Optimizing the Remote Detection of Tropical Rainforest Structure with Airborne Lidar: Leaf Area Profile Sensitivity to Pulse Density and Spatial Sampling
, Juliana Schietti 3, Bruce Walker Nelson 3, Carlos Alberto Silva 4, Eric Bastos Gorgens 5, Ruben Valbuena 6,7, Daniel de Almeida Papa 1,8 and Pedro Henrique Santin Brancalion 1
Round 1
Reviewer 1 Report
Summary and contribution The authors explore the effect of airborne laser scanning (ALS) pulse density and binning resolution on the estimation of leaf area index (LAI) and leaf area density (LAD) in a tropical rainforest in Brazil. They compare LAD acquired via ALS, destructive sampling, and portable canopy profiling lidar (PCL). High-density ALS was thinned to simulate ALS of differing pulse densities. The authors found that pulse density and binning resolution did indeed affect LAI and LAD estimates (Figure 5). As expected, the authors found that higher pulse densities are required to estimate LAD at finer spatial resolution, and that a spatial resolution of 10 m works well for estimating LAD with lower pulse density lidar. To the authors’ knowledge, no previous studies have investigated the effect of ALS pulse density and binning resolution on the estimation of LAI and LAD in a tropical rainforest. LAI and LAD are useful parameters for measuring forests; as such, this paper is a novel and useful contribution that merits publication. Major comments The manuscript is written fairly well written, although rewriting in several areas would make it more clear to readers. The introduction is appropriate in length and provides context for the study, the methodology is sound, and the conclusions follow from the results. Minor comments Lines 101-102: The phrase beginning with “, with higher pulse numbers...” seemed incorrect to me. Wouldn’t higher pulse density better sample the lower extent of the canopy, while the upper extent of the canopy might be sampled just as well by lower pulse density? The confusion if perhaps in not knowing what is meant by “the upper extent of the canopy”. I recommend rewriting or clarifying this statement. Lines 133-137: I recommend rewriting this run-on sentence. Line 138: “and” should be inserted after “26℃,”. Line 139: “month” should be plural. Line 163: “the” should be inserted before “R”. Line 177: Perhaps insert “horizontal” before “grain sizes” here for clarification. Line 178: I’m not sure what “Dz” stands for. I recommend clarifying. Lines 192-193: This sentence needs to be rewritten for correct grammar. Line 194: I recommend changing “can” to “could” to maintain verb tense. Equation 1: I recommend defining Dz. Lines 226-228: I recommend elaborating on the statement that “care was taken to ensure that the corresponding conditions were as close as possible”. I don’t understand what this statement means. Also, do the authors mean “component”, rather than “compartment” in this sentence? Line 229: I recommend changing “lidar surveys” to “lidar survey is”. Line 365: Starting a sentence with a bracketed number has always seemed odd to me; not sure what the publishing rules are with this. I recommend changing the beginning of this sentence to read “Kukenbrink et al. [38]”, although the publisher might disagree with this. Line 375-384: I recommend rewriting this paragraph, which is difficult to understand and has some grammatical errors. Particularly confusing is the meaning of “down-facing ALS occluded” - is there a word missing here? - as well as the very complex last sentence beginning with “Nonetheless,”. I think I follow the authors’ logic, which makes sense, but I think it could be much clearer with rewriting. Line 408: I don’t understand why the caption reads “Constant K here has a value of 5.7”, when the figure shows differing values of K ranging between 0.2 - 1.2. Figure 9 and Line 398: If this figure is adapted from McWilliam et al. 1993, this should be made more clear; I don’t understand what is meant by “(the latter from [13])”. Line 402: “specie” should be “species” Line 413: I recommend changing this sentence to read “...when evaluating ALS accuracy in tropical forests with dense understory vegetation (Figure 6)”, or something similar, since many forests with sparser understories won’t have this same issue to deal with. Line 469: “such” should be inserted before “as”
Author Response
Summary and contribution
The authors explore the effect of airborne laser scanning (ALS) pulse density and binning resolution on the estimation of leaf area index (LAI) and leaf area density (LAD) in a tropical rainforest in Brazil. They compare LAD acquired via ALS, destructive sampling, and portable canopy profiling lidar (PCL). High-density ALS was thinned to simulate ALS of differing pulse densities. The authors found that pulse density and binning resolution did indeed affect LAI and LAD estimates (Figure 5). As expected, the authors found that higher pulse densities are required to estimate LAD at finer spatial resolution, and that a spatial resolution of 10 m works well for estimating LAD with lower pulse density lidar.
To the authors’ knowledge, no previous studies have investigated the effect of ALS pulse density and binning resolution on the estimation of LAI and LAD in a tropical rainforest. LAI and LAD are useful parameters for measuring forests; as such, this paper is a novel and useful contribution that merits publication.
Major comments
The manuscript is written fairly well written, although rewriting in several areas would make it more clear to readers. The introduction is appropriate in length and provides context for the study, the methodology is sound, and the conclusions follow from the results.
Minor comments
Lines 101-102: The phrase beginning with “, with higher pulse numbers...” seemed incorrect to me. Wouldn’t higher pulse density better sample the lower extent of the canopy, while the upper extent of the canopy might be sampled just as well by lower pulse density? The confusion if perhaps in not knowing what is meant by “the upper extent of the canopy”. I recommend rewriting or clarifying this statement.
RESPONSE: We agree. We changed to “lower extent of the canopy”. Thanks for your correction.
Lines 133-137: I recommend rewriting this run-on sentence.
RESPONSE: Thanks for your suggestion. We split into two sentences and reduced the overall length.
Line 138: “and” should be inserted after “26℃,”.
Line 139: “month” should be plural.
Line 163: “the” should be inserted before “R”.
Line 177: Perhaps insert “horizontal” before “grain sizes” here for clarification.
Line 178: I’m not sure what “Dz” stands for. I recommend clarifying.
Lines 192-193: This sentence needs to be rewritten for correct grammar.
RESPONSE: Changed to: “Since we used a maximum of 5o off-nadir view in the ALS survey, we could work under the assumption that each lidar pulse is vertically incident.”
Line 194: I recommend changing “can” to “could” to maintain verb tense.
Equation 1: I recommend defining Dz.
RESPONSE: All implemented. Thanks for your corrections and suggestions.
Lines 226-228: I recommend elaborating on the statement that “care was taken to ensure that the corresponding conditions were as close as possible”. I don’t understand what this statement means. Also, do the authors mean “component”, rather than “compartment” in this sentence?
RESPONSE: Thanks for your suggestion. We have rewritten this sentence to: “Although some of the 1-ha ALS plots and PCL transects are not spatially paired, care was taken to ensure similar conditions, particularly the hillslope position (elevation).”
Line 229: I recommend changing “lidar surveys” to “lidar survey is”.
RESPONSE: Implemented.
Line 365: Starting a sentence with a bracketed number has always seemed odd to me; not sure what the publishing rules are with this. I recommend changing the beginning of this sentence to read “Kukenbrink et al. [38]”, although the publisher might disagree with this.
RESPONSE: We have rewritten these sentences so that they do not start with a bracketed reference.
Line 375-384: I recommend rewriting this paragraph, which is difficult to understand and has some grammatical errors. Particularly confusing is the meaning of “down-facing ALS occluded” - is there a word missing here? - as well as the very complex last sentence beginning with “Nonetheless,”. I think I follow the authors’ logic, which makes sense, but I think it could be much clearer with rewriting.
RESPONSE: The paragraph has been rewritten as follows:
“We here assume that LAD of a visible voxel at any given height provides an unbiased estimate of the LAD of occluded voxels at that height. In forests with many gaps the upper canopy surface, which is never occluded to the ALS, dips to lower heights. The canopy surface generally has high LAD as it is well illuminated by the sun, so these low and always visible voxels in gaps could have higher LAD than occluded voxels in the dark understory at the same height. Therefore, though forests with a lot of gaps should have a high mode of LAD in the lower mid-canopy, this mode may be exaggerated. The PCL provides a test of this bias because the LAD in shaded understory is not occluded from the perspective of the PCL. Whether using upward looking PCL or downward looking ALS, in gap-rich forest we observe an LAD profile of the same shape with mode in the same position[4]. This suggests that no significant bias is present.”
Line 408: I don’t understand why the caption reads “Constant K here has a value of 5.7”, when the figure shows differing values of K ranging between 0.2 - 1.2.
RESPONSE: Thanks for your correction. The caption now refers to a constant target LAI, not a constant K value. The full caption follows “Appropriate values for K coefficient to obtain a constant target LAI from the MacArthur-Horn equation, as a function of pulse density (x-axis) and grain size (line colors). The target LAI value was derived from direct destructive measurement (LAIsite = 5.7; [13])”
Figure 9 and Line 398: If this figure is adapted from McWilliam et al. 1993, this should be made more clear; I don’t understand what is meant by “(the latter from [13])”.
RESPONSE: The field destructive measurement was collected from McWilliam et al. 1993. We changed this sentence to:
“Figure 9 shows the appropriate K values for maintaining agreement between ALS-derived LAI and a field-measured mean LAI of 5.7 (field measurements from [13])”.
Line 402: “specie” should be “species”
Line 413: I recommend changing this sentence to read “...when evaluating ALS accuracy in tropical forests with dense understory vegetation (Figure 6)”, or something similar, since many forests with sparser understories won’t have this same issue to deal with.
Line 469: “such” should be inserted before “as”
RESPONSES: All three recommendations are implemented.
Reviewer 2 Report
The determination of a sampling grain size is effectively an important condition for accurate studies.
Line 54 : ALS comes from Airborne LiDAR Scanner in opposition to TLS for Terrestrial LiDAR Scanner. Then lidar is Li D A R for Light Detection And Ranging and we are supposed to write it LiDAR. I rather use Airborne or Terrestrial LiDAR since ALS and TLS are also brand naming conventions. But ALS is correct.
Line 58: scanline network spacing also counts... (juste a comment)
Line 70: I agree that non-destructive techniques are preferable but it is very interesting to evaluate the robustness of your technique with real data.
Line 83-97: Of course there is always simplifications so it is effectively important to recall them. Any local concentration can change the results which explain why we should always presents next to their basic assumptions. But other studies using ray tracing exist.
Line 98: trees with strong density of leaves at the top of their canopy exit. There is no universal technique. So, even LiDAR can fail to detect all the volume of a canopy form the top like from the bottom. It depends on the leave distribution.
Line 103: of course horizontal binning resolution counts like the vertical as stated by your reference to the voxel size.
Line 104: “inherent relationship between lidar pulse density and the horizontal” is also applicable to vertical.
The grain size as to be compatible with the leave density. It will change from tree studies to lower plant studies. It is a basic sampling theorem. You are looking for a grain size guarantying unbiased results. That’s all.
Line 108: OK clumping effect is effectively important.
Line 110-114: OK. So you are looking for the best grain size sampling in tropical forest and its typical leave density distribution.
Line 117: the destructive surveys is one of the main interest of your work which is based on real data. Experimental remote sensing is not so common so I would support your work.
Line 118: PCL is a Terrestrial LiDAR
Line 125: repetition of line 110?
Line 152: foot print are important but down track spacing is also a crucial parameter for airborne studies. It can be 10 times the footprint. A FOV of 10° is quite narrow but very good to insure NADIR data.
Line 156: 42 pulses m-2 OK, but along scanlines with which spacing?
Line 158: 77% first returns imply a poor penetration in the canopy.
Line 187: So finally you are working with full voxel (so what was your point in line 104?).
Line 201: This K coefficient looks arbitrary at this stage. Can’t you estimate it with your destructive data?
Figure 4: There is a lot of laser beam interception at the top of the canopy and a lot less laser beam passing down. Don’t you have an effect on the data? I guess that the main interest of complementary TLS data. It would be interesting to have a similar figure for TLS data.
Figure 5: So I guess that the mean size of tree crown is 10m…
Figure 6: % of LAI are stable and the best fitting between ALS, TLS and destructive data is at 5m
Why did you put all results description in discussion? Comments on results are not a discussion but a presentation of results.
Line 309: 15, 15, and 10 … ?
Line 310: The description of a stabilization at 10m is a result not a discussion.
Line 313: What do you mean?
Line 314: Better agreement … is also an important result which could be given next to Figure 6 in results. Than 5m seems to be the best grain size.
Line 320-325: Why don’t you consider the typical size of tree crown? Can we say that tree crown have a mean size between 5m and 10m?
Line 354: occluded voxels are inevitable…
Line 385: Calibrating the K coefficient seems to be more important
Line 398: K=5.7 is known since [13] so why don’t you use it earlier? You could have used this value in all your work instead of 1. So can you explain why did you choose to set K to 1 and only tell us that 5.7 would be the best value in discussion?
Line 412-426: lower vegetation below tree canopy can bias the results below 4m. Is it right?
Line 433: your choice of 10° FOV (field of view) was perfectly right for unbiased voxel calculation.
Line 436: there is not only footprint but scanline spacing effects…
Line 441: I would definitely not recommend a FOV of 60°. This is used in town to digitize building walls with large over lapping between flight line for 3D city modelling. A FOV below 10° should be a recommendation of your work. The quasi perfect match between ALS and TLS data in Figure 6 grain = 5m is the most impressive result for me. Why don’t you insist on this result?
Line 464-468: Your scan angle is right and give very good results in Figure 6 grain 5m (it would have been easier to have a reference of panel c). About flight parameter just remember that they can be set in function of your ground experiments. It should work in this way, from ground experiment based on real data to flight setting insuring unbiased results.
Line 469-473: And multispectral LiDAR are also very interesting …
Line 474-482: I would also recommend to have a look to full-waveform LiDAR which are very promising
Author Response
The determination of a sampling grain size is effectively an important condition for accurate studies.
Line 54 : ALS comes from Airborne LiDAR Scanner in opposition to TLS for Terrestrial LiDAR Scanner. Then lidar is Li D A R for Light Detection And Ranging and we are supposed to write it LiDAR. I rather use Airborne or Terrestrial LiDAR since ALS and TLS are also brand naming conventions. But ALS is correct.
RESPONSE: We agree that a single convention is desirable. “LiDAR”, “LIDAR” and “lidar” are all now commonly seen in the literature. Among these "lidar" is most consistent with other publications of our group, such as Shao et al., (2019). The terrestrial Portable Canopy profiling Lidar (PCL; a 2D profiling scanner) used in this study is different from a TLS (3D terrestrial laser scanner). For more details about PCL please see Parker et al., 2004; Hardman 2011, 2013; Stark et al., 2012; Almeida et al., 2016.
Shao, G.; Stark, S. C.; de Almeida, D. R. A.; Smith, M. N. Towards high throughput assessment of canopy dynamics: The estimation of leaf area structure in Amazonian forests with multitemporal multi-sensor airborne lidar. Remote Sens. Environ. 2019, doi:10.1016/j.rse.2018.10.035.
Parker, G. G.; Harding, D. J.; Berger, M. L. A portable LIDAR system for rapid determination of forest canopy structure. J. Appl. Ecol. 2004, 41, 755–767, doi:10.1111/j.0021-8901.2004.00925.x.
Hardiman, B. S.; Bohrer, G.; Gough, C. M.; Vogel, C. S.; Curtis, P. S. The role of canopy structural complexity in wood net primary production of a maturing northern deciduous forest. Ecology 2011, 92, 1818–1827, doi:10.1890/10-2192.1.
Hardiman, B. S.; Gough, C. M.; Halperin, A.; Hofmeister, K. L.; Nave, L. E.; Bohrer, G.; Curtis, P. S. Maintaining high rates of carbon storage in old forests: A mechanism linking canopy structure to forest function. For. Ecol. Manage. 2013, doi:10.1016/j.foreco.2013.02.031.
Stark, S. C.; Leitold, V.; Wu, J. L.; Hunter, M. O.; de Castilho, C. V.; Costa, F. R. C.; Mcmahon, S. M.; Parker, G. G.; Shimabukuro, M. T.; Lefsky, M. A.; Keller, M.; Alves, L. F.; Schietti, J.; Shimabukuro, Y. E.; Brandão, D. O.; Woodcock, T. K.; Higuchi, N.; de Camargo, P. B.; de Oliveira, R. C.; Saleska, S. R. Amazon forest carbon dynamics predicted by profiles of canopy leaf area and light environment. Ecol. Lett. 2012, 15, 1406–1414, doi:10.1111/j.1461-0248.2012.01864.x.
Almeida, D. R. A. de; Nelson, B. W.; Schietti, J.; Gorgens, E. B.; Resende, A. F.; Stark, S. C.; Valbuena, R. Contrasting fire damage and fire susceptibility between seasonally flooded forest and upland forest in the Central Amazon using portable profiling LiDAR. Remote Sens. Environ. 2016, 184, 153–160, doi:10.1016/j.rse.2016.06.017.
Line 58: scanline network spacing also counts... (juste a comment)
RESPONSE: Thanks.
Line 70: I agree that non-destructive techniques are preferable but it is very interesting to evaluate the robustness of your technique with real data.
RESPONSE: Yes, we fully agree that destructive validation of LAD and LAI for long forest profiles would be useful. This would, however, be a massive undertaking and is beyond the specific objectives of this paper.
Line 83-97: Of course there is always simplifications so it is effectively important to recall them. Any local concentration can change the results which explain why we should always presents next to their basic assumptions. But other studies using ray tracing exist.
RESPONSE: Yes, we agree. An example of simplification in our study is the acceptance of off-nadir scan angles up to 5 degrees, yet still treating all pulse traces as being vertical and parallel, as assumed by the MacArthur Horn equation. Thanks for your comment.
Line 98: trees with strong density of leaves at the top of their canopy exit. There is no universal technique. So, even LiDAR can fail to detect all the volume of a canopy form the top like from the bottom. It depends on the leave distribution.
RESPONSE: We agree, that’s for your comment.
Line 103: of course horizontal binning resolution counts like the vertical as stated by your reference to the voxel size.
RESPONSE: Actually, there is also a small effect of Dz (voxel vertical resolution). At beginning, we also examined this effect (Figure below). However, given the advantage of standardizing to 1-m Dz (higher resolution), we decided to remove these analyses.
Figure 4 – *please, see attached file.
Line 104: “inherent relationship between lidar pulse density and the horizontal” is also applicable to vertical.
RESPONSE: Yes, please see response above.
The grain size as to be compatible with the leave density. It will change from tree studies to lower plant studies. It is a basic sampling theorem. You are looking for a grain size guarantying unbiased results. That’s all.
Line 108: OK clumping effect is effectively important.
Line 110-114: OK. So you are looking for the best grain size sampling in tropical forest and its typical leave density distribution.
RESPONSE: Yes, we agree with all your comments. Thanks.
Line 117: the destructive surveys is one of the main interest of your work which is based on real data. Experimental remote sensing is not so common so I would support your work.
Line 118: PCL is a Terrestrial LiDAR
RESPONSES (Lines 117 & 118): Yes, the destructive sampling of LAI, which is from reference [13] is extremely important for calibrating both the ALS (as shown in our Figure 9, for example) and the PCL for obtaining LAI. We also suggest that once the PCL is calibrated, with its very high pulse density it can then serve as a kind of LAI calibration standard itself. Note also that the terrestrial Portable Canopy profiling Lidar (PLC; 2D profiling scanner) used in this study is different from the more widely used TLS (3D terrestrial laser scanner). For more details about PCL please see Parker et al., 2004; Hardman 2011, 2013; Stark et al., 2012; Almeida et al., 2016.
Line 125: repetition of line 110?
RESPONSE: There is some information overlap, but at the line 110 we are introducing and justifying our study subjects while at the line 125 (closing paragraph of the Introduction) we are succinctly listing our objectives, before moving on to the Methods.
Line 152: foot print are important but down track spacing is also a crucial parameter for airborne studies. It can be 10 times the footprint. A FOV of 10° is quite narrow but very good to insure NADIR data.
RESPONSE: Yes, the along-track spacing of footprints is taken into consideration during the data provider’s flight planning to obtain, ideally, a homogeneous spacing of pulse returns in the along-track dimension. We also rejected some non-homogeneous portions of the coverage. It is also good to know that our off-nadir threshold is well within the recommendation of the reviewer. Thanks.
Line 156: 42 pulses m-2 OK, but along scanlines with which spacing?
RESPONSE: The flight planning takes this into account to provide homogenous pulse densities in both along-track and cross-track directions. When choosing our 1ha plots we also checked for homogeneity. Due to this, some PCL transects, and ALS plots are not perfectly spatially paired (see Figure 1)
Line 158: 77% first returns imply a poor penetration in the canopy.
RESPONSE: We had ground-returns with a mean density of 0.9 returns m-2, which is reasonable for a good estimation of the Digital Terrain Model. We address the problem of occlusion at some length in the Discussion. There we note that both upward-looking PCL and downward looking ALS gave very similar Leaf Area Density profiles, suggesting that there was no appreciable bias in our method for dealing with occluded voxels.
Line 187: So finally you are working with full voxel (so what was your point in line 104?).
RESPONSE: Line 187 is a figure. It’s purpose is merely to depict the voxel concept and the concept of horizontal binning resolution.
We have changed line 104 to: “There is an inherent relationship between lidar pulse returns per voxel and the horizontal grain size”. Previously the underlined part of the sentence read “lidar pulse density”. Thank you for pointing out this discrepancy.
Line 201: This K coefficient looks arbitrary at this stage. Can’t you estimate it with your destructive data?
RESPONSE: We used K=1 for simplicity and consistency when examining the effect of bin size and pulse density on LAD and LAI (Figure 5). In the Discussion we turn this around (Figure 9) and examine how, in order to obtain a fixed target LAI of 5.7 (which is from the destructive data) different value of K are required, depending again on the bin size and return density. Finally, in the main part of the paper, we show how the problem of choosing a proper K-value (because it is sensitive to grain size and pulse density) can be largely attenuated if we express the LAD in each segment of a canopy profile as a percent of total LAI in that profile (Figure 6).
Figure 4: There is a lot of laser beam interception at the top of the canopy and a lot less laser beam passing down. Don’t you have an effect on the data? I guess that the main interest of complementary TLS data. It would be interesting to have a similar figure for TLS data.
RESPONSE: Figure 4 is only an example. What happens with PCL is exactly the opposite having voxels with NA values at the top of the canopy.
Figure 5: So I guess that the mean size of tree crown is 10m…
RESPONSE: This is an intriguing hypothesis. However, our objective, as expressed in the results shown in Figure 5 is to provide an empirical report on effects of grain size and pulse density on LAD profiles using a fixed K in the MacArthur Horn equation. We do cover some possible causes in the Discussion, but this paper does not aspire to formally tease apart the different possible mechanisms underlying the changes in mean LAD profile shapes with spatial scale and pulse density that are shown in Figure 5. These mechanisms may include canopy structure effects, such as branch scale clumping, crown scale clumping, and the changing relative importance of these two with height in the canopy.
Figure 6: % of LAI are stable and the best fitting between ALS, TLS and destructive data is at 5m
RESPONSE: Yes, though considering the uncertainty, all estimates at grain sizes <= 5m show good agreement between these three data sources for LAD profiles.
Why did you put all results description in discussion? Comments on results are not a discussion but a presentation of results.
RESPONSE: This is a matter of style. We follow the example of other authors who report their results in a succinct manner and keep all (or most) explanation of the possible causes of those results in the Discussion. As noted above, the paper’s objectives do not include an attempt at a formal breakdown of the roles of different mechanisms behind the patterns in Figures 5 a 6. Therefore, we have moved all discussion of these possible mechanisms into the Discussion section.
Line 309: 15, 15, and 10 … ?
RESPONSE: The full sentence is: “At small grain sizes of 1m, 2m, and 5m, absolute LAD profiles became stable at respective pulse densities of 15, 15, and 10 pulses m-2.” This should be clear.
Line 310: The description of a stabilization at 10m is a result not a discussion.
RESPONSE: In this part of the Discussion we are pointing out the best combinations of grain size and pulse density, so this is an interpretation of the results and a recommendation based on those results.
Line 313: What do you mean?
RESPONSE: “Sweet spot” is an idiomatic expression. We have changed to "ideal combination”. Thank you for pointing this out.
Line 314: Better agreement … is also an important result which could be given next to Figure 6 in results. Than 5m seems to be the best grain size.
RESPONSE: We agree. As mentioned earlier all grain sizes <= 5m were consistent with the profile obtained from destructively measurement. So have cnanged “less than 10m” to “<= 5m”.
Line 320-325: Why don’t you consider the typical size of tree crown? Can we say that tree crown have a mean size between 5m and 10m?
RESPONSE: Both tree crowns and within these, the branches, cause clumping of leaves. We did not investigate which is most important, nor how their relative importance (and size) change with height in the canopy. So we are reluctant to draw a conclusive inference about crown size and its effect on clumping artifacts.
Line 354: occluded voxels are inevitable…
RESPONSE: We agree, but it seems to also be important to discuss their role. Here we review the possibility that our manner of dealing with occluded ALS voxels might cause bias (an overestimate of understory LAD), and some evidence we have that this bias is small, namely that upward-looking PCL gives LAD profiles with the same shape as the LAD profiles from downward-looking ALS.
Line 385: Calibrating the K coefficient seems to be more important
RESPONSE: Thanks for your comment.
Line 398: K=5.7 is known since [13] so why don’t you use it earlier? You could have used this value in all your work instead of 1. So can you explain why did you choose to set K to 1 and only tell us that 5.7 would be the best value in discussion?
RESPONSE: Sorry if we were unclear. K is not equal to 5.7. The expected mean LAI = 5.7. Using K=1 gives an LAI value near to 5.7 when grain size is large, as can be seen in our Figure 9. For clarity, we have rewritten “Figure 9 shows the appropriate K values for maintaining agreement between ALS-derived LAI and a field-measured mean LAI of 5.7 (the latter from field measurements from [13]). For large grain sizes, a K value close to 1.0 gives the expected field-derived LAI value, independent of pulse density. Calibrated K is clearly sensitive to both grain size and pulse density and much more so at lower grain sizes and lower pulse densities.”
We have also added a sentence to the Methods section where we introduce K, “The K coefficient effect is addressed in detail in the discussion section. Suffice it to say here that K=1 does produce an LAI close to the field measured LAI, when grain size is large.”
Line 412-426: lower vegetation below tree canopy can bias the results below 4m. Is it right?
RESPONSE: Yes, ground returns and voxel occlusion could affect LAD profiles ALS-derived at low canopy heights.
Line 433: your choice of 10° FOV (field of view) was perfectly right for unbiased voxel calculation.
RESPONSE: Thanks.
Line 436: there is not only footprint but scanline spacing effects…
RESPONSE: Thanks for adding this. We have rewritten as “Many other factors (e.g., instrument type, beam divergence, scan angle, flight altitude, along-track and cross-track footprint spacing).”
Line 441: I would definitely not recommend a FOV of 60°. This is used in town to digitize building walls with large over lapping between flight line for 3D city modelling. A FOV below 10° should be a recommendation of your work. The quasi perfect match between ALS and TLS data in Figure 6 grain = 5m is the most impressive result for me. Why don’t you insist on this result?
RESPONSE: Yes, we agree. But, since this paper did not test scan angle effects we cannot say here that FOV of 10° or less is correct. Nonetheless, thanks for your comment. We are happy to see that our of nadir threshold matches with the reviewer’s experience.
Line 464-468: Your scan angle is right and give very good results in Figure 6 grain 5m (it would have been easier to have a reference of panel c). About flight parameter just remember that they can be set in function of your ground experiments. It should work in this way, from ground experiment based on real data to flight setting insuring unbiased results.
RESPONSE: As noted above, though our scan angle limit gave good results we do not know about the effect of other scan angle values. We intend to produce another study investigating this issue. Thanks for your comment.
Line 469-473: And multispectral LiDAR are also very interesting …
RESPONSE: Yes, multispectral lidar is very interesting, useful for land use classification.
Line 474-482: I would also recommend to have a look to full-waveform LiDAR which are very promising
RESPONSE: Thanks, we included two refences to studies using full wave form for LAI estimates.
Author Response File: 
Author Response.docx
