Modeling Actual Evapotranspiration with MSI-Sentinel Images and Machine Learning Algorithms

Round 1

Reviewer 1 Report

General comments:

The manuscript, in general, is well written and with the appropriate manuscript structure. Nevertheless, some point must be addressed in order to be ready for publication. In particular:

1. Please use passive voice within the manuscript.

2. The introduction at this point is very poor. The authors should enrich the list of references within this section.

3. Since the authors are using many abbreviations within the manuscript, it will be very helpful to provide a table with all these abbreciations.

4. Please provide some more information about the study area and why this was your choice as a study area. It will be very helpful for the readers to provide some extensive meteorological information in order to justify the selection of the study area.

5. Please separate the results-discussion sections. Each of then should be a separate section. Results should provide a concise and precise description of the experimental results, their interpretation as well as the experimental conclusions that can be drawn. Discussion should be provide the connection/review of your methodology to other works.

6. Also, the authors should enrich the last section of their manuscript, which is the Conclusions. At this point this section is very poor.

7. Last but not least, the authors should provide a a flow chart of their methodology.

Author Response

Agradecimento ao editor

Caro editor

Agradecemos a oportunidade de ter este artigo revisado por especialistas sugeridos pelo Atmosphere, que deram contribuições valiosas ao trabalho, tornando-o mais atraente e legível para o público da revista.

Gostaríamos de informar ao estimado editor que submetemos este artigo a uma revisão aprofundada do inglês após as correções.

Agradecimento aos revisores

Caros Revisores

We deeply appreciate your availability and contribution to this paper. The information and questions raised were necessary to clarify some important points, making the paper more comprehendible for the readers. Below we have detailed our changes/answers to any questions raised. In the text itself the changes are green-highlighted.

We would like to inform the esteemed reviewers that we have submitted this paper to an in-depth review of English after the corrections

Response to Reviewer #1:

The manuscript, in general, is well written and with the appropriate manuscript structure. Nevertheless, some point must be addressed in order to be ready for publication. In particular:

1. Please use passive voice within the manuscript.

R: The manuscript underwent a thorough revision from English to passive voice.

2. The introduction at this point is very poor. The authors should enrich the list of references within this section.

R: This section underwent reformulation and the addition of new references and are highlighted in green between lines 46-50 and 100-116. In total, 29 more references were added.

3. Since the authors are using many abbreviations within the manuscript, it will be very helpful to provide a table with all these abbreviations.

R: The list of abbreviations was inserted in the supplementary material.

4. Please provide some more information about the study area and why this was your choice as a study area. It will be very helpful for the readers to provide some extensive meteorological information in order to justify the selection of the study area.

R: Na linha 125-128 foi acrescentado o motivo das escolhas da região, assim como foi inserida a Figura 2, que mostra os dados climáticos do local.

5. Por favor, separe as seções de discussão de resultados. Cada um deles deve ser uma seção separada. Os resultados devem fornecer uma descrição concisa e precisa dos resultados experimentais, sua interpretação, bem como as conclusões experimentais que podem ser tiradas. A discussão deve fornecer a conexão/revisão de sua metodologia com outros trabalhos.

R: Essa sugestão foi atendida e os resultados foram separados das discussões. Os resultados são da linha 335 e as discussões da 507.

6. Além disso, os autores devem enriquecer a última seção de seu manuscrito, que são as Conclusões. Neste ponto, esta seção é muito pobre.

R: A seção de conclusão foi alterada conforme sugerido e enriquecido.

7. Por último, mas não menos importante, os autores devem fornecer um fluxograma de sua metodologia.

R: Conforme sugerido, o fluxograma foi inserido e pode ser visto na Figura 3, linha 137.

Author Response File: Author Response.docx

Reviewer 2 Report

This is a very interesting study attempting to reproduce the ETrF estimations coming from Landsat imagery using the METRIC model based on Sentinel 2 imagery and machine learning. The results obtained are important and interesting. The methodology also seems to be sound and reasonable, while the manuscript is well written and easy to understand. There are a few language issues, so, the manuscript will benefit from a thorough check.

I only have two main comments that I believe that should be considered by the authors.

1. While the study in reality investigates if ETrF estimations coming from Landsat using METRIC model are corelated with Sentinel 2 data and can be reproduced by them using machine learning, in many parts of the manuscript the ETrF estimations coming from METRIC are used and named as “observed” values while in reality they are also predictions. In this study there aren’t any real observations (e.g. by eddy covariance, lysimeters, water balance or something). The manuscript still has important merit, but it should be clear that METRIC predictions are not measurements and may also have errors. It should be clearer through the manuscript that what it is attempted is the reproduction of METRIC predictions by Sentinel 2 data.

2. In many instances it is mentioned that ETrF is equivalent to the Kc when using the ETr of the alfalfa. According to the FAO-56 methodology and terminology, this is the case only when there is an absence of stress (e.g.  due to low soil water content, high salinity) and in this case ETa is equal with ETp (potential evapotranspiration). In this study, it isn’t clear if the studied crop is always under well-watered conditions and no stress. The existence of stress could be for example the reason for the lower ETrF values presented in Figure 8 for the mid-season period. If this is the case the Kc values won’t be equivalent with the estimated ETrF values. It is important this to be clarified in the manuscript because otherwise future studies or applications may use the Kc values mentioned here inducing also stress to their cultivations.

Apart from the above two comments, I believe that it would be informative for the readers to have a better idea of how this method can be extended to other regions and other cultivations (e.g, what would be the future research needed for this to happen).

Author Response

Acknowledgement to the editor

Dear Editor

We appreciate the opportunity to have this paper peer-reviewed by experts suggested by Atmosphere, who have made valuable contributions to the work, making it more attractive and readable for the journal’s audience.

We would like to inform the esteemed editor that we have submitted this paper to an in-depth review of English after the corrections.

Acknowledgement to the reviewers

Dear Reviewers

We deeply appreciate your availability and contribution to this paper. The information and questions raised were necessary to clarify some important points, making the paper more comprehendible for the readers. Below we have detailed our changes/answers to any questions raised. In the text itself the changes are green-highlighted.

We would like to inform the esteemed reviewers that we have submitted this paper to an in-depth review of English after the corrections

Response to Reviewer #2:

This is a very interesting study attempting to reproduce the ETrF estimations coming from Landsat imagery using the METRIC model based on Sentinel 2 imagery and machine learning. The results obtained are important and interesting. The methodology also seems to be sound and reasonable, while the manuscript is well written and easy to understand. There are a few language issues, so, the manuscript will benefit from a thorough check.

I only have two main comments that I believe that should be considered by the authors. Please use passive voice within the manuscript.

1. The While the study in reality investigates if ETrF estimations coming from Landsat using METRIC model are corelated with Sentinel 2 data and can be reproduced by them using machine learning, in many parts of the manuscript the ETrF estimations coming from METRIC are used and named as “observed” values while in reality they are also predictions. In this study there aren’t any real observations (e.g. by eddy covariance, lysimeters, water balance or something). The manuscript still has important merit, but it should be clear that METRIC predictions are not measurements and may also have errors. It should be clearer through the manuscript that what it is attempted is the reproduction of METRIC predictions by Sentinel 2 data.

R: The observed variable is the one to be modeled in this work. Let's understand that it can generate misunderstanding in readers that can come to be interpreted as a measured variable.

In the response variable section we are treating the observed variable with a model and it can be seen on line 172 .

2. In many instances it is mentioned that ETrF is equivalent to the Kc when using the ETr of the alfalfa. According to the FAO-56 methodology and terminology, this is the case only when there is an absence of stress (e.g. due to low soil water content, high salinity) and in this case ETa is equal with ETp (potential evapotranspiration). In this study, it isn’t clear if the studied crop is always under well-watered conditions and no stress. The existence of stress could be for example the reason for the lower ETrF values presented in Figure 8 for the mid-season period. If this is the case the Kc values won’t be equivalent with the estimated ETrF values. It is important this to be clarified in the manuscript because otherwise future studies or applications may use the Kc values mentioned here inducing also stress to their cultivations.

R: This is an interesting point that agrees with our thinking. In the base publication by Allen et al. 2007 he considers it as Kc, however he could not, because the satellite captures the information that is occurring in the field whether the culture is well nourished and hydrated or not.

Between lines 547 and 556 of the discussion section, we address that the ETrF by the satellite corresponds to the product between Kc and Ks (stress coefficient)

3. Apart from the above two comments, I believe that it would be informative for the readers to have a better idea of how this method can be extended to other regions and other cultivations (e.g, what would be the future research needed for this to happen).

R: In line 580-588 we insert a paragraph with suggestions for application and improvement of the models.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments are attached below

Comments for author File: Comments.pdf

Author Response

Acknowledgement to the editor

Dear Editor

We appreciate the opportunity to have this paper peer-reviewed by experts suggested by Atmosphere, who have made valuable contributions to the work, making it more attractive and readable for the journal’s audience.

We would like to inform the esteemed editor that we have submitted this paper to an in-depth review of English after the corrections.

Acknowledgement to the reviewers

Dear Reviewers

We deeply appreciate your availability and contribution to this paper. The information and questions raised were necessary to clarify some important points, making the paper more comprehendible for the readers. Below we have detailed our changes/answers to any questions raised. In the text itself the changes are green-highlighted.

We would like to inform the esteemed reviewers that we have submitted this paper to an in-depth review of English after the corrections

Response to Reviewer #3:

The authors have done a good amount of work to justify the estimation of ET with machine learning approach, the amount of work conducted in this work is enormous and further, I believe that there are some problems after addressing those can be considered for publication.

Currently, many of the statements are not supported by published works. Authors may like to find studies in line with their statements to add scientific weight to their observations. I believe that after duly ddressing the comments authors can improve the quality of the manuscript substantially to make it more insightful.

Extensive English editing is required as there many problems with sentence restructuring, grammatical errors, punctuations. I suggest authors to consider the English editing a serious concern in this manuscript and with the help of native speaker they can improve this version of the manuscript adequately.

The discussion does not have a proper discussion. There is no citation and comparison with the literature. I recommend them to compare their study with a very recent papers described later in this field

Introduction

1. In the introduction, research gaps should be identified better.

R: The manuscript underwent a thorough revision in English and substantially improved the references, in addition to separating the results of the discussion for clarity.

2. I have a big concern in the Introduction, as the authors have missed providing detailed discussion on the important aspect of different classification of ET estimation methods. There is a vast literature on this I would like to suggest few lines following this which author should add is “The ETo estimation models available in the literature may be broadly classified as (1) fully physically-based combination models that account for mass and energy conservation principles; (2) semi-physically based models that deal with either mass or energy conservation; and (3) black-box models based on artificial neural networks, empirical relationships, and fuzzy and genetic algorithms”. I would recommend adding these recent references to add more scientific weight in their Introduction. https://doi.org/10.1061/(ASCE)IR.1943-4774.0001199

Almorox, J., and GiresserJ., (2016). Calibration of the Hargreaves–Samani method for the calculation of reference evapotranspiration in different Köppen climate classes. Hydrology Research, 47(2): 521-531.

 R: The reviewer's suggestion was added between lines 46 to 50.

3. Literature review for the ML techniques is very narrow please try to include some recent studies with the application of ML in ET estimations. Also it would be clear if the authors can describe the novelty of their ML technique than the previous studies. Authors need to provide the justification of this statement by using METRIC framework of water limited ecosystems. R: In lines 100-111, we inserted quotes referring to the ML and evapotranspiration algorithms, bringing a novelty to our work.

4. Line correct the symbols of degree at 132 and 134

R: I can't identify the error suggested here

**Data and Methods**

5. I would suggest authors to provide a flowchart describing the preprocessing of raw datasets obtained from different sources followed by their application.

R: The process flowchart was inserted in line 137.

6. Also, provide a table detailing all hydrometeorological information with their sources and duration.

R: In line 129 Figure 2, a graph with the precipitation and temperature conditions in the region was inserted

**Results and Discussion****

7. Authors are required to add a plot showing the sensitivity of the PM equation to climatic variables. It is important to analyze which variable has a significant effect on ETo estimation for different variables. The amount of change in ETo (mm day−1) with respect to a unit change in each climate variable should be presented in a graphical plot with a daily variation of sensitivity coefficients. In the discussion section, I would recommend authors to benefit from the given literature where they have used Landsat based ET estimation and crop coefficient estimation and its suitability at catchment scale (https://doi.org/10.1007/s12524-021-01367-w)

R: Authors understand that such a suggestion is not relevant in this study, as the objective is to train and evaluate machine learning algorithms to predict the evapotranspiration fraction. The suggestion in question requests an evaluation of the meteorological data in relation to the PM-FAO model, but the PM-FAO model is not the object of study, but the fraction by satellite image.

**Conclusion****

8. Improve the conclusions based on objective and rewrite it in the points. I would recommend the authors are suggested to provide the future scope of this study either at the end of the Discussion or in the Conclusion section.

R: Results were modified interior and in the last paragraph (line 580-588) added what can be advanced to have models with better performance.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

I would like to thank the authors for addressing my point issues.

Reviewer 2 Report

I am convinced by the responses and satisfied by the revisions made to the manuscript. Congratulations for your work!

Reviewer 3 Report

I want to thank the authors for addressing previous comments and for their constructive work. I found all replies satisfactory and the changes made to the manuscript significantly improve the quality of the paper. Overall, I think the resubmitted manuscript is almost ready for acceptance.