Next Article in Journal
The Joint Effects of Precipitation Gradient and Afforestation on Soil Moisture across the Loess Plateau of China
Next Article in Special Issue
Large Area Forest Yield Estimation with Pushbroom Digital Aerial Photogrammetry
Previous Article in Journal
Effects of Spatial Pattern of Forest Vegetation on Urban Cooling in a Compact Megacity
Previous Article in Special Issue
Modelling and Predicting the Growing Stock Volume in Small-Scale Plantation Forests of Tanzania Using Multi-Sensor Image Synergy
Article
Peer-Review Record

Assessing the Ability of Image Based Point Clouds Captured from a UAV to Measure the Terrain in the Presence of Canopy Cover

Forests 2019, 10(3), 284; https://doi.org/10.3390/f10030284
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Forests 2019, 10(3), 284; https://doi.org/10.3390/f10030284
Received: 21 February 2019 / Revised: 15 March 2019 / Accepted: 20 March 2019 / Published: 22 March 2019

Round  1

Reviewer 1 Report

Minor comments are reported on the pdf file

Comments for author File: Comments.pdf

Author Response

The authors would like to thank the reviewer for taking the time to review our manuscript. Our responses to to the reviewers comments (bold) follow.

Page 1 wrong affiliation  

We have updated this affiliation  

Line 158 – Please check this notation 

We have updated the text to explicitly state that sigma is standard deviation 

Line  173 – I noted that sometimes the reference authors are implicit and  sometime explicit (see e.g. at lines 60, 69 and 76). Please harmonise it all over the manuscript 

We have checked all references to ensure a consistent approach is used  

Table 2 – Please, add the unit of each parameter 

We have updated table 2 to include the units 

Line 190 - I  think that this operation can solve a practical problem but will  introduce a negative bias. In fact, if you assume that the height error  is both positive and negative you should better consider retaining the  nearest point to the average or the median and not the minimum in the  considered window of 0.01 m. I understand that this alternative filter  maybe won't affect the overall results and discussions, but I consider  it sounder from a statistical point of view 

While we agree that this approach may introduce some slight negative bias, the  size of the area under consideration (a 25px radius circle) and the  properties of the area in some cases (significant low vegetation) meant  that the use of a measure of central tendency is likely to introduce more positive bias than using the minimum filter approach. Please note  that the approach to remove noise has already filtered out a  significant portion of the points likely to introduce negative bias. 

L197 – delete 

Deleted 

L198 I cannot understand this part very well. Particularly I cannot identify what a quadrant can be precisely. 

We have added the following text to clarify this section  

To achieve this,  the angle and distance between the point being assessed and every other  point was determined. Each point was then assigned to one of four 90° quadrant (with quadrants defined between 1-90, 91-180, 181-270 and 271-360°) based on the calculated angle. The minimum distance in each quadrant was then determined, with dquad being determined as the mean of these four distances 

L229 It is not clear how did you estimate the % of covered area. From the Tab. 4 caption and since the cloud is voxeled,  I can suppose that you assumed a sampling representativeness of each  ground/vegetation point on the whole plot area (in x e y dimensions  only). In any case, please, explicit in the Methods (section 2.5) the  used approach. This is important to allow the experiment replication. 

The following sentence has been added to the method to clarify this issue:  

The area observed was calculated as the percentage of DTM grid cells containing at least one point identified as ground. 

L251 – Redundant 

The following sentence has been modified: 

[from]  At the low canopy cover sites (ENS01 and PNT01) the RMSE calculated  using the DTMs produced using the filtered ground points did not  increase markedly above the RMSE found using the manually identified  ground points (see Table 4 and Table 5). 

To: 

At  the low canopy cover sites (ENS01 and PNT01) the RMSE calculated using  the DTMs produced using the filtered ground points did not increase  markedly above the RMSE found using the manually identified ground  points (see Table 4 and Table 5). 

L258 – Add the unit 

We have added the necessary units and checked the manuscript for other omissions 

Reviewer 2 Report

The study was well designed and interesting. The manuscript was a pleasure to read. Some editing to remove typos and grammatical errors is needed. Well done!

Author Response

We thank the reviewer for their kind comments. We have carefully checked the manuscript for typos and grammatical errors.

Reviewer 3 Report

This manuscript represents an attempts to assess the height of the canopy from photogrammetry-based point clouds derived from UAV imagery. The work has a very simple design and the idea is not really novel. First of all, there were plenty of similar papers using LIDAR point clouds and quite a few relevant studies using photogrammetry and comparison LIDAr and photogrammetry (see for example https://www.mdpi.com/2072-4292/10/10/1562; https://www.mdpi.com/2072-4292/11/3/233/htm; https://www.mdpi.com/2072-4292/10/8/1266/htm among many other similar publications and cross-references). I guess this paper would have sufficient novelty for publication a few years ago in 2010-2012, but it does not look interesting in 2019. The paper is entirely empirical approach and closely linked to a particular forest in Chile. I do not see if these results can be applicable to other forests, I would say that the focal problem can probably be treated more theoretically. In addition, I concern that Forests is not a right destination for this manuscript, my impression is that another MDPU journal, Remote Sensing, would be a better fit for this article. I also concern that the experimental design does not employ another "true" terrain map, I would expect that the methodological study like this should be used another data, and the results obtained by authors, in particular, ground estimates must be compared with true ground measurements the "composite" image sets does not resolve a problem of having understory objects, wood debries,vegetation etc. Overall, the authors collected quite a lot of data in the field, reconstructed it photogrammetrically and analyzed it in a standard way, and I guess it is possible to publish this article, as the amount of work is substantial, but it does not look like an impressive work. I think the major revision is needed to clarify the novelty and major findings. 

Author Response

We thank the reviewer for there time in reviewing our manuscript. We have responded below with the reviewer comments (in bold). 

This manuscript represents an attempts  to assess the height of the canopy from photogrammetry-based point  clouds derived from UAV imagery. The work has a very simple design and  the idea is not really novel. First of all,  there were plenty of similar papers using LIDAR point clouds and quite a  few relevant studies using photogrammetry and comparison LIDAr  and photogrammetry (see for example  https://www.mdpi.com/2072-4292/10/10/1562;  https://www.mdpi.com/2072-4292/11/3/233/htm;  https://www.mdpi.com/2072-4292/10/8/1266/htm among many other similar  publications and cross-references). I guess this paper would have sufficient novelty for publication a few years ago in 2010-2012, but it does not look interesting in 2019. 

We thank the reviewer for their comments. As stated in section 1.3,  the objective of this paper was an attempt to resolve the issue of  terrain representation and not canopy height. The two key novelties presented in this paper as highlighted in the abstract and discussion  include: 

The use of a very high accuracy reference terrain for assessing the accuracy and coverage of the ground provided by the UAV photogrammetry. 

The inclusion of off-nadir imagery in an attempt to achieve a greater terrain representation. 

We find it curious that the reviewer would highlight similar papers from 2018 and 19 and  then suggest that our paper is only novel in 2010-2012. Nevertheless,  these papers do highlight the accuracy of the terrain representation as a  key limitation in their studies. This limitation is addressed in this study. 

The paper is entirely empirical approach and closely linked to a particular forest  in Chile. I do not see if these results can be applicable to other  forests, I would say that the focal problem can probably be treated more  theoretically.  

The  reviewer is incorrect here. Section 2.1 outlines the six distinct  forest types for which results were completed. These native, plantation  forests as well as forests that have been recently disturbed by fire and  or landslide. In particular, similar pine plantations can be found in South Africa, New Zealand, Australia and Spain. While the temperate forest is similar to those found in USA, Canada and Northern Europe.   

In  addition, I concern that Forests is not a right destination for this  manuscript, my impression is that another MDPU journal, Remote Sensing,  would be a better fit for this article.  

This paper has been submitted to the special issue of forests Applications of LiDAR and Photogrammetry for Forest Inventory and Management. In the authors opinion this is the best place for the manuscript.  

I  also concern that the experimental design does not employ another  "true" terrain map, I would expect that the methodological study like  this should be used another data, and the results obtained by authors,  in particular, ground estimates must be compared with true ground  measurements the "composite" image sets does not resolve a problem of  having understory objects, wood debries,vegetation etc.  

The reviewer appears to have misunderstood the method used to establish a reference terrain map for this study. Section 2.1 outlines the true terrain map used in this study.  This map was collected using a total station as a grid of spot heights.  Total station measurements are more precise GPS as a source of reference which has been the source of reference in than previous studies. This point of distinction for this manuscript is outlined in the discussion. 

The composite image set is only intercompared to the other image sets and is at no point used as reference. 

Overall,  the authors collected quite a lot of data in the field, reconstructed  it photogrammetrically and analyzed it in a standard way, and  I guess it is possible to publish this article, as the amount of work  is substantial, but it does not look like an impressive work. I think  the major revision is needed to clarify the novelty and major findings. 

We thank the reviewer for recognizing the effort put into this paper.

Round  2

Reviewer 3 Report

The authors provided some explanations to my previous comments. I do not see any change sin the manuscript though. This is an empirical study that is not novel or really advanced. However, the authors collected data and carefully analysed them, I do not see any reasons do not publish this work to let scientific community evaluate this publication.

Back to TopTop