Review Reports

Round 1

Reviewer 1 Report

My comments are in three parts – very general comments, specific comments, and general comments.

Very general comments

Title and throughout the manuscript – what is a ‘pollen grain segmentation method’? I have been involved in palynological research for nearly 60 years and quantitative palaeoecology for nearly 50 years and I have not before heard or read anything involving segmentation of pollen grains. Figure 3 and lines 133–138 do not say what segmentation is. Is it a technical term used in the machine-learning or computing worlds? Or is it what in applied statistics is called discrimination or partitioning? I find it difficult to review a manuscript when I do not understand the title, abstract, or one of the keywords.

 

Specific comments

Line 22: better to say ‘high demands on time’ than ‘time consumption’

Line 41: Sinapsis should be in italics; Brassicacaeae should not

Line 46: what is the ‘sediment of --- PDO’?

Line 48 and throughout: ‘segment’

Line 65: Urticacaeae should not be in italics

Line 72 and elsewhere: Gaussian with a capital G

Line 86: ‘thresholded’ does not really work as a verb

Line 86: I am not sure what ‘applied … operations’ means after ‘thresholded’

Line 96: what is loll?

Line 111: does ‘certified by a botanist’ mean identified by or checked by a botanist?

Line 117: perhaps change ‘closing’ to ‘covering’

Line 118: what is the mounting medium? Glycerine, glycerine jelly, other?

Line 128: why not remake the slide by spreading the pollen mix carefully over the area on which the cover-slip would be placed?

Line 158, 162: what is ‘salt and pepper noise’?

Line 186: grains not ‘grain’

Fig 4: a scale is needed. The pollen of Asphodelus (number 16) looked very odd to my palynological eye.

Table 2: what does N/A mean?

Line 235: ‘correctly segmented’ and ‘poorly segmented’ – I wish I knew what segmentation meant!

Line 236: see my comment for line 128.

 

General comments

As I have outlined above, I find this manuscript difficult to review as I do not know what ‘segmenting’ really means. I suspect it is a technical term in computing or machine-learning. Assuming it is, the manuscript is interesting and a useful addition to the growing literature on automated identification of pollen.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors present a detailed methodology to capture and isolate pollen grains from microscope images by segmentation. The article is well written and the workflow is clearly described, and the results are robust. I think the article is generally ready for publication, but have a few minor comments below:

• The article and method is clearly focused on an application in melisso-palynology, however, there may be more broad interest from aero- and paleo-palynology. For this reason, it would be useful to have a little more information on the methods, to permit further application and/or replication:
  • What software was used to construct the pipeline? Was this custom built or was existing image processing software used (e.g. ImageJ)?
  • How much time does this take to segment a single image?
• Line 24: The wording of the final sentence is a little odd. Maybe “it is necessary to provide effective tools that can help ensure it’s origin.”
• Line 33: Is Pefkothymaromelo Kritis, PDO specific to Crete or does this represent a larger geographical region?
• Line 109-110: Does the method to prepare the sample slides represent the method used to test a sample of honey? Are there any differences that might affect the application to a “real-world” sample?
• Line 125: As the goal of the paper is to produce a set of useful segmented images, the decision to remove ‘noisy’ or complex images is understandable. However, are these noisy images likely to occur in samples of honey? If so, the values given in table 3 (and any resultant predictive skill) are likely to be over-optimistic
• Figure 3: This is a great figure to illustrate the workflow. No further comment
• Table 3: It is unfortunate that the type with the highest percent poorly segmented grains is also the type that is the most characteristic of the honey product of interest. It would be useful to comment on this in the discussion and if the authors think that it might impact the application

Author Response

Please see the attachment.

Author Response File: Author Response.pdf