Review of Remote Sensing Applications in Grassland Monitoring
Round 1
Reviewer 1 Report
Your work was well prepared and shows a high amount of quality research. It will be very useful for the scientific comunity. I have some comments, though, below:
Please add "ALOS" before/after "advanced land observation satellite" and specify if PALSAR data are used. Also, comment if SAR data from Sentinel 1 are also considered in some of the articles.
Please review tables formatting (in my copy of the manuscript the last columns were not visible).
As a general comment, in my view, the dry or senescent part of grassland vegetation should be considered and cited as one of the main causes of the, sometimes, poor results found. Perhaps because you considered only recent articles (from 2012) this matter was not included in objectives of the articles considered. For arid or semi arid grasslands, the vegetation is dry or senescent most part of the year and cannot be detected by NDVI or indices constructed mainly to detect photossinthetic green part of grasses. I imagine you know this, but you failed to address this important issue in your review. It is only a question of searching for one or two articles that focus at this issue, explain it and then review again the articles you already worked on and enhance their mentions to dry, senescent vegetation or NPV - non photossinthetic vegetation (I can think about Dr Dar Roberts from UCSB/US that worked on that for Californian grasslands). Some other indices, such as the CAI (Cellulose Absorption Index) could be cited too, as it is usually highly correlated to grassland vegetation. I think adding comments about these issues will improve a lot your good review.
Author Response
First of all, thanks for your helpful comments.
1. Abbreviations, table formatting problems, and comments on SAR data usage
Response:
We have made corrections according to the reviewer’s comments. In detail, we added "ALOS" after "advanced land observation satellite" (see Line 66) and reduced the font size of problematic tables for viewing (see Table 2-9). The comments on SAR data usage were also listed in Section 4.1 (see Line 1208-1215), and we introduced the utilization of these data in some studies and described their main problems.
2. The effect of dry or senescent parts of grassland vegetation on obtaining poor results
Response:
We added a new paragraph to Section 4.4 (see Line 1373-1395) to discuss NDVI, EVI and other greenness-based vegetation indices in detail. In this paragraph, we summarized the impact of environmental and human factors on these vegetation indices. According to the work of Dr Dar Roberts you mentioned, we cited some of his highly relevant and significant research (Citation 212-215) in our manuscript and described the main contributions of them in detail. Besides, other factors such as soil background, vegetation density and human activities were also discussed. Finally, to make our manuscript more valuable, we discussed some possible solutions in Section 4.5 (mainly see Line 1411-1422).
3. Some articles referring to dry, senescent vegetation or NPV - non photossinthetic vegetation and some new vegetation indices.
Response:
We mentioned some of the work associated with senescent vegetation in Section 2.3 (see Line 424-432, 498-511) and the work related to dry vegetation in Section 3.3.2 (mainly involved in Paragraphs 2-4). As for the CAI (Cellulose Absorption Index) you mentioned, it had already been cited in Section3.2.1 (see Line 762). In addition, other valuable and innovative indices were presented in the tables or detailed in some paragraphs (for example, see Line 501, 861, 867, 880, 971).
Reviewer 2 Report
The manuscript by Zhaobin Wang et al. reviewed the latest remote sensing estimation methods for the estimation of grassland key attributes (for example, biomass, growth rate, or the proportion of incoming radiation absorbed by active growing vegetation). I believe this topic is of interest to Remote Sensing readers. However, three main aspects prevent me from providing better feedback, and I would not recommend the publication of this article.
First, and since this is a review article, there should be a clear difference between this article from previous similar ones. for example from: Hunt Jr, E.L. Applications and research using remote sensing for rangeland management. Photogrammetric Engineering and Remote SensingVolume 69, Issue 6, Pages 675 - 693.
Second, the Authors mentioned NDVI, EVI, and SAVI as spectral indices to be used. However, from looking in SCOPUS, and just by typing NDVI as the research word none of the key articles associated with these indices are mentioned. For example, the most cited article is the one entitled: Overview of the radiometric and biophysical performance of the MODIS vegetation indices which in SCOPUS has 5263 citations. However, this review did not cite this key article. This one is one example of many key articles a REVIEW in this field should provide.
Third and finally, one of the key messages of the review is the idea of a need for more complex statistical methods to improve remote sensing estimations of forage attributes. Even though I might be in accordance with the statement, the review should provide examples where readers can learn the improvement these techniques could generate. For example, by typing in SCOPUS, the words NDVI grassland and radiation use efficiency, the most cited paper is: ANPP estimates from NDVI for the central grassland region of the United States, by Paruelo et al. 1997 (above 400 citations). In that paper, the linear regression model used for the estimation of growth rate with NDVI had an R2 of 0.93. In other words, NDVI explained 93% of ANPP variations. Would any other type of statistical approach get a better estimation? If so, of how much would be the improvement? Would it really matter to increase a model that already explained 93% of the dependent variable with only a linear regression?
Author Response
First of all, thanks for your helpful comments.
1. The difference between our manuscript from other previous similar ones.
Response:
Firstly, regarding your mention of the article "Applications and research using remote sensing for rangeland management", this article was outdated and focused only on the potential and operability of remote sensing technology for rangeland monitoring, which pointed out that the utilization of some remote sensing sensors needs more research to develop useful tools for monitoring. In contrast, our manuscript presented a comprehensive review of estimation methods for some grassland parameters and some monitoring applications from the perspective of the utilization of remote sensing data. Therefore, the perspectives of the two articles were different, and our manuscript provided a comprehensive summary of the latest technology. Finally, we listed some articles that is similar to our manuscript in Section 1 (see Line 117-127 and Table 1) and the difference between our manuscript from them was also illustrated.
2. Some key articles a REVIEW in this field should provide.
Response:
Thanks to the reviewer for pointing out some of the problems we missed. Firstly, we cited the article "Overview of the radiometric and biophysical performance of the MODIS vegetation indices" in our manuscript (Line 87, Citation 37). However, this article was also outdated that only focused on the characteristics and applications of NDVI and EVI. Nowadays, a large number of new vegetation indices have been invented to address some of the deficiencies of NDVI or to adapt to different environmental conditions. Therefore, for the purpose of detailing this situation, we combined the article you mentioned with other involved articles to present the development of some vegetation indices in Section 1 (Line 83-98). In addition, we have cited some of the latest key review articles (Citation 52-56) related to grasslands remote sensing and described them (Line 117).
3. Doubts about the necessity of more sophisticated statistical methods to improve remote sensing estimates of forage attributes.
Response:
We are very sorry for our negligence of this issue. These doubts and suggestions are very constructive and meaningful for our manuscript. To explain and discuss them in detail, we added some new paragraphs in Section 4. Specifically, in Section 4.2.1 (Line 1245-1262), we described the same situation you pointed out that some linear regression model based on NDVI or EVI could reach a highly precise results, but we also pointed out that models driven only by NDVI or EVI generally failed to obtain robust results for alpine and arid grasslands and sometimes even obtain the worst results. The reasons for the failure of NDVI, EVI, and other greenness-based vegetation indices were discussed in detail in Section 4.4 (Line 1373-1395). It can be seen that the variability of these vegetation indices during the dry season was mainly influenced by solar illumination effects rather than the changes in vegetation. Other factors that influenced the effectiveness of these indices were also listed in this part based on some relevant citations. In addition, we have re-written some paragraphs in Section 4.4 (mainly see Line 1335-1345) to account for the main shortcomings of the univariate models. Finally, we also summarized and made some suggestions to address the above issues in Section 4.5, some paragraphs in this Section have been rewritten (mainly see Line 1411-1422).
Reviewer 3 Report
The authors have been researching the use of remote sensing technology in grassland management for several decades.
Compared to traditional ground-based measurement, remote sensing-based technology has the overall advantage of convenience, efficiency and cost-effectiveness, especially over large areas.
The latest remote sensing methods for estimating several critical grassland cover parameters including aboveground biomass (AGB), primary productivity, fractional vegetation cover (FVC), and leaf area index (LAI) are summarized in a comprehensive review.
Remote sensing applications are assessed in terms of their use for these parameters as well as for some of the parameters of remote sensing data evaluation. Disaster monitoring, impact analysis and carbon cycle monitoring are also addressed.
Empirical modelling and statistical regression models are used in most of the papers, while the number of machine learning approaches is on the rise. In addition, other models are also emerging nowadays, for example, some specialized methods have been widely used, such as lightweight efficiency approaches for primary productivity and mixed pixel decomposition methods for FVC.
However, all the above methods have some limitations. For future work, it is recommended that most applications adopt advanced estimation methods rather than simple statistical methods of regression models. In particular, the potential of deep learning in high-dimensional data processing and determining nonlinear relationships should be further explored. The fusion of images from multiple sources is also of interest.
This work is deeply focused on the investigation of existing methods for remote sensing evaluation. It is very thorough and, as a novelty, provides an evaluation of the methods and recommendations for the direction of research in image evaluation methods.
Author Response
Special thanks for your positive comments and high recognition to our work and manuscript. We have refined some parts of the manuscript to make our work more convincing. Firstly, we added some influential and key articles to the references (see Citation 37, 52-56, 212-215) and introduced their main contributions with our perspectives and discussions (see Line 87, 117, 1373). Secondly, we added some new paragraphs to demonstrate the necessity of more sophisticated statistical methods to improve remote sensing estimates of forage attributes. We described the situations and limitations faced by the greenness-based vegetation indices and univariate models in Section 4 (see Line 1245-1262, 1335-1345, 1373-1395). Finally, based on these refinements, we have modified Section 4.5 (mainly see Line 1411-1422) to provide a more comprehensive description.
Round 2
Reviewer 2 Report
Please find my comments in the attached file.
Comments for author File: 
Comments.pdf
Author Response
1. Problems about radiation use efficiency model and some key articles that should be added in manuscript
Response: Thanks for providing valuable supplementary information for our manuscript. Following the reviewer's comments, we have added some of the articles mentioned in the comments in Section 2.2 (Citation 84-87). Since our manuscript was focused on the latest technologies, we have selected references that are highly relevant to net primary productivity after 2018. We have also detailed the contribution of these articles in Section 2.2 (Line 359-369, 386). Meanwhile, we have modified paragraphs 4 (Line 337) and 5 (Line 350) in Section 2.2 to focus on describing some of the methods used to calibrate radiation use efficiency models by remote sensing data.
2. Difference between growth rate and above ground biomass
Response: It is worth noting that a similar distinction has already been mentioned in our manuscript. It can be seen that Section 2.1 (Line 150) focused on the estimation methods of above-ground biomass while Section 2.2 (Line 293) focused on the estimation methods of primary productivity. The concept of primary productivity (See Line 295) is highly similar to the concept of growth rate mentioned in the comments, which is also a rate and is usually estimated based on the dynamics of biomass. It also can be seen that most of articles provided by the first comment focused on estimating the production or the net primary productivity of grasslands, which is highly consistent with the content of Section 2.2. In addition, we have added some key references provided by you to Section 2.2 and modified some paragraphs to make them more explicit. In particular, we have read the reference (Citation 88) provided in the second comment in detail and then modified the last paragraph of Section 2.2 (see Line 414-420).
