Death and Regeneration of an Amazonian Mangrove Forest by Anthropic and Natural Forces
Round 1
Reviewer 1 Report
This paper unravels one plot of Amazonian mangrove forest dynamics, especially in Braganca Peninsula. The authors tried to understand the anthropogenic and natural environmental changes to the development of mangrove dynamics by utilizing long-term remotely sensed datasets. The methods used in the paper are not new but rather standard techniques in this field. At first, the authors conducted the mangrove classification mapping from the medium-resolution optical satellite imagery (LANDSAT and Sentinel products) and high resolution (google earth pro and Pleiades). They used annual classification mapping with on average 7 years difference from 1986 to the latest 2019. From these datasets, the classified maps were created: land cover class, mangrove cover class, and mangrove species class. Next, the Unmanned Aerial Vehicle-based Digital Elevation Model (DEM) was generated. The DEM was corrected with field measurement. Based on these spatiotemporal analyses, the authors deduced the dynamics related to the mangrove forest.
The study itself is interesting, especially considering how infrastructure development should consider environmental conditions. Not only to show the effect of new road construction on the changing mangrove dynamics but also to show to some extent the effect of the new construction on the ‘disruption’ of the mangrove forest. The study suggested that the disruption of hydrodynamic circulation, nutrient, sediment availability, and all related environmental properties that are carried by tidal circulation are an important factors in determining the mangrove dynamics. However, given no field dataset support the statement, for instance, water level record, bed level measurement, current measurement. Additional studies such as hydrodynamic modeling can be conducted to support the argumentation. So does the effect of climatic changes. I would avoid a strong statement if the authors could not show the correlation with field data. However, if the authors can provide solid references and show the correlation with a graph or table, rather than a description, I would agree to add the argumentation.
I tend to use Canopy Height Model (CHM) rather than Digital Vegetation Height Model (DVHM). However, if you can find references that suggested the use of DVHM I will have no complaint.
Section 2.2-2.4 : How did you classified the land cover? Random Forest, or any other technique.. please explain
Line 192-198 : A table that explains the flight plan, i.e., height, overlap (front-lateral), acquisition time, camera angle settings will be better.
Additional question: did you conducted the camera calibration procedure?
Line 213 : Did you mean point cloud classification?
Figure 6 : Orthophoto maps will be better than the oblique maps. I see that oblique photo will create different sense than the orthophoto.
Figure 7 : Why there is a difference of med and hires area in 2015?
Author Response
Reviewer 1
General comment
This paper unravels one plot of Amazonian mangrove forest dynamics, especially in Braganca Peninsula. The authors tried to understand the anthropogenic and natural environmental changes to the development of mangrove dynamics by utilizing long-term remotely sensed datasets. The methods used in the paper are not new but rather standard techniques in this field. At first, the authors conducted the mangrove classification mapping from the medium-resolution optical satellite imagery (LANDSAT and Sentinel products) and high resolution (google earth pro and Pleiades). They used annual classification mapping with on average 7 years difference from 1986 to the latest 2019. From these datasets, the classified maps were created: land cover class, mangrove cover class, and mangrove species class. Next, the Unmanned Aerial Vehicle-based Digital Elevation Model (DEM) was generated. The DEM was corrected with field measurement. Based on these spatiotemporal analyses, the authors deduced the dynamics related to the mangrove forest.
The study itself is interesting, especially considering how infrastructure development should consider environmental conditions. Not only to show the effect of new road construction on the changing mangrove dynamics but also to show to some extent the effect of the new construction on the ‘disruption’ of the mangrove forest. The study suggested that the disruption of hydrodynamic circulation, nutrient, sediment availability, and all related environmental properties that are carried by tidal circulation are an important factors in determining the mangrove dynamics.
Authors: We appreciate the analysis and comments. We have answered each comment individually below.
Reviewer 1
However, given no field dataset support the statement, for instance, water level record, bed level measurement, current measurement. Additional studies such as hydrodynamic modeling can be conducted to support the argumentation. So does the effect of climatic changes. I would avoid a strong statement if the authors could not show the correlation with field data. However, if the authors can provide solid references and show the correlation with a graph or table, rather than a description, I would agree to add the argumentation.
Authors: This issue is crucial to this manuscript and deserves to be highlighted. Ecohydrological models based on field dataset with topography, tidal inundation frequency, porewater salinity and vegetation distribution in the Bragança Peninsula were previously published [25,30-32]. These works have indicated that the topographically higher areas of the tidal flats have a lower tidal inundation frequency and greater exposure to evaporation, causing an increase in the porewater salinity (see Figure 3 in the new manuscript version). This pattern causes zonation in the mangrove and associated vegetation. For example, hypersaline zones (tidal inundation frequency <40 days/yr, porewater salinity of 80 - 100‰, 3 - 4 m above mean sea-level, amsl) are dominated by herbaceous vegetation. The herbaceous flats are surrounded by Avicennia trees 1 - 10 m tall in their lower limit (~80‰, ~3 m amsl). The frequently flooded tidal flats (240 days/yr, ~30‰, ~1 m amsl) are predominantly occupied by Rhizophora [25,30-32].
This database involving the topography and salinity relationship was used to support our interpretations of topographic gradient effects on the vegetation of the study area. Therefore, we think it unnecessary to obtain new water level records to discuss a hydrodynamic model again by correlation and tables because previous publications have widely discussed this issue and presented these data. It should be emphasized that fieldwork was carried out to validate topographic and vegetation maps obtained by aerial photogrammetry and satellite and drone images (lines 108 – 124).
Reviewer 1
I tend to use Canopy Height Model (CHM) rather than Digital Vegetation Height Model (DVHM). However, if you can find references that suggested the use of DVHM I will have no complaint.
Authors: We have used Digital Vegetation Height Model (DVHM) according to previous publications [24,26,30].
Reviewer 1
Section 2.2-2.4 : How did you classified the land cover? Random Forest, or any other technique.. please explain
Authors: The classification algorithm used for land cover prediction was the Nearest Neighbor, part of the eCognition object-oriented paradigm. It assigns each object to the class closest to it in the feature space. This method is based on segmentation, dividing the image into meaningful, spatially continuous, and spectrally homogeneous objects or pixel groups (lines 157 – 161).
Reviewer 1
Line 192-198 : A table that explains the flight plan, i.e., height, overlap (front-lateral), acquisition time, camera angle settings will be better.
Authors: We added the table 2 to clarify the flight plan.
Reviewer 1
Additional question: did you conducted the camera calibration procedure?
Authors: The camera and vision sensors were calibrated by the DJI Assistant 2 software before the fieldwork, while the Inertial Measurement Unit (IMU) and compass were calibrated once a day before the missions. (lines 217 – 220).
Reviewer 1
Line 213 : Did you mean point cloud classification?
Authors: Yes, thanks, the term has been replaced (lines 244-245).
Reviewer 1
Figure 6: Orthophoto maps will be better than the oblique maps. I see that oblique photo will create different sense than the orthophoto.
Authors: In addition to allowing a better perspective of the vegetation units, the oblique images made it possible to condense five images into a single figure. The ortho images occupied much space, requiring converting them into two figures, missing the complete visualization of the spatial-temporal evolution of the degraded area.
Reviewer 1
Figure 7: Why there is a difference of med and high area in 2015?
Authors: Differences in the mid and high-resolution degraded areas, for instance, in 2015, may be attributed to the mid-resolution sensors of the Landsat (30 m) that tend to unify the spectral response of various objects (mangrove shrubs, roads, the floodplain) in one pixel, while the high-resolution dataset allows identifying precisely, for instance, mangrove shrubs and seedlings (lines 309 – 313). Therefore, the difference in the quantification of the results may be related to the low capability of the mid-resolution sensors to detect isolated mangrove shrubs or trees, incipient mangrove, and low-density vegetation due to the spatial resolution of the selected images. On the other hand, high-resolution imagery has pixel sizes between 0.5 m and 2 m, which combined with field observations, allowed the identification of the incipient mangrove class (lines 420 – 422).
We hope to have addressed all comments accordingly. We look forward to your positive evaluation and acceptance of the revised manuscript.
Reviewer 2 Report
The authors discussed the death and regeneration of Amazonian mangrove forest and the influences from road construction, extreme climate events, sea level. It’s meaningful for government decision-making.
However, the paper needs some further explanations.
1) The Abstract must be reorganized to cover all the content for this manuscript, especially the methodology.
2) The area (in Table 2) for the degraded mangrove from mid- and high-resolution are clearly different (223.08—151.60 for 2015, 181.12—72.90 for 2019), so please specify these differences much clearly.
3) All the figures in this manuscript are not clear enough.
4) Please specify the sensor for the SAR data in Figure 13.
5) The Conclusions section can be improved.
6) Please correct the unit (km^2, m^2) for area.
Author Response
Reviewer 2
The authors discussed the death and regeneration of Amazonian mangrove forest and the influences from road construction, extreme climate events, sea level. It’s meaningful for government decision-making.
Authors: We appreciate the analysis and comments. We have answered each comment individually below.
Reviewer 2
However, the paper needs some further explanations.
1) The Abstract must be reorganized to cover all the content for this manuscript, especially the methodology.
Authors: We rewrote part of the abstract to emphasize the methodology (lines 17 – 21).
Reviewer 2
2) The area (in Table 2) for the degraded mangrove from mid- and high-resolution are clearly different (223.08—151.60 for 2015, 181.12—72.90 for 2019), so please specify these differences much clearly.
Authors: Differences in the mid and high-resolution degraded areas, for instance, in 2015, may be attributed to the mid-resolution sensors of the Landsat (30 m) that tend to unify the spectral response of various objects (mangrove shrubs, roads, the floodplain) in one pixel, while the high-resolution dataset allows identifying precisely, for instance, mangrove shrubs and seedlings (lines 309 – 313). Therefore, the difference in the quantification of the results may be related to the low capability of the mid-resolution sensors to detect isolated mangrove shrubs or trees, incipient mangrove, and low-density vegetation due to the spatial resolution of the selected images. On the other hand, high-resolution imagery has pixel sizes between 0.5 m and 2 m, which combined with field observations, allowed the identification of the incipient mangrove class (lines 420 – 422).
Reviewer 2
3) All the figures in this manuscript are not clear enough.
Authors: The manuscript has 15 high-resolution figures. The preparation of manuscripts with figures in the word file may decrease the figures resolution. However, these figures will be available with maximum resolution. Figure captions have been rewritten to reinforce the message of figures 1 (lines 94 – 96), 7 (lines 305-306), 8 (line 308), and 10 (lines 328-330). Additionally, we modified figure 10 and added figure 3.
Reviewer 2
4) Please specify the sensor for the SAR data in Figure 13.
Authors: We modified the description of the figure to clarify this. …(1972, X-band, Side-Looking Airborne Radar) and after (2019, Sentinel-1 C-band Synthetic Aperture Radar).. (lines 502-504).
Reviewer 2
5) The Conclusions section can be improved.
Authors: The Conclusion has been rewritten (lines 564 – 581)
Reviewer 2
6) Please correct the unit (km^2, m^2) for area.
Authors: Ok, thanks.
We hope to have addressed all comments accordingly. We look forward to your positive evaluation and acceptance of the revised manuscript.
Reviewer 3 Report
Major revision
This manuscript presented the dynamic changes of an Amazonian mangrove over the Bragança Peninsula using a multi-source dataset and analyzed the change factors from anthropic (highway construction) and natural (extreme climatic events) forces. It is important for mangrove conservation, especially for balancing mangrove conservation and urban development. The major problems of this manuscript are as follows.
1) The introduction is curt and should be improved. Specifically, the related works on analyzing the impact of human construction (not limited to the highway) on mangroves should be reviewed. Besides, the reason why so many multi-source datasets (mid-resolution, high-resolution satellite images, SAR, and drone-based imagery) are necessary is not clarified.
2) The relationship between the changes in mangroves and highway construction or extreme climatic events is speculated without convincing evidence. For example, the tidal flood was interrupted by the PA 458 road can be observed in figure 13, but we cannot obtain the result of increased porewater salinity (lines 450-452).
Besides, the following problems should also be paid attention to.
1) The revised methodology flowchart (Fig 2.) should highlight the differences with the source (Cohen et al., 2018).
2) The Caeté estuary and Taperaçu channel should be labeled in figure 1 (lines 84-87) since they are closely related to mangroves in the study area.
3) Why is incipient mangrove class necessary? According to the definition of incipient mangrove, low-height mangrove sprouts and shrubs (lines 396-397), and most of them are the seedlings of Avicennia (lines 312-313). It may lead to misclassification if they are with no distinguishing features.
4) Uncertainty of classification from multi-source datasets should be discussed.
5) Please keep the format of reference consistent (see Table 1 in Line 124 and Tab. 1 in line 156)
6) The class level, classes, satellite image, photograph, and description in figure 3 should be presented when the class level first occurs and are suggested to move forward.
7) We cannot find any information on incipient mangroves from Figs. 8 and 9 to identify them (lines 395-396). Please check.
8) The bar charts in figure 9 are suggested to be stacked, and the change will be presented.
Author Response
Reviewer 3
The introduction is curt and should be improved. Specifically, the related works on analyzing the impact of human construction (not limited to the highway) on mangroves should be reviewed. Besides, the reason why so many multi-source datasets (mid-resolution, high-resolution satellite images, SAR, and drone-based imagery) are necessary is not clarified.
Authors: We added two paragraphs to the introduction. The first paragraph adds a framework for the impact of human construction on mangroves (lines 66-73). The second paragraph shows the importance of multi-source remote sensing products in mangrove forest analysis (lines.48-57).
Reviewer 3
“The relationship between the changes in mangroves and highway construction or extreme climatic events is speculated without convincing evidence. For example, the tidal flood was interrupted by the PA 458 road can be observed in figure 13, but we cannot obtain the result of increased porewater salinity (lines 450-452).”
Authors: We added a map (Figure 3) with sediment porewater salinity in the Bragança Peninsula based on regressions between measured porewater salinities (obtained in 2002 and 2017 by Cohen and Lara [31] and Cohen et al. [30], respectively), tidal inundation frequency, and estuarine salinity gradient (modified from Lara and Cohen [32]). In this figure, the effect of the road construction on the porewater salinity is clear, causing a porewater salinity contrast between the NW (degraded mangrove with >70‰) and SE of the studied area (Avicennia Forest with < 50 ‰) (lines 488 – 492).
Reviewer 3
The revised methodology flowchart (Fig 2.) should highlight the differences with the source (Cohen et al., 2018).".
Authors: The methodology flowchart is original and has little relation with the flowchart of Cohen et al. (2018), then we have removed this citation (Figure 2).
Reviewer 3
The Caeté estuary and Taperaçu channel should be labeled in figure 1 (lines 84-87) since they are closely related to mangroves in the study area.
Authors: We modified Figure 1 to add the Taperaçu Bay’s and Caeté estuary’s labels.
Reviewer 3
Why is incipient mangrove class necessary? According to the definition of incipient mangrove, low-height mangrove sprouts and shrubs (lines 396-397), and most of them are the seedlings of Avicennia (lines 312-313). It may lead to misclassification if they are with no distinguishing features.
Authors: Lines 420-423 were rephrased for clarity. According to the methodology used for the mangrove species discrimination, only mangrove trees were classified as Avicennia germinans or Rhizophora mangle. We could not identify very small mangrove sprouts and shrubs with the high-resolution sensors, so, for classification purposes, they remained in level 2, categorized as non-mangrove vegetation. However, during fieldwork, we observed that vegetation in the degraded areas of the centre of the Bragança Peninsula corresponded to the regeneration of very low-heigh Avicennia plants. For this reason, the incipient mangrove class is necessary to better understand that degraded areas are being colonized by mangrove plants and not by other vegetation.
Reviewer 3
Uncertainty of classification from multi-source datasets should be discussed.
Authors: We added a paragraph (lines 424-435) discussing the uncertainty of multi-source mangrove classification. Multiple source datasets offered the possibility of obtaining two perspectives on degradation/regeneration occurring in Bragança’s mangrove forests. The mid-resolution dataset (Landsat and SAR data) and the methodology for the degraded area estimation did not allow us to accurately-quantify those areas. This dataset permitted us to understand the degradation/regeneration tendency due to the historical availability. The high-resolution dataset, with the spectral, geometric, and texture features, helped to quantify mangrove dynamics. This reliable quantification is evident in the accuracy assessment results of the proposed classes. We did not combine different resolution datasets to avoid classification uncertainties. Hence, both were presented together for comparative purposes. Degraded areas or mangrove forest coverage estimations did not use drone-derived imagery. Drone data was restricted to obtain vegetation height and tidal flat topography information.
Reviewer 3
Please keep the format of reference consistent (see Table 1 in Line 124 and Tab. 1 in line 156)
Authors: Formatting was corrected in line 189.
Reviewer 3
The class level, classes, satellite image, photograph, and description in figure 3 should be presented when the class level first occurs and are suggested to move forward.
Authors: We moved this figure to the 2.4 section (line 175), where the classification levels are mentioned first.
Reviewer 3
We cannot find any information on incipient mangroves from Figs. 8 and 9 to identify them (lines 395-396). Please check.
Authors: The figure reference has been corrected (line 422).
Reviewer 3
The bar charts in figure 9 are suggested to be stacked, and the change will be presented.
Authors: We modified figure 10 (previously 9) to stack the bars.
We hope to have addressed all comments accordingly. We look forward to your positive evaluation and acceptance of the revised manuscript.
Reviewer 4 Report
Title: “Death and regeneration of an Amazonian mangrove forest by anthropic and natural forces”
The manuscript presents a digression in the study area of ​​Bragança Peninsula where changes in the Amazon Macrotidal Mangrove Coast are analysed following the ecological variations created by the construction of Highway PA-458 in 1973. In order to classify the various species of mangroves are used a combination of satellite images from Landsat 5 TM, Landsat 7 ETM +, Landsat 8 OLI, Sentinel-2 ° and Drone imagery acquisition. A confusion matrix is ​​used to evaluate the accuracy of the model and the values ​​obtained from Kappa indices were 0.82 (2003 and 2015), 0.85 (2017), and 0.79 (2019), while the Kappa per class for vegetation is above 0.93, and for degraded areas class oscillates between 0.77 and 0.89.
The cross-sections show how degradation has been more extensive in the zones with tidal channels disconnected from the Caeté estuary.
The results obtained are widely supported by the analysis carried out.
Mangrove regeneration occurs on surfaces subject to a regular tidal flood regime. Alternately, depressions with tidal channels obstructed by the road tend to accumulate tidal water, leading to mangrove degradation. This degradation may occur in depressions permanently flooded due to porewater sulfides, a phytotoxin in wetlands that accumulates during permanent flooding conditions.
Author Response
Reviewer 4
The manuscript presents a digression in the study area of ​​Bragança Peninsula where changes in the Amazon Macrotidal Mangrove Coast are analysed following the ecological variations created by the construction of Highway PA-458 in 1973. In order to classify the various species of mangroves are used a combination of satellite images from Landsat 5 TM, Landsat 7 ETM +, Landsat 8 OLI, Sentinel-2 ° and Drone imagery acquisition. A confusion matrix is ​​used to evaluate the accuracy of the model and the values ​​obtained from Kappa indices were 0.82 (2003 and 2015), 0.85 (2017), and 0.79 (2019), while the Kappa per class for vegetation is above 0.93, and for degraded areas class oscillates between 0.77 and 0.89.
The cross-sections show how degradation has been more extensive in the zones with tidal channels disconnected from the Caeté estuary. The results obtained are widely supported by the analysis carried out. Mangrove regeneration occurs on surfaces subject to a regular tidal flood regime. Alternately, depressions with tidal channels obstructed by the road tend to accumulate tidal water, leading to mangrove degradation. This degradation may occur in depressions permanently flooded due to porewater sulfides, a phytotoxin in wetlands that accumulates during permanent flooding conditions.
Authors: We appreciate the analysis.
Round 2
Reviewer 3 Report
The authors have corrected some problems in prevision version. Still, some minor problems need to be resolved before being published.
1. The information on how to use SAR data for degraded areas mapping is missing, including SAR data preprocessing, features for classification, etc.
2. The conclusions should be clearly presented. Only the dynamic change of mangroves can be drawn from the results, and other descriptions, such as ‘This process caused a sediment deficit …, killing mangrove …’, cannot be drawn if no extra analysis was given.
3. The English language needs to be improved. For example,
· ‘Degraded areas or mangrove forest coverage estimations did not use drone-derived imagery’ in lines 70-71 on page 19.
· ‘Combining planialtimetric data with high and mid-spatial-resolution datasets indicated that the …’ in lines 246-247 on page 22.
Author Response
December 01, 2022
Dear Ms. Yulia Zhu
MDPI Remote Sensing Editorial Office
Ref: remotesensing- 1968595
Title: "Death and regeneration of an Amazonian mangrove forest by anthropic and natural forces”.
Please find attached the revised version of the paper entitled “Death and regeneration of an Amazonian mangrove forest by anthropic and natural forces”, by Cardenas et al.
We appreciate the suggestions. Changes are all indicated in the annotated version that we have attached. Below, we have answered each comment individually.
Editor
- The information on how to use SAR data for degraded areas mapping is missing, including SAR data preprocessing, features for classification, etc.
Authors: It has been done (lines 141 - 143, 153-157).
Editor
- The conclusions should be clearly presented. Only the dynamic change of mangroves can be drawn from the results, and other descriptions, such as ‘This process caused a sediment deficit …, killing mangrove …’, cannot be drawn if no extra analysis was given.
Authors: We have rewritten part of the Conclusion to emphasize only the data presented and discussed in this work (lines 194 – 208).
For instance:
A Radar image obtained in 1972 indicated that the road construction interrupted tidal channels previously connected with the Caeté estuary ( see Fig. 14) and caused topographic contrasts between both sides of this road (see Fig. 12), resulting in increased porewater salinity on the NW side (Fig. 3). Depressions permanently or semi-permanently inundated also degraded part of the original mangrove forests, preventing mangrove regeneration in some locations of the Bragança Peninsula (Fig. 9 and 12, transect C).
Editor
The English language needs to be improved.
Authors:
Our native English speaker colleague (Dr. Nicholas Culliganf) revised the manuscript again.
We hope to have addressed all comments accordingly. We look forward to your positive evaluation and acceptance of the revised manuscript.
Best regards,
Marcelo Cohen
Federal University of Pará
Rua Augusto Corrêa, 01
Cep: 66075-110, Bairro: Guamá
Belém-Pará
Tel: 91-3201-7478
Cel: 91-8031-1300
E-mail: [email protected]
www.ufpa.br/cpgg