Relationship between Lidar-Derived Canopy Densities and the Scattering Phase Center of High-Resolution TanDEM-X Data

Round 1

Reviewer 1 Report

The presentation of the article titled “Relationship between lidar-derived canopy densities and the scattering phase center of high-resolution TanDEM-X data”, includes the determination of the height of coniferous and deciduous trees using high-resolution Insar data from TanDEM- X and lidar-derived data at two locations. The article presents a good structure, and adequate writing, legible in its approach, very clear and concise. In general, I congratulate the authors, I consider that they have done a good job.

However, I also consider that it is necessary to address some observations that will improve the article. These are aspects that do not consider much modification, but that will help a better analysis.

The observations are as follows:

1.      Adjust to 250 words in the abstract

2.      The following phrase located on lines 26 and 27 "In the Free State of Thuringia, 26 Germany, forests constitute about 30% of the whole land surface" take you to the part where the specific case under study is discussed, or in In this case, its removal is recommended.

3.      Place a high-resolution base image on both maps to get an idea of the characteristics of the study area. In the coordinates, the decimals should be eliminated, and consider placing another frame of the coordinate grid.

4.      In line 205 it speaks of “statistical metrics”. Describe what specific metrics you are referring to.

5.      In Figure 5, both methods used in the article should be explained on the point layer (on the spatial data obtained in the field) similar to what is shown on the maps.

 

6.      In Figures 6, 7, and 8 it is necessary to add a statistical evaluation (eg ANOVA) to observe the differences and their statistical significance between groups. Consider within methodology and discussions.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

Summary

In this study, the authors seek to compare InSAR-derived heights with airborne lidar-derived height and density metrics in order to understand the relationship between tree structure and the scattering phase center. They use two study areas in Germany, one dominantly coniferous and the other dominantly deciduous to compare how forest type affects this relationship. Considering the work that has already been done in this realm, this study represents a relatively small contribution to the field in terms of novelty and potential impact. The study areas are small, which opens a question about the broad applicability of these results. There were some methodological choices along the way that have potentially important effects on the results. The writing lacks clarity, and the figures could certainly be improved. While I commend the authors for their efforts, I do not believe that this paper is publishable in its current form without extensive revisions, not only to the manuscript, but to the underlying analysis.

 

Major Comments

-          I recommend splitting the Introduction into Introduction (everything before section 1.1) and Background (sections 1.1 and 1.2).

-          The study areas are very small. In both cases, these are smaller than a typical tile of ALS data (1 x 1km). Having such small study areas detracts from the impact of the study, calling into question the broad applicability of the results. I understand having small study areas in situations where there is an extensive field campaign wherein it would be time or cost prohibitive to collect data over large areas. But this is a study purely based on remote sensing data with seemingly no such limitations. Why, then, wouldn’t the authors choose to conduct the study over a (much) larger area?

-          With regard to the elevation correction between ALS and SAR (L286), considering the importance of precision in this study, it seems a bit over-general to say “a height offset of about 46 m”. Why “about” and not an exact number? Is that offset the same throughout both study areas? I would imagine there would be at least some variation if not within study areas then certainly between them, which would, in turn, seem to have a somewhat important effect on the results.

-          In L307-309 you state “One can argue that analyzing only those pixels with positive differences could bias the data…”. I suppose I am one that would make that argument. Assuming a priori that SAR heights are always going to be lower than ALS heights seems like a pretty important and potentially flawed assumption. Can you prove that these are actually errors, as you suggest? Or was this simply done to get “better” results? And then if you do assume that these are errors, why would you then re-include them in a subsequent analysis (L317-318)?

-          Section 3.3.2. is confusing. It’s not clear how the exact position of SAR SPC was calculated, nor is it clear why this couldn’t be calculated on all cells. And how were the cross sections chosen? You state that they were “representative for the whole sites”, but was this just a qualitative evaluation? In Figure 1, it looks like there are two cross sections right next to each other running north-south, but they overlap the exact same pixels, so what is the value of having two? I would argue that these cross sections are not very representative, as they don’t capture the full range of h100 values throughout the study areas.

-          It seems that there needs to be some consideration of the actual canopy cover within each analysis pixel. Most of the results are discussed in the context of coniferous vs. deciduous. While surely C vs. D plays an important role, higher absolute vegetation density within a particular layer seems to be an important piece that is not explicitly quantified. For example, you could have two adjacent pixels, both of which have the same SPC, but that have very different vegetation densities. The one with a lower vegetation density will likely have a lower InSAR height than the one with a higher vegetation density, as there is a greater likelihood of the radar pulse interacting with vegetation in the denser vegetation. You have access to this information from ALS, why not use it? You make brief mention of this effect in L412-414, but it’s not clear why you wouldn’t try to quantify that, given its importance.

-          It seems like one of the major limitations of this study is the use of 5m vertical height intervals. Why not use a “height of median energy (HOME)” type of metric instead of enforcing the 5m interval voxels to identify the densest portion of the canopy? That would seemingly simplify your analysis and avoid the processing issues you mention.

-          The Conclusions section contains a lot of redundancy and needs to be shortened, distilling down the key findings and implications of the study.

 

Minor Comments

L3: not sure about the use of the abbreviated “e.g.” in the middle of a sentence. Consider revising. This style is also seen elsewhere throughout the paper (“i.e.” in L62, “e.g.” in L72, and in several other locations). In my experience, these are almost always found within parentheticals (as you’ve done in L116), rather than in the middle of sentences.

L4: no need to capitalize Synthetic Aperture Radar. Also, InSAR is provided as an abbreviation but not referred to again in the Abstract, so it may not be necessary to do so. Perhaps just abbreviate SAR.

L10: same comment as above – no need to abbreviate if you’re only referring to something once in the Abstract.

L25: seems a bit awkward to describe forests as gauges of forest health.

L26-27: your first sentence describes the importance of forests – a very broad statement. You next sentence discusses one very specific region of the world. I get that this is the region you are studying, but this sentence seems very out of place in the context of a general introduction.

L40: “over the last years” doesn’t tell us much… Either specify a time frame, or keep it more general and say “recently” or something similar.

L42: replace “cost-expensive” with simply “expensive”

L61-63: considering the importance of this sentence (defining the objective for the entire study), this should be stated more clearly. It took me a few reads to understand the main study objective, which should be made abundantly clear. Also, is this a single objective or multiple? Can it be described as a numbered list of objectives? I always appreciate when authors list objectives in this manner as it is easy to refer back to in order to understand the degree to which study objectives were (or were not) successfully accomplished.

L64: atypical to see subsections within an Introduction. Should 1.1 and 1.2 be placed in their own Background section?

L125-129: motivation for selecting these specific sites?

L132: replace “researches” with “studies”

L133: it is useful to provide Latin species names in situations like this, as you do in Section 2.1.2.

Figure 1: the histogram and the map don’t seem to align quantitatively. The legend in your map is bounded by 26 m at the low end, yet 26m appears to be nearly the median of the entire dataset in the histogram? Not clear. Also, those two north-south cross sections appear to be extremely close together – why was this done? Please provide the UTM zone to which the coordinates refer. Lastly, I’m sure I speak for many in the remote sensing community when I say it’s always helpful to see an aerial photo of a study area to help provide visual context for the vegetation type and structure.

Figure 2: exactly the same comments above about visual differences between map and histogram, added visual utility of an additional aerial photo map, need for the UTM zone in the caption, and questions about why the cross sections are so close together. Also, your labels on the y-axis of the histogram should include decimals since they are redundant.

L178: refer to the specific section rather than saying the previous section, as the previous section describes your study areas

L183: same comment as above

L186-188: not clear what this sentence means… DTMs and DSMs are raster models – that is, gridded surfaces that are generally created through interpolation of lidar points from a point cloud. So, how can a DTM and a DSM be downloaded as a point cloud?

L212: I think it’s worth being cautious with your terminology… You can’t “subtract ground points from the non-ground points”, as they are in different locations in x-y-z space. And this process doesn’t result in a CHM, it results in a height normalized point cloud. To generate a height normalized point cloud, you subtract the elevation of all points from some interpolated surface generated from ground points. You then create a CHM by gridding that point cloud up into a raster model.

L256-268: this phenomenon was described in Campbell et al. (2018) Remote Sensing of Environment as “overall relative density” vs. “normalized relative density”.

L275: is it fair to call this “overestimation”? Do you have alternative data sources (e.g., ground measurements of stand density) to support this statement? Or is it feasible that vegetation density truly is highest near the ground level? In my experience, there are forests in which canopy density is high and understory density is low and vice versa. You seem to be assuming the former and your point cloud measurements are pointing to the latter, but you are dismissing that as overestimation…

L281-283: I’m curious why you would even need to do this step? If your ALS CHM is based on ALS DSM-ALS DTM and your SAR CHM is based on SAR DSM-ALS DTM, then why not just compare ALS DSM to SAR DSM? I suppose it wouldn’t change your results, but would be a simpler workflow…

L296-299: this is not how you described generating a CHM from ALS earlier. Please clarify.

Figure 8: Not clear what the arrows represent

 

-        I recognize that the authors are at German institutions and, as such, may not be native English speakers. So, I certainly don’t fault them for this, but issues with the writing style (grammatical errors, occasional awkward phrasing, lack of flow, etc.) detracts from the quality of the paper. The paper’s clarity and readability could be significantly improved with a careful grammatical revision.

Author Response

Please see the attachment

Author Response File: Author Response.pdf