A Fruit Colour Development Index (CDI) to Support Harvest Time Decisions in Peach and Nectarine Orchards

Round 1

Reviewer 1 Report

The manuscript presents an interesting study for determining the CDI for fruit color determination. Below are my comments:

1. I suggest that the abstract should briefly describe the method used in the paper and the conclusion should list more detail about what the paper major achieved. The authors should rewrite the abstract and conclusion.
2. In the introduction, I suggest that the author should discuss in detail some major-related papers that the author aspired to or a great contribution to the field.
3. I am not sure about the importance of Fig. 3 and 4. which the authors should consider ignoring.
4. Fig. 5: the authors state that " detection fruit are shown with the red detection boxes", however, I found nowhere on the fig for the red direction boxes.
5. Line 283: what is the meaning of notation n square?
6. Fig. 6 legends: the authors state "Error bars represent 95% confidence intervals of the estimates". The author should explain more about this statement.
7. Fig. 7 legends: the authors state "Error bars represent standard errors the mean and different letters show significant differences..". The authors should explain the difference between of error bar from Fig. 6 and 7. Furthermore, the authors should add the average standard error of Fig. 7

Author Response

Dear reviewer,

Thanks for your valuable contribution. A full list of comments and answers to your queries (in black) is available (in red) below. We believe your contribution significantly improved the quality of our paper.

Kind Regards,

The manuscript presents an interesting study for determining the CDI for fruit color determination. Below are my comments:

1. I suggest that the abstract should briefly describe the method used in the paper and the conclusion should list more detail about what the paper major achieved. The authors should rewrite the abstract and conclusion.

Thanks for your comment. We think that the abstract provided a succinct but complete summary of the methodology. Nevertheless, we modified a couple of sentences to improve the description of the colour extraction methodology from machine vision. We have also updated the conclusions based on the reviewer’s suggestions.

2. In the introduction, I suggest that the author should discuss in detail some major-related papers that the author aspired to or a great contribution to the field.

Thanks for your feedback. Colour assessments through in situ applications of machine vision in ground-based platforms are a relatively new concept and little relevant literature is currently available. To our knowledge, we have cited all the literature that is relevant for the narrative of this paper. We are very keen to consider any additional reference that the reviewer thinks appropriate for the introduction of our paper.

3. I am not sure about the importance of Fig. 3 and 4. which the authors should consider ignoring.

The two figures are useful to introduce a spatial context of the orchards where the experiment was conducted. Our preference is to keep the two figures in the manuscript.

4. Fig. 5: the authors state that " detection fruit are shown with the red detection boxes", however, I found nowhere on the fig for the red direction boxes.

Thanks for this comment. Indeed, the detection boxes were not clearly visible in these images due to the image compression needed to fit 14 images in one figure. We have now updated Figure 5 to include a panel (H) with a zoomed-in cropped image that clearly shows detection boxes.

5. Line 283: what is the meaning of notation n square?

Eta-squared (η²) is a measure of effect size that is typically used after ANOVA models or General Linear Model procedures. η² measures the proportion (in percentage) of variance associated with an effect in the model. In our case, it’s the percentage of the variance of CDI that can be explained by cultivar, canopy side and time of measurement. Time of measurement had the lowest effect size. If a factor fully explains the variance of a predicted variable, η² is 1. If a factor explains no variance of the predicted variable at all, η² is 0. A sentence to describe η² has been added to the materials and methods paragraph “Statistical analysis”.

6. Fig. 6 legends: the authors state "Error bars represent 95% confidence intervals of the estimates". The author should explain more about this statement.

When presenting the mean of a quantitative variable in a graph, the dispersion of values around the mean can be presented with standard deviations (if the aim is to show the variability (or the spread) of values in the sample), standard errors (if the aim is to assess the potential difference between sample and population mean) or confidence intervals (strictly related to standard errors; for example, 95 % confidence intervals are calculated from standard errors using the 95% Z value (1.96), i.e., C.I. (95%) = +/- 1.96*standard error). These are standard statistical measures of dispersion around means.

7. Fig. 7 legends: the authors state "Error bars represent standard errors the mean and different letters show significant differences.". The authors should explain the difference between of error bar from Fig. 6 and 7. Furthermore, the authors should add the average standard error of Fig. 7.

Thanks for detecting the inconsistency. For consistency with Figure 6, we have now updated Figure 7 with 95% confidence intervals. Since our preference is to use confidence intervals, we believe there is no need to add average standard errors to Figure 7. Different letters describe statistically significant differences based on the Bonferroni test.

Reviewer 2 Report

The topic of the manuscript itself is interesting. The manuscript is well written. I really enjoy reading it. I have a minor concern related to the color index which is 0 and 1 currently. I would prefer more classification of the color index. However, the reason for choosing just 0 and 1 should be given in the manuscript. Standardization of these colors with reference to different types of cultivars will increase their applicability in the other regions of the world. I will also suggest putting the exact scientific name of the species in the manuscript. These types of studies also needed to be done for other food varieties to attend maximum output. 

The manuscript also missing citations in the introduction and methods against the information authors put there. 

The incorporation of the script in the supplementary material will increase the reproducibility of the manuscript. 

Author Response

Dear reviewer,

Thanks for your valuable contribution. A full list of comments and answers to your queries (in black) is available (in red) below. We believe your contribution significantly improved the quality of our paper.

Kind Regards,

The topic of the manuscript itself is interesting. The manuscript is well written. I really enjoy reading it. I have a minor concern related to the color index which is 0 and 1 currently. I would prefer more classification of the color index. However, the reason for choosing just 0 and 1 should be given in the manuscript. Standardization of these colors with reference to different types of cultivars will increase their applicability in the other regions of the world.

Thanks for your feedback, we appreciate that you enjoyed reading our paper. The colour index used in this work is not just either 0 or 1, but it ranges from 0 to 1. A very large number of values are present in the 0–1 range (several decimal places). The index conveniently ranges from 0 to 1 to represent no red colour development (0%) and full red colour development (100%). The way this index is calculated involves a standardisation so that it can be largely applicable in a number of crops and cultivars and applied in every region of the world. We hope our comments help.

I will also suggest putting the exact scientific name of the species in the manuscript.

Thanks for this comment, we have now added the scientific name to the “Experimental sites and cultivars” paragraph of our Materials and Methods”.

These types of studies also needed to be done for other food varieties to attend maximum output. 

That’s correct. In fact, we are currently studying CDI responses in apples, pears and some minor stone fruit with very positive preliminary results. This paper was tailored to peach and nectarine fruit and to measuring the effects of sunlight and time of day on CDI estimation.

The manuscript also missing citations in the introduction and methods against the information authors put there. 

Colour assessments through in situ applications of machine vision in ground-based platforms are a relatively new concept and little relevant literature is currently available. To our knowledge, we have cited all the literature that is relevant for the narrative of this paper. We are very keen to consider any additional reference that the reviewer thinks appropriate for the introduction of our paper.

The incorporation of the script in the supplementary material will increase the reproducibility of the manuscript.

The script is publicly available on-line at GitHub (cited in the text) and we don’t feel republishing it in the supplementary material is essential nor common practice.

Reviewer 3 Report

1.      Many parameters could affect the color of an object in an image, which time of day and fruit maturity, and health are among them. Would you please explain if and how you considered the effect of cloud cover on colors?

2.      Can you please provide more information on the fruit detection e.g., detection error and important features?

3.      Peaches and nectarines are very sensitive fruit when it comes to pests and diseases which directly influence coloration/pigmentation of the infested (attacked) area (fruit skin). How did you go about this in orchards you were measuring CDI?

4.      Related to question number 3: could you use some of the portable devices you mentioned in the first 2 paragraphs in pg 2, like portable spectrometers to calibrate your (CDI) model and make a connection to plant health/pests/diseases coloration. This is important to assess the quality of fruit/sales strategy (premium quality/ 2^nd class etc.)

Author Response

Dear reviewer,

Thanks for your valuable contribution. A full list of comments and answers to your queries (in black) is available (in red) below. We believe your contribution significantly improved the quality of our paper.

Kind Regards,

1. Many parameters could affect the color of an object in an image, which time of day and fruit maturity, and health are among them. Would you please explain if and how you considered the effect of cloud cover on colors?

Thanks for your feedback. Indeed, there are many factors possibly affecting fruit colour. Clouds covering the sun can buffer the CDI alteration observed in mid-morning and mid-afternoon measurements. In this experiment we only tested CDI estimation during a sunny day (Line 182 of Materials and Methods). We expect that in cloudy conditions (>70% cloud cover; it is particularly important that clouds cover the sun during measurements), the orchards can be scanned at any time of the day. Our expectations are for no impact of cloud cover as measurements were taken prior to dawn and after sunset as well as during the daytime. The significant alteration of CDI occurred when the sun was visible in the images as it caused a significant increase in image brightness that in turn caused an increase of CDI.

2. Can you please provide more information on the fruit detection e.g., detection error and important features?

Certainly, we have added a sentence in Line 170 – 173 of Materials and Methods. A zoomed-in cropped image was also added to Figure 5 in panel H, to show red detection boxes.

3. Peaches and nectarines are very sensitive fruit when it comes to pests and diseases which directly influence coloration/pigmentation of the infested (attacked) area (fruit skin). How did you go about this in orchards you were measuring CDI?

We are aware of the effects of pests and diseases on fruit colour. We conduct a thorough pest and disease monitoring strategy in our orchards and no significant outbreaks occurred during the season. We are confident that pests and diseases had no significant effects on colour detection. CDI will be able to detect changes in colour derived from pests or diseases or any other significant effect.

4. Related to question number 3: could you use some of the portable devices you mentioned in the first 2 paragraphs in pg 2, like portable spectrometers to calibrate your (CDI) model and make a connection to plant health/pests/diseases coloration. This is important to assess the quality of fruit/sales strategy (premium quality/ 2^nd class etc.).

Most definitely, CDI can be calculated from colour measurements using portable spectrometers and the data used to report fruit skin and flesh response to several experimental factors such as nutritional status, pests and diseases, storage disorders and so on. In another paper we successfully calibrated a relatively cheap and simple tristimulus colourimeter against a more traditionally used Konica Minolta spectrophotometer for hue angle and other CIELab parameters. Hue angle is the colour parameter used for CDI calculation, so I would say that virtually any colourimeter that can accurately detect RGB can be used to estimate hue angle and then CDI.