Next Article in Journal
Correction: Tang et al. Effects of Oasis Evolution on Soil Microbial Community Structure and Function in Arid Areas. Forests 2025, 16, 343
Previous Article in Journal
Tree Risk Assessment of Date Palms with Aerial Roots Using Minimally Invasive Technologies
 
 
Article
Peer-Review Record

Analysis of the Application of Machine Learning Algorithms Based on Sentinel-1/2 and Landsat 8 OLI Data in Estimating Above-Ground Biomass of Subtropical Forests

Forests 2025, 16(4), 559; https://doi.org/10.3390/f16040559
by Yuping Wang 1, Steven Hancock 2, Wenquan Dong 3, Yongjie Ji 4,*, Han Zhao 1 and Mengjin Wang 1
Reviewer 1: Anonymous
Reviewer 2:
Forests 2025, 16(4), 559; https://doi.org/10.3390/f16040559
Submission received: 14 February 2025 / Revised: 12 March 2025 / Accepted: 21 March 2025 / Published: 23 March 2025
(This article belongs to the Section Forest Inventory, Modeling and Remote Sensing)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper investigates the use of RF, SVR, and eXtreme gradient boosting combined with open-source remote sensing data from Sentinel-1, Sentinel-2, and Landsat 8 OLI for the estimation of above-ground biomass in subtropical forests in the Pu'er region of Yunnan, China. Below are my comments:

  • What is the rationale behind selecting specific bands from Sentinel-1, Sentinel-2, and Landsat 8 data (e.g., why certain bands from Sentinel-2 were chosen)?
  • Why particular features are more effective in AGB estimation? Can a comparative analysis of the effectiveness of different feature groups (e.g., texture vs. spectral features) be done?
  • Please explain the methodology behind the chosen hyperparameters for each machine learning models. A discussion on the choice of parameters for RF, SVR, and XGBoost is needed.
  • Why one leave out CV and not k-fold?
  • Why not using MAE for evaluation?
  • Correlation plots should be upper or lower 
  • Please compare different aspects of RF, SVR, and XGBoost such as computational efficiency or interpretability. I suggest to refer to “Fully component selection: An efficient combination of feature selection and principal component analysis to increase model performance”
  • Why does XGBoost perform better in some forest types compared to others?
  • The paper shows the resolution differences between Sentinel-2 and Landsat 8, but how these differences influence model performance?
  • where and why the models tend to over or underestimate? For instance, are there specific forest structures or AGB ranges where the models struggle?
  • The study focuses on the Pu'er region, please discuss the generalizability of the models to other subtropical or tropical forests.
  • This paper focuses primarily on remote sensing data and machine learning models. Please discuss how environmental and climatic factors (e.g., rainfall, temperature) could influence forest biomass, and how these factors were controlled or integrated into the analysis?

 

Author Response

Dear reviewers, thank you very much for your suggestions. We have placed your suggestion with our response in a Word attachment. Please download it for your review. Again, thank you very much for your suggestion.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The research topic, "An Analysis of Machine Learning Algorithms Based on Sentinel-1/2 and Landsat 8 OLI Data for Estimating Above-Ground Biomass in Subtropical Forests," requires the following modifications for improvement.

1- Lines 78 to 90: very long sentences restructure into short sentences.

Support your findings with references, especially from lines 78 to 87.

2-The authors should also include a review of literature that tackled AGB estimation using Deep Learning or ANN, such as:
-Zhang, L.; Shao, Z.; Liu, J.; Cheng, Q. Deep Learning Based Retrieval of Forest Aboveground Biomass from Combined LiDAR and Landsat 8 Data. Remote Sens. 2019, 11, 1459. https://doi.org/10.3390/rs11121459
The paper uses a Stacked Sparse Autoencoder network (SSAE) to estimate AGB from Lidar and Landsat 8 data
-Awad, M.M. FlexibleNet: A New Lightweight Convolutional Neural Network Model for Estimating Carbon Sequestration Qualitatively Using Remote Sensing. Remote Sens. 2023, 15, 272. https://doi.org/10.3390/rs15010272
The paper uses a new deep-learning model to estimate AGB for use in calculating carbon sequestration
-Zadbagher E, Marangoz AM, Becek K (2024). Estimation of above-ground biomass using machine learning approaches with InSAR and LiDAR data in tropical peat swamp forest of Brunei Darussalam. iForest 17: 172-179. - doi: 10.3832/ifor4434-017
Here, the authors use different machine learning models, including ANN, to estimate AGB the best was SVM.

3-There are two issues with the collected data: A. Why do Sentinel-2 and Sentinel-1 not have coincident dates? B. It is well-known that creating high-resolution Sentinel-2 images is challenging without the presence of a panchromatic band. Using the nearest neighbor method for upsampling is not acceptable.

4-The authors have replicated the same methods used in the literature for comparing RF, SVR, and XGBoost. The authors must highlight the innovative aspects of their research paper.

5-The use of the Coefficient of determination and RMSE to evaluate the result is not sufficient. The authors should use other reliable methods.

6-rephrase paragraph from lines 292 to 295 to the following: "Based on feature optimization, this paper constructs AGB inversion models for different forest types using RF, SVR, and XGBoost models under the conditions of S1, S2, LT8, and combined data. The results, obtained using leave-one-out cross-validation, are presented in Table 4."

7-The RMSE reported values are very high compared to some literature such as 
Li, L., Zhou, B., Liu, Y., Wu, Y., Tang, J., Xu, W., Wang, L., & Ou, G. (2023). Reduction in Uncertainty in Forest Aboveground Biomass Estimation Using Sentinel-2 Images: A Case Study of Pinus densata Forests in Shangri-La City, China. Remote Sensing, 15(3), 559.Where 

An RMSE of 1.733 t/ha was reported for AGB estimation using the Quantile Regression Neural Network (QRNN) model for Pinus densata forests in Shangri-La City, China.

Comments on the Quality of English Language

There are numerous grammatical errors, including issues with sentence structure and an overuse of passive voice.

Author Response

Dear reviewers, thank you very much for your suggestions. We have placed your suggestion with our response in a Word attachment. Please download it for your review. Again, thank you very much for your suggestion.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Well done!. No more comment. 

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have made all the necessary modifications to improve their research paper "Analysis on the application of machine learning algorithms based on Sentinel-1/2 and Landsat 8 OLI data in estimating above-ground biomass of subtropical forests" based on the reviewers' remarks. 

The authors are urged to improve the English language for some sentence structure.

Back to TopTop