|The resubmission addressed most of the previous comments. The manuscript structure is clear, and the writing is much better. However, I still have two major comments. |
It is unclear how G3Point conducts grain segmentation. First, how does G3Point select the initial points to start grain segmentation? I.e., do you select every point in a point cloud to find a path to the summit? Or do you randomly sample points to find their summits? If I understand correctly, based on the description of the Fastscape algorithm, only a path or a stream is found. The adjacent nodes in the path have the steepest gradient. How does the path with a list of nodes result in a segmented patch? It is important to clarify this point because the computing efficiency is related to the number of summits, which are the results of the grain segmentation.
The quality of this research is good for an ESurf publication. However, that’s only after significant improvement in scientific writing. I appreciate the proposed methodology, validation, and discussion. Although the writing of this resubmission has been improved, publishing this work still requires much more work. Please see the minor comments for some suggestions.
Title: 3D point clouds: 3D is redundant because it is implied in point cloud.
P1 L7: The grain-scale morphology and size distribution of sediments are important factors controlling the efficiency of erosion, sediment transport and the quality of aquatic ecosystems.
>> The grain-scale morphology and size distribution of sediments are important factors controlling the erosion efficiency, sediment transport, and aquatic ecosystem quality.
P1 L8: In turn, constraining the spatial evolution …. what do you mean by "constraining"? It does not make sense. Do you mean "obtaining"?
P1 L12: methodological approach >> methodology
P1 L13: geomorphological >> geometric
P1 L14: remove “applied here to 3D point clouds”. Redundant
P1 L15: remove “applied to each sub-cloud”. Redundant.
P1 L16: "conceived" is an inappropriate word here
P1 L23: I understand why the authors want to use this word, in-situ, but in-situ has a specific meaning when talking about sampling and measuring method. In this case, it is better to avoid such ambiguity. Please refer to google scholar results of in-situ: https://scholar.google.com/scholar?hl=en&as_sdt=0%2C3&q=in-situ&btnG=
P1 L23: remove “and grain cluster”. Redundant with grains
P1 L24: “The main limit of this method is that it is only able to detect grains with a characteristic size significantly greater than the resolution of the point cloud”. This limitation is very obvious. Authors should be more open to unique problems of the presented method. Or you can remove this sentence.
P2 L1: remove “on the”
P2 L20: potentiality >> potential
P2 L21: documenting >> inventorying
P2 L23: remove “most”. You need to conduct comparison or include reference to make “the most” statement.
P2 L23: “This method consists in measuring” >> This method measures. “consist in” is overly misused in the entire manuscript. Please modify them accordingly.
P2 L24: Suggested writing: The grid-by-number method is simple to implement and similar to a volumetric sampling xxx
P2 L29: Collection of a data set >> Collecting a data set
P3 L1: misused “consists in”
P3 L24: suggested writing: Structure-from-Motion (SfM).
P3 L35: Suggested writing: considering point clouds obtained from SfM to check its xxx.
P4 L8: remove “any type of”
P4 L9: remove “summarized by”.
P4 L9: remove the sentence, “A 3D point cloud …”. Point cloud is not new. It is unnecessary to explain.
P4 L11: remove “2”. It causes confusion.
P4 L13: remove “will”
P4 L14: how do you deal with background points such as dirt and mud?
P4 L14: this task >> this denoising task
P6 L7: remove “will”
P6 L14: remove “despite this main disadvantage”
P6 P16: is performed using >> uses
P6 P21: You may need a transition sentence before mentioning “the Fastscape algorithm”. It is unclear how the Fastscape algorithm is related to the point cloud algorithm you mentioned above.
P6 L21: order is ambiguous word. Do you mean “make the points in a sequence” ?
P6 L25: giver >> donor. You should have consistent terms through the entire manuscript.
P7 L7: “This algorithm only imposes one scale”. What do you mean by a scale? Scale is an ambiguous word.
P7 L23: Fig A1b. missing or you mean Fig S1b?
P7 L25: “Instead of xxx”. Long sentence. It’s difficult to follow.
P7 L30: what k value do you choose here?
P8 L8-10: If I understand this correctly, the number of summits is determined by the number of initial points. However, how initial points are selected is unspecified. This point is important because your algorithm efficiency is quadratic and thus limited by the number of summits.
P8 L17: consists in merging >> merges
P9 L10: “The most pertinent and simplest xxx”. how do you know this? have you done with comparison? or please cite references that draw this conclusion. Otherwise, be cautious about “the most” statement.
P9 L19: respected >> satisfied
P9 L30: consists in computing >> computes
P11 L10: “with an error less than 1.061% when p=1.6707”. how do you know? please clarify or add citation
P11 L33: “lab or natural environments”. Adding a data description at the beginning (after this paragraph and before 3.1), such as data quality and acquisition method, will help audience understand.
P12 L14: “For this purpose”. It is unclear what purpose you are referring to.
Section 3: Your writing in describing experiment processes is much more clear than describing method ideas. The validation work is solid.
P14 L30: “To segment grains, xxx”. This is why I asked about the dirt and mud? how can your algorithm manage them? If not, you should add an assumption at the beginning.
P20 L26: remove sentence, “G3Point can also perform grain size, xxx”, unless you had experiments on such large study areas.
P20 L30: “If G3Point can be directly applied to point clouds, xxx”. I don't understand why the authors mention this assumption in the context. also, how does a user without any G3Point experience know if G3Point can be applied to their studies?
Section 4.1. The authors seem to talk about calibration instead of validation. I think it's better to use a known grain size to calibrate the hyperparameters as other semi-automatic approaches do. Adding a calibration process overcomes the problem of your trial-and-error exploration.
P21 L14: “removing points”. how? does G3Point have any algorithms for this? or users should manually do this?
P21 L14: “which is a built-in option”. What do you mean by this?
P21 L20: “the point clouds processed by G3Point must be xxx”. mentioning this earlier as an assumption will help audience understand.