Next Article in Journal
An AI-Based Workflow for Fast Registration of UAV-Produced 3D Point Clouds
Previous Article in Journal
Coherent Accumulation for Measuring Maneuvering Weak Targets Based on Stepped Dechirp Generalized Radon–Fourier Transform
 
 
Technical Note
Peer-Review Record

Canopy Height Mapping for Plantations in Nigeria Using GEDI, Landsat, and Sentinel-2

Remote Sens. 2023, 15(21), 5162; https://doi.org/10.3390/rs15215162
by Angela Tsao 1,*, Ikenna Nzewi 2, Ayodeji Jayeoba 2, Uzoma Ayogu 2 and David B. Lobell 1
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Remote Sens. 2023, 15(21), 5162; https://doi.org/10.3390/rs15215162
Submission received: 18 August 2023 / Revised: 19 October 2023 / Accepted: 26 October 2023 / Published: 29 October 2023

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Comments for authors:

- The first two paragraphs of the introduction section are too generic, little related to the subject of the article. It is not until the third paragraph that the authors enter the subject.  Maybe those first two paragraphs should be simplified.

- In the third paragraph the authors say that LIDAR captures the vertical structure of the vegetation with a 25m footprint spatial resolution. This should be explained a little better.

- In the MATERIAL AND METHODS section, the authors demonstrate that they have experimented and trained many models to accurately estimate canopy heights in dispersed plantation environments.

- The CONCLUSIONS section may be somewhat scarce. Tell what strategies should be followed in the future to further improve the precision in canopy height estimates. Or refer to the new satellite materials that are currently available or will be available soon.

 

Author Response

- The first two paragraphs of the introduction section are too generic, little related to the subject of the article. It is not until the third paragraph that the authors enter the subject.  Maybe those first two paragraphs should be simplified.

Thank you to the reviewer for their suggestions about readability and clarity. In the manuscript, we have shortened some of the less-related background information. We do keep parts pertaining to the different importance of plantations compared to forests, as we believe that this is a main motivation for the work.

- In the third paragraph the authors say that LIDAR captures the vertical structure of the vegetation with a 25m footprint spatial resolution. This should be explained a little better.

Thank you for pointing out the improvement that can be made, clarified to “25m-radius footprint spatial resolution (size of an active LiDAR collection), spaced approximately every 60m along mission tracks with an across-track width of 4.2km” [Line 55-56]

- In the MATERIAL AND METHODS section, the authors demonstrate that they have experimented and trained many models to accurately estimate canopy heights in dispersed plantation environments.

We thank the reviewer for recognizing our experiments and modeling. 

- The CONCLUSIONS section may be somewhat scarce. Tell what strategies should be followed in the future to further improve the precision in canopy height estimates. Or refer to the new satellite materials that are currently available or will be available soon.

Given the feedback, we added some comments about future work and emerging data to the manuscript conclusion section: “Further improvements to the task of plantation mapping may continue to emphasize these short trees. Additionally, subsequent data collection from a renewed GEDI or other spaceborne LiDAR missions may allow for higher data availability and further improved performance in locally calibrated models, altering the trade-off between data availability and distributional representativeness. “ [378-383]

 

We thank the reviewer for their time spent reviewing this paper and for providing comments about how to tighten the manuscript's narrative.

Reviewer 2 Report

Comments and Suggestions for Authors

I am very glad to review this manuscript entitled “canopy height mapping for plantations in Nigeria using GEDI, Landsat, and Sentinel-2”. The authors developed the better locally canopy height models than the global ones. However, I think there is something more to explain: 1. I understand that the more the local samples were used, the higher prediction ability of the RF model could be. The authors explained the local calibration and spatially radiating sampling strategy in section 2.6. But it seems that they did not fully address what they have done in sampling procedure. I suggest to add some illustration figures to show what you have done here and explain more in the text. 2. Details of the study area should be addressed. It’s better to introduce more about the province name you used in the figures. Minors: R2 or r2 or km2 should be R2 or r2 or km2. Page 12 Line 332 (Healey, 2020) Line 334 (Dubayah, 2022). The citation should meet the journal requirements. Figures Figure 1 latitude and longitude frame are needed for the maps in (a) and (b). Legend should be added in the map (a) although you explained it in the title. Percentage will be much better than count in Figure (c). All scale bars are needed. Figure 3 please change the font size of the axis title in (b) and (c). Too small right now. Please delete the last sentence “However we also find…” in the figure title. The best place to illustrate your opinions in discussion section. Figure 4 latitude and longitude frame are needed for the maps in (a). The y-axis name should be percentage, not pixel counts. Tables Tables’ row line should meet the requirements of the journal. At least the last bottom line should be needed. <5 MAE in Table 3 should be MAE for Trees <5m.

Comments on the Quality of English Language

 Minor editing of English language required

Author Response

I am very glad to review this manuscript entitled “canopy height mapping for plantations in Nigeria using GEDI, Landsat, and Sentinel-2”. The authors developed the better locally canopy height models than the global ones. However, I think there is something more to explain:

We thank the reviewer for their feedback!

  1. I understand that the more the local samples were used, the higher prediction ability of the RF model could be. The authors explained the local calibration and spatially radiating sampling strategy in section 2.6. But it seems that they did not fully address what they have done in sampling procedure. I suggest to add some illustration figures to show what you have done here and explain more in the text.

Thank you for pointing out this description was unclear, we added to the manuscript some discussion of the sampling procedure:
“We took the union of each of these 62 buffered r-radius circles as the complete sampling region for a given sampling radius. The radii used were {0,5,10,20}.” [Lines 234-236]

  1. 2. Details of the study area should be addressed. It’s better to introduce more about the province name you used in the figures.

Thank you, this information was added in the manuscript to section 2.1. “We investigate and collect data over palm-growing Southern Nigeria, which is the largest oil palm producer in Africa. Through a community partner providing sustainable processing in the palm oil chain, and as a precursor to both the field work and GEDI analysis, we developed a non-exhaustive map of oil palm locations in the region(Figure~\ref{fig1}a). Based on partner capabilities, these efforts were focused on 5 main states with socioeconomic relevance and varying plantation qualities: Bayelsa, Benue, Edo, Ekiti, and Imo.”

  1. Minors: R2 or r2 or km2 should be R2 or r2 or km2.
  2. Page 12 Line 332 (Healey, 2020) Line 334 (Dubayah, 2022). The citation should meet the journal requirements.

Formatting corrected for these cases.

  1. Figures Figure 1 latitude and longitude frame are needed for the maps in (a) and (b). Legend should be added in the map (a) although you explained it in the title. Percentage will be much better than count in Figure (c). All scale bars are needed.

We appreciate the detailed comments about improvements to the figure! We have implemented them in the new figures.

  1. Figure 3 please change the font size of the axis title in (b) and (c). Too small right now. Please delete the last sentence “However we also find…” in the figure title. The best place to illustrate your opinions in discussion section.

Figure and caption both updated.

  1. Figure 4 latitude and longitude frame are needed for the maps in (a). The y-axis name should be percentage, not pixel counts. Tables Tables’ row line should meet the requirements of the journal. At least the last bottom line should be needed. <5 MAE in Table 3 should be MAE for Trees <5m.

Figure and tables updated according to the comments. We deeply appreciate the reviewer's suggestions and careful reading of the paper.

Reviewer 3 Report

Comments and Suggestions for Authors

The study deals with the estimation and mapping using GEDI, Landsat, and Sentinel. The manuscript was well written and organized. Several advices presented here for improving the manuscript for publication in remote sensing.

 

1.      In general, it is difficult for optical remote sensing images acquired the vertical structure like forest height and why the authors choose optical images like Landsat and Sentinel-2 for canopy height mapping and estimation? Would you please give some advantages of these datasets?

2.      Reference and contents related to novelty of the inversion methods and features extracted in this study need to further including in the introduction part.

3.      Table 1 is not complete.

4.      Line 140 to Line 144 is not the method.

5.      I suggested add the distribution the field collected points.

6.      In the method part, what is the use of field collected 62 points?

7.      How did you set the parameters of RF algorithms in this study?

 

 

Author Response

The study deals with the estimation and mapping using GEDI, Landsat, and Sentinel. The manuscript was well written and organized. Several advices presented here for improving the manuscript for publication in remote sensing.

We thank the reviewer for their suggestions and their time spent reviewing the paper.

  1. In general, it is difficult for optical remote sensing images acquired the vertical structure like forest height and why the authors choose optical images like Landsat and Sentinel-2 for canopy height mapping and estimation? Would you please give some advantages of these datasets?

We acknowledge that this is an important question to understand the importance of the work. GEDI is a sparse dataset, so any given sampled location likely does not have a GEDI footprint/reading over that specific point. Instead, more abundant optical datasets have the “wall-to-wall” coverage necessary to get features for plantation coordinates around the world. [lines 51-56].. In order to emphasize why optical data is a good feature, we also added in the line "Rather, wall-to-wall data such as optical imagery provide the features with the appropriate coverage to observe plantations, but lack a mapping of the image to a GEDI metric like relative height." [56-58

  1. Reference and contents related to novelty of the inversion methods and features extracted in this study need to further including in the introduction part.

Thank you for the comments. We cite previous literature that have suggested the strategy of GEDI fusion with wall-to-wall data for the broader task of canopy height mapping [including in particular citations from lines 59-66], and emphasize the contribution of our work is evaluation on a new ecological domain (plantation setting). 

  1. Table 1 is not complete.

Thank you, the table has been formatted to correct this.

  1. Line 140 to Line 144 is not the method.

These 4 lines about the importance of how data were distributed have been moved to the discussion section in manuscript.

  1. I suggested add the distribution the field collected points.

Thank you for this good observation, we added a histogram for field-collected heights to our figure 1 as section D, in order to compare with plantation vs. region height distributions.

  1. In the method part, what is the use of field collected 62 points?

Added to methods section of manuscript– “These data were held out as a test set to confirm the model’s performance on unseen data given the strategy of using GEDI as a label for supervised learning.” [lines 150-152]

  1. How did you set the parameters of RF algorithms in this study?

Updated to include specific settings. “Hyperparameters were set to default scikit-learn settings (number of trees=100, criterion=squared error, minimum sample-split=2, minimum samples per leaf=1, and no max depth) ”  [206]. We did not conduct further hyperparameter tuning.

Back to TopTop