Determination of Structural Characteristics of Old-Growth Forest in Ukraine Using Spaceborne LiDAR
Round 1
Reviewer 1 Report
The manuscript by Spracklen and Spracklen addresses a very interesting topic, namely the assessment of 3D structural characteristics from spaceborn GEDI laser scanning.
Overall, I think the work is worth being published, but some issues should be addressed beforehand:
- The manuscript does not state any hypothesis guiding the research. Even though there are to research goals presented (l.62) i miss some traditional hypothesis.
- The manuscript basically ignores almost any study using ground-based LiDAR for the task of assessing structures in OGF. There is several studies that I know that addressed the Uholka area in Ukraine using ground-based LiDAR.
- The manuscript focuses a lot on height, a stand measure that is rather "boring" when it comes to assessing structures. The authors discovered several issues with measured heights (large errors) by comparing their GEDI-based heights with other remote sensing data. Accordingly, several areas were removed from the analysis. This is legit, but I wonder how the authors can convnce the reader that the reminder of the measures derived (much more complex than height) is solid?
- Ground-thruthing is based on the WWF data, but since this is absolutely crucial for the interpretation of the findings I would expect that the process and data is explained and introduced to the reader in the present study. Just refering to the WWF study is not enough here.
- I miss a clear description of how the NOGF were managed, since this is decisive for their structures. See for example “A quantitative comparison of the structural complexity of managed, lately unmanaged and primary European beech (Fagus sylvatica L.) forests” for the effects of management in beech forests vs. old-growth.
Specific comments:
l.29: OGF are characterized by a lack of (detectable) human disturbance in general, not just recent.Consider deleting “recent”
l.29-33: Citation needed for this description of OGF
l.48: Airborne is one word right?
l.60: high-resolution is relative, I consider the GEDI data rather coarse resolution.
l.62: You should convert the research question into hypothesis.
l.76: sycamore “maple”
l.88: GEDI is mentioned here but not yet properly introduced.
l.94: Citation 26 is the paper that I think must at least be introduced in a little summary-like synthesis, so the reader must not read it to understand the content of the present study.
l.98: You suddenly talk about these shape-files, but the reader does not know where they come from or what they contain. The shape-file data must be introduced/explain.
l.106: To me, the GEDI scanning tracks issue remains unclear despite your describtion. Since the issue seems important please provide a figure illustration the 60m, 600m, the split beam, the power tracks and the cover tracks.
l.124: The error of the TandemX is ±8-10 m….this is a lot to me. How do you deal with this error when you derive measures like FHD with 5 m thick layers?
L.142: I suggest replacing “overwhelmingly” with “predominantly”
l.144: please replace “beech dominant” with “beech-dominated” throughout the manuscript.
l.192: How can you hypothesize that the presence of shrub or herb layer might be a useful indicator of undisturbed site and therefore OGF status? Earlier studies showed that it is not that simple. See for example “Assessing understory complexity in beech-dominated forests (Fagus sylvatica L.) in Central Europe—From managed to primary forests” and you will notice it depends on stand age, management type etc.
chapter 3.1.2 and 3.1.3. contain information that is better placed in the discussion.
Chaption of figure 5, l.362: Please add the structure for the mixture. The reader could understand a third class easily and it would help the reader to get a better picture. Its just box-plots, the reader can grasp that.
Figure 6: I suggest using exchanged axis here, so that the height is on the y-axis. This is more intuitive and would be more similar to what readers might be used to from “stand profiles”.
Table 2 and elsewhere: You split the data based on the 25 m threshold and I wonder what your motivation was to decide for this threshold.
l.412 to 416: To run the RF separately on the forest types seems awkward, since you must know them in advance before you could do that elsewhere, which means you need detailed forest mapping to apply your method. This should at least be discussed.
l.443 to 453: This is all about forest heights you derived vs. heights reported in other studies. What is the benefit of having that? In my opinion, you can’t compare the forest height of forest in Ukraine with a forest in Romania or Slovakia, since height is determined by site characteristics you did not address. So what is the benefit? Either explain better why you do so or remove (which I would strongly recommend).
l.468-472: You really can’t say that your coniferous sites exhibited old-growth characteristics to a much greater extent and on a wider range of metrics than the broadleaf. Coniferous and broad-leaved forest have naturally different structures. They look different when they are old-growth and therefore different measure apply to how “old-growthy” they look.
Author Response
REVIEWER 1:
We want to thank the Reviewer for this careful, thorough and knowledgeable review. We appreciate the time and effort it takes, and we think the suggested corrections will greatly improve the manuscript. We give our responses to the comments below, with each comment followed by our response in italics.
The manuscript by Spracklen and Spracklen addresses a very interesting topic, namely the assessment of 3D structural characteristics from spaceborn GEDI laser scanning.
Overall, I think the work is worth being published, but some issues should be addressed beforehand:
- The manuscript does not state any hypothesis guiding the research. Even though there are to research goals presented (l.62) i miss some traditional hypothesis.
Response- Changed to “Here we use this free, publicly available spaceborne LiDAR data to test the following hypotheses:
- OGF will differ significantly from reference managed forest stands (non-OGF, NOGF) on a number of vertical structure metrics. We expect OGF to be have a more open canopy and to be structurally more complex.
- The differences in structure between OGF and NOGF will permit the effective use of Random Forest classification to classify OGF and NOGF.”
- The manuscript basically ignores almost any study using ground-based LiDAR for the task of assessing structures in OGF. There is several studies that I know that addressed the Uholka area in Ukraine using ground-based LiDAR.
Response: Thanks to the reviewer to pointing out these interesting studies. We make just a quick mention of some of them in the Introduction: “Ground based LiDAR studies of forest structure are common, with, for example, these recent papers [10–13] all studying an OGF site in the Ukrainian Carpathians.” The citations are
- Stiers, M.; Willim, K.; Seidel, D.; Ehbrecht, M.; Kabal, M.; Ammer, C.; Annighöfer, P. A Quantitative Comparison of the Structural Complexity of Managed, Lately Unmanaged and Primary European Beech (Fagus Sylvatica L.) Forests. Forest Ecology and Management 2018, 430, 357–365.
- Stiers, M.; Annighöfer, P.; Seidel, D.; Willim, K.; Neudam, L.; Ammer, C. Quantifying the Target State of Forest Stands Managed with the Continuous Cover Approach–Revisiting Möller’s “Dauerwald” Concept after 100 Years. Trees, Forests and People 2020, 1, 100004.
- Willim, K.; Stiers, M.; Annighöfer, P.; Ammer, C.; Ehbrecht, M.; Kabal, M.; Stillhard, J.; Seidel, D. Assessing Understory Complexity in Beech-Dominated Forests (Fagus Sylvatica L.) in Central Europe—From Managed to Primary Forests. Sensors 2019, 19, 1684.
- Willim, K.; Stiers, M.; Annighöfer, P.; Ehbrecht, M.; Ammer, C.; Seidel, D. Spatial Patterns of Structural Complexity in Differently Managed and Unmanaged Beech-Dominated Forests in Central Europe. Remote Sensing 2020, 12, 1907.
We also add a short paragraph into the Discussion taking into account these interesting papers, and covering your Comment 5 below, as follows:” Historically, management of forest in the Ukrainian Carpathians has largely consisted of clearfelling and replanting, with minimal interest in continuous cover forestry. This practice generally results in blocks of even-aged plantation forest. A ground-based LiDAR study [11] has noted that the structural complexity of continuous cover forestry is significantly greater than in even-aged monocultures. We therefore note that distinguishing OGF from forest managed over long periods of time with continuous cover precepts may prove to be a tougher proposition than the plantation forestry largely encountered in our study. Such difficulties and differences may prove an interesting avenue of further research.”
- The manuscript focuses a lot on height, a stand measure that is rather "boring" when it comes to assessing structures. The authors discovered several issues with measured heights (large errors) by comparing their GEDI-based heights with other remote sensing data. Accordingly, several areas were removed from the analysis. This is legit, but I wonder how the authors can convnce the reader that the reminder of the measures derived (much more complex than height) is solid?
Response- In response to this comment, we make some significant alterations. We look at structural parameter ‘canopy cover’, a stand variable for which we have field study values from ground surveys in Uholka, Ukrainian Carpathians. We have added a third filter to our study, using the canopy cover. Accordingly we have expanded and rewritten Section 3.1 (Error Measurement section). The addition of the third canopy cover filter changes our results slightly, but the findings of our paper still stand. Much of the explanation of this Section has been moved to the Discussion, in compliance with a comment below.
- Ground-thruthing is based on the WWF data, but since this is absolutely crucial for the interpretation of the findings I would expect that the process and data is explained and introduced to the reader in the present study. Just refering to the WWF study is not enough here.
Response: We add a paragraph outlining the main criteria for OGF identification as follows: “Forests in the Ukrainian Carpathians have recently been field-surveyed by the World Wildlife Fund (WWF) to identify any remaining areas of OGF [27]. Resulting from this survey were shapefiles of identified OGF, with each shapefile including detailed information on its tree species composition. Criteria used in the survey for identification of forest as OGF included abundant lying and standing dead wood, a complex structure with a high diversity of tree sizes and ages, and a lack of visible anthropogenic influence. All the stands in our study were surveyed between 2010 and 2018. “
- I miss a clear description of how the NOGF were managed, since this is decisive for their structures. See for example “A quantitative comparison of the structural complexity of managed, lately unmanaged and primary European beech (Fagus sylvatica L.) forests” for the effects of management in beech forests vs. old-growth.
Response: We add the following to Section 2.1 (Materials and Methods, Study Area): “NOGF areas have been primarily managed in a system of large-scale clearfelling and replanting, resulting in a forest of even-aged stands of plantation forest. In 2009 it was estimated that in the Ukrainian Carpathians, 72% of harvested timber derives from clearfelling, 24% shelter-belt cutting and just 4% as a result of selection thinning. [30] While in recent years there has been an increased emphasis on sustainable forest management practices, [31] with the ultimate aim of producing more uneven-aged stand structures and continuous forest cover, much of the NOGF in the study area will be of an even-aged plantation structure, with a simplified age, species and spatial structure.”
Citations are: 30 Keeton, W.S.; Angelstam, P.K.; Bihun, Y.; Chernyavskyy, M.; Crow, S.M.; Deyneka, A.; Elbakidze, M.; Farley, J.; Kovalyshyn, V.; Kruhlov, I.; et al. Sustainable Forest Management Alternatives for the Carpathian Mountains with a Focus on Ukraine. In The Carpathians: Integrating Nature and Society Towards Sustainability; Springer, 2013; pp. 331–352.
- Krynytskyi, H.T.; Chernyavskyi, M.V.; Krynytska, O.H.; Deineka, A.M.; Kolisnyk, B.I.; Tselen, Y.P. Close-to-Nature Forestry as the Basis for Sustainable Forest Management in Ukraine. Науковий вісник НЛТУ України 2017, 27.
Specific comments:
l.29: OGF are characterized by a lack of (detectable) human disturbance in general, not just recent.Consider deleting “recent”
Response: changed “recent” for “detectable”
l.29-33: Citation needed for this description of OGF
Response: added citation for structure: “Franklin, J.F.; Van Pelt, R. Spatial Aspects of Structural Complexity in Old-Growth Forests. Journal of Forestry 2004, 102, 22–28.”
and for dead wood:
“Keren, S.; Diaci, J. Comparing the Quantity and Structure of Deadwood in Selection Managed and Old-Growth Forests in South-East Europe. Forests 2018, 9, 76.”
l.48: Airborne is one word right?
Response: changed to “airborne”
l.60: high-resolution is relative, I consider the GEDI data rather coarse resolution.
Response: changed “high resolution” for “accurate”
l.62: You should convert the research question into hypothesis.
Response: see Comment 1 above
l.76: sycamore “maple”
Response: changed to “sycamore maple”
l.88: GEDI is mentioned here but not yet properly introduced.
Response: changed “GEDI shots” to “areas”
l.94: Citation 26 is the paper that I think must at least be introduced in a little summary-like synthesis, so the reader must not read it to understand the content of the present study.
Response: We have added relevant material to the Methods section, including criteria for OGF identification.
l.98: You suddenly talk about these shape-files, but the reader does not know where they come from or what they contain. The shape-file data must be introduced/explain.
Response: We change to: “For comparative purposes we selected Non-OGF shapefiles. These NOGF shapefiles had been surveyed by the WWF, but had not met the criteria for classification as old-growth. The GEDI shots lying within these shapefiles were used for comparison to the OGF.”
l.106: To me, the GEDI scanning tracks issue remains unclear despite your describtion. Since the issue seems important please provide a figure illustration the 60m, 600m, the split beam, the power tracks and the cover tracks.
Response: Added descriptive figure as Figure 2.
l.124: The error of the TandemX is ±8-10 m….this is a lot to me. How do you deal with this error when you derive measures like FHD with 5 m thick layers?
Response: Apologies for our lack of clarity, we mean that the error in ground elevation values above sealevel given by the TANDEMX data is ±8-10m. We’ve slightly altered this section to make clear: “The ground elevation of each GEDI shot (above sea-level) is produced by the GEDI’s LiDAR instrument. Additionally, the elevation given by the TanDEM-X digital elevation model is also provided in the GEDI products. TanDEM-X elevation values are produced using X-band Synthetic Aperture Radar (SAR) on two satellites, have a resolution of 0.4 arcseconds and in forested, mountainous terrain similar to our study area has a mean error of ±8-10 m. [31,32] “
L.142: I suggest replacing “overwhelmingly” with “predominantly”
Response: changed as requested
l.144: please replace “beech dominant” with “beech-dominated” throughout the manuscript.
Response: changed to “beech-dominated” and “spruce-dominated” here and on lines 148, 151 and 152.
l.192: How can you hypothesize that the presence of shrub or herb layer might be a useful indicator of undisturbed site and therefore OGF status? Earlier studies showed that it is not that simple. See for example “Assessing understory complexity in beech-dominated forests (Fagus sylvatica L.) in Central Europe—From managed to primary forests” and you will notice it depends on stand age, management type etc.
Response: Thanks for pointing out this interesting study. We have rewritten this paragraph (including the citation) as: “A LiDAR study [33] in Central and Eastern Europe (including the Ukrainian Carpathians) of the difference in understory complexity of beech stands managed by thinning compared to unmanaged OGF found significant differences between the Ukrainian OGF study site and some of the managed sites, dependent on stand age and presence of tree regeneration.”
chapter 3.1.2 and 3.1.3. contain information that is better placed in the discussion.
Response- We move most of the content from Sections 3.1.1- 3.1.3 ( GEDI Error sections) to the start of the Discussion.
Chaption of figure 5, l.362: Please add the structure for the mixture. The reader could understand a third class easily and it would help the reader to get a better picture. Its just box-plots, the reader can grasp that.
Response: as suggested, mixed forest category added to Figure 6
Figure 6: I suggest using exchanged axis here, so that the height is on the y-axis. This is more intuitive and would be more similar to what readers might be used to from “stand profiles”.
Response: axes switched as suggested
Table 2 and elsewhere: You split the data based on the 25 m threshold and I wonder what your motivation was to decide for this threshold.
Response: We add explanation to Section 2(Methods) as follows: “. We noted forest structure differed between areas of shorter and taller canopy. We therefore examined canopy structure separately for canopy heights of less than and greater than 25m. This specific value was chosen as it is roughly half the maximum canopy height, and has been used in European forest studies as the definition of the upper canopy layer [39] , or to divide mature and middle-aged woodland from younger stands. [40]”
Citations are: 39. Rugani, T.; Diaci, J.; Hladnik, D. Gap Dynamics and Structure of Two Old-Growth Beech Forest Remnants in Slovenia. PloS one 2013, 8, e52641.
- Parpan, T.V.; Parpan, V.I. Ecological Structure of Beech and Coniferous/Beech Mountain Climax Forest Stands of Ukrainian Carpathians. Науковий вісник НЛТУ України 2017, 27.
l.412 to 416: To run the RF separately on the forest types seems awkward, since you must know them in advance before you could do that elsewhere, which means you need detailed forest mapping to apply your method. This should at least be discussed.
Response: We incorporate your point into the text, as an explanation for why we feel it necessary to both examine the accuracy with separated forest types and without separating the forest types. We feel this explanation fits best in the Random Forests Materials and Methods section (2.4), to which we add the following: “However, we recognize that to run the RF classification separately on the different forest types requires detailed knowledge of the area, which is not always as readily available as it is in our study area. We therefore also carry out the RF classifications without separating the forest types and examine how this impacts the classification accuracy.”
l.443 to 453: This is all about forest heights you derived vs. heights reported in other studies. What is the benefit of having that? In my opinion, you can’t compare the forest height of forest in Ukraine with a forest in Romania or Slovakia, since height is determined by site characteristics you did not address. So what is the benefit? Either explain better why you do so or remove (which I would strongly recommend).
Response: Removed as suggested
l.468-472: You really can’t say that your coniferous sites exhibited old-growth characteristics to a much greater extent and on a wider range of metrics than the broadleaf. Coniferous and broad-leaved forest have naturally different structures. They look different when they are old-growth and therefore different measure apply to how “old-growthy” they look.
Response: Deleted this line, now reads: “. In our study area, these characteristics were present in both conifer (predominantly Norway spruce) and broadleaf (predominantly beech) OGF (Fig. 5 and 6).”
Reviewer 2 Report
The study focuses on using Machine Learning Algorithm as Random Forest and fused data to predict old-growth forests. The successful use of GEDI data in combining into a MLA based on optical satellite data offers various perspectives in other fields of studies which deals with quantifying the forest structures (e.g. hydrological modelling, forest management etc.)
Minor suggestions:
line 26 - please use deforestation instead of forest clearence
lines 38,41 - missing reference to support the statement
line 433 - change the term "very little" with other such as : scarce, poor etc.
line 517 - term shapefile is designed for naming the file containing vectors. Maybe you want to say areas or zones
Author Response
Reviewer 2:
We would like to thank the Reviewer for their time. We give our responses to the comments below, with each comment followed by our response in italics.
The study focuses on using Machine Learning Algorithm as Random Forest and fused data to predict old-growth forests. The successful use of GEDI data in combining into a MLA based on optical satellite data offers various perspectives in other fields of studies which deals with quantifying the forest structures (e.g. hydrological modelling, forest management etc.)
Minor suggestions:
line 26 - please use deforestation instead of forest clearance
Response: changed as requested
lines 38,41 - missing reference to support the statement
Response: Added supporting reference “Omar, H.; Misman, M.A.; Kassim, A.R. Synergetic of PALSAR-2 and Sentinel-1A SAR Polarimetry for Retrieving Aboveground Biomass in Dipterocarp Forest of Malaysia. Applied Sciences 2017, 7, 675.”
line 433 - change the term "very little" with other such as : scarce, poor etc.
Response: Replaced with “insufficient”
line 517 - term shapefile is designed for naming the file containing vectors. Maybe you want to say areas or zones
Response- Apologies, we weren't very clear in our Methods section as to our use of shapefiles showing OGF areas supplied by the World Wildlife Fund Ukraine. We amend the lines in Section 2 so that they are hopefully clearer: "Forests in the Ukrainian Carpathians have recently been field-surveyed by the World Wildlife Fund (WWF) to identify any remaining areas of OGF [31]. Resulting from this survey were shapefiles of identified OGF, with each shapefile including detailed information on its tree species composition. "
Round 2
Reviewer 1 Report
The authors responded to all my comments and I see no more issues.
