Submit to this Journal Review for this Journal Propose a Special Issue

Article Menu

Share Help Cite Discuss in SciProfiles

Open AccessArticle

Peer-Review Record

Mapping and Estimating Blue Carbon in Mangrove Forests Using Drone and Field-Based Tree Height Data: A Cost-Effective Tool for Conservation and Management

Forests 2025, 16(7), 1196; https://doi.org/10.3390/f16071196

by Ali Karimi¹, Behrooz Abtahi¹ and Keivan Kabiri^2,*

Reviewer 1:

Sergei Shevyrev

Reviewer 2: Anonymous

Reviewer 3:

Aristeidis Georgakis

Reviewer 4: Anonymous

Forests 2025, 16(7), 1196; https://doi.org/10.3390/f16071196

Submission received: 27 May 2025 / Revised: 15 July 2025 / Accepted: 17 July 2025 / Published: 20 July 2025

(This article belongs to the Special Issue Mangrove Forest Ecosystems: Present Status, Challenges, and Future Directions)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Critical response on a manuscript entitled “Mapping and Estimating Blue Carbon in Mangrove Forests Using UAV and Field-Based Tree Height Data: A Cost-Effective Tool for Conservation and Management” written by Ali Karimi and others and submitted to Forests

The manuscript under consideration is a well deserved work dedicated to mangrove forest biomass assessment with UAV surveillance supported by in-situ management. Term “blue carbon” taken into a manuscript title emphasize problem of climate changes mitigation. However, paper is not related to climatic changes problem and it could seem biased to some extent. So, why wouldn’t use “carbon stored in a biomass” or “captured carbon”?

Authors just used methodology of “blue carbon” resources computing with an equation taken from a work of predecessors. They don’t do analysis of any climatic data, so is it really relevant and necessary to use “blue carbon” term here?

Apart of that observation, manuscript is well organized and written in a concise and comprehensive language. Critical text referenced commentaries are provided below. Hopefully, addressing them could make paper more valuable for the international reader.

Line 33. One reference may not be enough to proof that point clearly. Also, if “blue carbon” means carbon dioxide, the form of its presence in a sediment has to be clearly explained. Is it buried organic matter? Root systems of mangrove forest?

L34. “carbon storage capacity” … “makes mangrove one of the most important ecosystem in climate change mitigation”. It seems a bit pretentious. Can mangrove (that are actually connected to the tidal zone) really challenge equatorial forests or forested areas of the moderate zone? I hardly think so, they obviously can’t, basically on their existence in a very specific geographic zone. However, it doesn’t mean its carbon storage can not or must not be studied. Moreover, its harsh conditions make use of UAVs important for facilitating assessment of landscape characteristics, so, authors study is important. I just ask to make it less biased and shift emphasis to facts.

L42. AGB was already expanded (L15).

L53. Need exact reference on a study where that determination coefficient (not accuracy!) was derived, regression model and extend of that model application (meaning that model has limits of the application).

L76-82. Despite citing studies that used ensemble learning machine learning models, authors have limited their scope with simplest MS Excel linear regression models. That a bit frustrating. Authors used original facts and in-situ samples and conducted model completion in a most naive instrument ever. Also, it worth to build and alternative model and compare two model with the accuracy metrics.

L103-104. For what period of observations? Also, Google Earth service shows satellite image mosaics of different acquisition time and sources. The reliability of that satellite image source is questionable.

L106-107. You have outlined site manually. So, you don’t need to reference predecessors and capable to measure anything retrospectively on you own using image archives (e.g. Earth Explorer project).

Figure 1.At glance it is unclear to what picture scalebar is relevant, only with the scale size. Map insets should be marked with letters or better lay-outed. Top side image, is it true or pseudocolor RGB? Date when it was taken? Also, UAV payload has to be explayned/referenced.

Figure 2. What is the meaning of different colors of actions rectangle? If there are no different meaning colors have to be the same. Better to update flowchart making it complaint with ISO9001 standard flowcharts. Also, it is impossible to address actions in a text/discussion, because they don’t have numbers. Flow seems exactly linear, while it is not. E.g. if regression analysis (which model exactly?) fails with accuracy metrics it will proceed to “height group classification” anyways… How it comes? Model should be reselected/readjusted/reassessed over again.

Figure 3. Point of picture taking has to be pinpointed at Figure 1. Also, date of imaging and picture plane orientation is important, have to be provided.

L161. Formula has to have a number for future referencing. Also, source where formula was taken should be cited.

L167. Software have to be referenced data processing procedures must be clarified.

L174. Imagine, not every (potential) paper reader is this software user. So, please use algorithmic names instead of software tools names and reference them. Please, keep in mind that proprietary tools impedes reproduction of the results and could devalue paper to readers.

L194. Symbol after “1.185” is not typed. Also, is that regression equation applicable on your site?

L205. “Other studies” Which, exactly?

L210. These instruments aren’t really convenient to share (in comparison with research software scripts). Won’t you provide Excel workbook to make other prove accuracy of formulas?

L216. Power two is missing.

L235. What is the meaning of that?

Figure 4. How these exact values of absolute heights were derived and which height model is taken (b)?

Besides of the questions raised manuscript presents good looking research based on the original facts and field data. I recommend minor revisions.

Author Response

Reviewer 1:

Comments and Suggestions for Authors

Comment:

Response:

We sincerely appreciate your thoughtful comments and critical observation regarding the use of the term “blue carbon” in the title and throughout the manuscript. While we acknowledge that the manuscript does not directly include climate variable datasets (e.g., temperature or CO₂ flux measurements), the study is strongly grounded in the blue carbon framework, which encompasses the estimation of carbon stored in coastal vegetated ecosystems such as mangrove forests.

Our primary objective was to quantify above-ground biomass and convert it into carbon stock using globally accepted allometric and carbon conversion factors (e.g., the 0.48 factor suggested by Kauffman & Donato), a practice that is a central component in blue carbon assessments. Numerous studies such as Shaltout et al. and Walden et al. adopt similar approaches without including climatic parameters, and are nonetheless categorized under the blue carbon theme.

Therefore, we respectfully propose to retain the term “blue carbon” in the title, as it accurately reflects the nature and context of our study and aligns with established terminology in this field. We are, however, happy to clarify this scope further in the Introduction section, if the editorial team finds it helpful.

Comment:

Response:

We have revised the sentence and added two additional references. We note that the conversion factor 0.48 specifically applies to above‑ground biomass (AGB), whereas a factor of 0.39 is generally used for below‑ground biomass (BGB). In our study, we apply the 0.48 factor only to AGB to calculate carbon stored in living biomass. We do not estimate below‑ground or sediment carbon in this work.

Comment:

Response:

We agree with your observation and have revised the sentence to better reflect the specific.

Comment:

L42. AGB was already expanded (L15).

Response:

Done. The full term was removed from subsequent mentions, and only the abbreviation "AGB" is used after its first introduction.

Comment:

Response:

We clarified that the reported R² = 0.98 refers specifically to the drone-derived tree height estimation and field measurements, and not to biomass prediction. The text was revised accordingly to specify the regression context and to avoid any ambiguity.

Comment:

Response:

We sincerely thank the reviewer for this insightful comment. The use of advanced ensemble learning models such as Random Forests or other machine learning approaches was beyond the scope of the current study, which primarily aimed to test the applicability of simple empirical models using in-situ and UAV-based measurements. However, we fully agree that implementing and comparing machine learning-based models can provide deeper insights and potentially improve prediction accuracy. We consider this an excellent suggestion and will certainly explore and integrate such approaches in our future research.

Comment:

Response:

We have revised the text to report a nearly half mangrove loss based on the satellite-derived data for 2015–2022. Additionally, a disclaimer has been added to clarify that Google Earth imagery reflects variable acquisition dates and sources, which may introduce uncertainty in the exact change estimate.

Comment:

Response:

We appreciate the reviewer’s observation. Indeed, since we manually delineated the site using UAV-derived imagery, it would be feasible to estimate vegetation coverage directly from our own orthomosaic. The cited value (50%–75%) was initially included to support site selection rationale and is consistent with what we visually observed in the imagery. However, as the reviewer correctly noted, referencing external sources may be unnecessary here. Accordingly, we will revise the sentence to reflect our own site-based estimation.

Comment:

Response:

Figure 1 is re-shaped

Comment:

Response:

Thank you for your suggestions regarding the flowchart design. The current version of Figure 2 uses color-coded blocks to visually group different types of processes: green for field and tree-level data inputs, light green for raw UAV data acquisition, yellow/orange for image processing, dark green for biomass estimation, blue for blue carbon conversion, and grey for statistical analysis. While this structure does not strictly follow ISO 9001 flowchart standards, it was intentionally designed for clarity and logical grouping of research stages. Additionally, as the process was primarily linear in nature (with no decision loops or model iterations in this version), we aimed to present it in a straightforward format for readability.
We have now updated the figure caption to clarify the color scheme, and we will also describe the regression model and its outcomes in the corresponding section of the text to avoid ambiguity.

Comment:

Figure 3. Point of picture taking has to be pinpointed at Figure 1. Also, date of imaging and picture plane orientation is important, have to be provided.

Response:

We revised it.

Comment:

L161. Formula has to have a number for future referencing. Also, source where formula was taken should be cited.

Response:

We revised it.

Comment:

L167. Software have to be referenced data processing procedures must be clarified.

Response:

We have included full references for the software tools used—Agisoft Metashape Professional and ArcGIS Desktop. We have expanded the description of the image processing workflow in Section 2.5, including details of the settings used in Agisoft (e.g., alignment quality, depth filtering) and the steps followed to generate DSM, DTM, DEM, CHM, and orthomosaic. Additionally, we clarified how ground points were classified, how elevation differences were calculated, and how vegetation structure was derived from raster layers.

Comment:

Response:

We fully agree with your point regarding software-agnostic description of data processing. In the revised manuscript, we have replaced tool-specific names with generic algorithmic terms wherever applicable. For instance, instead of referencing software tools like “Raster Calculator” or “Classify Ground Points,” we now describe the processes in terms of their underlying geospatial methods (e.g., raster subtraction for CHM, ground point classification based on slope and elevation patterns). This enhances clarity and reproducibility regardless of the software used.

Comment:

L194. Symbol after “1.185” is not typed. Also, is that regression equation applicable on your site?

Response:

We revised it. Yes, it has been used in national reports. But unfortunately it is not accessible for all. We attached it in our resubmission email and highlighted the paragraph related to your question on page 53 (53 in Persian).

Comment:

L205. “Other studies” Which, exactly?

Response:

The sentence has been revised to explain that some regional assessments (e.g., national reports) have proposed alternative conversion factors (e.g., 0.42). However, for the sake of consistency with global practices, the commonly used value of 0.48 was selected. The alternative factor of 0.42 was reported in a Persian-language national report, which we opted not to cite directly, as it is not accessible to an international audience.

Comment:

L210. These instruments aren’t really convenient to share (in comparison with research software scripts). Won’t you provide Excel workbook to make other prove accuracy of formulas?

Response:

We attached our Excel files to our resubmission email

Comment:

L216. Power two is missing.

Response:

We revised it.

Comment:

L235. What is the meaning of that?

Response:

To clarify the meaning of "outputs," we have updated the section title to “Drone-Derived Geospatial Data Products,” which more accurately reflects the nature of the four geospatial layers (DSM, DTM, CHM, and orthomosaic) generated through UAV photogrammetric processing.

Comment:

Figure 4. How these exact values of absolute heights were derived and which height model is taken (b)?

Response:

The absolute height values presented were derived from a raw Digital Elevation Model (DEM) exported directly from Agisoft Metashape, not from the DSM (Figure 4b). As clarified earlier in the Methods section, for each tree, the highest visible point of the crown and the closest ground point beneath it were manually identified within the DEM. The elevation difference between these two points was then used to calculate tree height. This approach allowed us to isolate the true vertical structure of each tree without relying on automated segmentation from the DSM or CHM layers.

Besides of the questions raised manuscript presents good looking research based on the original facts and field data. I recommend minor revisions.

Reviewer 2 Report

Comments and Suggestions for Authors

Find my comments in the pdf file. Best of luck.

Comments for author File: Comments.pdf

Author Response

Reviewer 2:

Comments and Suggestions for Authors

Dear Authors,

The topic of UAV-derived data for tree analysis is interesting and relevant. However, the manuscript suffers from several issues that affect its overall contribution and clarity. The originality and novelty of the methodology are quite limited, especially due to the lack of statistical modeling and automation. The main focus is on estimating tree height using UAV data. The study relies on a small sample (30–60 trees) and lacks methodological rigor in key areas, including crown delineation, which is not automated or clearly explained.
As you state in your conclusion, “Our analysis revealed strong correlations between tree height and crown diameter.” However, the manuscript primarily centers around height prediction, with AGB estimation based on existing allometry and manual or unclear estimation of crown diameter. A major limitation of the study is that biomass estimates are only provided at the individual tree level. There is no extrapolation or upscaling to the entire area.

Below are detailed comments that could help improve the manuscript, if considered acceptable by the editor.

# Abstract
Line 24: Please clarify what is meant by “non-invasive estimates.” Are you referring to UAV data being collected without disturbing the forest structure?

Response:

Thank you for the helpful suggestion. We clarified the sentence to define “non-invasive estimates” as methods that do not require cutting, harvesting, or physically disturbing trees, in contrast to destructive sampling approaches often used to determine actual biomass. The revised sentence now reads accordingly.

# Introduction
Line 53: Clarify to which parameter the reported high R² value (0.98) refers.

Response:

We clarified that the reported R² = 0.98 refers specifically to the drone-derived tree height estimation and field measurements.

# 2. Materials and Methods
2.1 Study Site
What is the total area of the study site? Please include this information.

Response:

In the revised manuscript, we have added the total area of the study site (i.e., the red polygon shown in Figure 1) based on the georeferenced drone survey. This area was calculated from the orthomosaic using GIS tools and has been reported in Section 2.1.

Lines 106–108: Why suitable characteristic? Provide in on sentense the improtance of the choice.

Response:

We revised and clarified it.

Reference Figure 1 in the text where appropriate.

Response:

In the revised version we have referenced Figure 1 two times.

Figure 1. The study area description is vague. Clarify the extent of the study area. Is the red line in the upper part of the panel?

Response:

We clarified in the manuscript text that the red polygon in Figure 1 delineates the exact study area analyzed using UAV data. We also updated the figure caption to indicate that the red line represents the extent of UAV coverage used for individual tree assessments.

Figure 2: Describe how the DEM was generated and how was used in the study.

Field data was used only as validation for UAV-derived height?

Response:

In the revised version of the manuscript, we have added additional explanation regarding the Digital Elevation Model (DEM) in the Methods section, as suggested in earlier comments. To clarify, the DEM used in this study was an unclassified elevation surface exported directly from Agisoft Metashape prior to the generation of DSM and DTM. In other words, it represents the raw elevation model before any ground or object classification is applied.

Regarding the second part of your question, yes, the field data were collected solely for the purpose of validating UAV-derived tree height estimates. The field-measured heights were used as ground-truth references in the regression analysis and paired t-test, but not for constructing or modifying the DEM or CHM. This separation of datasets allowed for an unbiased evaluation of the accuracy of UAV-derived measurements.

2.3 Field Data Collection
The methodology for measuring tree height may have introduced errors. Were standard tools (e.g., clinometers, laser rangefinders) used?
Reference other studies and discuss how they typically measure tree height for comparison.

Response:

In our study, we used a leveling staff and ground-based photography, where the staff was placed adjacent to each tree and photographed to visually record the height. This approach was selected due to the muddy, intertidal, and physically constrained terrain of the site, which made tripod-based clinometers or laser rangefinders difficult to deploy effectively.

While tools like clinometers, hypsometers, and laser rangefinders are indeed widely used in forest measurements, several studies have also validated photo-based measurement methods in small-stature mangrove environments as being reasonably accurate and practical under harsh field conditions [e.g., Shaltout et al., 2021; Kabiri, 2020]. We have now expanded the Methods section to discuss this issue, added supporting references, and acknowledged the limitations of our approach more explicitly.

2.6 Biomass and Blue Carbon Estimation
Line 194: Number the equation and correct the exponational part.

Response:

We revised it.

Lines 199–201: Clarify how crown area was extracted from the orthomosaic. Was this done manually or using an algorithm? Since crown area is a predictor variable, the method used for extraction is important.

Response:

We have clarified in the revised manuscript that crown area was calculated using accurate diameter measurements obtained via the Measure tool in ArcMap, which offers sub-meter precision. For each tree, the major and minor crown diameters were extracted from the orthomosaic and used to calculate crown area under the assumption of elliptical crown shape (π × a × b). This approach provided reliable measurements due to the high spatial resolution and clear individual crown outlines in the UAV imagery.

2.7 Statistical Analysis
Describe how crown diameter was calculated. Did you use the average of minor and major diameters?

Response:

Yes, the average crown diameter was calculated as the arithmetic mean of the major and minor crown diameters, both of which were measured using the Measure tool in ArcMap.

Line 216: Correct the R².

We revised it.

The correlation analysis between height and different crown diameters (minor, major, average) needs justification. Was this step necessary for model selection or just exploratory?

Response:

The correlation analysis between tree height and various crown diameter metrics (minor, major, average) was conducted as an exploratory step to better understand the relationships between structural tree parameters. These analyses were not used for model selection or AGB estimation but served to support the descriptive component of the study.

State clearly which crown diameter metric was ultimately used for AGB estimation.

Response:

None of the crown diameter metrics were directly used in AGB estimation. Instead, they were used to calculate the elliptical crown area (CA), which served as one of the predictor variables in the allometric biomass model.

Address whether statistical assumptions were tested (e.g., normality of residuals, homoscedasticity), if regression model appied.

Response:

In this study, regression analysis was conducted primarily to assess the empirical relationship between variables at the tree level, rather than to build a predictive model. As such, no formal statistical tests (e.g., Shapiro–Wilk for normality or Breusch–Pagan for heteroscedasticity) were applied to the residuals. However, residual patterns were visually inspected through scatterplots to ensure no major deviations from linearity or constant variance. We have added this clarification in the revised Methods section.

Line 220: Why were 30 additional tree heights derived from the DSM instead of the CHM? The CHM is typically more accurate for height.

Response:

In our study, we originally used the unclassified DSM (referred to as a DEM in our text) for manually measuring individual tree heights in the supplementary drone sample for the following reasons: our goal was to ensure manual, tree-specific measurement by selecting the highest elevation point of each crown and subtracting the corresponding ground level near the trunk (rather than relying on raster-wide automated height values). in some cases, edge effects, crown overlap, or missing pixels in the CHM (due to point cloud misclassification over water or shadows) introduced artifacts that reduced confidence in automated CHM-based height values, by using the raw DSM and referencing local ground elevation manually, we ensured greater control over height calculation for each individual tree.

Nevertheless, we recognize the advantages of CHMs for automated large-scale canopy structure analysis. We have clarified this decision in Section 2.5 and acknowledged it as a potential source of variability in the Discussion (Section 4.2).

# 3. Results
3.1 Remote Sensing Outputs
Figure 4: Consider removing unrealistic terrain values (e.g., areas with DTM < 0, like -106 m).

We removed values in revised figure 4.

3.2 Field Measurements
Line 253: Clarify what is meant by “To evaluate the field measurements and spatial distribution of individual trees.” We usuall evaluate remote sensing data, no field measurements.

Response:

We agree that “evaluate the field measurements” may lead to misunderstanding. Our intention was to highlight the use of field-measured tree heights for spatial visualization and classification. We have rephrased the sentence to better reflect this purpose.

Line 257: Consider providing a table of descriptive statistics for the tree heights, both overall and by the height classes you defined, or referring to details in Figure 6, which more clearly illustrates individual tree heights.

Response:

We provided a new table (table 1) based on this comment.

Figure 5: The figure is very large. Consider reducing its size.

Response:

We revised it and now is smaller.

3.3 Field and UAV Data Comparison
Line 273: Embody the p-value in the text and explain the result.

We revised it.

Line 280: Embody the p-value in the text and interpret the statistical results more thoroughly. What does a p-value of 0.1159 imply in the context of your hypothesis?

We revised it.

Figure 8: Explain the observed differences between field measurements and UAV-derived samples, especially with regard to crown dimensions. Provide more information in the text.

Response:

We added some explanations to the previous paragraph of Figure 8.

3.4 Allometric Modeling Results
State clearly which type of crown diameter (minor, major, average) was used in the biomass model.

Response:

In the allometric biomass model, crown area was used as a predictor variable. The crown area was calculated using both the minor and major crown diameters assuming an elliptical shape, but the diameters themselves were not directly included in the model.

Tables 1, 2 and 3: Define all abbreviations used in the tables (e.g., AGB, DBH, R²).

Response:

In the initial version, we defined all abbreviations in a table at the end of the manuscript, yet in revised version based on the suggestion of reviewer 3 we used the full term the first time it appears, and then refer to it using only the abbreviation thereafter.

3.5 Tree Height Grouping and Carbon Distribution
Figure 9: This figure is very large. Consider summarizing the key findings in text.

Response:

We have revised the manuscript to include a textual summary of the key findings from the pie chart. The new sentence clarifies the proportion of carbon stored in each height class, highlighting the dominant role of tall trees in blue carbon storage.

The figure is now smaller.

3.6 Total Estimated Biomass and Blue Carbon
A major limitation of the study is that biomass estimates are only provided at the individual tree level. There is no extrapolation or upscaling to the entire area.

Response:

Indeed, this study was designed as a pilot effort to validate the integration of UAV-based tree height measurements with allometric models for individual-tree-level biomass estimation. Our primary aim was to assess the feasibility and accuracy of using UAV-derived data at tree scale. We have now added a statement to the discussion acknowledging this limitation and explaining that extrapolation to the entire area requires further data collection on tree density and spatial structure, which will be considered in future studies.

Lines 320–324: You do not provide sufficient evidence for "UAV-derived data, showed significantly higher totals". The section is too brief and lacks statistical validation or discussion.

Response:

We have revised the text to include the actual total biomass and blue carbon values obtained from the supplementary UAV sample group.

# 4. Discussion
4.1 Overview of Key Findings
Avoid repeating results in detail. This section belong primary to the Results section. Focus on interpreting the findings in the context of other studies and literature.

Response:

We removed details that belong to results such as R² and p values.

4.2 Evaluation of UAV Accuracy and Sources of Error
This section overlaps with the Results.

Response:

We sincerely appreciate your thoughtful comment. For this section we highlighted the factors that actually reduced accuracy of our study which is beyond the results. So we believe this section should remain in the manuscript.

# 5. Conclusion
Clarify that the study focuses on individual tree-level prediction, not area-level biomass estimation.
Acknowledge the small sample size (30 trees for prediction and 30 for UAV comparison) as a major limitation.
Give quantitative evidence that your mehtod works equally well as field data alone.

Response:

We have now clarified in the revised conclusion that the scope of our study was limited to individual tree-level estimation, without upscaling to the area level. We also explicitly acknowledged the limited sample size (30 field-measured and 30 UAV-sampled trees) as a key constraint in the generalizability of the findings. Moreover, we included quantitative evidence supporting the reliability of UAV-based estimates by highlighting that the total blue carbon from UAV-derived data (841.0 kg) differed by only 88 kg from field-based estimates (929.0 kg), with no significant difference observed at the individual tree level (as shown in Figure 7). These changes are reflected in a new paragraph added to the Conclusion section.

Reviewer 3 Report