Retrieval of Remotely Sensed Sediment Grain Size Evolution Characteristics along the Southwest Coast of Laizhou Bay Based on Support Vector Machine Learning

Round 1

Reviewer 1 Report

Dear Authors,

I've examined your work with great interest! This is example of very strong contribution. It deals with novel, advanced methodology and internationally-important information; it is well-written, well-structured, well-illustrated, and well-referenced. This can be used as a kind of template for remote sensing applications in sedimentological studies. I strongly recommend this manuscript for publication in "JMSE', although I also see the urgency of some changes/additions.

• Title: please, make it more attractive and internationally appealing.
• Subsection 2.1: the geographical setting (landforms, climate, hydrology, etc.) of the study area should be characterized more extensively and with the larger amount of citations.
• Subsection 2.2.1: this part should bear more sedimentological knowledge. For instance, which grain size classification have you used? Please, cite the basic sources.
• Table 4: what is the source of the information about discharge and precipitation?
• Discussion: how possible is that the sea-level changes were also a factor of the registered grain size changes?
• The language needs certain polishing. For instance, Line 264: corrAlation -> corrElation.
• Section 6: this section is unnecessary in this manuscript.

Author Response

Dear Authors,

I've examined your work with great interest! This is example of very strong contribution. It deals with novel, advanced methodology and internationally-important information; it is well-written, well-structured, well-illustrated, and well-referenced. This can be used as a kind of template for remote sensing applications in sedimentological studies. I strongly recommend this manuscript for publication in "JMSE', although I also see the urgency of some changes/additions.

Response: Thanks very much for your positive comments.

 

Title: please, make it more attractive and internationally appealing.

Response: The title was replaced by “Retrieval of remotely sensed sediment grain size evolution characteristics along the southwest coast of Laizhou Bay based on support vector machine learning”

 

Subsection 2.1: the geographical setting (landforms, climate, hydrology, etc.) of the study area should be characterized more extensively and with the larger amount of citations.

Response: We had followed your comment and revised the statement of the geographical setting (landforms, climate, hydrology, etc.)

 

Subsection 2.2.1: this part should bear more sedimentological knowledge. For instance, which grain size classification have you used? Please, cite the basic sources. Response: Grain size classification used here was from the mentioned reference (Lanzhou and Nanjing, 2006) in the manuscript. We replaced the reference by the basic source(Folk and Ward 1957).

 

Table 4: what is the source of the information about discharge and precipitation? Response: We provided the the source of the information about discharge and precipitation in the revised manuscript.

 

Discussion: how possible is that the sea-level changes were also a factor of the registered grain size changes?

Response: The sea-level changes were also a factor of the registered grain size changes. However, the effect may be more weak than those of discharge and precipitation. Thus, it was ignored here. 

 

The language needs certain polishing. For instance, Line 264: corrAlation -> corrElation.

Response: Thanks very much for your positive comments. We had polished the language throughout the manuscript.

 

Section 6: this section is unnecessary in this manuscript.

Response: We removed Section 6 from the manuscript.

Reviewer 2 Report

The authors did a very heavy work of sampling and laboratory analysis, as well as image processing preceeded by an excellent atmospheric correction. Unfortunately, all the work is undermined by the fact that the water content in the sediments strongly determines the spectral signature. If we consider that the studied environment is the intertidal zone, where the sediments follow very rapid soaking and drying cycles, the work is extremely weak. I am sorry that the work done cannot be used, but I really don't know what to suggest to overcome these problems. I hope that the other referees can be more proactive (or less critical).

39	Change in grain-size should indicate change in rate ..
45	I don’t agree: sampling ust, and generally follow a well defined strategy. It’s nor random.
62	‘property’ is too generic.
87	Van and Herman paper was accepted, but R2 = 0.72 is a poor result!
127	Is this the significative wave height? Hs?
Fig 1	Not easy to see sample’s location. How tidal area was mapped?
151	You don’t measure properties, you derive them from grain size distribution
164	Accuracy 1% ?
174	Use correct Folk & Ward symbols.
Fig 2	Sample number is artificially attributed by authors. Better to indicate the distance from the river mouth.
196	If water content was not measured any further analysis is compromised
Fig 3	Since one of the goal is to look for differences, putting the mean of 238 samples as value is meaningless. And again samples where wet? As at the time of satellite overpass (impossible to have measures). I think that the correlation you find at the end id between reflectivity and humidity. The latter being correlated with sediment size. Therefore you find a co-correlation!
Fig 4a	In x you have the band number, not band reflection
Fig 4b	You demonstrate that there is no correlation1
267	Explain in words what SI, DI, … means (even if we can understand it) and, more important, explain why you selected those indices. Are thrre some physical reasons for which they could give you the requested answer?
Tab 2	On which bands you make these operations?
352-357	Reducing the water dirchange, river velocity reduces and is less able to transpost coarser sediments. Here it seems the opposite: can you give the rational of this?
393-396	From which sources all these remotely sensed images come? They are not mentioned in Materials & Methods.
419	This is the main deficiency of this paper, as previously observed, since water content is the main factor influencing soil reflectivity, as widely known. And this had to be considered before starting the study.
421-426	Higher resolution and overpass frequency do not solve the problem.

Author Response

The authors did a very heavy work of sampling and laboratory analysis, as well as image processing preceeded by an excellent atmospheric correction. Unfortunately, all the work is undermined by the fact that the water content in the sediments strongly determines the spectral signature. If we consider that the studied environment is the intertidal zone, where the sediments follow very rapid soaking and drying cycles, the work is extremely weak. I am sorry that the work done cannot be used, but I really don't know what to suggest to overcome these problems. I hope that the other referees can be more proactive (or less critical).

Response: Thanks very much for your positive comments for our heavy work. As the comments mentioned, the water content in the sediments strongly affects the spectral signature. The effect of water on the reflectance was not considered in this paper. The reason is that moisture content is mainly dominated by grain size parameters. To some extent, there was an inverse correlation between grain size and water holding capacity (Goldberg et al., 1976; Mohammadi and Vanclooster, 2011). The smaller the grain size of the soil is, the greater its water holding capacity. The effect of water content on the reflectance was inherent; however, it was dominated by grain-size parameters. Thus, the yielded co-correlation could act as a proxy of the correlation between grain size and reflectance to some extent.

39	Change in grain-size should indicate change in rate .. Response: We revised the statement following your comment.
45	I don’t agree: sampling ust, and generally follow a well defined strategy. It’s nor random. Response: We revised the statement following your comment.
62	‘property’ is too generic. Response: We replaced ‘property’ by ‘types’. Then the meaning was clearly stated.
87	Van and Herman paper was accepted, but R2 = 0.72 is a poor result! Response: We think the fitting accuracy is not high enough but acceptable.
127	Is this the significative wave height? Hs? Response: The mentioned wave heights were from the observations and they showed local tidal height conditions.
Fig 1	Not easy to see sample’s location. How tidal area was mapped? Response: We provided a new sample’s location map. Tidal area was mapped by visual interpretation.
151	You don’t measure properties, you derive them from grain size distribution Response: We revised the statement following your comment.
164	Accuracy 1% ? Response: We revised the statement as “overall error is less than 1%”
174	Use correct Folk & Ward symbols. Response: We revised the statement following your comment.
Fig 2	Sample number is artificially attributed by authors. Better to indicate the distance from the river mouth. Response: Sampling points were distributed almost perpendicular to the coastline and were difficult to be expressed by distance.
196	If water content was not measured any further analysis is compromised Response: The similar comment was answered above.
Fig 3	Since one of the goal is to look for differences, putting the mean of 238 samples as value is meaningless. And again samples where wet? As at the time of satellite overpass (impossible to have measures). I think that the correlation you find at the end id between reflectivity and humidity. The latter being correlated with sediment size. Therefore you find a co-correlation! Response: The effect of atmospheric correction is demonstrated (Fig 3) by matching measured reflectance with remote sensing reflectance. The similar comment was also answered above.
Fig 4a	In x you have the band number, not band reflection Response: We revised Fig 4a following your comment.
Fig 4b	You demonstrate that there is no correlation1 Response: The correlations between the reflectance of the Landsat image bands and sediment grain-size parameters are not high enough. That’s the reason why we try to retrieval sediment grain-size based on SVM.
267	Explain in words what SI, DI, … means (even if we can understand it) and, more important, explain why you selected those indices. Are thrre some physical reasons for which they could give you the requested answer? Response: what SI, DI, … means were shown in Table 2. Because correlations between the reflectance of the Landsat image bands and sediment grain-size parameters are not high enough. Thus, we selected those indices, which were generally used to enhance the correlations by reducing the influence of atmospheric path radiation on reflectance.
Tab 2	On which bands you make these operations? Response: The reflectance of the Landsat image bands (b1, b2, b3, b4, b5 and b7) were used.
352-357	Reducing the water dirchange, river velocity reduces and is less able to transpost coarser sediments. Here it seems the opposite: can you give the rational of this? Response: The water dischange was reduced. Then transport of seawater removed fine sand, leading to coarsening.
393-396	From which sources all these remotely sensed images come? They are not mentioned in Materials & Methods. Response: Sources all these remotely sensed images come were stated in Section 2.2.3.
419	This is the main deficiency of this paper, as previously observed, since water content is the main factor influencing soil reflectivity, as widely known. And this had to be considered before starting the study. Response: We had answered the similar commets above. The main deficiency of this paper was also mentioned in the Discussion section(line 410-421).
421-426	Higher resolution and overpass frequency do not solve the problem. Response: This statement provided here aimed to improve the temporal and spatial resolution of observations in further studies, it is not to solve the error caused by water content on grain size retrieval.

 

 

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript entitled “Retrieval of remotely sensed sediment grain-size parameter distribution based on SVM in the southwestern coast of Lai- zhou Bay, China” is of a very interesting topic. The authors used Landsat images from 1989 to 2015 to map the distribution of the sediments according to their grain size along the southwestern coast of Laizhou Bay, China. They used support vector machine (SVM) to develop a nonlinear calibration model, considering the low correlations between the measured reflectance and grain-size. The topic of the paper is suitable for the “Journal of Marine Science and Engineering”. In general the paper is well written. I am not a native speaker but I think that the manuscript reads quite well!

One of the positives of the paper is that the proposed remote sensed methodology has been tested and calibrated in the field with sediment sampling and grain size analyses. However, the authors should discuss the uncertainties and the accuracy of the methods used. Does the resolution of the Landsat images restrict the accuracy of the methodology? A further explanation of the results (especially those regarding the relations among the discharge of the river and the gain size of the sediments) in terms of physical processes is required.

Here are some comments and suggestions:

In the title “….SVM…..” should be replaced by “….Support Vector Machine (SVM)……”

line 17: I am not sure if the grain size terminology (clay silt and sand) – I mean the characterization of a sediment according to the size of the grains -  is a grain size parameter. I agree that mean grain size is a parameter.

In the “Introduction” section, where the authors describe (based on literature) the various remote sensed techniques used by various researchers to investigate the grain size distribution, they should mention about the accuracy and the validation of the results of these techniques. What is the main way of validation of the results? Field measurements or other ways?

lines 106-107: “Study area” should be a separate section before the “Materials and Methods” and after the “Introduction”.

line 109: what do you mean with “…..precipitation…..”? deposition? usually precipitation is a term used for rainfall. Please clarify.

line 119: what is “…..M2 tidal wave…..”?

lines 125-130: You need to add references here.

In the “Study Area” section a location map showing the relief (topography) of the area is required. An inset map showing the location of the study area within the Yellow River delta as well as another one showing the location within chine is necessary. The map of Figure 1 which shows the location of the samples could be a separate Figure (without the inset map of China).

A flow diagram of the methodological steps would be ppreciated by the readers of the final version of the paper.

line 152: I insist that silt clay and sand are not sediment properties!

In the “Results” section at the figures which show the temporal and spatial distribution of the various grain size classes what are the non colored areas along the coastline? These areas are gravel or coble beaches? Please explain.

In Table 4 what is “Preciptation”? you mean precipitation of sediment? If yes I suggest to change it with Deposition.

I think that the way references are made within the manuscript is not the one proposed by the journal. The same stands for the references list at the end of the paper.

Author Response

The manuscript entitled “Retrieval of remotely sensed sediment grain-size parameter distribution based on SVM in the southwestern coast of Lai- zhou Bay, China” is of a very interesting topic. The authors used Landsat images from 1989 to 2015 to map the distribution of the sediments according to their grain size along the southwestern coast of Laizhou Bay, China. They used support vector machine (SVM) to develop a nonlinear calibration model, considering the low correlations between the measured reflectance and grain-size. The topic of the paper is suitable for the “Journal of Marine Science and Engineering”. In general the paper is well written. I am not a native speaker but I think that the manuscript reads quite well!

Response: Thanks very much for your positive comments.

 

One of the positives of the paper is that the proposed remote sensed methodology has been tested and calibrated in the field with sediment sampling and grain size analyses. However, the authors should discuss the uncertainties and the accuracy of the methods used. Does the resolution of the Landsat images restrict the accuracy of the methodology? A further explanation of the results (especially those regarding the relations among the discharge of the river and the gain size of the sediments) in terms of physical processes is required.

Response: The accuracy of the methods used were provided in Figure 5 and 6. The uncertainties was provided in the Discussion section(line 410-421). The accuracy of the methodology might be restricted by the resolution of the Landsat images to some extent. We would try to improve the temporal and spatial resolution of observations in further studies.

 

Here are some comments and suggestions:

In the title “….SVM…..” should be replaced by “….Support Vector Machine (SVM)……”

Response: We revised the statement following your comment.

 

line 17: I am not sure if the grain size terminology (clay silt and sand) – I mean the characterization of a sediment according to the size of the grains - is a grain size parameter. I agree that mean grain size is a parameter.

Response: We revised the statement following your comment.

 

In the “Introduction” section, where the authors describe (based on literature) the various remote sensed techniques used by various researchers to investigate the grain size distribution, they should mention about the accuracy and the validation of the results of these techniques. What is the main way of validation of the results? Field measurements or other ways?

Response: The various remote sensed techniques and their accuracy studied by various researchers to investigate the grain size distribution were mentioned(line 64-95). The main way of validation of the results is generally based on the match-up between the retrieval results and field measurements.

 

lines 106-107: “Study area” should be a separate section before the “Materials and Methods” and after the “Introduction”.

Response: To our best knowledge, the position of “Study area” was generally inside of the “Materials and Methods”. After careful consideration, we decided that this would be a better location.

 

line 109: what do you mean with “…..precipitation…..”? deposition? usually precipitation is a term used for rainfall. Please clarify.

Response: Precipitation is also a term used for rainfall here.

 

line 119: what is “…..M2 tidal wave…..”?

Response: M2 tidal wave was the terminology of a lunar semi-diurnal with the symbol M2.

 

lines 125-130: You need to add references here.

Response: We followed your comment and added add references here.

 

In the “Study Area” section a location map showing the relief (topography) of the area is required. An inset map showing the location of the study area within the Yellow River delta as well as another one showing the location within chine is necessary. The map of Figure 1 which shows the location of the samples could be a separate Figure (without the inset map of China).

Response: We followed your comment and peovided a new Figure 1.

 

A flow diagram of the methodological steps would be ppreciated by the readers of the final version of the paper.

Response: We followed your comment and peovided a new Figure showing a flow diagram of the methodological steps.

 

line 152: I insist that silt clay and sand are not sediment properties!

Response: We followed your comment and revised the statement.

 

In the “Results” section at the figures which show the temporal and spatial distribution of the various grain size classes what are the non colored areas along the coastline? These areas are gravel or coble beaches? Please explain.

Response: The non colored areas along the coastline were underwater shoals.

 

In Table 4 what is “Preciptation”? you mean precipitation of sediment? If yes I suggest to change it with Deposition.

Response: Precipitation is a term used for rainfall here.

 

I think that the way references are made within the manuscript is not the one proposed by the journal. The same stands for the references list at the end of the paper.

Response: We followed your comment and revised the way references made within the manuscript, so did the references list at the end of the paper.

 

Author Response File: Author Response.docx