Monitoring the Structure of Regenerating Vegetation Using Drone-Based Digital Aerial Photogrammetry
Round 1
Reviewer 1 Report
Dear authors, thank you for a refreshingly detailed, but still concise manuscript that gave careful consideration to its data collection, processing, and analysis approach. I would suggest that you consider whether all of the text descriptions of the results are necessary, and trim those that do not directly lead to an insight or a discussion point. My feeling was that the value of the manuscript lies primarily in the (excellent) demonstration of the design for a study, and insight into the current and potential capabilities of the technology, rather than answering any wider ecological questions, and therefore you could probably stand to lose some of the descriptions of the results. However, I would also understand if some of those results were being stated to set up ecological hypotheses for future studies, so I am definitely not insisting the results be reduced. All of the analytical techniques and their interpretations seem perfectly reasonable, and are supported by competent visualizations. The drone imagery was collected with careful attention, and some commendable redundancies, to maximizing the quality and accuracy of the photogrammetry, and the study sets a good example for others.
I do have a list of suggested minor corrections and cosmetic changes to the language, and also note several linguistical inconsistencies:
Line 23 “of shortest” to “of the shortest”?
Line 28 “which requires” to “that requires”?
Line 36 “natural sources” to “natural resources”?
Line 37 “implications on” to “implications for”?
Line 45 “policies some” to “policies, some”?
Line 76 “sensor” to “sensors”?
Line 82 “data is” to “data are” and “joined” to “joint”?
Line 88 “mesh” to “meshes”?
Line 89: “Advantages” to “The advantages”?
Line 95: “frequent” to “more frequent”?
Line 99 “data was” to “data were”
Line 147 “which covered” to “and which covered”
Line 152 “slope and elevation” to “slopes and elevations”?
Line 195 “short” to “short-“?
Line 196 “where” to “were”?
Lines 199-200 Is a camera mounted if it is integrated?
Table 3 “pts m-2” needs superscription of “-2”
Line 295 “which evaluates for” to “which evaluates, for”?
Line 308 “on” to “to” and “meters” to “m” for consistency?
Line 347 “clusters (k) selected” to “clusters (k), selected”?
Line 351 “clustering, the final” to “clustering. The Final”?
Line 352-353 Rearrange to clarify that it is the branches in which pre-clusters are more similar, not the dendrogram
Line 337 “cluster” to “Cluster” for consistency? Or drop the capitalizations elsewhere.
Figure 6 caption “bi-plots” to “bi-plot”?
Line 381-381 “first consisting cluster” to “first, consisting of cluster”? And also, decide whether you are capitalizing Cluster when referring to your numbered clusters.
Line 388 “is characterized” to “was characterized” for consistency of tense?
Line 412 “Spp.” Probably doesn’t capitalization or italicization.
Figure 8 caption “0.5 x 5 m2” would be “0.5x5 m2” elsewhere. Should “class” and “cluster” be capitalized here as elsewhere? Or is this a more general reference and you only capitalize references to specific classes and clusters?
Line 438-441 Consider rephrase as “followed” implies Site A is being presented as the highest, but the language at the beginning of the sentence doesn’t state this.
Line 442 Now “class one” has lost its capitalization! It’s the new Cluster! Same again in lines 443-445.
Line 451 “is shown” to “are shown” and “indicates” to “indicate”? Or “distributions” to “distribution” instead?
Line 482 “followed Class” to “followed by Class”, and this sentence has a similar problem to the sentence in lines 438-441 that “followed” implies the first component is being presented as the most or least of something, but the language doesn’t reflect this.
Line 503 “site A” to “Site A”!
Line 505 “site B” to “Site B”!!
Line 508 “highest” to “the highest”
Line 523 “to ROW’s” to “to the ROW’s”?
Line 533 “5” to “five”? You have been pretty dedicated to writing out integer numbers less than 10 up until this point.
Line 540-541 “allowed the detailed” is not compatible with “than studies”. It’s probably meant to be “allowed more detailed”?
Line 543 Take one of the “in”s in “in in” out out.
Line 547 “the Cloud” to “cloud processing” or “cloud computing” or something similar?
Line 559-560 “In our study image” to “In our study, image”?
Line 560-561 “imagery, high” to “imagery, and high”
Line 561-562 “site” to “Site” twice!
Line 577 “succesional” to “successional”
Line 590-591 implies “erosion” is a “restoration objective”. Is that the intended meaning, or should it be “erosion mitigation” or similar?
Line 614 “in relation with” to “in relation to”?
Throughout:
Inconsistency in the use of Oxford commas.
Inconsistency of hyphenation of short-stature
Use of “which” for non-parenthetical statements where “that” would be appropriate.
“ROWs” is established as the abbreviation for Rights of Way, but is used as ROW while still in reference to the plural form.
“in situ” is italicized throughout, but “et al.” is not (line 307, this is also the only time an author and date is included, even when a reference is addressed directly, which happens several times in the discussion) and “post hoc” is not (Figure 7 caption).
Several references to “dendogram” that I believe are supposed to be “dendrogram”, although maybe I’m behind the times?
Author Response
Dear reviewer,
Thank you for your detailed review, which helped us solve many issues and improve the manuscript overall. We were very pleased with your positive feedback and kind response. Great you found particularly the demonstration of the unsupervised learning approach and insights in in the current potential and limitation of DAP interesting. This was indeed our primary aim. You can find our response to your comments below and the edited manuscript attached.
Kind regards,
Rik Nuijten on behalf of all authors
_______________________________________________________________________________
_______________________________________________________________________________
Dear authors, thank you for a refreshingly detailed, but still concise manuscript that gave careful consideration to its data collection, processing, and analysis approach. I would suggest that you consider whether all of the text descriptions of the results are necessary, and trim those that do not directly lead to an insight or a discussion point. My feeling was that the value of the manuscript lies primarily in the (excellent) demonstration of the design for a study, and insight into the current and potential capabilities of the technology, rather than answering any wider ecological questions, and therefore you could probably stand to lose some of the descriptions of the results. However, I would also understand if some of those results were being stated to set up ecological hypotheses for future studies, so I am definitely not insisting the results be reduced. All of the analytical techniques and their interpretations seem perfectly reasonable, and are supported by competent visualizations. The drone imagery was collected with careful attention, and some commendable redundancies, to maximizing the quality and accuracy of the photogrammetry, and the study sets a good example for others.
We reviewed section 3.3 which includes the descriptions of the “final structure classes”. We made some small edits (Class 5 was described two times) and changed as follows :
Line 422: “Class four is absent of vegetation structures and characterized by exposed soil, mosses and short grass. Class five includes mosses, short grasses and small seedlings (approximately 30—40 cm). Class six is representative of short graminoids and perennials, that sometimes grew together with small seedlings, including aspen, black spruce and willow.”
We have chosen to avoid stating a hypothesis related to successional processes including plant-soil interactions as that would require many assumptions as well as extensive soil measures (microtopography, organic matter, texture etc.). For example, we could hypothesize that vegetation can to some extent be an indicator of the level of disturbance (plants regenerating from propagules present in undisturbed soil vs. light-seeded species?). We leave this for further studies utilizing both structural and multispectral data, in which we are likely able to discriminate plant communities closer aligned with such more functional plant communities.
I do have a list of suggested minor corrections and cosmetic changes to the language, and also note several linguistical inconsistencies:
Line 23 “of shortest” to “of the shortest”?
The suggestion is implemented.
Line 28 “which requires” to “that requires”?
The suggestion is implemented.
Line 36 “natural sources” to “natural resources”?
The suggestion is implemented.
Line 37 “implications on” to “implications for”?
The suggestion is implemented.
The same grammatical error was found in Line 111 and corrected.
Line 45 “policies some” to “policies, some”?
The suggestion is implemented.
Line 76 “sensor” to “sensors”?
The suggestion is implemented.
Line 82 “data is” to “data are” and “joined” to “joint”?
Both suggestions are implemented.
Line 88 “mesh” to “meshes”?
The suggestion is implemented.
Line 89: “Advantages” to “The advantages”?
The suggestion is implemented.
Line 95: “frequent” to “more frequent”?
The suggestion is implemented.
Line 99 “data was” to “data were”
The suggestion is implemented.
Line 147 “which covered” to “and which covered”
The suggestion is implemented.
Line 152 “slope and elevation” to “slopes and elevations”?
The suggestion is implemented.
Line 195 “short” to “short-“?
The suggestion is implemented.
Line 196 “where” to “were”?
The suggestion is implemented.
Lines 199-200 Is a camera mounted if it is integrated?
Line 191: “mounted on” has be replaced by “carried by”. The current sentence states: “…a 20-megapixel camera measuring in red, green and blue carried by a DJI Phantom 4 Pro (integrated camera) or DJI Matrice 200 v2 (Zenmuse X5S camera) system.”
Table 3 “pts m-2” needs superscription of “-2”
The suggestion is implemented.
Line 295 “which evaluates for” to “which evaluates, for”?
The suggestion is implemented.
Line 308 “on” to “to” and “meters” to “m” for consistency?
The suggestion is implemented.
Line 347 “clusters (k) selected” to “clusters (k), selected”?
The suggestion is implemented.
Line 351 “clustering, the final” to “clustering. The Final”?
The suggestion is implemented.
Line 352-353 Rearrange to clarify that it is the branches in which pre-clusters are more similar, not the dendrogram
Line 351: We now state “The dendrogram consists of stacked branches which partition the pre-clusters in more homogeneous clusters.”
Line 337 “cluster” to “Cluster” for consistency? Or drop the capitalizations elsewhere.
The suggestion is implemented, I would like to keep capitalizations for numbered clusters and classes.
Figure 6 caption “bi-plots” to “bi-plot”?
The suggestion is implemented.
Line 381-381 “first consisting cluster” to “first, consisting of cluster”? And also, decide whether you are capitalizing Cluster when referring to your numbered clusters.
The suggestion is implemented.
Line 388 “is characterized” to “was characterized” for consistency of tense?
The suggestion is implemented.
Line 412 “Spp.” Probably doesn’t capitalization or italicization.
This should indeed not be capitalized nor italicized, this is changed.
Figure 8 caption “0.5 x 5 m2” would be “0.5x5 m2” elsewhere. Should “class” and “cluster” be capitalized here as elsewhere? Or is this a more general reference and you only capitalize references to specific classes and clusters?
The suggestion is implemented.
Line 438-441 Consider rephrase as “followed” implies Site A is being presented as the highest, but the language at the beginning of the sentence doesn’t state this.
Agreed. we now say (Line 439) :“Total cover by relatively large and complex structures, presumably woody vegetation, was highest at Site A (Lower Boreal Highlands) with 35.2% followed by Site C (Upper Foothills) and B (Lower Boreal Highlands) with 27.6% and 13.9% respectively.”
Line 442 Now “class one” has lost its capitalization! It’s the new Cluster! Same again in lines 443-445.
The suggestion is implemented and the document has been searched for similar cases.
Line 451 “is shown” to “are shown” and “indicates” to “indicate”? Or “distributions” to “distribution” instead?
The suggestion is implemented.
Line 482 “followed Class” to “followed by Class”, and this sentence has a similar problem to the sentence in lines 438-441 that “followed” implies the first component is being presented as the most or least of something, but the language doesn’t reflect this.
We now state (Line 477):
“The southwestern half of the ROW, which included a power line, had large amounts of Class four (vegetation structure absent) and six (very short but dense vegetation structures).”
Line 503 “site A” to “Site A”!
The suggestion is implemented.
Line 505 “site B” to “Site B”!!
The suggestion is implemented.
Line 508 “highest” to “the highest”
The suggestion is implemented.
Line 523 “to ROW’s” to “to the ROW’s”?
The suggestion is implemented.
Line 533 “5” to “five”? You have been pretty dedicated to writing out integer numbers less than 10 up until this point.
The suggestion is implemented.
Line 540-541 “allowed the detailed” is not compatible with “than studies”. It’s probably meant to be “allowed more detailed”?
We now state (Line 531)
“The incorporation of metrics derived over a regularized surface, describing vegetation height, cover, variability and surface complexity, allowed more detailed characterization of vegetation patterns than studies which simply utilized the canopy height model, such as [23].”
Line 543 Take one of the “in”s in “in in” out out.
The suggestion is implemented.
Line 547 “the Cloud” to “cloud processing” or “cloud computing” or something similar?
The suggestion is implemented and also the words “integration with the” are removed.
Line 559-560 “In our study image” to “In our study, image”?
The suggestion is implemented.
Line 560-561 “imagery, high” to “imagery, and high”
We have decided to split the sentence, which results in (Line 552):
“In addition, image pitch angles from the DJI Phantom 4 acquired imagery were highly variable. High bidirectional reflectance was observed above exposed soil at Site A and B and heavy shadowing was found at site C.”
Line 561-562 “site” to “Site” twice!
The suggestion is implemented and the document has been searched for similar cases.
Line 577 “succesional” to “successional”
The suggestion is implemented.
Line 590-591 implies “erosion” is a “restoration objective”. Is that the intended meaning, or should it be “erosion mitigation” or similar?
We have changed “erosion” to “soil stabilization”, which better represents the action/objective.
Line 614 “in relation with” to “in relation to”?
The suggestion is implemented.
Throughout:
Inconsistency in the use of Oxford commas.
We have removed all Oxford commas but have retained commas if we say something in the lines of “Cover percentages and, if applicable, heights of….”
Inconsistency of hyphenation of short-stature.
Short-stature is now hyphenated throughout
Use of “which” for non-parenthetical statements where “that” would be appropriate.
This makes many sentences clearer. We have replaced “which” by
“that” everywhere the clause provides essential information.
“ROWs” is established as the abbreviation for Rights of Way, but is used as ROW while still in reference to the plural form.
We changed rights-of-way (ROWs) to rights-of-ways (ROWs).
“in situ” is italicized throughout, but “et al.” is not (line 307, this is also the only time an author and date is included, even when a reference is addressed directly, which happens several times in the discussion) and “post hoc” is not (Figure 7 caption).
We have removed the full reference and italicized “post hoc” for consistency.
Several references to “dendogram” that I believe are supposed to be “dendrogram”, although maybe I’m behind the times?
This was a very consistent typo, we have changed it throughout the manuscript.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Comments on “Monitoring Structure of Regenerating Vegetation using Drone-based Digital Aerial Photogrammetry” by Nuijten et al.
Monitoring vegetation structure is an essential task for evaluating the efforts of vegetation restoration. In recent years, DAP has shown great potentials in quantifying vegetation structure with very low cost. In this study, the authors conducted a site-specific but very detailed study on evaluating the capability of DAP on monitoring vegetation structure in areas with ongoing anthropogenic disturbance in temperate coniferous and boreal forests. Overall, the manuscript is very well written, and easy to follow. The experiment is appropriately designed and described in the manuscript. The results can inform forest managers on how to use DAP to evaluate vegetation restoration in this region. I think the manuscript is ready for publication. I have only a few comments for the authors to consider.
First, the whole analysis was on the basis of DAP-derived vegetation structure information. Although I do agree that the clustering analysis based on the DAP-derived vegetation information is very informative on the vegetation regenerating status, I still kept thinking whether can further include the spectral information in the analysis to further differentiate vegetation compositions. I know it might be a little bit too much for this manuscript, but it would still be nice to discuss this in the discussion section.
Second, it would be great if the authors can spend some space to discuss the differences of vegetation status in disturbed and undisturbed areas, and how they are differentiated in different study sites, if the authors have data.
A small mistake in Line 366: 54 should be 54%.
Author Response
Dear reviewer,
Thank you for your detailed review, which helped us solve to incorporate some major improvements such as a new “future outlook” paragraph at the end of the discussion. We were very pleased with your positive feedback and kind response. Great you found particularly the demonstration of the unsupervised learning approach and insights in in the current potential and limitations of DAP interesting. This was our primary aim. You can find our response to your comments below.
Kind regards,
Rik Nuijten on behalf of all authors
Monitoring vegetation structure is an essential task for evaluating the efforts of vegetation restoration. In recent years, DAP has shown great potentials in quantifying vegetation structure with very low cost. In this study, the authors conducted a site-specific but very detailed study on evaluating the capability of DAP on monitoring vegetation structure in areas with ongoing anthropogenic disturbance in temperate coniferous and boreal forests. Overall, the manuscript is very well written, and easy to follow. The experiment is appropriately designed and described in the manuscript. The results can inform forest managers on how to use DAP to evaluate vegetation restoration in this region. I think the manuscript is ready for publication. I have only a few comments for the authors to consider.
First, the whole analysis was on the basis of DAP-derived vegetation structure information. Although I do agree that the clustering analysis based on the DAP-derived vegetation information is very informative on the vegetation regenerating status, I still kept thinking whether can further include the spectral information in the analysis to further differentiate vegetation compositions. I know it might be a little bit too much for this manuscript, but it would still be nice to discuss this in the discussion section.
Second, it would be great if the authors can spend some space to discuss the differences of vegetation status in disturbed and undisturbed areas, and how they are differentiated in different study sites, if the authors have data.
We have decided to add a new paragraph at the end of the discussion presenting a future outlook for the technology, which includes the use of multispectral data, improved flight protocols and insights into data consistency as well as further potential in the field of forest restoration. We emphasize on the possibility to improve effective discrimination between plant communities by using both the spectral and structural information during analysis. The new paragraph is as follows:
"Future studies should investigate the potential for the application of off-the-shelf multispectral and hyperspectral drone based sensors, in combination with spectral vegetation indices that capture unique characteristics of plant reflectance, for improving effective discrimination between plant communities. Where a structural or spectral dataset alone may not able to resolve the boundaries of a plant community, a combination of the two may improve discrimination as demonstrated by [76]. In addition, it is important to investigate the consistency of the derived information under varying flight parameters and develop novel flight protocols considering simplicity, cost effectiveness and quality of derived information. Considering that there is a growing push for restoring and monitoring of sites impacted by resource extraction, as well as interest for monitoring of present-day reference sites within the ecological restoration community [77], we suggest drones will play an increasingly important role in forest restoration projects in the future."
Second, it would be great if the authors can spend some space to discuss the differences of vegetation status in disturbed and undisturbed areas, and how they are differentiated in different study sites, if the authors have data.
We thought it would be a good idea to mention that there is growing interest in monitoring reference sites within the restoration community in addition to an increased push for restoring and monitoring sites following resource extraction, which allows drones to fulfill an important role in such projects.
At the moment we have not acquired data at present-day references but we are planning to collect more data again this summer at sites with similar characteristics. Constructing and describing a reference would likely be out of the scope of this manuscript. We would like to avoid making many assumptions based on complex ecological processes that may be different between the disturbed and undisturbed sites. For example, the level of disturbance (root layers still present, compaction, soil mixing) plays a role at the site-level and can perhaps be “revealed” by presence of young woody vegetation vs. opportunists.
A small mistake in Line 366: 54 should be 54%.
The suggestion is implemented.
Author Response File: Author Response.docx
Reviewer 3 Report
Introduction
Line 89 - What are the main limitations of DAP-UAV in relation to LiDAR? Does it work well and in areas with the dense forest? And the terrain, can it influence the data collected?
What can possible problems be seen in restoration areas? Can the representation of the vertical structure be a problem? Is this due to the smaller size of the vegetation being regenerated (height <2 m)? What does the literature say about it?
Methodology
Table 1 - Elevation and aspect? Slope and aspect?
Line 176 - What is the average RMSE of the location of the points?
Line 179 - Why this amount? Any reference?
Line 208 - What is the average RMSE of the processing of the 20 GCPs?
Line 214 - What is the RMSE in X, Y, and Z at the end of processing?
Table 3 - Why such a high HAGL? Can this not hinder the reconstruction of the height of vegetation? Some studies recommend flying lower in these conditions.
Figure 2 - Shouldn't the point cloud come before height normalization?
Line 240 - What are the distance and angle values used? Inform.
Line 256 - Why the maxim? Why not P99% or P95%? Explain.
Results of
Figure 3 - It is essential to inform the RMSE%.
Discussion
Line 551 and 554 - It is vital to inform the type of vegetation analyzed in the mentioned study [51].
Line 555 - What is the reason? Does this usually happen? Do the flight characteristics interfere? Relief? Type of vegetation? Could the automatic DTM construction/normalization process have interfered with this?
Line 556 - Explain the reason for the difficulty encountered by the DAP.
Author Response
Dear reviewer,
We are pleased with your positive feedback and would like to thank you for your detailed review, which helped us make some major improvements to the manuscript. The biggest change can be found in the discussion. The rewritten paragraph in the discussion now provides a comprehensive overview of potential factors influencing accuracy of DAP height by drawing lines between our own results and other studies. The second major improvement is the way RMSEs are calculated, which improved transparency and comparability with others. You can find our response to your comments below and can find the edited manuscript attached.
Kind regards,
Rik Nuijten on behalf of all authors
________________________________________________________________________________
________________________________________________________________________________
Comments and Suggestions for Authors
Introduction
Line 89 - What are the main limitations of DAP-UAV in relation to LiDAR? Does it work well and in areas with the dense forest? And the terrain, can it influence the data collected?
What can possible problems be seen in restoration areas? Can the representation of the vertical structure be a problem? Is this due to the smaller size of the vegetation being regenerated (height <2 m)? What does the literature say about it?
We have added the following section highlighting the key differences between DAP and ALS and potential challenges with characterizing structure of small plants:
"The advantages of DAP reconstructions are their high level-of-detail and the ability to incorporate spectral information [25–27]. Unlike ALS, drone based DAP primarily characterizes the outer canopy envelope [28] limiting the ability to accurately model terrain which is a prerequisite for measuring vegetation height. In spite of that, others have shown that DAP is capable of characterizing structure of small plants depending on vegetation density, slope of terrain, careful ground filtering and implemented geometric control [25,26,29–31]."
The last sentence provides new references, although most are used in the discussion as well:
25. Cunliffe, A.M.; Brazier, R.E.; Anderson, K. Ultra-fine grain landscape-scale quantification of dryland vegetation structure with drone-acquired structure-from-motion photogrammetry. Remote Sens. Environ. 2016, 183, 129–143.
26. van Iersel, W.; Straatsma, M.; Addink, E.; Middelkoop, H. Monitoring height and greenness of non-woody floodplain vegetation with UAV time series. ISPRS J. Photogramm. Remote Sens. 2018, 141, 112–123.29. Chen, S.;
29. McDermid, G.J.; Castilla, G.; Linke, J. Measuring vegetation height in linear disturbances in the boreal forest with UAV photogrammetry. Remote Sens. 2017,
30. Zahawi, R.A.; Dandois, J.P.; Holl, K.D.; Nadwodny, D.; Reid, J.L.; Ellis, E.C. Using lightweight unmanned aerial vehicles to monitor tropical forest recovery. Biol. Conserv. 2015, 186, 287–295. à NEW
31. James, M.R.; Robson, S. Mitigating systematic error in topographic models derived from UAV and ground-based image networks. Earth Surf. Process. Landforms 2014, 39, 1413–1420.
Methodology
Table 1 - Elevation and aspect? Slope and aspect?
The column name is changed to “Elevation (m)” reflecting the data.
Line 176 - What is the average RMSE of the location of the points?
We do not have an estimate of the Trimble base station's RMSE however we know the errors, were in a small number of cases up to 1m AS a result we used the measured location of neighboring trees and shrubs for further transect alignment:
We now state “To verify transect alignment with the acquired imagery location of known neighboring trees or large shrubs were used.”
Line 179 - Why this amount? Any reference?
This was not based on a reference but instead on the following goal and assumption. Our goal was to characterize dominant vegetation types and species by grid cell, thus any sprouting plants or minor edge effect would be irrelevant. On the other hand, we have sampled species and not only general vegetation types to be able to investigate the possibility of estimating species richness in future research incorporating multispectral data. Lastly, we assumed vegetation covering less than 10 x 10 sq cm would not be detectable on drone imagery.
Line 208 = 196 - What is the average RMSE of the processing of the 20 GCPs?
Unfortunately this was unavailable.
Line 214 - What is the RMSE in X, Y, and Z at the end of processing?
I derived the RMSE from Agisoft by site and provided them at the end of section 2.3 (Line 213), after we describe the processing workflow and report GSD and point densities
“Root mean square errors (RMSE) of the models were 0.344, 0.43 and 0.521 pixel for Site A, B and C, respectively, using the GCPs as check points.”
Table 3 - Why such a high HAGL? Can this not hinder the reconstruction of the height of vegetation? Some studies recommend flying lower in these conditions.
In the methods section we have added the following line to explain why HAGL was high:
Line 199 “High height above ground level (HAGL) was necessary to comply with visual line-of-sight rules and allow rapid data collection.”
In the discussion we acknowledge the influence high HAGLs likely had on the accuracy of 3D reconstructions. We referred to others who have flown at lower altitudes. We recognize further studies should find the optimal flight parameters for measuring short-stature vegetation structure, incorporating both technical and practical factors. We need to know the consistency of such measures and provide recommendations on fight parameters. We are starting a related project this summer.
Relevant sections in the discussion are highlighted below:
"In our study, DAP heights were underestimated and show reduced height variability, which is in line with findings of [30]. This can be explained by height error introduced during data collection, DAP processing or ground filtering, being close to short-stature vegetation height. Observed RMSE values ranged from 38—58 cm for various life forms, while [29,30] found values in the range of 20—46 cm for two height strata (0—0.5 m and 0.51—2.0 m) at comparable sites with respect to forest type, disturbance type and successional stage. Different flight parameters could have caused such a discrepency. Investigation by [73] found that small image pitch angle, high solar elevation and low flight altitudes ideally between 25—50 m (while retaining high image overlap) improved structure measurements of tree crops. We acknowledge that high flight altitudes (90—180m), chosen to comply with visual line-of-sight requirements and limit data acquisition time, may have reduced data quality in our study. In addition, image pitch angles from the DJI Phantom 4 acquired imagery were highly variable. High bidirectional reflectance was observed above exposed soil at Site A and B and heavy shadowing was found at site C. Our study suggests that errors can partially be explained by vegetation characteristics. The large height offsets found for conifers suggest that DAP has difficulties resolving seedlings that ideally stand out more clearly and form an unambiguous structure class. This could be caused by a combination of flight altitude, illumination conditions and, as a result, issues with identification of treetops during image matching. Automated ground filtering can become problematic in densely vegetated areas [30,74] which could explain the large normalized RMSE found for graminoids. Novel flight protocols including oblique imagery such as developed by [25,31] can produce very detailed reconstructions of graminoid and shrub canopy as well as terrain, however this will require multiple consecutive flights. We recommend further investigation of optimal flight parameters for short-stature vegetation, which should also consider battery performance and flight time."
Figure 2 - Shouldn't the point cloud come before height normalization?
That would indeed make more sense, thank you.
We have modified the figure.
Line 240 - What are the distance and angle values used? Inform.
We have reviewed the ground classification process and parameters in LASTOOLS again and made changes in the manuscript with the aim to be more transparent.
“In short, the TIN densification method creates a sparse triangular surface model based on the lowest points within a search window, after which spikes are removed and points added iteratively based on certain criteria such as maximum distance away from the surface and angle [57]. Set parameters include step (5 m), spike (50 cm), offset (10 cm) and bulge (0°).”
Line 256 - Why the maxim? Why not P99% or P95%? Explain.
We have added the following explanation after Line 2718 The main reason is to incorporate treetops in the metrics. Treetops, which are basically a few outliers, would be ignored if P99% or P95% are used.
“All metrics were tested for multicollinearity and removed from analysis if inter correlation (r) > 0.8. Selected metrics include maximum height (Max), coefficient of variation (COV), Rumple index, percentage points between 2.5 and 25 cm (%P2.5_25) and total vegetation cover (%P25_200). Maximum height was preferred over percentile heights as it retains treetops.”
Results of
Figure 3 - It is essential to inform the RMSE%.
Thank you, we agree a clear explanation is warranted. The way we calculate mean height offset (bias), RMSE and normalized RMSE are now consistent with Chen S et al. (2017), thus allowing fair comparison. In the methods section we have added the following (Line 256):
“Second, for each 2-m cell the maximum DAP and in situ height were derived stratified by life form and compared using regression and Spearman correlation coefficients. In addition, mean height off-sets (i.e. bias), root-mean-square errors (RMSEs) and normalized RMSEs were calculated as described by [29].”
Discussion
Line 551 and 554 - It is vital to inform the type of vegetation analyzed in the mentioned study [51].
We have revised the comparison of RMSEs and provided a more detailed description of the similarities between the studies. The sentence was changed to:
“Observed RMSE values ranged from 38—58 cm for various life forms, while [29,30] found values in the range of 20—46 cm for two height strata (0—0.5 m and 0.51—2.0 m) at comparable sites with respect to forest type, disturbance type and successional stage. Different flight parameters could have caused such a discrepency. ”
Line 555 - What is the reason? Does this usually happen? Do the flight characteristics interfere?
Relief? Type of vegetation? Could the automatic DTM construction/normalization process have interfered with this?
Line 556 - Explain the reason for the difficulty encountered by the DAP.
We have rewritten the paragraph that discussed height inaccuracies. We discuss our flight parameters in relation to other studies, potential issues caused by flight parameters and conditions, resulting issues during DAP processing (treetops / leaders) and limitations due to ground filtering. Hopefully we addressed your suggestions sufficiently.
“In our study, DAP heights were underestimated and show reduced height variability, which is in line with findings of [30]. This can be explained by height error introduced during data collection, DAP processing or ground filtering, being close to short-stature vegetation height. Observed RMSE values ranged from 38—58 cm for various life forms, while [29,30] found values in the range of 20—46 cm for two height strata (0—0.5 m and 0.51—2.0 m) at comparable sites with respect to forest type, disturbance type and successional stage. Different flight parameters could have caused such a discrepency. Investigation by [73] found that small image pitch angle, high solar elevation and low flight altitudes ideally between 25—50 m (while retaining high image overlap) improved structure measurements of tree crops. We acknowledge that high flight altitudes (90—180m), chosen to comply with visual line-of-sight requirements and limit data acquisition time, may have reduced data quality in our study. In addition, image pitch angles from the DJI Phantom 4 acquired imagery were highly variable. High bidirectional reflectance was observed above exposed soil at Site A and B and heavy shadowing was found at site C. Our study suggests that errors can partially be explained by vegetation characteristics. The large height offsets found for conifers suggest that DAP has difficulties resolving seedlings that ideally stand out more clearly and form an unambiguous structure class. This could be caused by a combination of flight altitude, illumination conditions and, as a result, issues with identification of treetops during image matching. Automated ground filtering can become problematic in densely vegetated areas [30,74] which could explain the large normalized RMSE found for graminoids. Novel flight protocols including oblique imagery such as developed by [25,31] can produce very detailed reconstructions of graminoid and shrub canopy as well as terrain, however this will require multiple consecutive flights. We recommend further investigation of optimal flight parameters for short-stature vegetation, which should also consider battery performance and flight time.“
Author Response File: Author Response.docx
