Prediction of Forest Aboveground Biomass Using Multitemporal Multispectral Remote Sensing Data

Round 1

Reviewer 1 Report

1. Introduction- In this section, the paper has introduced Lidar and multi-spectral (Landsat, Sentinel-2, MODIS) for forest biomass estimation. There are lots of studies on forest biomass prediction more accurately using SAR on different wavelengths (C, S, L, P) around the world including the temperate countries like Italy. There are also non-parametric approaches highlighting more accuracy like Random forest and Entropy. This paper should include more information on SAR biomass relationships with different wavelengths ( Santoro et al 2020 Ningthoujam et al. 2017 Saatchi et al. 2011 Baccini et al. 2012 Thurner et al. 2014 ). 
2. 1. Study area: Mention the satellite data in Figure 1. Are those Sentinel-1. 2.2. Field Data: Provide tree density, tree height, DBH for all the plots (88) in a table.  4.2. Spectral analysis: Include RS predicted biomass values from Thurner et al. 2014 and Santoro et al. 2020 for your study sites for validation against the Field estimated biomass in Figures 7 and 8 to support your finding. 4.3. Spatial analysis: Since the field plot biomass were calculated at 30 m diameter circular plot and trying to relate against 3m, 5m and 10m RS products. Demonstrate these 3m, 5m and 10m against biomass in these similar scales for both the biomass and RS datasets (Drone, rapid Eye and sentinel-2) (Figures 9-12). Because the RS biomass are always predicted at plot scale (see Saatchi et al. 2011- Impacts of spatial variability of tropical forest structure on radar estimation of forest biomass). This means there should be more than 88 plots if you compare at 3m and 5m spatial scales. Improved these figures after improving the sample sizes. Describe in detail how this has been done. Include and compare against Thurner et al. 2014 and Santoro et al. 2020 biomass products. 
3. Discussion: Improve the Discussion and Conclusion section based on the corrections  with above comments particularly Spectral and Spatial analysis.  

All the best

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This is an interesting study on the estimation of Above Ground Biomass (AGB) of temperate forests using Remote Sensing (RS) Data. Although there are numerous studies attempting to estimate AGB using RS data and methods there is still a long way to go until a consensus is reached and accurate AGB estimations can be made using RS data and methods alone. So, any study on this subject is welcome. In this study the accuracies achieved in the various models and using various data sets are not particularly high, but as the authors state in their discussion this study does not aim to provide a ready to use tool but rather a guideline.

The manuscript is well written and reads well throughout. The introduction, although too long, provides the necessary background information. The methods are adequately described and can be transferred to other areas with similar data. With one exception which is also my main concern about this study. The authors assess the accuracy of their AGB estimation based on field measurements of DBH and Tree height and an allometric equation and a relevant reference is provided (59), which is published in Italian. Perhaps the most important step when RS data and methods are evaluated against fields measurements is the accuracy of the latter. I believe the authors should provide more convincing information, including of course the equation used, and the ecosystems where this was developed so the reader can be confident that the ground truth estimations are accurate. Furthermore, in the material and methods section perhaps the authors could provide the formulas for the various indices used in a table including the relevant references.

The results are presented with a good combination of tables and although lots of the information provided in tables are repeated in the text, it reads well. The discussion is in accordance with the results with no unnecessary speculations.

With all the above being said I would like to recommend major revision on this manuscript which I sincerely believe it will improve the manuscript.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Estimating forest AGB by remote sensing is a popular domestic and international research topic, and has progressed substantially in recent years. In this paper, the authors used different spaceborne multispectral remote sensing data to predict forest AGB, and evaluate the effects of temporal, spectral and spatial capacities of satellite multispectral data for AGB prediction. However, the innovation of the method is weak in the manuscript, and using high-precision remote sensing data to improve estimation accuracy of AGB is small, which may also be related to the model used. to have it published on the remote sensing, substantial improvements are still needed. Below please find some questions and suggestions that are major.

Main comments:

1. How to match sample plots of 30m diameter with remote sensing data (spatial resolution of 3, 5 and 10m)?
2. I suggested that the variable calculation formula in this paper can be put into a table in the Section 3.2, which is more intuitive.
3. Line 231, one more “.” Symbol added, and deleted it.
4. Lines 338-342, the variables were frequently selected multi-times, for example, the GLI variable was selected 4 times. Are the values of selected variables different times the same, or are its different in time series?
5. In spatial analysis of Section 4.3, the authors only analyzed the relationship between vegetation index and AGB at different resolutions (3, 5 and 10m), but did not analyze the spatial change of AGB. I suggest that authors should analyze the spatial heterogeneity of AGB.
6. Figures 9-12, the coordinate axis is not clear.
7. The AGB sample plot investigated is 2014 for Pellizzano, while the remote sensing data is 2016 and 2017. The time mismatch may also affect the predicted AGB, which should be analyzed.
8. The results indicated that using the multitemporal data of DOVE and RapidEye to improve estimation accuracy of AGB is smaller than Sentinel, which may also be related to the model selected.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

1. Improve the abstract by adding the main message from the sensitivity of Vegetation Indices that you have found be useful for predicting biomass.

2. In the Result section, introduce 4.4. to include spatial biomass maps for 2 sites (Lavarone and Pellizzano) at Sentinel-2 (10 m) scale to highlight the spatial biomass distribution and its accuracy assessment. Provide detail in the text especially in those under-sampled sites (refer to fig 1- 88 plots distribution).  

3. Provide the limitations of this study (field data, biomass estimation, remotely sensed data) and recommendation for future work in the discussion section.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors have addressed all my comments in an informative and convincing manner and especially the part concerning the estimations of the ground truth AGB values. I believe the manuscript is now more robust and will constitute a significant contribution to the relevant literature.  I would like to congratulate the authors for their work and encourage them to continue their work in the topic.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

In section 4.3, I suggest that author mapping the AGB of  research area.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf