Influence of GEDI Acquisition and Processing Parameters on Canopy Height Estimates over Tropical Forests

Round 1

Reviewer 1 Report (Previous Reviewer 1)

The following lists should be considered: 

1. line 259 timespan， please relpace it by time span；

2. Line 409，a Stepwise Multilinear Regression (SRH),  is it SMR, or SRH? please confirm. 

3. Line 456, for the best performance of algorithm setting group number 5, the  RMSE is 11.6m, general speaking, it is too big. Why did the six algorithm setting groups all underestimate canopy heights? please explain detaildly.

4. I strongly suggest a figure should be added about the four study areas and their GEDI data and ALS data.  Hence, we can clearly see how the GEDI data and ALS data cover the areas.   

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report (Previous Reviewer 4)

Dear authors,

All my comments and doubts have been conveniently addressed and answered with remarkable clarity and exhaustiveness. Congratulations for your good work.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Question 1.  GEDI data was acquired after 2019, while in line 221 and 223, the reference data was obtained in 2008 and 2016, how do you deal with the canopy heights difference extracted from Airborne LiDAR in 2008 and GEDI data in 2019 and 2021? The main problem about the methodology is the acquired times difference between the reference data and the GEDI data. There are ten years difference while the canopy heights have changed greatly. When you use the canopy heights derived from airborne data acquired ten years ago to verify the canopy heights from GEDI from 2019 to 2021, how do you deal with the heights difference?

Question 2.   Do you consider the topographic factors when extracting canopy height from GEDI data？

Question 3.  Most of the cited references are out of date, I suggest the author reduce some references especially published before the year of 2016.

 

 

 

 

Reviewer 2 Report

1，Properly adjust the font size in the picture, and it should not be too far away from the font of the text.

2，The contribution and motivation of the paper need to be clearly defined in the manuscript.

Reviewer 3 Report

This paper presents an analysis of GEDI canopy height metrics across 4 locations, with 'reference' ALS data, in order to quantify and understand the implications of processing metrics and choice of power etc on the retrieved height.

There is some interesting work here, particularly in the technical aspects of characterizing the impact of power and the various algorithm choices on the height metrics from GEDI. The comparison to ALS is also potentially interesting and useful. However, in the current form I don't think there is sufficient clarity in the data presentation or analysis to support key aspects of the work.

In particular, there are a number of areas where the sites are grouped together into "similar" eco-climatic regions, but no evidence is presented to justify this. Is this in any way true and if so in what sense? To say something like that you’d need to provide MAT, rainfall etc. and show what you mean by ‘similar’. This seems a big stretch to me, particularly eg given the coastal v inland regions (and elevation, geology) in the French Guiana case. I think it's well-known that Paracou and Nouragues are fairly different climatically. Given the analysis lower down is v often lumped, then why make this claim of eco-climatic similarity?

The other major issue is that the ALS 'reference' data for the FG sites, are up to potentially 8 years or more prior to the GEDI data. Again, at the very least we would expect this to introduce a potential height bias into any comparison. Maybe GEDI is actually understimating more than is thought because the 'reference' data are from canopies which were a m or more lower when covered in the ALS. This is just ignored and really can't be - it needs to be dealt with somehow. This is potentially the most difficult issue to deal with for the work as it is.

The other aspect is then the scatter plot analysis - why not just do it per site so that we can see whether there are site-specific differences, or biases, and if so where they may arise? Then you don't have to make the pretty unlikely assertion about eco-climatic similarity and it may also help quantify why things work or don't.

I think if these issues can be addressed properly then the paper could be published, but it would need some significant work.

Reviewer 4 Report

Nice paper dealing with the evaluation of the influence of various GEDI (full waveform LiDAR) space sensor signal acquisition and processing parameters in very dense equatorial rainforest study sites to estimate canopy heights. Reference heights (CHM) are extracted from ALS LiDAR data. The document is clearly presented, while the methods applied and the results obtained appear to be scientifically sound. In addition, the results obtained are valuable for the scientific community, since they refer to two study sites that are quite similar in terms of climatic and environmental conditions, biomass density and forest structure, but located on two different continents. In this way, the experiment dealing with the transferability of the regression models obtained from both study sites (ie, cross-training validation experiments) is highly appreciated.

However, some issues arise that should be clarified. For example, there is a significant time gap between the reference data capture date (ALS data taken in 2008, 2012, and 2016) and the GEDI data acquisition (2019 to 2021). The reader may wonder if this excessive time lapse could affect the results obtained in this work. Please consider that the rainforest has a high annual growth rate. In addition, there may be notable changes in some GEDI points as a result of phenomena linked to deforestation or afforestation that should be previously detected and removed from the data set when considering this large time lap between reference data and GEDI data. 

Regarding taking the maximum value of the CHM raster cells contained within the GEDI footprint as the reference height (ground truth), I'm not sure that's the best option. You write that the maximum value of the CHM raster cells was chosen because it is theoretically closer to the canopy top signal obtained from the GEDI waveforms, but no reference citation is provided. Why not try other proxies for reference, such as the average value (or mean value) of CHM within the GEDI footprint? Note that GEDI waveforms have been smoothed during processing. Please clarify this point conveniently in the manuscript. 

Finally, and for the sake of clarity, a quantitative value has been applied to the beam sensitivity variable in the case of RF regression, with sensitivity being a key explanatory variable in this case. However, you have treated the sensitivity variable as categorical in the linear regression case (values = 0 or 1, depending on the 98% threshold value). I wonder if a fair comparison is being made between both regression models (linear and RF). Please clarify this point because some reasonable doubt may arise, especially considering that RF showed better results than linear regression. Could it not be played fair?