A Semi-Automated Machine-Learning Tool for Assessing Building Phases: Discriminant Analysis of Mortars from the 2022 Excavation at the Sarno Bath Complex in Pompeii
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsDear authors,
follow just some comments.
The manuscript considers a set of 15 mortar samples recovered at the Sarno bath Complex / Pompei. The research is supported by previous works, and a rich dataset was already available. This allowed a clear chrono-technological contextualization of samples within a specific temporal period.
Goals: Goals are clear and well explained.
Archaeological contextualization: Clear and well explained.
Structure: The manuscript is well-conceived. Every section is clear, well explained, and the most significant information are reported. Every part is easy to read and connected to the next one.
Figures: Illustrations show good quality, clear, and the reader can easily link them to the text.
Methods: The methods applied are appropriate. Q-XRPD is not commonly applied, so in my opinion is a plus. The combination Q-XRPD/statistic worked really well.
Results: I am wondering if authors applied any kind of normalization to the dataset before the application of PCA/Discriminant analysis. PCA and discriminant analysis results are coherent. Discriminant analysis is generally applied in charred seed morphometry, and hundred of observations are normally employed. Nevertheless, It is not possible to do the same with mortars. Thus, I consider, it was a good application, showing that PCA can be considered just one of the possible approaches.
Discussion: Fluid and fully supported.
No more comments.
Author Response
Comment:
Dear authors,
follow just some comments.
The manuscript considers a set of 15 mortar samples recovered at the Sarno bath Complex / Pompei. The research is supported by previous works, and a rich dataset was already available. This allowed a clear chrono-technological contextualization of samples within a specific temporal period.
Goals: Goals are clear and well explained.
Archaeological contextualization: Clear and well explained.
Structure: The manuscript is well-conceived. Every section is clear, well explained, and the most significant information are reported. Every part is easy to read and connected to the next one.
Figures: Illustrations show good quality, clear, and the reader can easily link them to the text.
Methods: The methods applied are appropriate. Q-XRPD is not commonly applied, so in my opinion is a plus. The combination Q-XRPD/statistic worked really well.
Results: I am wondering if authors applied any kind of normalization to the dataset before the application of PCA/Discriminant analysis. PCA and discriminant analysis results are coherent. Discriminant analysis is generally applied in charred seed morphometry, and hundred of observations are normally employed. Nevertheless, It is not possible to do the same with mortars. Thus, I consider, it was a good application, showing that PCA can be considered just one of the possible approaches.
Discussion: Fluid and fully supported.
No more comments.
Response: We really thank the reviewer for the comments and compliments to our work.
Reviewer 2 Report
Comments and Suggestions for AuthorsAfter reviewing this manuscript I think that it should be published as it is, without any revisons.
Topic of this research is excelent.
There is not much published work about composition of arhaeological mortars from ancient site in mediteranian region, so this work certainly fills the gap in the field.
Methodology that authors use is inovative and up to date.
Conclusions are consistent with evidence provided in the paper.
References are apropriate for this work.
Author Response
Comment:
After reviewing this manuscript I think that it should be published as it is, without any revisons.
Topic of this research is excelent.
There is not much published work about composition of arhaeological mortars from ancient site in mediteranian region, so this work certainly fills the gap in the field.
Methodology that authors use is inovative and up to date.
Conclusions are consistent with evidence provided in the paper.
References are apropriate for this work.
Response: We really thank the reviewer for the comments and compliments to our work.
Reviewer 3 Report
Comments and Suggestions for AuthorsComments and suggestions are included into the word document.
Comments for author File: 
Comments.pdf
Author Response
Comment 1:
Abstract: No comments.
Introduction: The introduction approaches the problem simply and logically. I have no comments. The cited bibliography and the methods used are sufficiently presented.
The previous research: The presentation is simple and clear, orienting the reader to the purpose of the research in the article.
The 2022 excavations at civ. 21: Same comment.
The 2022 sampling: The chapter systematically presents the archaeological and architectural structure data, providing a clear picture of the studied samples through the table.
Response 1: We thank the reviewer for the comments.
Comment 2:
Analytical workflow and data processing:
The analysis is well presented. However, the sensitivity of the method could have been specified; detection limits for groups of compounds which could be determined.
Response 2:
The program used for both PCA and DA was reported in the methodological part (Statgraphics Centurion PRO 19). The specific sentence has been rephrased in a clearer way. A sentence reporting the limit of the technique has been reported where appropriate
Comment 3:
5.1 PCA on Mineralogical data:
The justification and benefits of applying the method are well presented. However, the program on which the analyzes were run should be specified. The interpretation of the results is correct, and the conclusion from this phase of the research justifies the completion of the following stages. Perhaps some ambiguities in assigning samples to a group also come from the limitations of the methods (analytical and predictive).
5.2 Discriminant analysis on Mineralogical data:
Same comments.
Response 3: The limit of detection and the ambiguity in assigning samples, after PCA and DA, in explaining the dataset variability has been reported in a new sentence.
Comment 4:
5.3 OM and SEM-EDS analyses (data verification):
The chapter presents in detail the additional analyzes and how they can be interpreted. It also provides new clues on the correlation of samples that is more difficult to assign to a group.
Discussion:
The discussions gradually involve the multidisciplinary combination of information. The division into sub-chapters supports this gradual transition from observations in the context of samples, to observations at the level of large events in the studied historical interval, with an emphasis on volcanic activities in connection with earthquakes.
Conclusive remarks:
It summarizes well all the research and identifies the usefulness of combining data from experimental analyzes, with techniques related to mathematical statistics.
Response 4: We thank the reviewer for the comments.
Reviewer 4 Report
Comments and Suggestions for AuthorsThe article presents a semi-automated approach that employs discriminant analysis as a machine learning tool to identify ancient construction phases, marking a novel contribution to the field of archaeology.
By integrating advanced analytical methods (QPA-XRPD, PCA, OM, SEM-EDS), the study enables a comprehensive analysis of mortars and construction materials, providing a robust scientific framework. Statistical results are corroborated by complementary techniques, enhancing the reliability of both the data and its interpretations.
The findings significantly advance our understanding of the construction chronology of the Sarno Baths Complex, resolving stratigraphic ambiguities and supporting efforts in restoration and conservation.
Moreover, the mortar analysis offers valuable insights for the historical reconstruction of Pompeii, a site of global archaeological importance.
Suggestions for improvement:
- While the methods employed are effective, the attribution of certain samples (e.g., sample M332) remains uncertain, underlining the necessity for further analyses or more extensive reference datasets. Are there additional complementary techniques that could help address these uncertainties? The authors should briefly discuss possible approaches to manage such cases.
- The findings rely heavily on the local context of Pompeii and the datasets available, which limits their applicability to other archaeological settings. It would be beneficial for the authors to explain how this model could be adapted or generalized for use in different contexts.
- The manuscript lacks a detailed discussion of how the results might influence related disciplines, such as structural engineering or preventive archaeology. Expanding on this point with a few sentences would strengthen the paper's interdisciplinary impact.
-
Some suggestion about the English Language
Abstract:
“These results confirmed the reliability of the semi-automated sample processing proposed in this research, adopting Discriminant Analysis as a machine-learning-based tool for defining construction phases in Pompeian contexts.
“Confirm” instead of “confirmed”
Introduction:
“Archaeometric research on ancient mortars, investigated adopting a diverse array of analytical techniques, have become increasingly common in the last decade."
“Has become” instead of “have become”
Section 6. Discussion:
"The analyses on 2022 mortars allowed to attribute with certainty the sampled structures to a specific building phase on the basis of the characteristics of the mortar employed."
Better: "The analyses of 2022 mortars allowed the sampled structures to be attributed with certainty to a specific building phase based on the characteristics of the mortar employed."
Author Response
Comment 1: While the methods employed are effective, the attribution of certain samples (e.g., sample M332) remains uncertain, underlining the necessity for further analyses or more extensive reference datasets. Are there additional complementary techniques that could help address these uncertainties? The authors should briefly discuss possible approaches to manage such cases.
Response 1: We thank the reviewer for this suggestion. We added a sentence in the manuscript discussing other possible approaches to better define the attribution of certain samples.
Comment 2: The findings rely heavily on the local context of Pompeii and the datasets available, which limits their applicability to other archaeological settings. It would be beneficial for the authors to explain how this model could be adapted or generalized for use in different contexts.
Response 2: We thank the reviewer for this suggestion. We added a sentence in the manuscript explaining possible application of the protocol to other case studies.
Comment 3: The manuscript lacks a detailed discussion of how the results might influence related disciplines, such as structural engineering or preventive archaeology. Expanding on this point with a few sentences would strengthen the paper's interdisciplinary impact.
Response 3: We thank the reviewer for this suggestion. We added some sentences in the manuscript about the influence of the research on other disciplines.
Comment 4:
Some suggestion about the English Language
Abstract: “These results confirmed the reliability of the semi-automated sample processing proposed in this research, adopting Discriminant Analysis as a machine-learning-based tool for defining construction phases in Pompeian contexts.
“Confirm” instead of “confirmed”
Introduction:
“Archaeometric research on ancient mortars, investigated adopting a diverse array of analytical techniques, have become increasingly common in the last decade."
“Has become” instead of “have become”
Section 6. Discussion:
"The analyses on 2022 mortars allowed to attribute with certainty the sampled structures to a specific building phase on the basis of the characteristics of the mortar employed."
Better: "The analyses of 2022 mortars allowed the sampled structures to be attributed with certainty to a specific building phase based on the characteristics of the mortar employed."
Response 4: We edited the manuscript according to the suggestions of the reviewer and we rephrased the suggested sentences.
