Estimating Completely Remote Sensing-Based Evapotranspiration for Salt Cedar (Tamarix ramosissima), in the Southwestern United States, Using Machine Learning Algorithms
Round 1
Reviewer 1 Report
The authors used machine learning algorithms to estimate remote sensing Evapotranspiration (ET) for Salt cedar in desert areas of the United States. The English language used needs extensive editing throughout with a high to moderate level of change required as there are many grammatical errors.
Major Comments
The authors should rewrite abstract sections as their repetition of lines 16-17 in lines 21-22. Also, the abstract lacks methods, results, findings, and significance of the study.
Please write the objectives of the study clearly at the end of the introduction section and how this study contributes to existing knowledge.
Please redraw the map with the geographic coordinate system in high quality (at least 600 dpi).
The authors should also use existing MODIS-based ET products to study differences between developed methods and existing products.
The study has serious flaws as the authors did not use any statistical measure to quantify the errors between modeled ET results and in-situ ET values.
I am doubtful about the methodology to estimate ET based on only 4 predictors using machine learning methods as ET is a quite complex process that requires details of atmospheric conditions.
There is no need to draw too many figures 4-10, instead a table can be used to summarize results.
Minor Comments
Please use the abbreviation of ‘BRBE’ in the abstract.
Please also write keywords related to the article.
Line 18: Please rewrite as it is not clear.
Author Response
Reviewer 1
The authors used machine learning algorithms to estimate remote sensing Evapotranspiration (ET) for Salt cedar in desert areas of the United States. The English language used needs extensive editing throughout with a high to moderate level of change required as there are many grammatical errors.
Major Comments
- The authors should rewrite abstract sections as their repetition of lines 16-17 in lines 21-22. Also, the abstract lacks methods, results, findings, and significance of the study
Accurate estimation of evapotranspiration (ET) is a prerequisite for water management in arid regions. Field based methods estimate point-wise ET accurately, but the challenge is in estimating ET over a region with high accuracies. Machine learning based approaches were taken to estimate ET over a large spatial scale using Bowen Ratio Energy Balance (BREB) technique. BREB method depends on terrestrial energy balance equation to estimate ET. Thus, remote sensing based parameters representing variables in the energy balance equation, and vegetation index representing plant health conditions were used in model. The study was conducted in the arid areas of the southwestern United States, where dense patches of Salt cedar consume water from the primary source of water. The preliminary model used enhanced vegetation index (EVI), global horizontal irradiance (GHI), surface temperature (TS), and relative humidity (RH) as parameters. The k-nearest neighbor method consistently generated poor accuracies. When all the parameters were used, accuracies of the other models varied within 90 – 94%. When one predictor parameter was dropped, the best model produced accuracies between 90 – 93%, which dropped to 87 – 92% when a second variable was dropped. Random forest and support vector machine with radial kernel consistently produced the best predicting accuracies.
- Please write the objectives of the study clearly at the end of the introduction section and how this study contributes to existing knowledge.
The following to be added at the end of introduction to define objectives more clearly:
From the above discussion, it is evident that remote sensing ET estimating procedures rely heavily on the energy balance equation. However, energy balance-based procedures suffer from complexities limitations. Conversely, all the field observation based estimations have higher precision. Thus, the primary objectives of this research are – (i) develop purely remote sensing based models that would project field scale ET estimation over entire region; (ii) the models should not have the difficulties and complications of the existing purely remote sensing based procedures; and, (iii) the accuracies should be as high as accuracies of field based estimations. Bowen ratio energy balance (BREB) base ET estimations were used as ground truth. In the proposed method remote sensing based assessment of parameters representing the variables of energy balance equation as the predictors is developed. The hypothesis behind this investigation is that the machine learning based model(s) can estimate ET with the same high precision at the ground truth pixels as well as outside.
- Please redraw the map with the geographic coordinate system in high quality (at least 600 dpi).
Following figure to be added:
- The authors should also use existing MODIS-based ET products to study differences between developed methods and existing products.
The only established MODIS-based ET products are available via MOD16 products. However, the primary objective of this study was to develop models that would use remote- sensing based estimations of different components of energy balance equation as predictors and compare the results with Bowe ratio based ET estimates. The major objective is to scale field measurements, and in this particular case, Bowen ratio estimations, over the large spatial scale.
- The study has serious flaws as the authors did not use any statistical measure to quantify the errors between modeled ET results and in-situ ET values.
Statistical comparison has been made. The following to be added in the materials and method to clarify the evaluation process:
The machine learning modeling procedures were adopted from [66, 67]. A 10-fold cross validation was calculated with five-repetitions, and the model accuracies were estimated by comparing correlations between estimated and predicted estimations of the validation data. The entire procedure (including the cross validation) were repeated 500 times via “for” loop for each model. For each model, the median values of 500 correlations were compared to identify the optimum model. Additionally, for each median correlation value, the range of accuracies, the maximum and minimum correlation values out of the 500 iterations, were obtained to emphasize the reliability of the model.
The statistical evaluation was adopted from:
Adak, A.; Murray, S.C.; Božinović, S.; Lindsey, R.; Nakasagga, S.; Chatterjee, S.; Anderson, S.L., II; Wilde, S. Temporal Vegetation Indices and Plant Height from Remotely Sensed Imagery Can Predict Grain Yield and Flowering Time Breeding Value in Maize via Machine Learning Regression. Remote Sens. 2021, 13, 2141. https://doi.org/10.3390/rs13112141.
- I am doubtful about the methodology to estimate ET based on only 4 predictors using machine learning methods as ET is a quite complex process that requires details of atmospheric conditions.
Applications of energy balance equation is a quite a common approach in estimating ET. It is also the principle used in many purely remote sensing based ET estimating models such as - SEBAL, METRIC, SEBI, I-SEBI, etc.. These models are completely dependent on the energy balance equation. The ground truth data using Bowen Ration method also depends, essentially on this energy balance equation, the components of which are net radiation, soil heat flux, sensible heat,and latent heat. The primary objective of this study is to develop a purely remote sensing based model that would use the components of energy balance equation and would estimate ET precisely close to Bowen ratio ET estimations. In fact, this model we use vegetation indices as proxy to crop phenology, which is in addition to all energy balance components in the ET estimation. Ideally, all the four variables are expected to be used. We have also explored models that would use lesser number of parameters in cases of users not having access to all variables.
- There is no need to draw too many figures 4-10, instead a table can be used to summarize results.
These figures are to be confined within a table.
Minor Comments
- Please use the abbreviation of ‘BRBE’ in the abstract.
Fixed
- Please also write keywords related to the article.
Added Remote Sensing, Machine Learning, Evapotranspiration
- Line 18: Please rewrite as it is not clear.
Fixed
Author Response File: 
Author Response.docx
Reviewer 2 Report
The manuscript uses deep learning to estimate regional evapotranspiration is of great scientific value. However, the structure of the manuscript is confusing, the verification method is too simple, and the calculation results are not enough to prove the validity and reliability of the method.
L71 In Section Introduction, formula should not appear here.
L108 This paragraph can be deleted.
L112, The climate and hydrology of the study area should be introduced.
L114 A schematic diagram of the topography and vegetation coverage of the study area should be provided.
L127 This part should be placed in the Section Method.
L138 In Study Area, formula should not appear here.
L170 The title does not match the content.
The method introduction part is confusing.
Analysis of Figure 2 and 3?
L283 Figur 4-9 are not mentioned in the text.
Author Response
Reviewer 2
The manuscript uses deep learning to estimate regional evapotranspiration is of great scientific value. However, the structure of the manuscript is confusing, the verification method is too simple, and the calculation results are not enough to prove the validity and reliability of the method.
L71 In Section Introduction, formula should not appear here.
Moved
L108 This paragraph can be deleted.
Deleted
L112, The climate and hydrology of the study area should be introduced.
The following has been added:
CNWR is located within the Sonoran Desert region in the southwestern United States, that has an extreme arid climatic condition. CNWR receives very low precipitation with an annual mean of about 100 mm). The soil is predominantly sandy and silt loam, with little clay. The soil is basic in nature (pH > 8), with high percentage of salt (NaCl). The soil salinity gradually increases with distance from the Lower Colorado River [45].
L114 A schematic diagram of the topography and vegetation coverage of the study area should be provided.
An image from the google earth (satellite) has been added to Figure 1
L127 This part should be placed in the Section Method.
Moved
L138 In Study Area, formula should not appear here.
Moved
L170 The title does not match the content.
The method introduction part is confusing.
Changed
Analysis of Figure 2 and 3?
Provided in the section 3.1
L283 Figur 4-9 are not mentioned in the text.
Figures replaced by table (as per recommendation by another reviewer)
Author Response File: Author Response.docx
Reviewer 3 Report
The manuscript evaluates a number of machine learning techniques regarding their power to estimate the evapotranspiration in the US southwest arid areas populated by Tamarix means of remote sensing. The results are adequately presented and provide some insights. At the current stage, the manuscript needs some tidying up. For example, a paragraph in the abstract is identically duplicated. Some sentences are incomplete in the manuscript. Some of the figures need improvement. For instance in Figure 1 the 3 different colours used are hard to distinguish. Once the manuscript has been tidied I see no obstacle to its publication.
Some sentences are incomplete.
Author Response
Reviewer 3
Comments and Suggestions for Authors
The manuscript evaluates a number of machine learning techniques regarding their power to estimate the evapotranspiration in the US southwest arid areas populated by Tamarix means of remote sensing. The results are adequately presented and provide some insights. At the current stage, the manuscript needs some tidying up. For example, a paragraph in the abstract is identically duplicated. Some sentences are incomplete in the manuscript. Some of the figures need improvement. For instance in Figure 1 the 3 different colours used are hard to distinguish. Once the manuscript has been tidied I see no obstacle to its publication.
The manuscript has been revised. Revised manuscript attached.
Comments on the Quality of English Language
Some sentences are incomplete.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors have significantly improved the manuscript by incorporating my comments. I think it should be published now.
Author Response
Comment (Reviewer 1): "The authors have significantly improved the manuscript by incorporating my comments. I think it should be published now."
Response:
Dear sir/madam
Thank you for your time and effort for reviewing this paper on two rounds. I sincerely appreciate your comments, which significantly improved the quality of this article.
Thanking you,
Sincerely,
Sumantra Chatterjee
Reviewer 2 Report
I thank the author for spending considerable time and effort in revising the manuscript. After revision, the quality of the manuscript has been greatly improved. But there are still some minor problems, as shown below.
L188 The calculation formulas should not appear in Section Study area.
Delete Appendix A and B.
Author Response
Comments (Reviewer 2): "I thank the author for spending considerable time and effort in revising the manuscript. After revision, the quality of the manuscript has been greatly improved. But there are still some minor problems, as shown below.
L188 The calculation formulas should not appear in Section Study area.
Delete Appendix A and B."
Response:
Dear sir/madam
Thank you for your time and effort for reviewing this paper on two rounds. I sincerely appreciate your comments, which significantly improved the quality of this article. The following modifications have been made to address your two comments:
Comment 1: "L188 The calculation formulas should not appear in Section Study area" Response: Separate sections have been made under the main heading "Methodologies", where separate subsections have been Bowen ratio data collection and remote sensing data collection. The equations have been moved within the subsection for Bowen ratio data collection.
Comment 2: "Delete Appendix A and B". Response: These sections have been deleted.
Thanking you,
Sincerely,
Sumantra Chatterjee
