Integration of Color Analysis, Firmness Testing, and visNIR Spectroscopy for Comprehensive Tomato Quality Assessment and Shelf-Life Prediction
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsI have had the pleasure of reviewing your manuscript titled " Postharvest tomato fruit shelf life and quality assessment based on color, firmness and visNIR spectroscopy combined with chemometrics." I must commend your comprehensive and robust approach to this pertinent issue in the realm of postharvest.
Your study design, especially the inclusion of DLP NIR Nano Scan, enabling a holistic understanding of the variables at play. However, there are some points to be considered to get your MS in more suitable form
Point1: The suggested title of the MS should accurately represent the performed experiments without unnecessary information. I would suggest the title to be “Integration of Color Analysis, Firmness Testing and Multi-Spectral visNIR Spectroscopy for Comprehensive Tomato Quality Assessment and Shelf Life Prediction”
Point 2: Please remove the detailed instrument specifications from the abstract section (line 13-16). Specific information about spectroradiometer models, manufacturers, and technical specifications (PSR+ 3500, Spectral Evolution and DLP NIR Nano Scan, Texas Instruments) should be moved to the Materials and Methods section.
Point 3: My next comment pertains to line 83-88 where you illustrated the main objective of the study. The hypothesis of your work needs to be more justified. Please rephrase.
Point 4: My next comment pertains to line 95-100 where you described the classification of 1280 fruits into 4 different ripening stages. The clarity of this part of M&M is very crucial to the reader. The criteria, fruit numbers per treatment, and methodology used for this classification remain ambiguous. Please justify
Point 5: in Line 176 you mentioned that each value represents the mean of 80 fruits in each ripening stage and storage day, please double check this number !!
Author Response
Please see our responses in the file attached.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe initial experimental work is interesting and detailed, with a large number of samples and the objective of examining the different evolution of tomatoes throughout their shelf life based on their initial ripening stage. As a starting point, I find it highly relevant. However, the development of the paper includes many aspects and perhaps remains superficial. In fact, I believe it would be more appropriate to divide it into two complementary articles:
- Article 1: This would include the destructive tests and their relationship with the non-destructive measurements of high-resolution, broad-spectrum NIR.
- Article 2: Focusing on the NIR and its different predictive capabilities when reducing the spectral range of high-resolution NIR and transitioning to the low-resolution minispec.
By trying to include everything in the same paper, it lacks depth, and the content presented gives the impression of corroborating already known aspects rather than providing new knowledge to the field.
Article 1
To properly support Article 1, I would emphasize and justify in the introduction the lack of studies on the different evolution of tomatoes during their shelf life based on their initial ripening stage. I would identify the spectral wavelengths most related to the physicochemical parameters determined, such as color, firmness, pH, and SSC/TA ratio, and establish a predictive model capable of classifying tomatoes based on their shelf life. In the current paper, the presentation lacks key statistical values for the PCR and PLSR models. It should include not only the R2 values but also other statistical metrics that define the predictive capacity of the models, such as R2, RMSEP, RMSEC, and RMSECV.
Additionally:
- The classification errors of the model should be reported, including false positives, false negatives, etc.
- It should be justified whether a predictive model based on an initial tomato color classifier at the horticultural center is more appropriate, or whether a single predictive model should be applied regardless of the initial sample color.
Article 2
A specific NIR-focused study could be developed to evaluate predictive capacity across different spectral ranges and resolutions. This has already been extensively studied, so the specific contribution must be clearly defined, either in the area of tomato shelf life (if this has not yet been done and can be properly justified) or by leveraging more novel or advanced chemometric methods.
Conclusion
I believe that as a chapter of a thesis, it is a good starting point, but the novelty and scientific contribution are not sufficiently justified for publication in its current form.
I am attaching the document with some notes and formatting corrections in case they are of interest to the authors
Comments for author File: Comments.pdf
I think the English could be improved, although I fully understood the content. However, I am not a native English speaker, so I cannot be more specific
Author Response
Please see our responses in the file attached.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors1. English abbreviation should be enclosed in a parenthese after the corresponding full name when it first appear. Please check the entire manuscript.
2. The sorting of references is chaotic.
3. What is the number of wavelengthbands in the 900-1700nm range for the portable spectroradiometer?
4. The statement on line 147 is inaccurate,
5. What is the difference between DFx and DF in Table 1? and what is the difference between SSy and SS? is the symbol z *** the same as *** z and ***?
6. Which subgraph in Figure 1 is Figure 1A? and which subgraph is Fig.1B? What does the @ 25C in the horizontal axis title mean? is it 25℃, but it is different from the storage temperature of the sample, which is 22℃.
7. Which subgraph in Figure 2 is Figure 2A? and which subgraph is Fig.2B?
8. What is the RMSECVs of the prediction model in Figure 3 and Figure 4? What are the harvest stages corresponding to each of the four subgraphs?
9. In Figure 5, Figure 6 and Figure 7, What are the harvest stages corresponding to each of the four subgraphs?
10. What is the RMSECV of the prediction model in Tables 2 and 3?
Author Response
Please see our responses in the file attached.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsThis manuscript focused on the external and internal changes in tomato fruit harvested at various ripening stages monitored by visNIR. The color, firmness, pH and soluble solids to titratable acidity ratio during storage were analyzed. The work is meaningful for the tomato industry and the applications of NIR spectroscopy. If more analysis were addressed on the results, the work would be greater.
Specific comments:
1. The NIR spectroscopy has been wildly applied in the quantity analysis of fruit and vegetables for decades. The introduction section could be more powerful by focusing on the NIR measurement on tomato or the applications related to shelf-life.
2. Whether the changes of color, firmness, pH and soluble solids to titratable acidity ratio were consistent or similar with other published researches?
3. The Regression coefficients of cross validated data of the shelf life prediction in the table 2 were not that good as well as other published works.
4. The Regression coefficients of cross validated data of the pH and SSC/TA prediction in the table 3 were less than 0.90.
5. More analysis would be addressed on whether the regression coefficients could be increased to 0.90 and above.
6. The establishment of regression model should be introduce in the method section.
Author Response
Please see our responses in the file attached.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsAfter careful examination of the revised manuscript, I am pleased to report that the authors have thoroughly addressed my previous comments and implemented substantial improvements. The revisions have significantly enhanced the quality and clarity of their work. Their research makes a valuable contribution to the field of agronomy, particularly in the areas of tomato postharvest and self life evaluation
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for Authors1. Figure 1 is not displaying correctly.
2. On the x-axis of Figures 1 and 2, I still don't understand what the @ symbol is meant to represent. If it indicates temperature, please use T instead.
3. Include the numerical value of the LSD in the figure captions, as indicating this value with the length of a vertical line is not precise enough.
4. To get an idea of the data dispersion at each point in Figures 1 and 2, error bars indicating either the standard deviation or the confidence interval should be included at each point.
5. Figures 1 and 2 show that, except for green tomatoes, the evolution of the variables over time is not linear in the other categories. Therefore, using a linear regression predictive model will not yield good results and simply demonstrates the non-linearity between the variable and shelf life. Fitting the data to a Logistic or Gompertz-type model (increasing or decreasing depending on the variable under study) would result in better fits. The asymptote to which these models tend, and the inflection point on the curve, would indicate the day of shelf life after which that variable ceases to be useful for estimating the tomato's shelf life.
6. In Table 2 for NIR Nano scan data if you are estimating the shelf life of tomatoes, which ranges between 0 and 12 days, and the RMSECV is 4.499 days, means the model, on average, is off by approximately 37.5% of the total range.
This error represents more than one-third of the possible range, which can be considered high in many contexts.If you need precise predictions for supermarkets or logistics planning (e.g., to prevent tomatoes from spoiling on shelves), an error of 4.5 days is high. It could lead to economic losses or selling spoiled products.
To better interpret the RMSECV, standardize this error by calculating the Coefficient of Variation of the RMSE (CV(RMSE)), which expresses the error as a percentage of the average shelf life. If assume the average shelf life of tomatoes is between 6 and 12 days. If the average shelf life is 6 days:
A 75% error indicates that the model's predictions deviate significantly from the actual shelf life, suggesting the model is highly inaccurate in this scenario.
If the average shelf life is 12 days: A 37.5% error is still high, but more acceptable compared to the previous case. However, it would still indicate room for improvement, especially if precise shelf life predictions are critical for decision-making.
7. A similar analysis could be done in Table 4 with the Nano scan data to predict pH and SSC/TA.
8. I still believe that the article is too long, that it could actually be divided into two separate approaches, and that it has some methodological flaws that need improvement. In fact, the estimation models using only NIR, and especially Nanoscan, require applying another approach to improve the estimation results.
Comments on the Quality of English Language
..
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsAccept in current form.
