Study on Real-Time Water Demand Prediction of Winter Wheat–Summer Corn Based on Convolutional Neural Network–Informer Combined Modeling

Round 1

Reviewer 1 Report

This manuscript combines convolutional neural networks with Informer models to study the water absorption model of winter wheat summer corn planting mode, and verifies the effectiveness of the model, providing a basis for predicting crop water demand.This is an interesting study, but there are still some issues that need to be addressed.

1.In section 2.1 of the research area overview, there is a lack of description of the hysical and chemical properties of the soil taken for research. It is suggested that the author add a table to introduce the physical and chemical properties of the soil.

2.The manuscript lacks a description of field experiment settings. It is recommended that the author add detailed information about field experiments to help readers understand the research content of the manuscript.

Author Response

请参阅附件。

Author Response File: Author Response.docx

Reviewer 2 Report

Note: Without line numbers, it is much more difficult for me to review the paper and make corrections as I see fit.

1 “In this context, accurate prediction of crop water demand holds paramount importance in ensuring the sustainable development of agriculture [1,2].”

In such an important statement, which is intended to justify much of the content developed by the authors in this paper,... have the authors not found other "more international" bibliographical references? I am surprised, in a negative way.

2 “Achieving this necessitates the development of efficient and precise prediction models for crop water demand combinations [4-6]”.

It seems to me that it is redundant with what the authors have expressed above and... again, the bibliographical references, with the exception of one, are certainly local.

3 “This study holds immense practical significance in addressing the challenges of water scarcity, enhancing grain production, and mitigat-ing agricultural water shortages.”

In the first paragraph? Let's see, the interest of the article, according to the authors, should appear once its "opportunity" and "justification" has been "conveniently" justified, on a solid base of publications that support it.

I think that here, it turns out... pretentious, ostentatious, vain?

The second paragraph is too long.

4 “FAO Penman-Monteith equation”. For this international formula, the authors do not add a bibliographic citation: too bad!

“The incorporation of surface resistance in this equa-tion enhances its calculation accuracy and regional applicability without requiring exten-sive parameter calibration [7]”.  And yet, for a modification of this one, they do add a quote, too... "national". I do not understand this.

In this second paragraph, the authors are discussing (presenting) different methodologies. I assume they will lead us to the justification of the one they intend to apply. I read on.

5- “the model proposed in this paper holds significant promise in enhancing the accuracy of crop water demand pre-diction and facilitating the efficient utilization of agricultural water resources.”

I beg the authors to "stop praising their own work". They should leave it to the international community to judge the outcome of their work on "other papers".

The objective(s) of the authors in this paper... ARE NOT CLEAR TO ME. And... what will the readers think? Other scientists? I think they should rephrase it specifically and synthetically.

6 The reference [19] seems to me to be poorly formatted and inappropriate for the authors' purpose.

7 “The Encoder part of the Informer model receives ultra-long sequence inputs, and the input feature vector of the model consists of a feature scalar ɑ𝑈𝑖𝑡, a fxed position embed-ding PE, and a learnable stamp embeddings SE, expressed as:

𝑋𝑓𝑒𝑒𝑡[𝑖] 𝑡=ɑ𝑈𝑖𝑡+𝑃𝐸(𝐿𝑋∗(𝑡−1)+𝑖)+Σ[𝑆𝐸(𝐿𝑥∗(𝑡−1)+𝑖)]𝑝𝑝 (2)

In the formula:i∈{1 LX}, and ɑ is the factor balancing themagnitude be-tween the scalar projection and local/globalembeddings”.

Possible typographical errors?

I'm seeing more typos but... as I don't have the line numbers, I can't point them all out.

8 I don't quite understand Figure 3. Could you clarify the caption and the x-axis further?

9 “The accuracy of the CNN-Informer crop water demand prediction model was evaluated using the Nash-Sutcliffe Efficiency Coefficient,Mean Absolute Error, Root Mean Square Error and Mean Absolute Percentage Error. The accuracy of the CNN-In-former crop water demand prediction model was evaluated using the Nash coefficient (NSE), mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE)”

Repetition with acronyms? Purge!

10 Figure 4: The titles of some axes and their units are "unreadable".

11 “As observed from the figure, there is a general similarity in the overall change pattern between the extracted values of meteorological data features for winter wheat and sum-mer maize, and the measured values. For winter wheat, the water demand is high during the sowing-tillering stage, slows down during the overwintering period, and then in-creases as the crop begins to grow roots and leaves, and tillers. During the greening-nod-ulation period, when growth is at its fastest and largest, water demand peaks. In the later stages of winter wheat fertility, particularly during the reproductive period, high temper-atures lead to increased water demand, and extreme weather conditions can lead to data outliers. Overall, the data exhibit cyclical changes over the three-year period, with extreme data points most often occurring in the summer.”

I don't know what other reviewers will think, and I don't want to think what readers will think, but... I find it "very difficult" to see in the graphs of the figures, everything that the authors say can be seen. I recommend a clarification or simplification, so that the most important of all that is going on can be seen clearly/simplified.

12 The heading in Table 3 is the editors' staff. Please specify according to its content.

13 “This demonstrates the effectiveness of the model in solving long sequence pre-diction problems.” This statement... what does it mean? Doesn't it seem to the authors to be a "conclusion"; well, if they agree with me, they should remove it from the results section and put it in the appropriate section.

14 In Figure 5 it is not easy to identify the measured values (i.e. what actually happened), so that we (the readers and this humble reviewer) can visually assess the fit of the predictive models to what actually happened.

15 In Figure 6: abnormal or outlier value? Or do the authors mean that these values do not fit a normal distribution? So? It seems to me that this needs to be properly clarified.

16 In the conclusions section:

“This paper presents an innovative approach to predicting crop water requirements using a long series prediction model. This model combines a CNN feature extraction mod-ule with an optimized Informer based on meteorological data and deep learning techniques. The results demonstrate that the combined model can effectively capture the long time series information of crop water demand. This improvement in accuracy enables pre-cise regulation of crop irrigation water use, ultimately alleviating issues such as agricul-tural water scarcity.”

Where are the conclusions here? Does it not seem to the authors that, firstly, they give a brief summary of their work and, secondly, they praise it?

I think that: (1) the summary is not done in the conclusions section; and (2) the general praise of their work should not be written by the authors in the paper itself, but should be done by "other researchers" in other papers, citing it as deserving the corresponding justifications. Fix this!

17 “First, it extracts global features from meteorological data, reducing the length of in-put sequences for each layer”

How have the authors specified this in the results and/or discussion section? It is not at all clear to me!

In general, the description of the model used was not entirely clear to me and, in this sense, I do not know how another scientist could replicate this same approach (but with other input data).

Conclusion number 2: it does seem to me to be a "real conclusion".

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Well, I think the paper has improved and, although I could make more comments, I will not do so to speed up the publication process. I ask you to stop praising your own work in future papers. I would have liked more clarity when explaining the implemented model (attach the code as a supplementary file?). Attention in the figure of the explanation of the model that I think I have seen a typographical error.

Author Response

Please see the attachment.

Author Response File: Author Response.docx