Remote Sensing of Pigment Content at a Leaf Scale: Comparison among Some Specular Removal and Specular Resistance Methods
Round 1
Reviewer 1 Report
In this manuscript, means to compensate for effects of leaf reflectance anisotropy on leaf chlorophyll and carotenoid content retrieval using leaf spectral reflectance are evaluated. Four main approaches (and hybrid methods comprising more than one of these methods) are compared: (a) vegetation indices specifically designed to mitigate impacts of leaf surface reflectance on the relationship between leaf pigments and its spectral response; (b) derivative spectra; (c) continuous wavelet transform; and (d) Physical based approaches. These approaches were compared with (e) raw reflectance information and multiple linear regression was used to select features and to perform final estimates for multivariate implementations for items (a-c and e). This material present valid contribution to the field and considerable changes were made from the first material presented for publication. However, some doubts still remains considering the outputs presented. In particular, there is still a concern regarding at which extent the physical and empirical methods are comparable under the modelling scheme adopted in the current version. Since no spectral band selection or comparable multivariate modelling approach is adopted for the physical based retrieval methods their performance might be compromised. Also, the size of the dataset might have affected the outputs obtained, and this aspect should be better discussed. In addition, the text should be revised concerning the language (I have given some opinions but I am not native speaker, someone else should be consulted) and specially the introduction should be better formulated. Some further specific recommendations follows below:
Title: replace “Remote...” by ‘Proximal...’
Title: replace “...among some...” by ‘...of selected...’.
Title: replace “...specular resistance...” by ‘...specular resistant...’
Abstract:
Line 14: replace “...is disturbed...” by ‘... can be negatively affected...’
Line 14-15: replace “To alleviate this disturbance,...” by ‘To mitigate the influence of this factor,...’
Line 15-16: replace “...some specific techniques that consider the specular reflectance or specular effect have been proposed”. by ‘...some specific techniques taking into account these effects have been proposed’.
Line 16-19: Indicate all the approaches used, not only a few of them.
Line 19-20: Some RT approaches presented relatively good accuracy, please indicate the specific cases in which prediction was considerably affected by off-nadir measurements.
Line 21-23: What criteria indicate if an approach is effective or not? Improvements corresponding to increase in R2 values from 0.95 to 0.99? Indicate the levels of retrieval performance (R2 or RMSE) in each case, perhaps including relative error and a short definition for this concept. This might help the reader to have an idea about the level of accuracy and the improvements reached by each method tested without the necessity to look in the text.
Line 23-27: It might be useful to indicate the size of the dataset used, and the fact that in such small dataset multivariate approaches have a certain advantage over RTM-based methods, since they can ‘fit very well’ the specific data being modelled.
Line 29-31: Indicating the narrow and broad band regions might be of interest to the reader.
Keywords: I would remove ‘interference alleviation’ from the list for being too broad.
Line 38 (suggestion): replace “...to nitrogen,...” by ‘...to nitrogen content,...’
Line 38-39 (suggestion): replace ”...plant nitrogen content...” by ‘...plant nitrogen nutrition status...’
Line 39: I would suggest to change “Car also contributes to light-harvesting for photosynthesis, and can protect the photosynthetic system by dissipating the excessive light energy when the energy exceeds the needed [3]”. by ‘Car also contributes to general photosynthesis process, protecting the photosynthetic system by dissipating the excessive light energy when the energy exceeds the needed [3]’. In this case I would recommend the following reference to justify this affirmation: 10.1016/j.rse.2008.10.019.
Line 42: replace “...measurement of pigment content...” by ‘...estimation of pigment content...’.
Line 44: replace “One method to measure pigment content is remote sensing”. by ‘A possible alternative for pigment estimation is remote sensing’.
Line 45: replace “So it can monitor...” by ‘So it can be used for monitoring...’.
Line 61: replace “In comparison...” by ‘In contrast...’.
Line 68-70: replace “The specular removal type indicates in these approaches the specular reflectance or specular effect is explicitly estimated (accurately or roughly), and then is removed or separated out”. by ‘The specular removal type comprises explicit estimation (accurately or roughly) of specular reflectance or specular effects, which are removed or separated from the diffuse component’.
Line 70-72: replace “On the contrary...” by ‘On the other hand,...’
Line 74: replace “...estimated...” by ‘...estimate...’
Line 76: include ‘In this case,...’ before “The specular effect...”
Line 86: replace “...in retrieval”. by ‘...in the retrieval’.
Line 90: replace “So they may be called the specular disturbance alleviation techniques in general”. by ‘So they may be generally called specular disturbance alleviation techniques’.
Line 103-104: replace “...under two different specular circumstances, the conventional nadir and an oblique viewing direction”. by ‘...under two different viewing geometries, the conventional nadir and an oblique orientation, resulting in distinct specular effects on the reflectance measurements’.
Line 104-105: replace “Specifically, the performance of wavelet analysis in specular disturbance alleviation was focused. By ‘Special attention was given for evaluating the performance of wavelet analysis in specular disturbance alleviation’.
Line 108: replace “The experiment was detailed in our other paper [17]. So it was just outlined here.” by ‘The dataset used was obtained in an experiment already described in another study [17]. So it is just outlined here’.
Line 112: replace “...had...” by ’...has...’ (and so forward for general aspects of leaves.)
Line 116: More details about the measurements should be provided (e.g.: distance to the samples (leaves), area of the leaf measured (IFOW of the sensor), if measurements included central leaf veins, etc.)
Line 155: replace “...in comparison with [26,36,44]”. by ’...similarly to other studies [26,36,44].’
Line 360: replace “...the retrieval with PROSPECT-D decreased a lot compared to the nadir...” by ’... the retrieval accuracy with PROSPECT-D decreased considerably in comparison with measurements in the nadir...’
Line 361: replace “All other models improved the retrieval...” by ’ All other approaches had a better retrieval performance...’
Line 363: check possible double spacing between “...35° and...”
Line 445: replace “...high errors...” by ‘...relatively high errors...’
Line 447-448: what do you mean by “At such a high angle (> 20°), BRF was highly anisotropic in directions”. Should probably be rephrased to convey the idea that with such contrasting view angles the anisotropic behaviour of leaf reflectance was considerable.
Line 451: replace “...so the retrieval was terrible...” by ‘...so the retrieval performance is expected to be poor...’
Figure 5: adding the corresponding graphs for the best approaches in each case may be of interest to illustrate how these methods improved the prediction accuracy.
Line 470-471: What do you mean by “...(may also together with wavelengths)...”?
Line 470-472: If the model coefficients and wavelengths adjustment provided the potential better prediction it means that fitting individual models for each species may result in more comparable outputs between RTM and empirical methods. This should be considered even if it cannot be tested due to the reduced amount of data. Also, what are the main differences between species that make the specular reflectance behaviour so characteristic? This aspect is not taken into account in the RTM models? For example, if the bspec parameter is adjusted for each species considering a different interval of possible values, results would not be better? Those aspects, coupled with the small size of the dataset, which benefits empirical approaches, should be taken into account throughout the paper in order to provide truly comparable conditions for both cases. Also, selecting spectral bands and fitting multivariate models to simulated data is possible (e.g.: [1]). Should not this be taken into account as well since using the complete spectra against selected wavelengths might be not completely fair. Please indicate how these points are tackled in the current version, and if not what are your propositions to make a more meaningful comparison between these approaches in this paper or next studies.
Line 546: replace “...in a direction...” by ’...in a given direction...’
Line 547-548: what do you mean by: “...and the adj. R² values were very close for a pigment in a direction”.
Line 549-552: Selection of regions less affected by specular effects when necessary may be the main reason? With alterations in the number of cases from each species in the calibration and validation dataset in the k-fold modelling scheme, especially considering the small number of samples. This should be considered in other parts of the discussion as well, if the selection of data for cross validation did not followed the proportional number of observations for each species.
Line 591-596: Please consider all points indicated in the comment concerning Lines 470-472. Not only a larger dataset should be considered, but also RTM approaches could be more specific, with wavelength selection, inversion based on multivariate approach, etc.
Author Response
Thank you very much for your time and consideration on our manuscript. Your comments are greatly valuable to improve our work. We revisied the manuscript accordingly. We hope this version will meet the standard of publication in Remote Sensing.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The manuscript has been greatly improved compared with an earlier version. I would like to thank the efforts that the authors have made. I only have a few minor comments.
Line 39: Delete “also”.
Line 42: Change “fields” to “applications”.
Line 45: Delete “So”.
Line 59: Replace “The incident radiation” with “The reflected radiation”.
Line 86, 212, 429: Change “erase” to one of these words “remove/minimize/reduce”.
Line 92: Remove “on the same data set”.
Line 149: Change “advised” to “suggested”.
Line 451: Change “terrible” to “poor”.
Line 455: Delete “certainly”.
Line 593: Delete “just”.
Author Response
Thank you very much for your time and consideration on our manuscript. Your comments are greatly valuable to improve our work. We revisied the manuscript accordingly. We hope this version will meet the standard of publication in Remote Sensing.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments concerning manuscript 443778:
This review is for the manuscript entitled “Remote retrieval of Leaf Pigment Content: Comparison among Some Specular Removal and Specular Resistance Methods”. The research concerns “remote” sensing of certain leaf pigment content via methods that purport to alleviate a supposed confusion in content caused by specular reflectance.
First, the paper is in serious need of rewriting due to the poor “English” that causes awkward sentences that are difficult to understand. In addition, numerous examples of improper tense for verbs adds to the difficulty. The manuscript that I received consists of text of different colors (black, green, and red) that also makes for difficult reading. All of this then makes it very difficult to ascertain the worth of the effort.
A few specific examples:
Line 14 – I think that the word “disturbed” would be better changed to “modified” or “contaminated”
Line 16 – the sentence beginning on this line is awkware/incomplete
Line 25 – the acronyms Chl and Car are not defined
Line 28 – “the first derivative” of what?
Line 29 – the sentence that begins here is awkward and not understandable…
Line 40 – last word “needed” – needed what?
Line 46 – the phrase “extended at a spatial scale” has no meaning…
Line 73 – the two acronyms mSR705 and mND705 are not defined…
And, so on…
It would also help the reader if an example of explicit significant contamination of a reflectance measurement is shown and, then, use that example to show how the current measurements also suffer such a contamination.
Author Response
Thank you very much for your time and consideration on our manuscript. Your comments are greatly valuable to improve our work. We revisied the manuscript accordingly. We hope this version will meet the standard of publication in Remote Sensing.
Author Response File: Author Response.pdf
Reviewer 4 Report
The approach is very interesting , however, the authors start at the knowledge that was published one or two decades ago. Add recent literature.
The language needs severe revision, e.g. "comparable" are all data, but you probably mean "similar".
The presentation of results doesn't show the quality necessary for an international, science journal. All figures need revision. Recheck the tables: Data should be presented to the point, not simply showing all data available.
Author Response
Thank you very much for your time and consideration on our manuscript. Your comments are greatly valuable to improve our work. We revisied the manuscript accordingly. We hope this version will meet the standard of publication in Remote Sensing.Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Considering the changes implemented so far, in comparison with the first version submitted, the manuscript content considerably improved. I have made some suggestions regarding the comparability between the RTM-based retrieval and empirical methods implemented by the authors, notably those relying on MLR. The authors have chosen to follow an standard procedure for retrieval based on RTM and MLR, which differ considerably between methods - LUT-based inversion in the RTM case and model fitting in the case of MLR. Therefore they compare an approach that is not optimized for each species (RTM in the general configuration adopted by the authors) with another that allows a better fitting to the specific dataset available.
We understand the choice made (evaluate the methods under a general application scenario) and accept the current implementation. We also understand that most of the differences between the retrieval methods outputs might be related to limitations on the RTMs to model off-nadir information (according to the authors, I am not an expert on this topic). Therefore I agrre in general with the answers given by the authors although some small final adjustments should be made in my opinion to improve the results presentation. More specifically, in figure 5, besides the RTM inversion outputs, results for the optimal methods obtained (e.g., RTM + correction based on R455, MLR with CWD values, etc.) should be provided for a better understand on how the improvements occured (it was for all species?). Beside that, concerning corrections proposed in the abstract, the authors indicate that they could not define which spectral regions influenced more the predictions. However I still think that it is feasible to at least give a broad indication since the authors have found some specific scales to be more important, as well as regions with high correlation with a given pigment (e.g. Figure 8). However we leave that to the authors to judge the pertinence. In addition, we still recommend a good language review for the complete document.
Best regards,
Author Response
Thank you very much for your detailed reviewing and comments on our manuscript. The comments are greatly valuable to improve our manuscript. We studied them carefully and revised the manuscript accordingly.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments concerning second review of manuscript 443778:
This is a review of the second version of the manuscript titled: “Remote Sensing of Leaf Pigment Content: Comparison among Some Specular Removal and Specular Resistance Methods”. Again, the authors are apparently attempting to show how many different methods are able, or unable, to discern the diffuse reflectance of three different species of leaves by overcoming the specular component of the reflectance that is inherent in most reflectance measurements (if I understand properly). This version is not much different from the earlier version that I reviewed regarding the proper use of “English” as well as “organization”. It would seem as though there should be some way in which to report the results that greatly decreases the page count of this manuscript. In any case, the “English” issues still make it difficult to understand the manuscript. Some of the specific comments in my earlier review were ignored; I will attempt to list a few line numbers where there are issues. I will not go to further detail because to do so would take an inordinate amount of time and it is not known to what avail would be made of this detail.
Line 20-21 – awkward sentence
Line 26
Line 39-41 – incomplete sentence..
Line 42 – awkward …
Line 46 – meaning of terminology…
Line 55 – sentence refers to what?
Line 67-71 – awkward sentence and improper tenses…
Line 73 – undefined terminology – for example, what is mSR705?
Line 93 – missing article…
…several other similar examples…
Line 182 – sentence is awkward and does not make sense…
Line 214-6 – sentence is more of a fragment and seems out of place..
And many more. The reference list contains some typographical errors (paper titles capitalized sometimes and not sometimes…)
Author Response
Thank you very much for your time and consideration on our manuscript. Your comments are greatly valuable to improve our manuscript. We revisied the manuscript accordingly. We hope this version will meet the standard of publication in Remote Sensing.
Author Response File: Author Response.pdf
Reviewer 4 Report
Some contents were improved in the revision. Some language issues remain. Consequently the support of a native speaker is recommended. In example: "comparable" are all data that can reasonably be compared. While "similar" is the term you are looking for.
The amount of data presented could be reduced to the point, but may be okay in an exclusively electronic publishing.
Author Response
Thank you very much for your time and consideration on our manuscript. Your comments are greatly valuable to improve our manuscript. We revisied the manuscript accordingly. We hope this version will meet the standard of publication in Remote Sensing.
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
Review
This manuscript describes results of statistical modelling of leaf chlorophyll and carotenoid content from leaf-level spectral measurements taken from the nadir and off-nadir (specular) position. The authors tested several statistical approaches including five vegetation indices and stepwise multiple linear regression models that used full spectral between 400-1000nm, their first derivative and continuous wavelet transformation.
Understanding of leaf specular reflectance behaviour is interesting and important for remote sensing community, yet this paper does not add that much new knowledge.
This paper is just another from many that tested different statistical models to predict leaf pigments from leaf-level spectra. Their results showed that vegetation indices were outperformed by multiple linear regression models. Between the models using either first derivative or continuous wavelet transformation were almost no differences and there were also negligible differences in RMSE obtained from nadir and specular reflectance measurements. I find it a bit uneasy to give conclusions based on 68 leaf samples collected only from three species and on statistical models build from nadir and off-nadir measurements.
The title looks a bit misleading as it promises removal of interference of specular reflectance ... I would just say that the approaches tested here are insensitive to specular behaviour as no removal technique is actually proposed.
Paper Li et al. (in press), to which you are heavily referring to, should have been made available to the reviewers. Details of the experiment are described there, as well as some results that are relevant for this paper. It's hard to judge what will be published in Li et al. (in press) and how it is different from this manuscript.
The paper is not even listed in the references, no title, no journal where the readers can look for it. If it is in press, likely it will be online soon. Provide the proper reference and more details in the corresponding sections of your manuscript.
2.1 Experiments - Specify here how many samples did you measure. Suddenly at line 150 you tell something about 50 and 68 samples and finally in Table 1 I can see that you measured in total 68 samples.
Illustration or photo with the measurement set-up would be useful.
How far away was the ASD sensor from the leaf, what was the field of view?
Did you apply some post-processing to the spectra, smoothing?
2.2.3 Indices - please, clarify which specific wavelengths were used to compute the vegetation indices when only indication NIR, red, blue was used. Was it a single wavelength or average over a broader region?
Rdiff-reSR index ... how the diffuse component was extracted? Please describe.
2. Methods in general, the paper should be about the removal of specular effects, yet in the method section there is nothing described in this sense. So, how did you remove the specular effects?
Computing several linear models with different spectral inputs (nadir vs. specular reflectance) is not according to my opinion the removal of the specular effects, it is just comparison of two types of input data.
3. Results
Table 1. I was surprised to see that some ginkgo samples had very low Chl content. These must have been yellow leaves. Did you sample the leaves at the end of their vegetation season? Please, explain.
Lines 201-203. Starting from "Based on these ... " can be removed until the end of the paragraph. It seems inappropriate here.
Figure 2. All Mucunna samples behaved like this that the specular reflectance was so much different from the nadir one? It looks like wrongly measured spectrum.
Please, provide a new figure (with 3 subplots for each species) with mean and standard deviation for measured nadir and specular reflectances of all samples per species.
Tables 3-6, provide also the value of average R2 for 10 repetitions.
Table A1. It is not clear what "entire" dataset means and how the consistency was computed.
Author Response
We deeply appreciate your detailed and constructive comments which improve our manuscript greatly. We carefully studied these comments and tried our best to revise the manuscript accordingly. We think this new version should meet the standard of publication in Remote Sensing. Thank your again for your time and detailed reviewing. Look forward to hearing from you soon.
Author Response File: Author Response.pdf
Reviewer 2 Report
In this manuscript, means to compensate for effects of leaf reflectance anisotropy on leaf chlorophyll and carotenoid content retrieval using leaf spectral reflectance are evaluated. Three main approaches are compared: (a) vegetation indices specifically designed to mitigate impacts of leaf surface reflectance on the relationship between leaf pigments and its spectral response; (b) derivative spectra; and (c) continuous wavelet transform. These approaches were compared with (d) raw reflectance values and multiple linear regression was used to select features and to perform final estimates for multivariate implementations (b-d). This material present valid contribution to the field, however overlap with already existing articles needs to be further evaluated. In any case, some modifications are recommended before a potential publication. In particular, some authors indicate that additive and multiplicative scattering effects may potentially be corrected using some relatively simpler spectral pre-treatments [1,2], which motivated in some cases vegetation indices formulations. The authors might consider to add some additional spectral pre-treatments for comparison (e.g., Standard Normal Variate, SNV; Barnes et al., [3]). Besides that, the dataset used in this research is relatively limited which demands some careful consideration in terms of the approach used for evaluating the models performance. In this case, instead of using a single division between calibration and validation dataset, maybe a repeated k-fold (e.g., 100 times 10-fold validation) approach might be implemented to avoid basing all conclusions on a single or on only a few random draw(s) from such a small dataset. Also, the text should be revised concerning the language (I give some opinions but I am not native speaker, someone else should be consulted) and specially the introduction should be better formulated. Some further specific recommendations follows below.
Abstract
Line 14: Define mirror-like. For example: ”in an angle similar to that described by the illumination geometry but in the forward direction in the principal plane.”, or something similar.
Line17-25: Include RMSE values, with units for each pigment and method cited. It might help the reader to identify the improvements made by the methods tested. Indicate the baseline prediction accuracy as well.
Introduction
Line 30-31: These two paragraphs may be better merged together. Instead of: “Leaf chlorophyll (Chl) and carotenoid (Car) is of great importance for green plants. Chl plays a crucial role in plant photosynthesis which directly determines plant productivity [1,2].”; use: “Leaf chlorophyll (Chl) and carotenoid (Car) are of great importance for green plants. Chl plays a crucial role in plant photosynthesis which and directly affect determines plant productivity [1,2].”
Line 31-32: Is the amount of nitrogen contained in chlorophyll molecules the important aspect or the fact that nitrogen is one of the components of chlorophyll and that it allows therefore one to evaluate plant N status based on chlorophyll content? Maybe the last information is the one you want to emphasize.
Line 33-36: Can be better elaborated. For example, focus on the protective functions of Car, which is its main function I believe. Also what aspects of physiology and stress these pigments concentration may indicate, give examples.
Line 36: Instead of citing remote sensing in general, better to specify proximal sensing and indicate the wavelengths used in the study. This helps to pass from general literature to your study case. In general, the complete paragraph might be improved.
Line 40-41: replace: ” The remote sensing method is usually conducted with two approaches: the empirical ones and the physical ones. The empirical approaches...”, by: “The proximal sensing-based method is usually conducted using either empirical or physical approaches. The former...”
Line 41: replace: “indices”, by: ”vegetation indices”. The same in the rest of the article to make clear to what are you referring.
Line 49: not clear what you mean by: “Of all these approaches, leaf reflectance is often a major data source.”
Line 56: replace: “in directions...”, by: “in all directions...”
Line 58: replace: “Given such situation...”, by: ”Given that...”
Line 60-84: In your discussion you might consider other studies that used wavelengths to reduce leaf reflectance effects or other nuisance factors on leaf pigment pigments/traits retrieval (e.g., Li et al., [4,5]).
Material and methods
Section 2.1: Maybe giving more information here about the experiment is of interest, even if already described in another paper. Also, I believe that an approximate citation for the article in press should be given (journal, title, authors, etc.). A better explanation of the experiment might help the reader to understand from where the variability on the dataset comes from.
Line 98-105: Provide better explanation concerning the measuring set-up. Mainly regarding the final information sampled (e.g., hemispherical-conical? [6]).
Line 113: Be more specific here. Modelling methods?
Line 2.2.3: replace: “indices”, by: “vegetation indices”
Section 2.2.3: Even if published in another article, an indication on how the diffuse reflectance was estimated should be given.
Line 135: remove “several”
Line 138: replace: “etc”, by: “for example”
Line 140: replace: “was”, by: “were”
Line 142: replace: “was as the criterion of”, by: “was used as the criterion for”
Line 143: replace: “The SMLR was conducted with the aid of the Matlab (Version 2016a, MathWorks) function stepwise.”, by: “The SMLR was conducted within Matlab (Version 2016a, MathWorks) using the function stepwise”
Line 151-157: Not clear what do you mean here. This paragraph need to better formulated. From what you explain, I understand that a minimum interval of 10 nm was kept between the bands included in each dataset sampled for the “baseline” estimations. Other parts are more difficult to grasp. What do you mean by this, for example: “The wavelengths selected from the entire data set were always used as independent variables to maximize the representativeness of models.”?
Line 159: replace: “conducted”, by: “derived”
Figure 1: “No processing” should be indicated in the same format as other datasets, since was used as input for the modelling approach in the same way. Also, ordinary linear regression should be indicate as modelling approach for vegetation indices.
Results
Line 194: replace: “confusion caused by great...”, by: “effects of...”
Line 195: start with: “In the nadir directions...” and finish indicating the Figure: “...(Figure 2).”
Line 196: replace: “range”, by: “wavelengths”; and replace: “absorption of...”, by: “absorption by...”
Line 197: replace: “...in green a peak appeared because of the relatively weak absorption.”, by: “...in the green wavelengths a peak appeared because of the relatively weaker absorption.”
Line 198: replace: “...owing to...”, by: “...related to...”
Line 200: replace: “...low speed...”, by: “...lower intensity...”. Next paragraph can be removed, or integrated to avoid repetition.
Line 202: replace: “The review about these indices can be found in [12].”, by: “A review about these and other vegetation indices can be found in [12].”
Figure 2: resolution should be improved. Difficult to distinguish colors.
Line 210-211; Table 2: Again, further information about the calculation of specular reflectance should be given, even if published elsewhere. This is important, in particular to evaluate the present manuscript before a potential publication. Approximate citation for the article in press should be given.
Line 214: replace: “...direction...”, by: “...trend...”
Line 215-216: This is related to the ill-posed problem of leaf properties retrieval based on leaf spectra. You may refer to other article(s) discussing that, for example those dealing with retrieval based on physically-based approaches.
Line 217: Use “Pigments” instead of “Pigment”
Line 219: replace: “indices” and “prediction power”, by: “vegetation indices” and “predictive performance”
Line 233: replace: “...the retrieval accuracy in forward 35° is lower than the nadir for both Chl and Car, owing to the interference by large amount...”, by: “...the retrieval accuracy corresponding to view angle 35° forward in the principle plane is lower than that for Nadir considering both pigments estimated, chl and car, due to interference from large amount...”
Line 235-236: replace: “...seriously disturbed...” and “...disturbance...”, by: “...intensely affected...” and “...effects...”
Line 238: replace: “...the strong specular disturbance removal effect of...”, by: “the better performance concerning specular disturbance removal provided by...”
Line2 45-246: When you say that: “...revealed that the derivative transformation plus MLR furtherly removed the specular interference...”, you not necessarily remove the effects but select wavelengths which considered together might mitigate the effects of specular reflectance. What is the message you want to pass to the reader?
Line 254: replace: “...pluc...”, by: “...plus...” or “...coupled with...”. Maybe better to indicate this combination by CWT+MLR or something similar.
Line 257: “The best retrieval in a given direction was achieved for Chl in comparison to Car...while for a given pigment, retrieval concerning 35° forward was slightly better than that observed for Nadir...”
Discussion
Figure 3: Better insert the Figure after appearing in the text.
Line 271: replace: “...is drawn...” and “...It showed...”, by: “...is illustrated...” and “...It indicates...”
Line 279: replace:”...in Car...than for Chl.”, by: “...for Car...than for Chl.”
Line 284: replace: “...the data at each...”, by: “...the information derived in each...”
Line 337: Isn’t the low consistency important? What were the main selection patterns in each case (e.g., spectral regions with greater importance, etc.), are they comparable? Less stability is related to the fact that large regions were well related to the components of interest after application of spectral treatments (derivative and wavelets)?
Line 344-345: The lack of consistency might be due to the small sample size, which may cause overfitting in some cases? Adopting a more robust calibration and validation approaches might improve that?
Line 354-355: Shouldn’t this index be included in the evaluation? It seems they were optimized for retrieval of pigments at leaf level. There are carotenoids specific index developed by the same authors (Gitelson et al.,). It might be worth to test.
Conclusions
Line 362: replace: “It demonstrated...”, by: “It was demonstrated...”
References
1. Yu, K.; Lenz-Wiedemann, V.; Chen, X.; Bareth, G. Estimating leaf chlorophyll of barley at different growth stages using spectral indices to reduce soil background and canopy structure effects. ISPRS Journal of Photogrammetry and Remote Sensing 2014, 97, 58–77, doi:10.1016/j.isprsjprs.2014.08.005.
2. Mohd Asaari, M.S.; Mishra, P.; Mertens, S.; Dhondt, S.; Inzé, D.; Wuyts, N.; Scheunders, P. Close-range hyperspectral image analysis for the early detection of stress responses in individual plants in a high-throughput phenotyping platform. ISPRS Journal of Photogrammetry and Remote Sensing 2018, 138, 121–138, doi:10.1016/j.isprsjprs.2018.02.003.
3. Barnes, R.J.; Dhanoa, M.S.; Lister, S.J. Standard Normal Variate Transformation and De-trending of Near-Infrared Diffuse Reflectance Spectra. Appl. Spectrosc., AS 1989, 43, 772–777.
4. Li, D.; Cheng, T.; Jia, M.; Zhou, K.; Lu, N.; Yao, X.; Tian, Y.; Zhu, Y.; Cao, W. PROCWT: Coupling PROSPECT with continuous wavelet transform to improve the retrieval of foliar chemistry from leaf bidirectional reflectance spectra. Remote Sensing of Environment 2018, 206, 1–14, doi:10.1016/j.rse.2017.12.013.
5. Li, D.; Wang, X.; Zheng, H.; Zhou, K.; Yao, X.; Tian, Y.; Zhu, Y.; Cao, W.; Cheng, T. Estimation of area- and mass-based leaf nitrogen contents of wheat and rice crops from water-removed spectra using continuous wavelet analysis. Plant Methods 2018, 14, doi:10.1186/s13007-018-0344-1.
6. Schaepman-Strub, G.; Schaepman, M.E.; Painter, T.H.; Dangel, S.; Martonchik, J.V. Reflectance quantities in optical remote sensing—definitions and case studies. Remote Sensing of Environment 2006, 103, 27–42, doi:10.1016/j.rse.2006.03.002.
Author Response
We deeply appreciate your detailed and constructive comments which improve our manuscript greatly. We carefully studied these comments and tried our best to revise the manuscript accordingly. We think this new version should meet the standard of publication in Remote Sensing. Thank your again for your time and detailed reviewing. Look forward to hearing from you soon.
Author Response File: Author Response.pdf
Reviewer 3 Report
The manuscript presents a study on minimizing the effects of specular reflectance on predicting leaf pigments with multiple linear regression and continuous wavelet analysis. The study is interesting to the community since specular reflectance often decreases the estimation accuracy of leaf biochemistry using leaf reflectance. However, there are some major issues that need to be addressed.
(1) The manuscript aimed to test the feasibility of multiple linear regression and continuous wavelet analysis on predicting leaf pigments, yet the methods and results did not well support the conclusions.
Non-parametric regression methods (e.g. multiple linear regression and continuous wavelet analysis) often produced more accurate estimation of vegetation variables than parametric regression methods (e.g. vegetation indices), because the former utilize the full spectrum information rather than a few wavelengths (see Jochem Verrelst et al. 2015, ISPRS, Optical remote sensing and the retrieval of terrestrial vegetation bio-geophysical properties – A review). Therefore, the conclusion that using multiple linear regression and/or continuous wavelet analysis more effectively minimized the specular reflectance than vegetation indices does not hold. This is not “cause and effect”.
In other words, the better performance of multiple linear regression and/or continuous wavelet analysis could be attributed to other reasons and not necessarily to minimizing specular reflectance. It is not fair to compare vegetation indices and multiple linear regression, which is not comparing “apples to apples”.
(2) The physical understanding of minimizing specular reflectance in multiple linear regression and/or continuous wavelet analysis is not clarified. To illustrate if the specular reflectance is addressed, it is suggested to obtain specular reflectance using polarization measurements and then calculate diffuse reflectance by subtracting specular reflectance from leaf reflectance. That can be found in the paper recently published by the authors “An Approach to Improve Leaf Pigment Content Retrieval by Removing Specular Reflectance Through Polarization Measurements” in IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (see: https://0-ieeexplore-ieee-org.brum.beds.ac.uk/document/8509605).
Specific comments:
Line 19: “index ones” can be replaced by “indices”.
Lines 31-32: “In plants much nitrogen is contained in Chl” is not correct. Nitrogen is primarily in rubisco, and it is 6.5% (by weight) of chlorophylls (see Kokaly et al. 2009, RSE, Characterizing canopy biochemistry from imaging spectroscopy and its application to ecosystem studies).
Lines 71-77: It is true that some vegetation indices were developed for minimizing the effects of specular reflectance. That did not mean that other approaches such as multiple linear regression were developed for the same purpose. The advantage of multiple linear regression over vegetation indices is the ability of exploiting the information of full spectrum.
Line 73: Change “is” to “are”.
Line 90: The reference of “Li et al. (in press)” is not given. Does it refer to this paper “An Approach to Improve Leaf Pigment Content Retrieval by Removing Specular Reflectance Through Polarization Measurements”?
Line 131: Same query as the previous comment. Since diffuse reflectance is critical in leaf biochemistry estimation (as stated in lines 54-57), more details of calculating diffuse reflectance are need for the readers. Also, the amplitude of specular and diffuse reflectance affects the estimation of leaf pigments, it is suggested to show their full spectrum rather than just three wavelengths as listed in Table 2.
Lines 184-185: This should be in section Methods.
Lines 187-192: It would be good show the correlation between chlorophylls and carotenoids. Also, the strength of relationship by species might be interesting to the readers.
Lines 210-211: Same as previous comment on line 131.
Tables 4-6: It would be good to have scatter plots of measure versus predicted pigments. It is important to see how they fall into the 1:1 relationship. R2 should be included.
Lines 285-286: The highest r between reflectance and Chl seems to locate at around 400nm.
Line 288: “The highest r value increased to 0.99.” The correlation and adjusted R2 (mostly higher than 0.9 in Table A1-A5) looks surprisingly high. Given the fact that three species have significant different levels of pigments, the high r might be attributed to the range of pigments in these three species. An interesting test would be calibrating models with two species and then validating with the third one.
Figure 5: More intervals of the legend are needed, so the readers would know the r values in the brightest and darkest regions.
Author Response
We deeply appreciate your detailed and constructive comments which improve our manuscript greatly. We carefully studied these comments and tried our best to revise the manuscript accordingly. We think this new version should meet the standard of publication in Remote Sensing. Thank your again for your time and detailed reviewing. Look forward to hearing from you soon.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Substantial changes were made in the content of the manuscript, notably by adding RT-based approaches for comparison with the empirical methods tested. Te authors found that empirical methods provided marginal benefits for the dataset used. However this dataset is quite small, which potentially favors empirical and data driven models performance. This fact should be better considered/ discussed throughout the paper, in particular in the discussion and conclusions. Beside that a better explanation about the approaches used and the relationship between then should be provided (Section 2.2.3). Finally, a considerable language editing would be recommended.
Title: replace “...some...” by “...a selection of...”.
Line 20: replace “...radiative transfer (RT) and hybrid models, vegetation indices (VIs)...” by “...vegetation indices (VIs), radiative transfer modelling (RT) and hybrid approaches...”
Line 21: replace “...the specular reflectance or effect, and continuous wavelet transform (CWT), in the nadir and near the mirror-like direction.” By “...the specular reflectance effects, coupling some of these methods with continuous wavelet transform (CWT), considering spectral response in the nadir and near the mirror-like direction.”.
Line 26: 30 or 35 degrees?
Line 28: Explain shortly what MLR means in this case (what inputs were used in contrast to single univariate modelling?).
Line 32: what “Wavelet analysis plus..” means?
Section 2.2.3: should provide a deeper description on how these techniques work and how they relate to each other.
Results: Despite the fact that the authors achieved better retrieval performance using empirical methods in comparison with RT-based approaches it is important to consider that the gain was marginal, e.g.: from R2 of 97 to 99, and considering that the dataset is very limited this scenario largely benefits empirical or data driven (MLR) approaches. This should be better considered in the discussion and conclusions.
Reviewer 3 Report
The manuscript has been largely improved compared with an earlier version. I appreciate the efforts that the authors have made. In this new version, the authors added two approaches, radiative transfer and hybrid models to estimate leaf pigments which may minimize the effects of specular reflectance. More details of the approaches are needed. Also, I find it difficult to follow the logic in several parts of the manuscript. Here are my comments with line numbers in the clean manuscript.
Line 24: Missing “MLR” after “plus”?
Line 42: Doublecheck the punctuation.
Line 68: Change “a same” to “the same”.
Lines 64-91: This paragraph needs editing. The authors are suggested to put the statements in lines 82-88 at the start of paragraph, and then to list previous studies in the two categories. More details are needed for PROSPECT, PROSPECT-5 and PROSPECT-D either here or in the Methods.
Line 88: Remove “However”.
Lines 149-151: Both CWT and first derivative are methods of spectra transformation, one of them may be enough. The authors need to clarify why CWT needs to be performed on first derivative of reflectance.
Table 1: The superscript in “Hybrid model” should be “d”. What does “bspec” stand for?
Line 222: What does “u” stand for? Is a square missing in the merit function? One often use the sum of squares in a merit function.
Lines 223-225: I am confused about the f() and g() functions in Eq. (5). For PROSPECT + R-445, is f() only applied to Smear? Also, is g() only applied to Ssimu in PROCWT? I thought the functions should be applied to both Smear and Ssimu. I feel it difficult to follow the five methods of radiative transfer and hybrid models, and suggest the authors revise this part and add more details.
Line 304: Change “Table 3” to “Table 5”.
Lines 303-304: This sentence is hard to follow and needs to be rephrased.
Line 363: Replace “furtherly” to “further”.
Lines 398-403: The discussion here is not convincing. How would “negative coefficients” alleviate the specular disturbance? I didn’t see the physical mechanisms.
Line 406: “in a direction”?
Line 407: “Lower content of Car” does not necessarily lead to lower correlation with reflectance. I suggest the authors look at the coefficient of variation (CV = std/mean) of Car.
Lines 408-409: I generally agree. But in this study, the Pearson coefficient between Chl and Car is as high as 0.96 (line 248). One should expect similar model performance for Chl and Car since they are almost the same and differ only in magnitude.
Line 426: This seems contradict with Figure 9 (b), in which noise is not observed.
