Adjudicating Perspectives on Forest Structure: How Do Airborne, Terrestrial, and Mobile Lidar-Derived Estimates Compare?
Round 1
Reviewer 1 Report
The topic of this paper is very timely, as a broad remote sensing-forestry community gets access to acquire lidar data from different platforms, questions regarding differences and similarities, as well as accuracy and easy to use/process data arise. The results from this work provide a good overview of advantages, limitations, and challenges of using ALS, TLS and MLS for forestry purposes.
The document has been well written, however there are a couple of things that need to be fixed: 1) line 123, change "an unique" to "a unique", 2) lines 361 going to 362, delete "other"; 3) needs consistency on how to present the units next to the numbers, most of the document shows "50m", but in some places shows up as "50 m" with a space in between the units and the number, and in line 218 shows “3-m”. As a reader I would prefer to have the units with a space in between the number and the units.
In Fig. 1, it would help to have a number of returns per unit area for each system. The image shows there is difference, so this information will provide more detail for the reader.
Section 2.3 It is not clear what year was the GeoSlam data collected. The section states the years of collection for the ALS and the TLS only.
In Fig. 6, the labels of both axis in the graphs are identical, the axis showing the predicted attribute should be identified as such. Although the 4 graphs are presented together, the reader starts with the (a) first, but the legend shows up later on in graph (d), so it would be better to insert a legend in each one of them.
In Fig. 7 the legend also shows up later in graph (b), the reader starts with (a) and has to look around for the legend to understand what is looking at. Again, it will be good to add legends to all graphs.
Author Response
Reviewer 1
Comments and Suggestions for Authors
The topic of this paper is very timely, as a broad remote sensing-forestry community gets access to acquire lidar data from different platforms, questions regarding differences and similarities, as well as accuracy and easy to use/process data arise. The results from this work provide a good overview of advantages, limitations, and challenges of using ALS, TLS and MLS for forestry purposes.
The document has been well written, however there are a couple of things that need to be fixed: 1) line 123, change "an unique" to "a unique", 2) lines 361 going to 362, delete "other"; 3) needs consistency on how to present the units next to the numbers, most of the document shows "50m", but in some places shows up as "50 m" with a space in between the units and the number, and in line 218 shows “3-m”. As a reader I would prefer to have the units with a space in between the number and the units.
Response: All of the suggested corrections were incorporated. Thank you for your review.
In Fig. 1, it would help to have a number of returns per unit area for each system. The image shows there is difference, so this information will provide more detail for the reader.
Response: Number of returns per meter for each platform were added to the figure caption.
Section 2.3 It is not clear what year was the GeoSlam data collected. The section states the years of collection for the ALS and the TLS only.
Response: Date added.
In Fig. 6, the labels of both axis in the graphs are identical, the axis showing the predicted attribute should be identified as such. Although the 4 graphs are presented together, the reader starts with the (a) first, but the legend shows up later on in graph (d), so it would be better to insert a legend in each one of them.
Response: The axis labels were corrected and the legend was moved to the first panel (from d to a). The authors have consistently used the exact same colors throughout the manuscript to assist reader interpretation and to repeat the exact same legend in every panel would not add anything to the figure. This also goes against current data visualization practices (i.e., repeating legends in panels).
In Fig. 7 the legend also shows up later in graph (b), the reader starts with (a) and has to look around for the legend to understand what is looking at. Again, it will be good to add legends to all graphs.
Response: Again, the authors have consistently used the exact same colors throughout the manuscript to assist reader interpretation. A legend was added to assist the reader in the first panel (a) and the box plot legend shows up in the first where box plots appear (b).
Reviewer 2 Report
L92-97: “… in this study, we compare methodologies for assessing a) individual-tree; b) stand-level; and c) canopy cover and landscape metrics attributes derived from rasterized canopy height models among dataset obtained from ALS, TLS and MLS platforms across a range of forest conditions in a ponderosa pine forest of northern Arizona, USA. Resulting tree and forest characteristics were compared to assess relative accuracies across scale.”
This study addresses the objective copied above. It is well-done, and the text well-written. I appreciate the discussion includes interpretation of the larger context (L472-483 for example), though it would be interesting if the authors would speculate directly if they think these results would generally hold true for very different forest types/structures or if they think more research is needed in other types.
A minor comment: there are a few typos in the document. For example, line 100, seems to be missing the word ‘on’ such as “For this study, we focused on a pure ponderosa pine”; Line 115 seems to be missing the word ‘tree’, as in “For all trees with…”; Line 114, the word “sued” is probably measured; there are a few other places also.
Author Response
Reviewer 2
Comments and Suggestions for Authors
L92-97: “… in this study, we compare methodologies for assessing a) individual-tree; b) stand-level; and c) canopy cover and landscape metrics attributes derived from rasterized canopy height models among dataset obtained from ALS, TLS and MLS platforms across a range of forest conditions in a ponderosa pine forest of northern Arizona, USA. Resulting tree and forest characteristics were compared to assess relative accuracies across scale.”
This study addresses the objective copied above. It is well-done, and the text well-written. I appreciate the discussion includes interpretation of the larger context (L472-483 for example), though it would be interesting if the authors would speculate directly if they think these results would generally hold true for very different forest types/structures or if they think more research is needed in other types.
Response: Thank you for the kind words. To address the reviewer’s suggestion, a sentence was added was added to the discussion (Lines486-492).
A minor comment: there are a few typos in the document. For example, line 100, seems to be missing the word ‘on’ such as “For this study, we focused on a pure ponderosa pine”; Line 115 seems to be missing the word ‘tree’, as in “For all trees with…”; Line 114, the word “sued” is probably measured; there are a few other places also.
Response: These corrections were made as well as other minor wording and grammatical mistakes corrected following edits from the authors.
Reviewer 3 Report
LiDAR has been widely adopted in various applications, forest is a vital part of the applications of LiDAR point cloud data, not only for inventory, but also for ecosystem studies and climate change issues. The dense point cloud is able to characterize the detailed structural parameters at high resolution, in terms of forest applications, different type LiDAR systems would make big differents in data sets, this manuscript touch the very topic with the comparision of point cloud acquired by ALS, TLS, MLS for forest tree characterization. The topic is of great interesting in various research and application community, and it also meet the scope of the journal well. However, the current version manuscript needs further revision for the final publication. Below please find the comments and suggestions that the authors should address the issues in the revision.
1. In my opinion, the density of forest plot would be a dominant factor for the single tree characterization, i.e. for sparse forest scene, the laser pulse of ALS could reach the ground surfaces under tree canopies, and TLS and MLS are able to detect the top of tree canopies. the authors illustrated the sparse scene in Figure 1, my concern is, for dense forest plot, it is difficult to get such full data sets of targets, it would be better to address the issues in the text.
2. For the DEM-normalization of point cloud, the TRUE ELEVATION values of points are difficult to identify, the normalization process could seriously affect the data quality for further application, my question is, did the authors evaluate the performance of the approaches?
3. For individual tree segmentation, the authors applied an open source approach, the problem is, is there any comparision between different approaches?
4. Generally speaking, it is difficult to detect tree top with TLS and MLS, but in Figure 5, most of sample trees are accurately measured with TLS and MLS,
I think it is not general situation, I am afraid that the authors selected "good" cases in the experiment, if so, the study would not have generative means.
5. Language issue: the manuscript could be improve with carefull language check, I find many ambiguous terms, for instance, "Lidar Acquisitions and Processing(LiDAR Data Acqusitions and Processing?)", " Field data (Collection of Field Data)", "As stated earlier", etc.
In summary, the topic of the manuscript would be interesting for audiences of RS,but the current version manuscript is not good enough for publication. I strong recommend the authors to make significant revision upon the current manuscript, and submit a revised version for further review.
Author Response
Reviewer 3
Comments and Suggestions for Authors
LiDAR has been widely adopted in various applications, forest is a vital part of the applications of LiDAR point cloud data, not only for inventory, but also for ecosystem studies and climate change issues. The dense point cloud is able to characterize the detailed structural parameters at high resolution, in terms of forest applications, different type LiDAR systems would make big differents in data sets, this manuscript touch the very topic with the comparision of point cloud acquired by ALS, TLS, MLS for forest tree characterization. The topic is of great interesting in various research and application community, and it also meet the scope of the journal well. However, the current version manuscript needs further revision for the final publication. Below please find the comments and suggestions that the authors should address the issues in the revision.
1. In my opinion, the density of forest plot would be a dominant factor for the single tree characterization, i.e. for sparse forest scene, the laser pulse of ALS could reach the ground surfaces under tree canopies, and TLS and MLS are able to detect the top of tree canopies. the authors illustrated the sparse scene in Figure 1, my concern is, for dense forest plot, it is difficult to get such full data sets of targets, it would be better to address the issues in the text.
Response: The caption for Figure 1 was edited to make sure the reader understands that this was just an example visualization, and not an illustration of a sparse scene (as the reviewer suggested). Furthermore, the authors clearly report the range of plot densities in Table 1 and go on to state “Data consisted of 209 ponderosa pine trees on 12 sample plots, with densities ranging from open conditions (25 trees ha-1) to dense, closed canopy, conditions (1361 trees ha-1). While denser conditions for contemporary forest have been reported for nearby sites e.g., [32], the conditions observed were a unique opportunity to explore patterns of tree detection and estimation accuracy for typical ponderosa pine forest conditions across western USA.” on lines 121-124 and 134-135. The authors also state that plots were placed to coincide with data collected from a previous study. The authors selected plots, which represent a gradient of forest density and tree size conditions, in locations that maximized the ability to compare amongst these three types of lidar platforms.
2. For the DEM-normalization of point cloud, the TRUE ELEVATION values of points are difficult to identify, the normalization process could seriously affect the data quality for further application, my question is, did the authors evaluate the performance of the approaches?
Response: On lines 108-109 we state that the site exhibited “slight variation in slope (<5%) and aspect”. The actual calculated slopes for our study site ranged from 2-3% in the area where field observations were made, and therefore we are confident that any error introduce from our DEM normalization approach would be inconsequential. This is evident in the assessed ALS-based estimate of tree heights as compared to our field-observed heights.
3. For individual tree segmentation, the authors applied an open source approach, the problem is, is there any comparision between different approaches?
Response: The study focuses on comparisons of lidar-derived estimates and field observations and to make the comparisons as valid as possible, the same methods were used (where appropriate). Furthermore, the authors provide a description as to how we arrived at the methodology utilized for the ALS platform as compared to the TLS and MLS platforms. Comparisons of various segmentation approaches (for which there are many) is beyond the scope of this study.
4. Generally speaking, it is difficult to detect tree top with TLS and MLS, but in Figure 5, most of sample trees are accurately measured with TLS and MLS,
I think it is not general situation, I am afraid that the authors selected "good" cases in the experiment, if so, the study would not have generative means.
Response: The authors clearly outline our approach to selecting sample locations in the methods (which were a function of the TLS sample sites), provide summary information on the resulting forest conditions, present plot-level error rates in table 2 and figure 4, and present all of the individual-tree data in figure 5 (d-f). While the reviewer is correct that “most” of the lidar-derived heights matched the field observed height with a good deal of accuracy, ALS still outperformed MLS and TLS and no effort to select “good” cases were made. Furthermore, we state that conditions are “typical ponderosa pine forest conditions” (see earlier response) and thus are confident that our results are generalizable for ponderosa pine and other open, conifer-dominated ecosystems. Anticipated results in other systems is discussed in the last paragraph of the Discussion.
5. Language issue: the manuscript could be improve with carefull language check, I find many ambiguous terms, for instance, "Lidar Acquisitions and Processing(LiDAR Data Acqusitions and Processing?)", " Field data (Collection of Field Data)", "As stated earlier", etc.
Response: Headings were expanded for clarity (as suggested) and the ambiguous phrase was clarified to state “As stated in the previous section…”
In summary, the topic of the manuscript would be interesting for audiences of RS,but the current version manuscript is not good enough for publication. I strong recommend the authors to make significant revision upon the current manuscript, and submit a revised version for further review.
Response: Thank you for your suggestions. The manuscript has been edited to incorporate reviewer comments and edits were made for clarity.
