Spatial and Temporal Patterns of Mangrove Forest Change in the Mekong Region over Four Decades Based on a Remote Sensing Data-Driven Approach

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Chaiyana and coauthors provide a straightforward and concise submission of mangrove forest change during the past 40 years using Landsat imagery. Their goal was clear (p.4 lines 148-149) “…to map mangrove forests from 1984 to 2023 at the pixel scale by integrating a time series of Landsat data and a conventional machine-learning approach.”

I gathered that their model, LandTrendr offers multiple advantages over existing tools used to monitor mangrove forest coverage over time. For instance, “The LandTrendr algorithm has been proposed by Kennedy et al. This algorithm can address missing data, cloud cover and shadow issues.” (p.3, lines 126-127).

In addition, their study “…focuses on assessing mangrove areas along the coast of Cambodia, Laos, Myanmar, Thailand, and Vietnam. These countries are known for their ecological significance and biodiversity.” (p.4 lines 161-162).

 

Specific comments are as follows:

Figures 1 and 2 are excellent.

p.10 line 311. Random Forest modeling. This is a sound approach using feature selection and a hold out sample data set for model validation.

p.11 Figure 4. This is an excellent figure showing both feature importance and the methodology for selecting the number of features.

p.14 Figure 7. The page is blank. There is no Figure shown.

p.18 Figure 9. This is a helpful figure. Based on your findings in this study, does global mangrove watch (GMW) need to adjust their reporting and any associated findings or conclusions? How does your study improve the collective knowledge of global mangrove forest monitoring?

Though I cannot see Figure 7, if I were able to inspect that figure, would I find any changes in mangrove coverage consistent with sea level rise during the past 40 years? Assuming an average annual SLR of ~2 mm per year, the total change over 40 years should be roughly 8 cm. Such a change might be sufficient to cause mangrove losses closest to the sea and increase near the inland boundary due to saltwater intrusion that favors mangrove vegetation. I am curious if you qualitatively observed such movement of mangrove forest comparing early imagery (40 years ago) compared to recent imagery.

Author Response

Please find the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

Major revision

This manuscript aims to map mangroves in Mekong region over four decades. The core method adopted, random forests, is a classic algorithm, and the preprocessing steps of creating stabilized imagery and water masking are proposed to improve mangrove classification, which is reasonable and practical. However, there are some critical issues that need to be addressed, which should have underscored the novelty of this manuscript, given the abundance of existing global mangrove products.

• Water is easily separated from mangroves. It is believed that water masking may not significantly contribute to mangrove classification. If this is the case, please present it in the manuscript. Similarly, cloud contamination is a common issue in mangrove mapping using optical imagery, and it is believed to be considered by other studies. No comparison with other methods cannot highlight the contribution and novelty of this manuscript.
• While the comparison between proposed results and existing products was provided, it is still recommended to offer an overview comparison of mangrove dynamic changes. This comparison should include the mangrove area change over the examined areas from 1996 to 2020, based on the Global Wetland Mapping (GWM) dataset.
• Based on GWM, reference data is manually collected using Collect Earth with a 3-meter PlanetScope. However, the reliability of the differences in mapping mangroves between the proposed and GWM methods is questionable without in-situ collection data.

Besides, there are some minor issues to be paid attention to.

• P131: it seems that the sentence is broken, ‘The of …’, please have a double check.
• P150-153: the second and third goals in this study seems to be same. Please consider use an appropriate word, such as analyze, rather than ‘monitor’ in the third goal to differentiate other goals.
• P269: how to create water mask based on the JRC dataset, occurrence with 100% or seasonality with a value greater than 6 months, please elaborate it as it will affect the final results.
• Figure 7: no figure is shown.
• Figure 9: please zoom in the major differences between proposed results and existing mangrove products to demonstrate its advantage.

Author Response

Please find the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

General comment:
The study addresses an important topic—mapping and analyzing mangrove distribution in the Mekong region using Landsat imagery and RF classification. The methodology is generally sound and the results are valuable. However, there are several issues that need clarification and correction to improve the clarity, consistency, and scientific rigor of the manuscript.

 

Major Comments:

 

1. Missing Figure 7:
   The manuscript refers to Figure 7 for satellite images showing mangrove classification in Myanmar, Thailand, Cambodia, and Vietnam at four time points. However, Figure 7 is missing in the manuscript and preview. Please provide this figure.
2. Method steps clarity:
   Section 2.2 mentions “three stages” (processing Landsat data with LandTrendr, constructing the RF model, and results validation), but the detailed Steps 1–3 do not fully correspond. For example, Step 2 (Spectral extraction and potential mangrove area) involves sample extraction and use of auxiliary topographic data, which is not included in the three stages, and Step 3 combines model construction and validation. This inconsistency may confuse readers about the workflow, reduce clarity, and affect reproducibility. Authors should clearly describe the relationship between stages and steps and unify terminology.
3. Feature selection clarity:
   In Section 3.1, it is stated that “12 out of 48 features were removed, leaving 26 variables,” but Figure 4 suggests that 22 features were actually removed. The quantitative criteria for retaining the 26 features are also unclear: it is not specified whether OA and F1 scores were jointly maximized or if another metric guided selection. Additionally, the discussion of OA and F1 scores using 15 features versus 48 features is not directly relevant to the choice of 26 features. Please check the numbers and clearly describe the quantitative basis for feature selection, including the OA and F1 scores metric used. Also, please consider removing or rewriting unrelated discussions about 15 features versus 48 features if they do not directly support the final selection.

4. Results and figure consistency:
   There are multiple inconsistencies between text descriptions in Section 3.3, Figure 5 and Figure 6:

• Myanmar mangrove area in 1984 is described as 586,509 ha, but Figure 5 shows over 610,000 ha.130,000 ha;
• Vietnam mangrove area in 1984 is described as 127,713 ha, but Figure 5 shows over 130,000 ha, and in 2023 is described as 157,123 ha, but Figure 5 shows over 155,000 ha;
• Proportions in Figure 6 for Thailand and Vietnam in 1990, 2010, and 2020 do not match textual descriptions; in 2010 and 2020, the order is reversed.
  These discrepancies may mislead readers and reduce confidence in the results. Please verify all values, ensure consistency between text and figures, and consider adding a table of annual mangrove areas.
1. Spectral reflectance interpretation:
    In Section 4.1, text states mangroves have lower red-band reflectance and higher SWIR reflectance, but Figure 8 shows the opposite. Please check data and revise text to ensure that the text description is consistent with the figure to ensure the scientific rigor of the article.
2. Logical basis of discussion conclusions:
    Section 4.1 claims that analysis provides insights into mangrove distinction from other vegetation. Since non-mangrove class includes bare land, water, settlements, etc., this conclusion is not fully supported. Please clarify the comparison groups or revise conclusions.

Author Response

Please find the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have addressed some issues, but two additional concerns remain.

• To validate the results, it is recommended to incorporate the dynamic change of mangrove areas based on the existing products into Figure 5.

 

• The authors have identified the misclassification of mangroves in existing products (as depicted in Figure 9) and have also stated that no in-situ field surveys have been conducted in various countries. Consequently, it is imperative to elucidate the methodology for determining or validating the accuracy of the proposed results.

Author Response

Please review from attached file.

Author Response File: Author Response.docx