to this model are indeed re-iterated in more detail in the “limitations” section of the discussion.

Figure 7: The strength of the relationship is strong, but I am wondering whether this is an artifact of the low number of field plots (4). It is possible that you just got lucky and more plots would actually reduce the correlation. The authors do touch on this, but it could be discussed more in the discussion section.

We certainly agree it is not statistically strong to have only four reference points. We have added more about this to the discussion, pointing forward to future work which should use the UAV data for more ground data.

Line 448: Add ‘direction’ after pass and remove ‘allowed’

Corrected

Line 454: Should ‘direction’ read ‘selection’?

Yes, corrected.

Line 461: Might be useful to highlight this area on the figure 10

We have given approximate coordinates to this area in the text. This should make it easy to find, but avoid cluttering the map.

Page 16: Much of this seems more like discussion than results / Section 3.5: This section could be added to the methods, not just the final paragraph of the intro. / The Discussion section is a little general and not much comparison with other studies is made. Should section headings here have numbers?

Thank you for highlighting these issues surrounding the structure and discussion. In response, we have rearranged material so that the evaluation of our biomass maps and their comparison to the Sentinel-1 study are now part of the discussion. We have also added section numbers to the discussion and included further references in the Importance of multiple geometries… section.

Lines 524-525: I am not 100% convinced by the ‘carbon emissions’ angle (or its relevance to this work). In your study, merchantable timber was extracted, presumably for construction, so the removal of ‘carbon’ is really a transfer from a live to dead pool (i.e., it remains locked in the timber).

You make a very important point here, indeed not all the carbon extracted is immediately emitted to the atmosphere. However, only a fraction of the biomass lost during selective logging will end up in a stable carbon pool, taking into account collateral damage, the discarded crowns and losses at the sawmill. This is probably why the IPCC guidelines state that total biomass extracted should be counted as an “immediate emission” – and on practical terms, these guidelines define what emissions are to countries aiming to meet their NDCs. We have added a sentence to clarify this in the discussion (now line 467).

Reviewer 3 Report

I would like to congratulate the authors on a great paper. My suggestions/comments in the attached pdf are really very minor.

Comments for author File: Comments.pdf

Author Response

Dear Reviewer 3,

Thank you for taking the time to review our paper and for your excellent eye to detail.

Re: manufacturers of UAV and TLS

These details are included in sections 2.3.2 and 2.3.3, respectively.

Re: section 2.4 confusion

This paragraph has been condensed to include only the relevant points.

Re: Figure 3 caption

We have clarified here that the two rectangles correspond to ascending and descending footprints.

Re: Figure 10 caption

The colours have been corrected to match the figure

Re: Line 643

Thank you for highlighting this sentence. It has been revised and should now make sense.

Round 2

Reviewer 1 Report

Dear authors,

I do acknowledge the improvements brought by the revisions, but the paper is still suffering from a too limited validation or verification extent. While the limitation of in situ measurements can be understood, the comparisons with other EO based methods (radar, optical or combined) could be performed, especially with respect to false alarm rate which must be better clarified. On that matter, these external methods should not be reduced to the sensor, as improperly done here with S1. Indeed, a more relevant comparison would be with the Bouvet et Al (2019) method which also combines ASC and DSC acquisitions. In addition, these comparisons should also include the total loss area, which is not only a way to deal with product resolution differences  but  also a key regional scale parameter for REDD policies.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Back to TopTop