Predicting Crop Evapotranspiration under Non-Standard Conditions Using Machine Learning Algorithms, a Case Study for Vitis vinifera L. cv Tempranillo
Round 1
Reviewer 1 Report
The manuscript presents interesting data on predicting crop evapotranspiration using machine learning.
However, some issues should be corrected.
The abstract lacks numerical results.
Keywords should be corrected to be different from the words in the title. More keywords should be added.
The introduction provides sufficient background.
Latin species names should be in italics.
Materials and methods are described in detail.
Although the results are promising, the discussion should be expanded and supplemented with more references.
Author Response
Dear Reviewer,
We must thank you for the constructive comments and suggestions for improvement.
Below you may find all the answers and changes implemented in the manuscript.
The abstract lacks numerical results.
Thank you. The correction was made in the article.
Keywords should be corrected to be different from the words in the title. More keywords should be added.
Thank you for the suggestions. New keywords were added.
The introduction provides sufficient background.
Latin species names should be in italics.
Thank you. The correction was made in the article.
Materials and methods are described in detail.
Although the results are promising, the discussion should be expanded and supplemented with more references.
Thank you for your suggestion. To the best of our knowledge, several efforts have been made to optimize the estimation of the basal crop coefficient (Kcb) in order to improve the estimation of crop evapotranspiration (ETc) for various crops. Furthermore, several previous studies focused on predicting reference evapotranspiration (ET0) using machine learning models. However, we were unable to find any references related to the estimation of ETc of vineyards using machine learning models. Therefore, in case there is any specific work that should be reviewed and discussed, we kindly ask you to indicate it to us.
Reviewer 2 Report
This is a clearly written, well-structured paper. I have attached a paper with a few suggested edits, but these are only minor. The findings would be of interest to decision makers working in irrigation, as it shows promise in being able to predict evapotranspiration and therefore improve the efficiency of water application for irrigation. I wonder how accessible the technology behind this would be, and how it might be able to be applied by irrigators in practice, but that is for further down the track.
Comments for author File: 
Comments.pdf
The English and grammar are fine in the paper. I have only made a few minor suggestions for improvement throughout the paper.
Author Response
Dear Reviewer,
We must thank you for the constructive comments and suggestions for improvement.
Below you may find all the answers and changes implemented in the manuscript.
Pag.1, Line 18: Can you add examples, i.e. such as ....
We added “One significant challenge is the accurate estimation of the basal crop coefficient (Kcb), which can be influenced by incorrect estimations of the effective transpiring leaf area and surface resistance.” to the text as suggested.
Pag.1, Line 20: the feasibility of utilizing…
Thank you for your comments and sugestions. We replaced by “their feasibility to be used ”.
Pag.2, Line 83-86: I haven't seen the conclusions presented in the introduction before. Does this match with the journal structure guidelines?
Thank you for your comments. The journal's guidelines require a highlight of the main conclusions in the introduction.
Pag.2, Line 93-94: Why was it chosen? Is it widely grown, representative, particularly significant from an ETc perspective?
The 1103 Paulsen rootstock was chosen for its tolerance to drought conditions and possible significant temporary spring humidity. In regions with limited water availability, this rootstock can access water from deeper soil layers, reducing the risk of grapevine water stress during dry periods. Is a well-suited rootstock for warm and hot climates, where grapevines are exposed to high temperatures and increased evapotranspiration rates. For this reason, is one of the most grown rootstocks in warm hot regions, namely in Portugal.
Pag.8, Line 244-246: Is it typical practice to remove outliers?
The IQR method is one of the different methods developed to detect outliers. Is a robust and widely used approach for outlier detection. This method was used to detect and avoid abnormal extreme data in the dataset without compromising the data structure.
Pag.8, Line 261: I'm unsure of the influence of removing outliers, and then modelling 'under non-standard conditions'. Perhaps more explanation of what non-standard conditions are would help.
According to Allen et al. (1998), crop evapotranspiration under standard conditions (ETc) refers to the evapotranspiration from a disease-free, well-fertilized crop that is grown under optimal soil moisture conditions, achieving full production potential within a given climate.
Pag.15, Line 433-449: Could this title be simplified so it isn't as wordy. e.g. Figure (A) Recorded ETc act meas value versus ETc act est value using GPR Exponential kernal function
Thank you for the constructive suggestion. The title was simplified as suggested.
Author Response File: Author Response.pdf
Reviewer 3 Report
It is a well-written paper for estimating actual ET of viticulture fields by using machine learning techniques and comparing its performance with traditional approach. Major comments: From these machine learning models, could the sensitivities or relative contributions of each factor to the actual ET be derived? How difficulties or what are the uncertainties for using the final trained ML models to unmeasured fields (i.e. other areas/regions)? Are there any differences in model performance between before irrigation days and after irrigation days? If there are, what is the driving force? Please add discussions in answering the above questions. Minor comments: Table 4 could be removed by just add full descriptions in the text. Line 204-206: what does this sentence mean? Fig.2: suggest add the irrigation information for the dates, i.e. before or after irrigation Line 150: What does "all measurements were taken under clear sky conditions" mean? Were the cloudy or rainy days being removed from this analysis?Author Response
Dear Reviewer,
We must thank you for the constructive comments and suggestions for improvement.
Below you may find all the answers and changes implemented in the manuscript.
Major comments:
From these machine learning models, could the sensitivities or relative contributions of each factor to the actual ET be derived?
It is indeed possible to derive the sensitivities or relative contributions of each factor to actual evapotranspiration (ET) using machine learning models. Sensitivity analysis of these models can be conducted by systematically perturbing each input variable while keeping the others constant and observing the resulting changes in ET predictions. Furthermore, future experiments are planned in different climate conditions and with various grapevine varieties. The acquisition of additional data under varying environmental conditions will facilitate the refinement of these machine learning models, bolstering their capacity for generalization. Furthermore, this endeavor will yield valuable insights into the potential fluctuations in the relative contributions of different factors across diverse scenarios.
How difficulties or what are the uncertainties for using the final trained ML models to unmeasured fields (i.e. other areas/regions)?
ML models rely on data for training, and the effectiveness of these models depends on the representativeness of the training data. If the characteristics of the unmeasured fields significantly differ from those in the training data, the model's predictions may be inaccurate. In fact, environmental conditions, soil types, microclimates, and other factors can vary greatly between regions. Consequently, ML models trained on data from one location may not capture the nuances of another region, leading to errors in predictions. These ML models can offer valuable insights and predictions; however, they should be applied to unmeasured fields or regions with caution, and their performance should be carefully evaluated and adjusted as needed to account for local conditions.
Are there any differences in model performance between before irrigation days and after irrigation days? If there are, what is the driving force?
There are differences in model performance between before irrigation and after irrigation days. The model performance has been tested and it was found that these differences are mainly related to gsw. An article exploring the effect of gsw on canopy surface evaporation is being written.
Minor comments:
Table 4 could be removed by just add full descriptions in the text.
Thank you for the suggestion. Table 4 has been removed and a full description of the ETc act meas measurements has been added to the text.
Line 204-206: what does this sentence mean?
This sentence means that with a confidence level of 95%, it can be stated that 90% of the observed values of ETc act exhibit an error magnitude of less than 0.002 mm.h-1.
Fig.2: suggest add the irrigation information for the dates, i.e. before or after irrigation
Thank you for the suggestion. The information was added.
Line 150: What does "all measurements were taken under clear sky conditions" mean? Were the cloudy or rainy days being removed from this analysis?
The mentioned expression is employed to emphasize that all measurements were conducted exclusively under clear sky conditions. There were no recorded instances of rain events during the analyzed period, and none of the measurements were taken under cloudy conditions.
Author Response File: Author Response.pdf
