Comparison between Artificial and Human Estimates in Urban Tree Canopy Assessments

Round 1

Reviewer 1 Report

I commend the authors for their work and recommend this article for publication. I have some recommended revisions, which are relatively minor.

Line 76 - I would suggest not using the term "human intelligence" as it refers to specific methods of intelligence gathering in the military. Other types are "imagery intelligence" and "signals intelligence."

Line 135 - Was an accuracy or agreement assessment done to gauge the quality of the human interpreters?

Line 150 - The use of a wetland class is confusing because, at the resolution you are mapping, wetlands do not exist. The individual component classes of wetlands (water and vegetation) are present. Some clarification on standards, such as how minimum mapping units vary by land cover class would be helpful.

Line 153 - The figure is nearly impossible to interpret. Please redo. Recommend decreasing or eliminating the line width of the outline.

Line 264 - The limits of AI here should be rethought. The advantage of AI is that it takes into account both spectral and spatial components. If the AI were to fail due to the reasons you stated it would most likely be due to training data that are not representative.

Line 288 - The increase makes sense to me. It is worth mentioning the influence of land use history. An area transitioning from agriculture to suburban can experience an increase in tree canopy with 1-2 decades as trees are planted in areas previously void of them.

 

 

 

Author Response

Thank you for your comments to address for improvement of the paper.

Line 76 - I would suggest not using the term "human intelligence" as it refers to specific methods of intelligence gathering in the military. Other types are "imagery intelligence" and "signals intelligence."  <<<< Left as is. While Human Intelligence is a method used by the military, human intelligence by definition is our mental quality to learn from experience, much like artificial intelligence but in this case a human who used their intelligence through experience to obtain a level of being able to identify landcovers. We do appreciate your suggestion but prefer to leave as is..

Line 135 - Was an accuracy or agreement assessment done to gauge the quality of the human interpreters? <<<< Yes, this is presented in 2.4 as our 10% accuracy assessment per community (100 of 1000 points were randomly assessed for agreement).

Line 150 - The use of a wetland class is confusing because, at the resolution you are mapping, wetlands do not exist. The individual component classes of wetlands (water and vegetation) are present. Some clarification on standards, such as how minimum mapping units vary by land cover class would be helpful. <<< We were able to see water and wetlands on the 1 m resolution imagery. Sample points over water were validated easily between two evaluators. Wetlands are another matter and a challenge. We included these two land classes as we were comparing to a system that included these.

Line 153 - The figure is nearly impossible to interpret. Please redo. Recommend decreasing or eliminating the line width of the outline. <<<< Left as is. The intent of the figure is to demonstrate sample communities occurred throughout the state of Wisconsin and this color rendition showed this best and the inset was used to provide some clarity. If the figure is a deal breaker, we will remove from the paper as not central to the study, just used to provide clarity to the study locations, which we agree is a challenge at the statewide scale. Will defer to the editor

Line 264 - The limits of AI here should be rethought. The advantage of AI is that it takes into account both spectral and spatial components. If the AI were to fail due to the reasons you stated it would most likely be due to training data that are not representative. <<<<We added … There are reasons explain AI challenges to identify UTC and no-represented training data may be a factor. Thanks for this point!

Line 288 - The increase makes sense to me. It is worth mentioning the influence of land use history. An area transitioning from agriculture to suburban can experience an increase in tree canopy with 1-2<<<< We added this to backup this important point …. Areas that transition from agricultural landcover to urban or suburban may experience an increase in UTC [64].

 

Author Response File: Author Response.docx

Reviewer 2 Report

Dear Authors,

Presented manuscript titled „Comparison Between Artificial and Human Estimates with Urban Tree Canopy Assessment”  is generally well-written. However, I have found some imperfections- which in my opinion- should be improved before an eventual publication.

1.       In my opinion in Abstract section the titles of sections (1) Bacgrounds, (2) Methods.. etc. might be omitted. Please use full names of investigated metrics such as UTC.

2.       In my opinion in chapter Introduction the brief review of current state of investigations on comparison between artificial and human estimates with urban Tree Canopy Assessment. Such section should contain information about similar studies perhaps in other areas (if they exist, if not they not occur-such information about lack of investigations should be highlighted as justification of study aims).

3.       Conclusions chapter looks like summary of results. In my opinion this section might be shorten and should contain the proposed future directions/areas of investigations.

4.       Please look into below listed publications. Perhaps they will be helpful in the manuscript corrections:

·         Oliver Baines, Phil Wilkes, Mathias Disney, 2020. Quantifying urban forest structure with open-access remote sensing data sets, Urban Forestry & Urban Greening, 50,126653.

·         Soojeong Myeong, David J. Nowak, Michael J. Duggin, 2006. A temporal analysis of urban forest carbon storage using remote sensing, Remote Sensing of Environment, 101(2), 277-282.

·         Blaz Klobucar, Neil Sang, Thomas B. Randrup, 2021. Comparing ground and remotely sensed measurements of urban tree canopy in private residential property, Trees, Forests and People, 5, 100114.

·         Nölke, N. Continuous Urban Tree Cover Mapping from Landsat Imagery in Bengaluru, India. Forests 2021, 12, 220. https://doi.org/10.3390/f12020220

 

Author Response

Reviewer 2

Thank you for your comments to address for improvement of the paper.

Comparison Between Artificial and Human Estimates with Urban Tree Canopy Assessment”  is generally well-written. However, I have found some imperfections- which in my opinion- should be improved before an eventual publication.  <<<< Thank you we made edits as listed below.

1. In my opinion in Abstract section, the titles of sections (1) Backgrounds, (2) Methods.. etc. might be omitted. Please use full names of investigated metrics such as UTC. <<<< Removed the section titles.  Spelled out all acronyms in the abstract, but kept acronyms in the paper for economy of space.
2. In my opinion in chapter Introduction the brief review of current state of investigations on comparison between artificial and human estimates with urban Tree Canopy Assessment. Such section should contain information about similar studies perhaps in other areas (if they exist, if not they not occur-such information about lack of investigations should be highlighted as justification of study aims). <<<< Left as is, in the discussion some limited examples are presented.
3. Conclusions chapter looks like summary of results. In my opinion this section might be shorten and should contain the proposed future directions/areas of investigations. <<<< Editing to add future research.
4. Please look into below listed publications. Perhaps they will be helpful in the manuscript corrections: Oliver Baines, Phil Wilkes, Mathias Disney, 2020. Quantifying urban forest structure with open-access remote sensing data sets, Urban Forestry & Urban Greening, 50,126653. <<<< Looked at the reference and found it applicable so added this to the paper. & added Soojeong Myeong, David J. Nowak, Michael J. Duggin, 2006. A temporal analysis of urban forest carbon storage using remote sensing, Remote Sensing of Environment, 101(2), 277-282.

Author Response File: Author Response.docx

Reviewer 3 Report

 

Review for “Comparison Between Artificial and Human Estimates with Urban Tree Canopy Assessment” #land-1982917

The authors presented a study that aims to evaluate the difference in UTC estimations between HI and AI methods.

 

Here, I provide a short, non-comprehensive review. This is largely due to incomplete methods. Others would be unable to reproduce this research given the methodological details provided in this draft. Since the methods are incomplete, it is not practical to provide a final review the work presented.

 

I left some comments in the marked-up PDF. Some of the more important or recurring questions/comments that I have are provided below.

 

Objective/goals/etc.:

The authors state in the abstract that “Our objective is to investigate and quantify the difference between HI UTC estimate and AI estimate within and outside regional AI training area.”

 

Then they state in the Intro:

“This study aims to assess whether AI and HI vary in UTC estimation. Specifically, the objective was to measure tree canopy using HI and test how it differs from values derived through an AI estimation approach.”

 

Then in their conclusions:

“Our objectives were to analyze whether there was a difference between human intelligence (HI) and artificial intelligence (AI) with urban tree canopy (UTC) assessment methods, whether there was a change in UTC between two time periods, and validate assessor estimates and their agreement.”

 

The objectives in the abstract, introduction, and conclusion do not match. Not trying to be pedantic, they are close, but consistency would help the reader.

 

The author present four questions, I am unsure of how these two questions align with the stated objective(s) above “3) What is the level of agreement in UTC estimates between two human assessors, and 4) was there a change in UTC in Wisconsin communities between 2013 and 2018 for the HI estimation method.”

 

Methods:

The foundational metric for this study, as portrayed by the Title and Introduction is UTC. The authors provide the following definition for UTC: “Tree Canopy Coverage or Urban Tree Canopy (UTC) is a metric to measure urban forests and is represented by the percentage of the total area directly covered by a tree crown comprised of leaves, branches, and stems [4, 6, 7].”

 

However, the authors do not estimate UTC as the “percentage of the total area directly covered by a tree crown comprised of leaves, branches, and stems.” Instead they determine if random points are one of several landcover classes. The authors do not provide a linkage between their methodology and the provided definition of UTC. I do not understand how the authors go from a title and intro focused on UTC to a method section focused on landcover classes.

 

Moreover, the methods are lacking the necessary detail to replicate this study. For example, there is no background provided for how they chose these landcover classes nor how they are classified as such. The AI data used is from a previous study, but additional description of this dataset within the manuscript is needed; simply providing a citation is insufficient given importance of the dataset within the study.

 

What was the spatial extent of the communities? Did point density affect the estimations?

 

Why did the authors use the water and wetland classifications to compare HI vs AI, if the water and wetland classifications were not derived from AI? How were they derived? Is it possible the “vector” maps were determined by HI?

 

Figure legends are incomplete.

 

There is inconsistent use of terminology: e.g., UTC, canopy, canopy cover, tree canopy; land, landcover, land-use

 

 

Comments for author File: Comments.pdf

Author Response

Thank you for your comments to address for improvement of the paper.

The authors presented a study that aims to evaluate the difference in UTC estimations between HI and AI methods.

Here, I provide a short, non-comprehensive review. This is largely due to incomplete methods. Others would be unable to reproduce this research given the methodological details provided in this draft. Since the methods are incomplete, it is not practical to provide a final review the work presented.

Objective/goals/etc.:

The authors state in the abstract that “Our objective is to investigate and quantify the difference between HI UTC estimate and AI estimate within and outside regional AI training area.” …. Then they state in the Intro:  … “This study aims to assess whether AI and HI vary in UTC estimation. Specifically, the objective was to measure tree canopy using HI and test how it differs from values derived through an AI estimation approach.” …. Then in their conclusions:  “Our objectives were to analyze whether there was a difference between human intelligence (HI) and artificial intelligence (AI) with urban tree canopy (UTC) assessment methods, whether there was a change in UTC between two time periods, and validate assessor estimates and their agreement.” …. The objectives in the abstract, introduction, and conclusion do not match. Not trying to be pedantic, they are close, but consistency would help the reader.  <<< Made edits to remove inconsistent language …. Changed abstract to remove the inconsistent language about objectives. …. Intro-: Changed intro to …. Our objectives are to investigate and quantify differences in estimates of Urban Tree Canopy (UTC) and other landcovers determined by Human Intelligence and Artificial Intelligence, evaluate the efficacy of AI algorithms for estimating landcover types outside of the training area, assess agreement between human assessors in estimates of UTC, and determine if there were changes in UTC between two time periods.. …. Conclusion Also removed the inconsistent objective language.

The author present four questions, I am unsure of how these two questions align with the stated objective(s) above “3) What is the level of agreement in UTC estimates between two human assessors, and 4) was there a change in UTC in Wisconsin communities between 2013 and 2018 for the HI estimation method.”  <<<< Made edits to read as …. Abstract Objective-Added- human assessors and comparing that to artificial assessments, assessing change in urban tree canopy between two time periods and analyzing assessment agreement accuracy …. Introduction Objective Specifically, the objective was to measure UTC using HA and test how it differs from values derived through an AA estimation approach, analyze change in UTC between two time periods and calculate assessor agreement.

Methods:

The foundational metric for this study, as portrayed by the Title and Introduction is UTC. The authors provide the following definition for UTC: “Tree Canopy Coverage or Urban Tree Canopy (UTC) is a metric to measure urban forests and is represented by the percentage of the total area directly covered by a tree crown comprised of leaves, branches, and stems [4, 6, 7].” <<<< Edited for clarity

However, the authors do not estimate UTC as the “percentage of the total area directly covered by a tree crown comprised of leaves, branches, and stems.” Instead they determine if random points are one of several landcover classes. The authors do not provide a linkage between their methodology and the provided definition of UTC. I do not understand how the authors go from a title and intro focused on UTC to a method section focused on landcover classes.  <<<< We edited for clarity with a point over a tree or shrub also synonymous with UTC.

Moreover, the methods are lacking the necessary detail to replicate this study. For example, there is no background provided for how they chose these landcover classes nor how they are classified as such. The AI data used is from a previous study, but additional description of this dataset within the manuscript is needed; simply providing a citation is insufficient given importance of the dataset within the study. <<<< Added some text for clarity but more so added the reference for the process used to import shapefles,, imagery, and record classification for the over 800,000 sample locations.

What was the spatial extent of the communities? Did point density affect the estimations? <<<< Point density was not used, 1000 sample points per community which is a common method used.

Why did the authors use the water and wetland classifications to compare HI vs AI, if the water and wetland classifications were not derived from AI? How were they derived? Is it possible the “vector” maps were determined by HI? <<<<We included water and wetland since they were used in the study we are comparing. We agree you could not include them, but we did and left in the paper.

Figure legends are incomplete.  We added information for each figure to explain the whisker plots.

There is inconsistent use of terminology: e.g., UTC, canopy, canopy cover, tree canopy; land, landcover, land-use <<<< We edited to just use UTC  for tree canopy but did leave reference to landcover were appropriate.

 PDF comments:

The title states “Urban Tree Canopy Assessment <<<< Changed to Assessments

 

Line 29- There is a rich literature to cite here. Consider adding some. If not, perhaps use a more recent review. <<< Added a more relevant and current reference and added additional newer references were appropriate.

 

Line 34- Awkward <<<< Changed to-Tree Canopy (UTC) assessments are a method used to measure urban forests and is represented by the percentage of the total area directly covered by a tree crown comprised of leaves, branches, and stems

 

Line 39-Would a non-expert in your field understand this sentence? “Landcover, including tree canopy.  <<<< edited  …. Changed to …  potentially highly accurate method to measure how land is used, including how many trees are present in cities

 

Line 53- For such a central part of your study, could you expand the citation list here? <<<< we added several more references  …. Thus, a UTC assessment is a measurement approach used to understand factors that drive forest change in cities [4, 5, 6, 7, 8, 9, 17, 23].

 

Line 65-You may enjoy the work from the Steve Frank Lab https://ecoipm.org/publications <<<< Thanks, interesting papers but we did not find any related to this topic.

 

Line 79- landcover? land-use? >>>> “Land” changed to landcover, throughout paper have now changed all to landcover

 

Line 82- It is confusing when the Authors' use many ways to refer to UTC. <<<< Changed for consistency

 

Line 84-Do the methods create algorithms? <<<< Changed create to use

 

Line 88-The authors do not provide a synthesis of the limits in their discussion.<<<< Changed to read …. “Ultimately, understanding the limits of HI 88 and AI as it applies to UTC estimates should lead to better predictions [43].”

 

Line 90-What are these? “NAIP and SPOT” <<< -imagery sources added

 

Line 98-Is that a goal of using AI to estimate UTC? “expenses to develop detailed and accurate landcover maps specifically to calculate UTC.” <<<< Yes, that is the reason the AI map was created in the paper we are comparing against.

 

Line 98- Is this addressed? The author's do not return to this question in the discussion.  <<<< Deleted the text in question.

 

Line 103- "Tree canopy" is still not defined at at this point in the manuscript. Unless it is referring to the definition of UTC. <<<< Changed to UTC

 

Line 109- Will the authors address the limits "of HI and AI as it applies to UTC estimates should lead to better predictions "  <<<< Yes we do throughout the paper

 

Line 112- What is a community?  <<<< A census designated place, changed to read as such

 

Line 123- Citation please .<<<< Added source 58 …. National Agriculture Imagery Program (NAIP) Available Online https://naip-usdaonline.hub.arcgis.com/

Line 129-What was the spatial extent of the communities? Did point density affect the estimations?   <<<< Not sure of the point, but each sample place was similarly samples with 1000 sample ponts.

Line 131- landcover? land-use? <<<< Made edit to landcover

Line 132- Where did these classifications come from?  <<<< Not sure of the point but we used these seven landcovers and provide reference to what they mean.

 

Line 134-Were the assessors naive to the questions asked for this study? “team of eight assessors classified all communities” Line 139 <<<< They were not naïve and we edited the sentence to better say they collectively classified communities

 

Line 136 There is insufficient detail to replicate this procedure. <<<< Added the citation Hilbert et al., 2019 that provided the method. For brevity did not include the method which documents the addition of shapefiles, imagery, and how the assessments were tabulated in ArcMap

 

Line 145- MMMM  <<<<, The AI system could not tell soil from impervious

 

Also, wouldn't the HI and AI use the same images? re: inconsistencies <<< Removed word images-some image quality is worse than others human eye may pick up on this.

 

Line 180- alpha? <<<< Yes, changes to α p≤ 0.05

 

 

Line 192&193& elsewhere- Please provide the test statistic value here and elsewhere. <<<< Left as is, the test statistic is presented  for these p values is Table 1 and Table 2, thus prefer to leave as is for simplicity in this paper.

 

Line 204- Figure 2 &figs 3,4,5 What do the "x"s indicate? Added Whisker Plot information for clarity.

 

Lines 250-252 Unclear. <<<< Edited for clarity

 

Lines 256-260- Not very clear.  Also, wouldn't the HI and AI use the same images? re: inconsistencies <<<< Edited for clarity and deleted Overall, the canopy assessments were significantly different in their interpretation of UTC throughout the state.

Line 279-281- Is that Significant >>>> Not sure since different methods between our work and that cited. The point is two works have numbers with 4% which may be the result of exclusion of water in the Nowak and Greenfield work.

Lines 312-315- This list is not parallel, awkward to read. <<<< Edited for clarity and to better align with the Abstract and introduction.

 

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

Review for revised “Comparison Between Artificial and Human Estimates with Urban Tree Canopy Assessments” #land-1982917 

The authors presented a study that aims to evaluate the difference in UTC and LANDCOVER estimations between HI and AI methods. The authors responded to most of my previous comments and concerns, thank you. My remaining concerns are below.

----

However, I do not think they have sufficiently explained the issue with their analyses not using UTC as they define it.

The authors still provided the following definition for UTC (Lines 37-41):

“Urban Tree Canopy (UTC) assessments are a method used to measure urban forests and is represented by the percentage of the total area directly covered by a tree crown comprised of leaves, branches, and stems [4, 6, 7]. Example methodologies of UTC assessment include on-the-ground assessments, visual interpretation of aerial imagery, and classification by artificial intelligence (AI) programs [8, 9, 10, 11].”

 

Citations 6 and 7 use the dot method like authors, but I am not sure they should used here with the statement that includes “Area”. Citation #4 – uses a different method than the authors for estimating canopy cover from aerial images (Page 66 – first full paragraph) and better fits the definition that includes “area”.

 

Then in the Methods section 2.2, they state (Lines 137-142): “Each community had 1000 randomly placed sample points which were used to estimate UTC, using the Create Random Points tool in ArcMap 10.8.1 (ESRI, Redlands, CA) using methods of Hilbert et al. [17]. Each point was identified as fitting within one of seven assessed landcover classifications used in the AI study. These classifications included 1) agriculture, 2) herbaceous & grass, 3) impervious, 4) soil, 5) tree and shrub, 6) water, and 7) wetland. The tree and shrub classification are also defined as UTC to calculate total % UTC in the results.”

 

I still do not see how UTC as defined as “percentage of the total area directly covered by a tree crown comprised of leaves, branches, and stems” is the same as the: (sum of points within a landcover category / 1000) * 100.

 

The authors’ metric has no quantification of “the total area directly covered by a tree crown comprised of leaves, branches, and stems”

 

For example, the dot method, from their citations #7 & 11 seem to fit what they did, but this does not match the definition provided for a UTC assessment. These citations do not include area in their estimates, just percentages. There is no “area” in the authors method.

 

Please explain and make this connection between your metric and the definition provided. The added "The tree and shrub classification are also defined as UTC to calculate total % UTC in the results." Does not address the above.

Other comments

----

Lines 119-120: cite US Census bureau?

 

----

Lines 194: “an α p≤ 0.05” – fix please.

 

----

The authors state that (Lines 320-321): “As the number of sample points for a landcover increases, error declines [8].” This citation, Parmehr et al. (2016), also states this, as listed in citation #6 page 4) “First, the number of points needed to reach a target level of accuracy (e.g. 95% confidence intervals) depends in part on the amount of canopy cover (Parmehr et al. 2016). Areas with less tree canopy cover need fewer points to achieve the same level of accuracy (Parmehr et al. 2016). If the area of tree canopy cover were already known a priori then one would not need to conduct random point sampling in the first place. This circuitous problem is often ignored or simply assumed away.

[…]

To detect a 5% change one needs more than 95% confidence (Parmehr et al. 2016).”

Do the authors address these or other concerns with random point sampling for estimating canopy cover that are discussed in citation #6 (like the excerpt above)? Perhaps they should at least be acknowledged?

 

 

Author Response

Review for revised “Comparison Between Artificial and Human Estimates with Urban Tree Canopy Assessments” #land-1982917 

The authors presented a study that aims to evaluate the difference in UTC and LANDCOVER estimations between HI and AI methods. The authors responded to most of my previous comments and concerns, thank you. My remaining concerns are below.

However, I do not think they have sufficiently explained the issue with their analyses not using UTC as they define it.

The authors still provided the following definition for UTC (Lines 37-41):

“Urban Tree Canopy (UTC) assessments are a method used to measure urban forests and is represented by the percentage of the total area directly covered by a tree crown comprised of leaves, branches, and stems [4, 6, 7]. Example methodologies of UTC assessment include on-the-ground assessments, visual interpretation of aerial imagery, and classification by artificial intelligence (AI) programs [8, 9, 10, 11].” >>>> We removed reference to tree parts and now as …. Urban Tree Canopy (UTC) assessments are a method used to measure urban forests and is represented by the percentage of the total area directly covered by a tree crown [4, 6, 7]. Example methodologies of UTC assessment include on-the-ground assessments, visual interpretation of aerial imagery, and classification by artificial intelligence (AI) programs [8, 9, 10, 11].”

 Citations 6 and 7 use the dot method like authors, but I am not sure they should used here with the statement that includes “Area”. Citation #4 – uses a different method than the authors for estimating canopy cover from aerial images (Page 66 – first full paragraph) and better fits the definition that includes “area”. >>>> We appreciate your suggestion and thought with and we used a standard method with estimating UTC which citations 6 and 7 are consistent with. We did not use the method of citation 4, however, we believe it gives excellent perspective of methods that can be used which we believe is what a literature review could and should develop around. In particular, from ref 4 it is page 62 that has relevance …. Canopy cover refers to the proportion of the forest floor covered by the vertical projection of the tree crowns. … We left as is.

Then in the Methods section 2.2, they state (Lines 137-142): “Each community had 1000 randomly placed sample points which were used to estimate UTC, using the Create Random Points tool in ArcMap 10.8.1 (ESRI, Redlands, CA) using methods of Hilbert et al. [17]. Each point was identified as fitting within one of seven assessed landcover classifications used in the AI study. These classifications included 1) agriculture, 2) herbaceous & grass, 3) impervious, 4) soil, 5) tree and shrub, 6) water, and 7) wetland. The tree and shrub classification are also defined as UTC to calculate the total % UTC in the results.” <<<< Yes this is the method we used

 I still do not see how UTC as defined as “percentage of the total area directly covered by a tree crown comprised of leaves, branches, and stems” is the same as the: (sum of points within a landcover category / 1000) * 100.  <<<< We removed the reference in the literature review to leaves, branches, and stems and hopes this removes the confusion.

The authors’ metric has no quantification of “the total area directly covered by a tree crown comprised of leaves, branches, and stems” <<<< Removed the reference in the literature review, hope this removes this concern for something we did not do

 For example, the dot method, from their citations #7 & 11 seem to fit what they did, but this does not match the definition provided for a UTC assessment. These citations do not include area in their estimates, just percentages. There is no “area” in the authors method. <<<< Again we removed the reference to leaves, branches, and stems from the literature review since we did not do this

 

Please explain and make this connection between your metric and the definition provided. The added "The tree and shrub classification are also defined as UTC to calculate total % UTC in the results." Does not address the above.  <<<<As stated above we removed reference to leaves, branches, and stems which should remove any confusion.

Other comments

Lines 119-120: cite US Census bureau? <<<< Made reference to  this agency as.  <<<< We had the citation on line 123. We also added this to line 119-120

The authors state that (Lines 320-321): “As the number of sample points for a landcover increases, error declines [8].” This citation, Parmehr et al. (2016), also states this, as listed in citation #6 page 4) “First, the number of points needed to reach a target level of accuracy (e.g. 95% confidence intervals) depends in part on the amount of canopy cover (Parmehr et al. 2016). Areas with less tree canopy cover need fewer points to achieve the same level of accuracy (Parmehr et al. 2016). If the area of tree canopy cover were already known a priori then one would not need to conduct random point sampling in the first place. This circuitous problem is often ignored or simply assumed away.  >>>> We added citation 6 to the paper. We also agree very much with your statement and interestingly have a study we have results on, but not yet published addressing sample size based on canopy cover  % and also total lands area.

To detect a 5% change one needs more than 95% confidence (Parmehr et al. 2016).” >>>> We are uncertain about this comment since we were reporting our accuracy assessment and comparing to other work using the method in this paper. We did not makes claims in the paper about being able to detect a 5% change.  While agree with your statement, however, we left as is since we did not address this in the paper.

Do the authors address these or other concerns with random point sampling for estimating canopy cover that are discussed in citation #6 (like the excerpt above)? Perhaps they should at least be acknowledged? <<<< Good point, we added this at line 321-322 location …. Further, the sampling intensity (e.g., number of sample points) is a function of the area covered by a landcover attribute and a pre-sampling estimate can be used to determine the sampling point intensity

 

Author Response File: Author Response.docx