Cascading Machine Learning to Monitor Volcanic Thermal Activity Using Orbital Infrared Data: From Detection to Quantitative Evaluation

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this paper, the authors propose a robust Machine Learning approach to accurately detect, recognize, and quantify the high-temperature volcanic features using Sentinel-2 MultiSpectral Instrument imagery. This is a good work to provide an important contribution to the monitoring, mapping, and characterization of volcanic thermal features using the thermal infrared data.

My main comments are as follows.

1. As the key outputs to train the network, how to identify NVA, ITA, or ETA from the images?

2. The contents in Section 5 are lengthy. I suggest to do the discussions with sub-title according to the scientific problems. Moreover, though the authors gave abundant descriptions of their work, the scientific significance of the expressions deserves to be further summarized.

3. The contents in Section 6 should be concise, e.g., using several items to give a brief description.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

In this work an innovative technique for monitoring thermal anomalies in volcanic areas is shown. the combination of two different machine learning approaches allows to  obtain information on the ongoing volcanic activity in an unsupervised manner and relatively quickly. The work is well written, although, in my opinion, the introduction, which is too long and repetitive, should be revised. Some notes:

1 the name of the Etna Observatory is INGV OE not EVO, as indicated in the introduction

 

2 Report the limits of anomaly detection algorithms based on simple thresholds

 

3 Add a flowchart showing the entire process from acquisition to validation of results

 

4 on page 21 line 352 there is an incorrect reference to a figure

 

5, figure 8 needs to be redrawn: the lines of the graphs are barely visible and the characters are too small

 

6 figure 9 needs to be redrawn: the lines of the graphs are barely visible and the characters are too small

 

some questions that should be answered in the manuscript:

 

1 Why didn’t you use a single DL model to classify and segment the image at once?

 

2 How much is the sensitivity of the algorithm in discriminating between the isolated and extended thermal anomalies ?

 

 

3 Is the tool available online and ready to be used?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript intitled "Cascading machine learning to monitor volcanic thermal activity using orbital infrared data: from Detection to Quantitative  evaluation"

The work is focusing on continuous advances in remote sensing thermal imaging and the growing capacity of Artificial Intelligence (AI) techniques bring extraordinary opportunities to volcano monitoring that can directly benefit from automatic procedures for processing huge amounts of satellite data.

The autors devliir that taking AI in operational monitoring as the Volcano Observatory centers may accelerate the response to volcanic hazardous events. Specifically, in the case of remote thermal image as massively large raw thermal images can be turned into useful information about the status of the volcanic activity in very short time. Infact the authors over the past decades thermal in- frared remote sensing observations have been effective in achieving extensive improvements in volcano monitoring from space. In particular, thermal infrared satellite-based.

The Auhors propose an new step using the potentiality of DL in automatically learning spatial and spectral attributes from images to distinguish the type of volcanic activity as either isolated volcanic thermal anomalies, due for instance to intra-crater activity, degassing, new vent opening, explosive activity, or extended volcanic effusive activity, including also the lava lakes, domes and  flow.s

In my opinion, I consider that the focus of manuscript is sound with the aims of the jornal. The organization of the paragraphs allows a good understanding of the methodological approach developed. All figures and tables included in the text are necessary and appropriate. The abstract accurately reflects the contents of manuscript and the purposes of study are stated clearly in the introduction section.  In general, I believe that the manuscript is suitable for publication but only after a revision of the English by a Native speaker.

Comments on the Quality of English Language

English requires review by a native speaker.

Author Response

The Reviewer #3 positively wrote: “In my opinion, I consider that the focus of the manuscript is sound with the aims of the journal. The organization of the paragraphs allows a good understanding of the methodological approach developed. All figures and tables included in the text are necessary and appropriate. The abstract accurately reflects the contents of the manuscript and the purposes of the study are stated clearly in the introduction section.  In general, I believe that the manuscript is suitable for publication but only after a revision of the English by a Native speaker.

Authors: The authors sincerely thank the reviewer for recognizing the usefulness of combining machine learning techniques and infrared satellite data to monitor volcanic thermal activity. The English of the manuscript was revised by a native English speaker.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors gave  a good response to my comments. The revised paper is well in written and the content is scientific. I have no new comments.