Dear authors,

Thank you for your revised submission to AMT. I asked a referee from the previous round to take a read through and based on that, and my own reading, feel that the study needs some further revisions before it can be accepted for publication. There are several points for clarification, and comments relating to tying the method back into techniques in the existing literature.

I have copied the reviewer's comments below (and please see the attached pdf file):

"The authors have significantly improved the clarity and readability of the manuscript by incorporating the comments from both previous reviews. As the authors emphasized in their reply to the last comment, this manuscript is supposed to focus on the method and a detailed analysis of the TSI long-term dataset is envisaged for future publications. To give due consideration to this purpose of the manuscript, it is necessary to further relate the presented method to the existing literature. For example:
• Gnanadesikan, R. (1977) Methods for Statistical Data Analysis of Multivariate Observations, Wiley. ISBN 0-471-30845-5 (p. 83–86)
• Alpaydin, Ethem (2010). Introduction to Machine Learning. MIT Press. ISBN 978-0-262-01243-0 (chapter 5, 10)
Please ensure that already established techniques are described using the proper technical terms where applicable. I would argue that the presented method to detect 22deg halos on TSI images is indeed a binary classification (as opposed to the statement on P16, L7) but based on statistical (multivariate) analysis rather than machine learning. However, there is considerable overlap between both fields (statistical analysis vs. machine learning), especially considering training and testing of the algorithm as well as the related technical terms.

In my opinion, there are 3 major points that have to be addressed before the manuscript can be published concerning training and testing of the algorithm, citing relevant literature, as well as the choice of the sky type categories. These points are further discussed in the referee report."

Please let me know if you have any questions; I look forward to receiving your next revision.

Best wishes,

Andrew

Dear authors,

Thank you for submitting this revision of your manuscript. After consideration of your Response to Reviewers and the revision, I feel that you have addressed the reviewer’s remaining concerns and so am happy to accept your manuscript for publication in AMT.

While going through the manuscript I noticed the doi provided for your code (10.5281/zenodo.8475) in the “code availability” section directs to https://zenodo.org/record/8475 , which appears to be incorrect. From the Boyd (2018) reference in the bibliography I believe the correct url is https://zenodo.org/record/2226125 which has doi 10.5281/zenodo.2226125. I would appreciate if you could make this correction before uploading your files for the article production process. Please let me know if you have any questions.

Best wishes,

Andrew