Automatic Mapping of Center Pivot Irrigation Systems from Satellite Images Using Deep Learning

Round 1

Reviewer 1 Report

This study attempts to map center pivot irrigation using a deep learning approach based on the PlanetScope images. Taking the two regions of Brazil as a case study, the authors used a modified CNN architecture for the automatic detection and mapping of center pivot irrigation systems. It is interesting and could contribute in literature. However this manuscript still needs some improvements before publication.

1. In introduction part, authors should add more literatures in other regions to clarify the importance of your method and study area. The second paragraph of the introduction quotes too much in one reference. I think authors also need to argue why irrigation requirement is important at regional scale. Please see this reference for more details. (Y.Q. Liu, W. Song, X.Z. Deng. Spatiotemporal patterns of crop irrigation water requirements in the Heihe River Basin, China. Water, 2017, 9(8), 616)

2.The “Materials and Methods” section needs to be reorganized. Although the study adopted the images with cloud cover below 1%, atmospheric correction remains an important preprocessing to reduce classification uncertainty. Why you used the median image mosaics? Why not choose random samples for training, validation, and testing? What are the advantages of your methods? Authors should elaborate on the principles, content, and selection basis of some key methods or models.

3. Authors should divide the “3. Results and Discussion” into two sections. Please remove all the explanations to the discussion part.

4. My primary concern is whether your detection method is robust in other regions, or data sources, or training samples. Thus, more data analysis or explanation is required in the manuscript to warrant further consideration for publication.

5. The manuscript could benefit from a careful editorial review on the language.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Line 112: It would be better if you cited the sources 18-20 not only by the figures in brackets, but in the style like “such as …. Implemented by …. [18], something else used for … by …. []19]’ and so on.

Line 130: An excessive article “the” after “the feature maps”

Line 116 – ReLU, and in line 132 – ReLu. So, what spelling is correct?

Lines 161-164: Seems that formula 2 and 3 are similar but are used for the calculation of different indices, and, therefore, have to provide different results. Please, check and correct or provide additional explanation regarding these equations.

Line 186: Seems to me that it should be “on several factors”

Line 191: Seems that it should be “well-developed crops”

Line 193: “proposed by Zhang et al.”

Line 242-243: Check the use of prepositions, verbs and their forms in the sentence.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Dear Editor,

The technical note "Automatic Mapping of Center Pivot Irrigation Systems from Satellite Images Using Deep Learning" presents a method to automate the detection of center pivot irrigation systems, as well as the quantification of the area under such irrigation systems using satellite images. The methodology is shown to work very well in the selected test area in Brazil.

The manuscript is very well written and might be suitable for publication in its current form. Notwithstanding this, I provide a short list of recommendations to further improve its quality in my comments to the authors.

I would like to stress the fact that the paper is very well written and structured. The definitions are clear and the logic statements are sound. It presents clearly and appropriately the methodology, including the use of very novel technical terms from the field of Deep Learning. It describes the dataset construction, the modifications in the network's architecture and the parameters used for training the model very precisely in the correct terms, making it easy to understand the technical details easily. Furthermore, it incorporates particular techniques of the field, like data augmentation, the use of the U-net itself, and different evaluation metrics.

Regarding the irrigation part, the manuscript shows the usefulness of the proposed system in the practice, also comparing with similar studies. It addresses the effect of soil properties, coverage and geometrical irregularities found in these systems.

I hereby recommend to ACCEPT AFTER MINOR REVISION the manuscript "Automatic Mapping of Center Pivot Irrigation Systems from Satellite Images Using Deep Learning".

Kind regards

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors have revised the paper according to all my comments. The present form can be accepted after the language editing service.

Author Response

Please see the attachment.

Author Response File: Author Response.docx