Review Reports
- Margherita Valentini 1,
- Paolo Brotto 2 and
- Rita Vecchiattini 1,*
- et al.
Reviewer 1: Josefina Garcia-Leon Reviewer 2: Mohamed S. Emara Reviewer 3: Anonymous
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors
The topic is very interesting and topical, but its depth could be improved by addressing the following issues:
- The state of the art (1.3.1) briefly comments on the situation without going into detail and leaves a sentence in Italian that is later in English. This must be corrected.
- There is a great deal of work being done in the city as part of the Genoa Action Plan related to the research of the article that I find relevant and is not mentioned in the bibliography.
- The figures do not clearly show the work. For example, in Figure 3, the legend is not legible; it is not possible to see what the colours blue, green, red, or yellow indicate. Or in Figure 11, it is not clear what the X-axis or Y-axis indicates. Does the X-axis represent days, hours of the day, or years? On the other han, the Figures 4 and 7 do not contribute anything significant; they may be removed or modified to contribute to the research.
- It would include as a result any query or result from the interactive digital twin in graphical form. Because in Figure 10 it is not possible to see any results.
Author Response
Si prega di consultare l'allegato.
Author Response File:
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for Authors The manuscript presents a comprehensive and methodologically robust workflow for environmental monitoring, degradation assessment, and digital management of built heritage through the EN HERITAGE platform. The study demonstrates strong interdisciplinary alignment, combining environmental sciences, remote sensing, photogrammetry, HBIM, and AI-based semantic segmentation. The workflow, from data acquisition to analysis and Digital Twin publication, is clearly presented and addresses the need for scalable, data-driven diagnostic systems. However, the interdisciplinary approach, although central to the study, is not entirely justified. The manuscript shows how environmental and architectural datasets can be combined to map degradation and support preventive conservation within a Decision Support System. Nonetheless, several aspects require clarification to improve scientific depth and repeatability.Major Issues
- Limited quantitative evaluation of the AI-based method
The manuscript should include performance metrics such as IoU, F1 score, and accuracy, together with confusion matrices, to demonstrate the reliability of the segmentation process. - Insufficient interpretation of environmental data
While environmental trends such as Black Carbon cycles are identified, the discussion should better relate them to: -
- specific material degradation processes
- pollutant impacts on roof materials
- relationships between environmental pressures and detected degradation patterns
- Underdeveloped discussion
The discussion lacks a deeper post-analysis of the interdisciplinary datasets and does not adequately justify the multidisciplinary approach. - Scalability claims not fully demonstrated.
The manuscript asserts scalability, but it does not demonstrate generalisation to other materials, climates, or building types. A brief reflection on expected limitations or necessary adaptations would improve transparency.
- Some sections are verbose and would benefit from more concise scientific writing.
- A more quantitative approach is recommended, particularly in the environmental monitoring results, with additional graphs and numerical summaries.
- Ensure all figures, including Figures 10 and 11, are clearly labelled and of sufficient resolution, as axis labels are difficult to read.
- A summary table would assist readers, including:
- sensors used
- environmental parameters measured
- survey outputs
- AI related parameters
- For reproducibility:
- Specify dataset size, number of annotations, and data split proportions.
- Consider providing a small sample of the annotated dataset if possible.
Author Response
Please see the attachment.
Author Response File:
Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors- The paper describes in detail the UNet-based semantic segmentation pipeline (RGB / RGB+NIR, tiling, TTA, augmentation, etc.), but no quantitative performance metricsare reported (e.g., IoU, mIoU, pixel accuracy, F1-score per class, confusion matrix). Please report the size of the training/validation/test sets (number of images/tiles, number of annotated roofs, class distribution). Provide quantitative metrics on a held-out test set and, if possible, compare with at least one or two baselines, such as: Plain UNet without TTA / without IR channel; A simple thresholding/edge-based or classical CV pipeline.
- The manuscript repeatedly refers to the platform as following the Digital Twin paradigm, but the description currently resembles a federated data integration and decision support systemrather than a full digital twin (which typically implies a continuously synchronized, bidirectional connection between the physical asset and its virtual counterpart, including predictive simulation and feedback loops).
- Please explicitly map your system to commonly accepted Digital Twin definitions(e.g., real-time data flow, state synchronization, predictive capabilities, “what-if” simulation and potential control).
- Clarify whether the platform is currently used only as an information viewer/DSS, or if there is any form of dynamic updating, scenario simulation, or automated decision support that would justify the term “digital twin” beyond a metaphorical sense.
- If the system is still closer to a “prototype DSS + HBIM + IoT platform”, consider tempering the claims or clearly stating that it is a step towardsa full Digital Twin.
- The State-of-the-Art section is rich and well structured, but the specific novelty of EN-HERITAGEcompared with existing HBIM+IoT+AI frameworks is not clearly highlighted.
- Please add a concise paragraph at the end of the Introduction explicitly stating:What is conceptually new (e.g., particular integration of BC monitoring with roof degradation mapping; a specific HBIM degradation data schema; a certain multi-source pipeline). What is technically new (e.g., any new data model, algorithmic improvement, or platform functionality that goes beyond the literature you cite).
- The environmental monitoring and modelling part describes advanced methods, including CatBoost regression and a combination of Bayesian models with Physics-Informed Neural Networks for Black Carbon prediction.
However, the Results section mainly offers qualitative statements (e.g., daily cycles, peaks consistent with Genoa) and one figure of BC temporal profiles. - Please provide numerical performance indicatorsfor the regression / PINN models (e.g., RMSE, MAE, R² for BC estimation, uncertainty intervals from Bayesian inference).
- Clarify how these models concretely feed back into heritage decision-making(e.g., threshold-based risk levels for materials, spatial maps of expected degradation rate)
- The HBIM part is conceptually sound, but the description remains high-level.
- Please provide a more precise description or a schematic diagram of the Property Setsand data schema for degradation instances (e.g., attributes, link to IFC entities, time parameters T_Survey_T0/T1, T_Validation_T0/T1).
- It would be useful to show example queries or usage scenarios from the end-user’s perspective (e.g., “select all roofs with exfoliation level > X and BC exposure above Y”).
- The platform is described as a Decision Support System for heritage authorities, but no user study, feedback, or real decision caseis presented.
- Even a small qualitative evaluation (e.g., interviews or questionnaires with a few professionals) or one documented example of how the platform helped prioritize interventions would significantly strengthen the paper.
- Sections 3 and 4 are well aligned, but often repeat descriptive contentwithout deep analysis (e.g., “the methodology proved effective…”).
- Consider condensing repeated descriptive phrases and focusing the Discussion on critical reflection: limitations, failure cases of the AI model, edge cases in environmental monitoring, and specific future improvements.
- In the keywords, “AI tecnology” should be corrected to “AI technology”.
- In Section 1.3.1 there is an Italian sentence left in the text. Like,“Il tema della conservazione ha assunto rilevanza anche a scala locale.”
followed immediately by the English translation. This duplication should be resolved (keep only English). - Please ensure that all figures include necessary elements: scale bars (for orthophotos), legends (for BC time series and maps), and clear indications of north/orientation where relevant.
- Some figures are only briefly referenced in the text; consider adding a bit more description in captions so they are self-explanatory.“2.4. Management platform for diagnostic in the conservation and use of historic buildings through digital twin technology” – the phrase “for diagnostic in the conservation” is slightly awkward. Consider “for diagnostics and conservation management of historic buildings using digital twin technology”. Ensure consistent capitalization in headings (e.g., “Environmental monitoring” vs. “enviromental monitoring”)
Author Response
Please see the attachment
Author Response File:
Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe text has been sufficiently improved to be published.
Reviewer 2 Report
Comments and Suggestions for AuthorsI have no further comments. The authors have adequately addressed all previously raised concerns. One minor remaining issue is that the font size in Fig. 13 should be reconsidered for improved readability. Additionally, if the authors are inclined to use American English, the term “modeling” is preferred over “modelling” for consistency.
Reviewer 3 Report
Comments and Suggestions for Authorsno other concerns.