Modeling the Missing DBHs: Influence of Model Form on UAV DBH Characterization
, Neal C. Swayze 2, Chad M. Hoffman 3, Lauren E. Lad 3 and Mike A. Battaglia 1Round 1
Reviewer 1 Report
The authors have performed an interesting methodology for taking remotely sensed samples of DBHs and predicting missing DBHs after testing different models. This study analyzes the UAV data collection steps for a specific forest area in Colorado, presents the processes to extract tree locations from the SfM dataset and then apply different models to extract DBH information based on SfM-derived parameters. However, some minor issues need to be reviewed and better explained.
Section 2.4. à Explain further how the Equation 3 (a model to represent the regional height to DBH relationship) was developed
Section 2.5. à Give a caption for the variables of equations 4 to 8
line 205: spelling issue, change “than” to “then”
Figures 2 and 3 are nice and clearly show the height size classes and errors
Figure 4. light yellow color for UAV Filtered does not appear in the plot. There is a general mismatch between the colors in the plot and the legend.
The discussion and conclusion parts are well-written and easy to follow. The outputs of this study seem really promising to provide near census level inventories of tree locations, heights, DBH and can be also tested in other case studies.
Considering the arguments presented, in reviewer’s opinion, the article can be accepted after the minor revisions regarding graphs, photos and some further explanations. Also, minor spell check is also suggested.
Author Response
The authors have performed an interesting methodology for taking remotely sensed samples of DBHs and predicting missing DBHs after testing different models. This study analyzes the UAV data collection steps for a specific forest area in Colorado, presents the processes to extract tree locations from the SfM dataset and then apply different models to extract DBH information based on SfM-derived parameters. However, some minor issues need to be reviewed and better explained.
Section 2.4. à Explain further how the Equation 3 (a model to represent the regional height to DBH relationship) was developed
We appreciated the reviewer’s concern about how this model was developed and agree that it is important to the filtered dataset’s performance. To explain how this was down we have expanded this paragraph to now state “We filtered the SfM dataset of tree crown locations with heights and one or more potential DBH values following methods adapted from Swayze et al. [14]. This filtering used local Forest Inventory and Analysis (FIA) data [27] from the study region to develop Equation 3 as a model representing the regional height to DBH relationship. This model used FIA data from Colorado with greater than 70% ponderosa pine by basal area, densities exceeding 10 m2 ha-1 of basal area to represented untreated conditions, and site indices within 3 m of study site’s estimated site index of 22 m at a base age of 100 years to maintain a similar height to DBH relationship. Based on results from Swayze et al. [14], the model was fit as second order polynomial using the stats R package [23]. Subsequently, Equation 3 was applied to predict a regionally expected DBH value for each tree height.” at lines 179--189. The revised passage explains how the model data was selected from the FIA database and how the subsequent model was fit.
Section 2.5. à Give a caption for the variables of equations 4 to 8
We thank the reviewer for catching this oversight and have added the following passage to clarify the variables “In Equation 4 through Equation 8, Height is the UAV extracted tree height in meters, Crown Area is the UAV extracted crown area in square meters, TPH is the tree ha-1 of UAV identified trees within a 5 meter radius of the target tree, Relative Height is the height of the target UAV tree divided by the tallest neighboring tree within a 5 meter radius times 100, and Min Neighbor Distance is the horizontal distance in meters between the target UAV tree and the closest other UAV tree.” at lines 223-228.
line 205: spelling issue, change “than” to “then”
We thank the reviewer for catching this and have made the necessary revision.
Figures 2 and 3 are nice and clearly show the height size classes and errors
The authors thank the reviewer for the comment.
Figure 4. light yellow color for UAV Filtered does not appear in the plot. There is a general mismatch between the colors in the plot and the legend.
We understand the reviewer’s concern about the color alignment between the legend and plot and would like to explain why this is occurring. Because we chose to use an overlapping histogram to allow for easier comparison of the three distributions, the plotting package (ggplot2 in the R statistical program) shifts the plotting color in areas of overlapping bars. This can best be seen with for the UAV Unfiltered class where ethe light blue appears by itself at the upper end of the distribution and a darker blue color appears where it overlaps with the purple Stem Map distribution. To clarify this in the manuscript we have modified the Figure 4 legend to represent both colors a distribution could use.
The discussion and conclusion parts are well-written and easy to follow. The outputs of this study seem really promising to provide near census level inventories of tree locations, heights, DBH and can be also tested in other case studies.
Considering the arguments presented, in reviewer’s opinion, the article can be accepted after the minor revisions regarding graphs, photos and some further explanations.
We are please the reviewer agrees with our conclusions.
Reviewer 2 Report
The paper entitled “Modeling the Missing DBHs: Influence of Model Form on UAV DBH Characterization” reflects the development of applied research, the topic is interesting and the manuscript has an approach innovative. However, some issues should be improved (see comments). Thus, minor changes are recommended.
Comments
1) Along the text (for example lines 22, 35, 49) – use number of trees instead trees ha-1 when referring
2) Line 22 – 22.6 trees ha-1?
3) Keywords – meaning of UAS
4) Line 49 – basal area (in m2)?
5) Line 102 – selective logging or selective system?
6) Lines 115, 117, 125 – reference of the software’s is missing
7) Line 120 – 4 m sec-1 or 4 m s-1?
8) Line 152 – why did the authors used 5 m radius? Please provide justification
9) Lines 160-161, 163 – reference is missing
10) Lines 189-191 – provide e more detail explanation of the method used
11) Line 201 – to fit multiple linear regression model R program was used?
12) 2.5 section – why did the authors not analysed the multicollinearity of the models with more than one explanatory variables?
13) Lines 217-224 – what about the 1.37-5 m and 5-10 m classes?
14) Figure 4 – y axis in the bottom left figure needs to be wider. Also, the colour of UAB filtered is different in the legend and in the figure
Author Response
The paper entitled “Modeling the Missing DBHs: Influence of Model Form on UAV DBH Characterization” reflects the development of applied research, the topic is interesting and the manuscript has an approach innovative. However, some issues should be improved (see comments). Thus, minor changes are recommended.
Comments
1) Along the text (for example lines 22, 35, 49) – use number of trees instead trees ha-1 when referring
We thank the reviewer for this suggestion, and we agree that there are instances where stating number of trees is more appropriate and have made this change as suggested at line 22. However, the authors believe that the use of trees ha-1 is a common standard in silviculture literature and treatment prescriptions, and that it is a common variable in many tree allometries. As such, we have retained trees ha-1 at many other instances.
2) Line 22 – 22.6 trees ha-1?
The authors are unsure of the reviewers meaning with this comment, if the reviewer would clarify their comment, we would be happy to address their concern.
3) Keywords – meaning of UAS
We thank the reviewer for pointing this out and have added “unmanned aerial system to the keyword list.
4) Line 49 – basal area (in m2)?
The reviewer is correct that this was an unstandardized way of stating these units and have chosen to simplify the passage to now read as simple “stand basal area”, as the units are not important in the context of the topic that is describing different metrics commonly found in allometries.
5) Line 102 – selective logging or selective system?
We agree with the reviewer that multiple terms can be found in the literature, however, we chose to retain the term “selective logging” as it better describes the combination of patch and single tree extraction that is believed to have happened.
6) Lines 115, 117, 125 – reference of the software’s is missing
The reviewer is correct that citations have not been given for these software packages. However, as the journal does not have a requirement for giving full citations for software, we have chosen to only provide the publishing company’s name and location. If the editor and journal have a desired citation format for this, we would happy revise accordingly.
7) Line 120 – 4 m sec-1 or 4 m s-1?
We thank the reviewer for catching this and the authors have made the change to “4 m s-1” as this is the proper SI unit abbreviation.
8) Line 152 – why did the authors used 5 m radius? Please provide justification
The reviewer is correct that this decision should be supported. The authors chose this radius as it can be a simple proxy for the scale of direct competition between two mature ponderosa pine as it represents slightly less than two times a mature tree crown radius. We describe this in the manuscript text as “The 5 m radius was selected as it represents a distance slightly less than the point at which two mature ponderosa pine trees in the region would begin to have interlocking crowns, representing a proxy for direct competition.” at lines 160-163.
9) Lines 160-161, 163 – reference is missing
We thank the reviewer for point this out and have added the following passage and the below citation to document the R package used to achieve the described steps. The passage reads as “The rest of the DBH detection process was completed using the R TreeLS package [26].” at lines 169-170.
-
- Conto, T. TreeLS: Terrestrial Point Cloud Processing of Forest Data. R package version 1.0. 2019. https://CRAN.R-project.org/package=TreeLS
10) Lines 189-191 – provide e more detail explanation of the method used
We agree with the reviewer that more detail is needed about how F-score was calculated and so we have revised the passage to now include an equation for its calculation.
11) Line 201 – to fit multiple linear regression model R program was used?
We thank the reviewer for pointing this out and while making our revision realized that all models in the study were fit using the R stats package and so we remove the sentence we had about fitting the simple linear and non-linear models and added the following “All models were fit using the R stats package [23].” at lines 221-222 at the end of the paragraph.
12) 2.5 section – why did the authors not analysed the multicollinearity of the models with more than one explanatory variables?
The reviewer is correct that this is an important step in model development, and we thank them for catching this oversight. During our analysis we did test for collinearity between the variables and discovered only weak correlations. We have added the following passage “Prior to model development, all variables were tested for collinearity with no variables exceeding r = 0.34.” at lines 211-213 to explain this.
13) Lines 217-224 – what about the 1.37-5 m and 5-10 m classes?
The reviewer is correct that additional description of the tree extraction performance could be provided. We have revised the paragraph to now describe the extraction rates across all height size classes. The paragraph now reads as “Extraction of tree locations and heights from the SfM data resulted in an overall F-Score of 0.74 but varied across different tree height classes from 0.51 to 0.93 (Figure 2). Tree extraction also resulted in a mixture of False Positive and False Negative rates that varied across height classes. The SfM data extracted 1,755 trees across all size classes, representing an underestimation of 3.9% (or 26.7 trees ha-1) compared to the stem map observations (1,827 trees). The 1.37 – 5 m, 5 – 10 m, and 20 – 25 m height size classes had the best tree extraction success, missing stem density by 3.3% (8.3 trees ha-1), -4.7% (3 trees ha-1), and -4.1% (2.5 trees ha-1), respectively. Conversely, the worst extraction was in the 10 – 15 m and 15 – 20 m height size classes, where the False Negative and False Positive rates partially offset resulting in underestimations of 21.2% (5.5 trees ha-1) and 29.5% (15.3 trees ha-1), respectively. Additionally, the overall mean height error was 0.04 m and ranged from -0.63 m in the 5 – 10 m height class to 0.76 m in the 15 – 20 m height class (Figure 3).” at lines 238-250.
14) Figure 4 – y axis in the bottom left figure needs to be wider. Also, the colour of UAB filtered is different in the legend and in the figure
We thank the reviewer for their suggestion and have recreated the Figure 4 with the axis extended to capture the broken bar. We also understand the reviewer’s concern about the color alignment between the legend and plot and would like to explain why this is occurring. Because we chose to use an overlapping histogram to allow for easier comparison of the three distributions, the plotting package (ggplot2 in the R statistical program) shifts the plotting color in areas of overlapping bars. This can best be seen with for the UAV Unfiltered class where ethe light blue appears by itself at the upper end of the distribution and a darker blue color appears where it overlaps with the purple Stem Map distribution. To clarify this in the manuscript we have modified the Figure 4 legend to represent both colors a distribution could use.
Reviewer 3 Report
Journal: Forests (ISSN 1999-4907)
Manuscript ID: forests-2055793
Title: Modeling the Missing DBHs: Influence of Model Form on UAV DBH Characterization
Overall Comments and Suggestions for Authors
Dear authors,
Regarding the model development of missing DBHs using UAV techniques especially in pine dominated stand, this manuscript might be interesting to the relevant researchers who deals with similar issues such as forest biometrician and investigator using remote sensing techniques. Still, some parts can be improved. I suggested a few comments to update the manuscript. I wish the results in this study will serve for tree allometric predictions in the region.
I hope that this manuscript can be improved based on peer-review’s comments. My specific comments were provided in detail as follows.
Kind regards,
Reviewer
Point 1.
Keywords: I am not so sure if “UAS” abbreviation can be directly mentioned in keywords. It could be written with full text. Also, “allometry” can be specified or changed, for example, tree allometry.
Point 2.
Line 32: the height for DBH measurement is usually specified to the countries. It differs by country. I guess 1.37 m is for US. Specify where it is. It would be better to the international readers.
Point 3.
It would be proper to write what you did in past tense. For example, in Line 90 evaluated, Line 92 we used, and Line 92 evaluated. Find more and fix it correctly.
Point 4.
Is the description about Materials and study area enough? I consider that adding stand characteristics, such as mean DBH, height, species composition, basal area, number of trees, and etc., will be informative to the other researchers, especially if any stand condition affects the output of this study. It will be helpful to future studies when one tries to compare the results or cites the outputs as a reference.
Point 5.
Line 151. Trees ha-1 or trees per ha. Not trees per ha-1
Line 198-200: Are those multivariate models? Or multiple nonlinear regression? It seems multiple, not multivariate. Correct the words appropriately.
Line 205: than -> then?
Point 6.
Why don’t you provide the parameter b1, b2 at least for the best fit model? If possible, authors can provide all. I believe the model output with parameters will be useful to the relevant researchers.
Point 7.
In Table 2, what are the values inside parenthesis in adj. R2? Are those p-values? I consider authors should not write it there. It may be related to F-statistic of regression model. Check the Table carefully and correct it as required.
Point 8.
In Discussion, the last sub-section comprised mainly of model limitation. Along with the point, why don’t author provide some information about application from a practical point of view? Authors can also mention some kinds of recommendation when one wants to apply this type of model. For example, those could be forest condition that can be applied, applicable measurement tools, recommended data processing, or authors’ other thought regarding the application in practice.
Author Response
Dear authors,
Regarding the model development of missing DBHs using UAV techniques especially in pine dominated stand, this manuscript might be interesting to the relevant researchers who deals with similar issues such as forest biometrician and investigator using remote sensing techniques. Still, some parts can be improved. I suggested a few comments to update the manuscript. I wish the results in this study will serve for tree allometric predictions in the region.
I hope that this manuscript can be improved based on peer-review’s comments. My specific comments were provided in detail as follows.
Kind regards,
Reviewer
Point 1.
Keywords: I am not so sure if “UAS” abbreviation can be directly mentioned in keywords. It could be written with full text. Also, “allometry” can be specified or changed, for example, tree allometry.
We thank the reviewer for their suggestions and have chosen to add the full text version of UAS to the keywords.
Point 2.
Line 32: the height for DBH measurement is usually specified to the countries. It differs by country. I guess 1.37 m is for US. Specify where it is. It would be better to the international readers.
The reviewer is correct that this could be further clarified for an international audience. We have expanded the statement to now read as “a tree’s diameter at breast height (DBH; 1.37 m above ground in United States)”.
Point 3.
It would be proper to write what you did in past tense. For example, in Line 90 evaluated, Line 92 we used, and Line 92 evaluated. Find more and fix it correctly.
We thank the reviewer for catching these tense issues and agree that they should be in the past tense. We have made the necessary revisions that the reviewer highlighted and have checked the rest of the manuscript for additional tense issues.
Point 4.
Is the description about Materials and study area enough? I consider that adding stand characteristics, such as mean DBH, height, species composition, basal area, number of trees, and etc., will be informative to the other researchers, especially if any stand condition affects the output of this study. It will be helpful to future studies when one tries to compare the results or cites the outputs as a reference.
The authors agree with the reviewer that adding additional description of the stand structure will provide more context for readers to better interpret the study. To accomplish this, we added the following passage at the end of Section 2.1 that reads as “The stand is 99.9% ponderosa pine by basal area with a Dq of 21.8 cm and density of 457 trees ha-1 and 17 m2 ha-1. Tree sizes tended to center around an understory canopy stratum of 340 trees ha-1 with a Dq of 8.2 cm and an average height of 4.1 m, and an overstory canopy stratum of 117 trees ha-1 with a Dq of 40.6 cm and an average height of 19.6 m.” at lines 113-117.
Point 5.
Line 151. Trees ha-1 or trees per ha. Not trees per ha-1
We thank the reviewer for catching this and have made the change to now read “trees ha-1”.
Line 198-200: Are those multivariate models? Or multiple nonlinear regression? It seems multiple, not multivariate. Correct the words appropriately.
The reviewer is correct, the model forms all only had a single dependent variable making them multiple regressions, we have made the necessary corrections in Section 2.5 and confirmed that it is correct throughout the manuscript.
Line 205: than -> then?
We thank the reviewer for catching this and have made the necessary revision.
Point 6.
Why don’t you provide the parameter b1, b2 at least for the best fit model? If possible, authors can provide all. I believe the model output with parameters will be useful to the relevant researchers.
We thank the reviewer for this suggestion and agree that providing the model coefficients will be valuable for the readers. Currently we present these values in Table 2 for all of the tested models and are sorry for any confusion. To make this clearer, we have modified the table caption to now read as “Model coefficients (b1, b2, etc.) for the five parametric model forms fit to the SfM DBHs before and after filtering as the dependent variable. Values in parentheses are P values.”
Point 7.
In Table 2, what are the values inside parenthesis in adj. R2? Are those p-values? I consider authors should not write it there. It may be related to F-statistic of regression model. Check the Table carefully and correct it as required.
We appreciate the reviewer’s comment and agree that additional clarity is needed about the table’s presentation. The values in parenthesis are P values which the authors believe provide additional nuance to help readers interpret the model performance and the true influence of each independent variable. We have chosen to retain the P values in the table but have clarified the table caption to now read as “Model coefficients (b1, b2, etc.) for the five parametric model forms fit to the SfM DBHs before and after filtering as the dependent variable. Values in parentheses are P values.”
Point 8.
In Discussion, the last sub-section comprised mainly of model limitation. Along with the point, why don’t author provide some information about application from a practical point of view? Authors can also mention some kinds of recommendation when one wants to apply this type of model. For example, those could be forest condition that can be applied, applicable measurement tools, recommended data processing, or authors’ other thought regarding the application in practice.
The reviewer is correct that this would be important information for managers to implement the evaluated techniques in practice. However, the current manuscript is focused on one step in a larger system of UAV-based forest monitoring and does not discuss or evaluate how other influences might impact this technique’s reliability, such as multiple species that often have differing height to DBH relationships or how species with greater crown ratios might impact the ability to detect a sufficient DBH sample. Because there are still large gaps in the application of UAVs for forest monitoring the authors do not think this manuscript is the appropriate place to outline how managers could implement these techniques.
