Mapping Areas Invaded by Pinus sp. from Geographic Object-Based Image Analysis (GEOBIA) Applied on RPAS (Drone) Color Images

Round 1

Reviewer 1 Report

Dear Authors,

This version of the manuscript is now stronger. The Authors improved their previous submission.

There are still some minor issues to be taken into account before the publication.

 

Line 174: Please pay attention to figure 1. It seems that the flight area falls outside of the PNMDLC boundary;

Lines 241-243: Please give more information about this step (software, time of processing, obtained GSD, etc…). Moreover, please define the abbreviation VARI;

Line 318: Please check typo errors;

Line 567: 15 m a.g.l.? Is it correct?

Author Response

 "Please see the attachment."

Author Response File: Author Response.docx

Reviewer 2 Report

In this study, areas invaded by Pinus sp. in restingas were mapped using images taken by a remotely piloted aircraft system (RPAS, or drone) to identify the invasion areas in great detail, enabling management to be planned for the most recently invaded areas, where management is simpler, more effective, and less costly. Experimental results show the good performance of the proposed method. However, some issues should be addressed.

Major issues:

1) The structure of the paper is not standardized. The content of the paper is replaced by a large number of appendixes, which makes the readability of the paper not good. The authors should embed the contents of some appendixes directly into the paper, such as method model and experimental results.

2) Although the authors study a specific area. But I think that the method proposed by the authors is universal. In other words, can the authors verify the effectiveness of the proposed method in other areas? It is recommended to add at least one more set of experimental data.

Minor issues:

1) The methods proposed by the authors is mainly based on remote sensing image. In particular, some classification methods are applied. In the introduction part, there is too little introduction to the latest mapping methods, such as the mapping based deep learning, the mapping based on subpixel and so on. A comprehensive and systematic background introduction and description related to the advanced and latest works, e.g.,

[1] A Simple Method of Mapping Landslides Runout Zones Considering Kinematic Uncertainties, Remote Sensing, 2022, 14(3): 668.

[2] Super-Resolution Mapping Based on Spatial-Spectral Correlation for Spectral Imagery [J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(3): 2256-2268.

2) The last part of the introduction needs to give chapter arrangement. In addition, there are some grammatical errors in the article, which need further careful proofreading.

Author Response

"Please see the attachment."

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Thanks for the authors' reply. I have no problem here

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Dear Authors,

The paper “Mapping Areas Invaded by Pinus sp. from Geographic Object-Based Image Analysis (GEOBIA) Applied on RPAS (Drone) Color Images” presents a methodology based on the use of RGB images taken from camera on board on UAV to perform an object-based classification for the detection of invasive species such Pinus sp. Moreover, the paper compares the use of two different machine learning algorithms such random forest (RF) and support vector machine (SVM).

The topic is interesting but there are some issues that should be solved before its publication.

 

GENERAL COMMENTS

• Methods should be better explained. Information about the first three proposed steps (flight surveys, photogrammetric process and layer stacking) are completely missing. Explanation of other steps should be improved;
• The accuracy is presented only for the classification process. Please consider also accuracy in the segmentation process like in doi: 10.1080/22797254.2021.1951623, doi: 10.1016/j.rse.2017.11.024 and doi: 10.1016/j.isprsjprs.2017.06.003;
• Study area PEST seems to be covered in most part by herbaceous species. Why did the author choose this area to reach the goal of Pinus classification?
• In results section the differences obtained between RF and SVM should be better highlighted;
• Discussions should be improved comparing the obtained results with ones obtained by similar studies;
• Please reduce the number of the tables, most of them presents aspect that can be written into the main text;
• If tables and/or figures are prepared by the authors, please delete source statements

SPECIFIC COMMENTS

• Lines 171-175: please add more information about the flights (e.g., fly height above ground level, overlapping %, weather conditions, n. of photos for each flight, and so on…);
• Lines 188-194: please add information about the number and the type of classes adopted for the classification result. Moreover, add information about the machine learning algorithms adopted during this step of the proposed method;
• Lines 204-211: please move tables 1 and 2 to supplementary material;
• Line 236: table 3 presents the percentage of training and validation points adopted for each study area. Are these polygons distributed proportionally between the classification classes?
• Lines 238-247: these lines explain what the mentioned tools do but not how they have been used in the presented work;

 

Author Response

 Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

General comments:

The paper: “Mapping Areas Invaded by Pinus sp. from Geographic Object-Based Image Analysis (GEOBIA) Applied on RPAS (Drone) Color Images” fits well with the scope of the special issue “Structure and Trend Monitoring of Forest Vegetation and Savanna Based on UAS Platform” proposing an operational workflow to monitoring forest environments inside conservation units (Sapiens Park, Rio Vermelho State Park, Lagoa da Conceição Dunes Municipal Natural Park, Serra do Tabuleiro State Park) in Brazil. These areas are invaded by invasive alien plants of Pinus genus causing several impacts on the local biodiversity and difficulties in managing. The authors produced vegetation maps, including Pinus spp. class, using remote sensing data derived by remotely piloted aircraft system (RPAS) with RGB sensor following the Geographic Object-Based Image Analysis approach (GEOBIA) in four subsets each of 9 ha of four conservation units considered. All classifications were produced using two machine learning algorithm Random Forest (RF) and Support Vector Machine (SVM) considering as variables the values extracted by attributes of LargeScaleMeanShift and ZonalStatics algorithms calculated using Orfeo Toolboxes (OTB) and attributes of feature extraction function calculated using GeoDMA plugin. For the classifications, the authors have considered five combinations in both machine learning algorithms: all attributes derived OTB and GeoDMA, the combination of attributes calculated by OTB and GeoDMA, and two attributes subsets filtered by correlation-based feature selection and a wrapper type. Finally, the accuracy assessment of all classifications were evaluated using Global Accuracy, Kappa Cohen, F-score, User Accuracy, and Producer Accuracy.

The obtained results indicated very high accuracy values for all classifications, and Pinus spp. class was well detected and discriminated in all four areas. Furthermore, the SVM showed a higher efficiency to detect and classified these environments compared to RF algorithm. Whereas, in the four areas, the use of different attributes showed divergences to improvement in classifications.

In conclusion, the approach of this study confirmed the importance of RPAS to monitoring the vegetation and to quantified the spread of invasion by Pinus spp..

In general, the manuscript is too long (29 pages) and some phrases could be reduced and/or eliminated. Furthermore, study evidences concerns for confusion of the manuscript and for gap in some steps, especially in “Materials and Methods” and “Results and Discussion”, and before being published the study requires important improvements. In particular, all steps indicated in the flowchart (paragraphs 2.3) are described too briefly and the reader not understands all results described in the Results and Discussion paragraph. The description of “filter” tools indicated in the 2.3. paragraph confuse the reader, given that the specific name of these tools are reported for the first time only in the Results and Discussion paragraph, while the adequate description, with the specific name were reported only in the Appendix A. The tables are not standardized with a unique formatting and the caption of images not reported all information to understand their contents.

In consideration of above, the study may be addressed in a major revision.

Specific comments

Lines 45-49: The human pressure have modified large part of natural ecosystems, please change humans in human pressure. Moreover, please cited: Pysek et al. 2017 Naturalized alien flora of the world: species diversity, taxonomic and phylogenetic patterns, geographic distribution and global hotspots of plant invasion. Preslia 89, 203-274. doi: 10.23855/preslia.2017.203

Lines 57-59: Pleas add a reference in which the impacts of this invasion are reported.

Lines 68-71: Please add a reference.

Lines 72-75: Please add a reference, for example a review or several researcher in which the RPAS are used to detect an invasive alien plant.

Lines 85-87: Is this phrase necessary to introduce your research? This phrase could be eliminated.

Figure 1: Indicate in the caption the reference system used and horizontal datum. Furthermore explained the acronym SRC.

Line 135: Eliminate.

Lines 177-179: Indicate date, hour, height and velocity of flights, and also the percentage of overlap between the RPAS images.

Lines 186-187: Describe in detail the step of the image adjustments.

Lines 188-189: Describe the approach of GEOBIA and the setup of parameters used for the segmentation.

Lines 192-193: The paper cited is write in Portuguese. Report all steps described in Gonçalves and Ribeiro in order to make this method accessible to everyone.

Figure 3: Describe in the caption the elements reported in cartographic information.

Line 197: Eliminate.

Lines 198-199: Describe the steps to convert the vector into a multipart feature, furthermore report and describe the machine learning methods and theirs parameters used.

Table 1 and Table 2: These tables are too long, please reduced their dimensions or move their in a supplementary material. Furthermore, what is the reason that same attributes were calculated by OTB and GeoDMA. In particular:

Number of pixels (OTB) - Count of Pixels (GeoDMA);

Mean (OTB) - Mean (GeoDMA);

Minimum (OTB) - Minumum value (GeoDMA);

Maximum (OTB) - Maximum value (GeoDMA);

Variance (OTB) - Variance (GeoDMA);

Standard deviation (OTB) - Standard deviation (GeoDMA).

Is not possible classified the vegetation using different attributes between OTB and GeoDMA?

Finally, lines: 205-206 I do not understand this phrase, please re-write it.

Lines 212-218: Please add a reference in which this approach was used. Furthermore describe the SVM and RF algorithms and the parameters used for the classifications.

Lines 224-226: How the labeling was executed using a visual interpretation or in a field campaign? Please report it in the text.

Figure 4: see Figure 3.

Line 229: Eliminate.

Line 237: Eliminate.

Lines 238-247: In these phrases, the description of “filter” tools is only partial. The name and the specific description of these filter tools is missed, however in the results you reported the specific name of these tools. Furthermore, in these phrases, and in general in the 2.3. paragraph, I not understand the exact classifications produced. Please write a new table in which insert all classifications reporting the attributes and algorithms used.

Lines 255-260: I not understand the reasons to not calculate the accuracy assessment with the attributes derived by GeoDMA. Furthermore, how you calculated the improvement of the classifications derived by GeoDMA attributes if you no calculated their accuracies?

Table 5: In this table there are acronym not indicated in the text (e.g. libsvm, weka_CfsSE_GreedyStepWise_cv10x1_80%, etc.). Please, describe these acronym in the Materials and Methods and in the caption of this table.

Line 270: Eliminate

Lines 271-279: Move these phrases in Materials and Methods paragraphs. It is not a result.

Table 6: See comment above.

Line 275: Eliminate.

Table 7: This table is not understandable. Please re-write and standardize similarly to the other tables.

Line 289: Eliminate

Lines 295-299: The wrapperSusbsetEval combined with classifier J48, CfsSubsetEvaluation, Genetic Search, BestFirst, GreedyStepWise are not described in the methods. Please write a paragraph in which describe these tools and insert reference in which these tools were applied.

Table 9: Is this table really important? Is it possible merge the table 7, 8 and 9?

Line 302: Eliminate.

Lines 316-319: Eliminate the bold.

Lines 345-348: The Random Forest results are not reported, why?

Figure 6: Describe in the caption the Cartographic information.

Line 361: Eliminate.

Lines 371-372: Indicate the name of classes.

Figure 7: See comment of Figure 6.

Line 395: Eliminate.

Figure 8: See comment of Figure 6.

Line 414: Eliminate.

Line 427: Eliminate.

Figure 9: See comment of Figure 6.

Line 434: Eliminate.

Figure 10: See comment of Figure 6.

Line 446: Eliminate.

Paragraphs 3.1., 3.2., 3.3., 3.4. are necessaries in the main text. These paragraphs could be move in a supplementary material and in the main text you can reported only the principal results and difference between the four areas.

Author Response

Please see the attachment.

Author Response File: Author Response.docx