Next Article in Journal
High-Precision Population Spatialization in Metropolises Based on Ensemble Learning: A Case Study of Beijing, China
Next Article in Special Issue
Estimation of Suspended Sediment Concentration in the Yangtze Main Stream Based on Sentinel-2 MSI Data
Previous Article in Journal
Novel Higher-Order Clique Conditional Random Field to Unsupervised Change Detection for Remote Sensing Images
Previous Article in Special Issue
Prediction of Total Phosphorus Concentration in Macrophytic Lakes Using Chlorophyll-Sensitive Bands: A Case Study of Lake Baiyangdian
 
 
Article
Peer-Review Record

A New Method for Calculating Water Quality Parameters by Integrating Space–Ground Hyperspectral Data and Spectral-In Situ Assay Data

Remote Sens. 2022, 14(15), 3652; https://doi.org/10.3390/rs14153652
by Donghui Zhang 1,2, Lifu Zhang 1,2,3,*, Xuejian Sun 1,2, Yu Gao 2,4, Ziyue Lan 2, Yining Wang 2, Haoran Zhai 2, Jingru Li 2, Wei Wang 2, Maming Chen 2, Xusheng Li 5, Liang Hou 6 and Hongliang Li 7
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Remote Sens. 2022, 14(15), 3652; https://doi.org/10.3390/rs14153652
Submission received: 14 July 2022 / Revised: 25 July 2022 / Accepted: 26 July 2022 / Published: 29 July 2022
(This article belongs to the Special Issue Hyperspectral Remote Sensing Technology in Water Quality Evaluation)

Round 1

Reviewer 1 Report

The authors have addressed all my concerns satisfactorily in their responses.
Although the manuscript presentation would benefit from an extensive revision by an experienced scientific English editor, I believe the article can be published in Remote Sensing.

I provide below some suggestions and comments for the consideration of the authors and the editor:

L63: What does "First" refer to? Is it the "working mode design of multi-platform sensors"? Please do not abuse the use of pronouns.

L78: "a calibration cloth *with dimensions of several meters*"

L79: "a new spectrometer product floating on the water *was introduced* / *appeared in the market*". Please, do also name that floating spectrometer (i.e. its manufacturer and model).

L84: "The third is selection of characteristic bands. We can compute these bands for various substances in water and *from the statistical analysis of data and derive a formula for concentration* [32]"

L93-94: "to build a machine learning model to *fit water quality parameters to spectral data*"

L96: "each water quality parameter in the whole *spectral range or from* individual bands [35-36]."

L109-110: "Compared with the first three problems, there is no unified evaluation standard for the *design and evaluation of the* algorithm *results*, so it is difficult to *arrive at an universal one*."

L153: "the spatial resolution is 0.2m". For AUV-borne sensors, spatial resolution must be given as an average value for a typical flight height; please, include this height in the details.

L158: "which *makes it a good platform* for most sensors"

L211-212: "*Water samples were collected in* a bottle of 500 mL *from* each sampling site, *and were kept at low temperature* in a box with an ice bag. *Chemical testing was completed within the following* 12 hours (Table 1)."

L279: "Here, the cross-correlation spectrum matching algorithm *(CCSM)* is introduced,"

Eq. (2): There is no m index on the right hand side of the equation; I guess it should appear as a wavelength/band shift in one of the reflectance spectra R_r or R_t of the cross-product sum. Unless this is included in the equation, the following paragraph is not clear enough to implement the method.

L297: "between the *spectral reference* and the *actually measured spectrum*"

Eq. (3): Please, do not use capital R to indicate a correlation coefficient; use a lower-case letter, perhaps with some decoration (a hat, a bar, etc.) above the letter. Capital R have been used above to denote reflectance spectra. Also, for clarity, do not introduce the number k, just write 2*m+1 in the denominator (which clearly is the number of terms in the sum on the numerator).

Eq. (4): Please, use log (lower case) to denote the logarithm and put the argument 1/R_i between brackets (not as an exponent), i.e. A_i = log_10(1/R_i)

L331: "After calculating *the absorbance using* formula 4, *values in the n spectral bands will be different for different samples*."

L332-333: Actually, after logarithmic transformation the differences will be less dramatic; nevertheless, they will be more linearly related with the pigment concentration.

Eq. (5): I am not yet convinced of the validity of this equation; I find it mathematically ill posed, because of the changes in the sign of S_i (not just for those bands where the denominator vanishes).

L377: "the sampling method will lead to the unreliability of the subsequent conclusions". Please, be more specific about this "sampling method", because you are using a type of "(re)sampling method", and that would entail that your conclusions are unreliable.

L385 and Fig. 6: "Take the spectral mean as the spectral value of each level.", but clearly, the right plot in Fig. 6 does not correspond to these averages (graphs are displaced, in order to improve visualization).

L422-423: "Support vector machine belongs to the black box algorithm, and the process from input data to output results is difficult to explain": NO! SVM are not "black-box" algorithms; they have a solid mathematical foundation. Please, delete this entire sentence.

L503: You probably mean "turbidity" (not "turbine").

Author Response

Thank you very much for your revision.

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This manuscript is trying to introduce an effective integration of aerial remote sensing data and ground multi-source data that is always one of the difficulties of quantitative remote sensing. A new monitoring mode is proposed to install the hyperspectral imager on the UAV and places a buoy spectrometer on the river. Water samples are collected simultaneously to obtain in situ assay data of total phosphorus, total nitrogen, COD, turbidity and chlorophyll measurements. The cross correlogram spectral matching (CCSM) algorithm is used to match the data of the buoy spectrometer with the UAV spectral data to reduce the UAV data noise significantly. An absorption characteristics recognition algorithm (ACR) is designed to realize a new method for comparing UAV data with laboratory data. The method is also taking into account the spectral characteristics and the correlation characteristics of test data synchronously. This new working mode of integrating spectral imager data with point spectrometer data will be useful for monitoring water quality in the future.  

Although the technical method is interesting, there are some points to be revised or improved as below:

1) all the yellow marks should be removed in the revision;

2) In Table 1, the "Turbidity (mg/L)" should be revised and improved;

3) Line 299, RMS should have a unit;

4) Line 356, y should have a unit;

5) Line 458, RMSE should have a unit;

6) In Figure 7, the noise at 900-1000nm should be explained;

7) In Figure 9, the noise should be explained; 

8) In Figure 11, the color meaning and scaled color range should be described or meaningful;

9) why Table 2 used four decimal digits while Table 1 used 1 or 2 decimal digits?

10) Line 611-615 and Figure 12 also used four decimal digits. Can you apply the same decimal digits in the same research work?

11) Figure 13 and line 666-671 have the same issue as 10) above;

12) Figure 14 and line 704-709 are same as 10) and 11);

13) latest references should be updated.

 

Author Response

Thank you very much for your revision.

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The revised version has improved well to be accepted for publication.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.


Round 1

Reviewer 1 Report

In mi opinion, the article draft must be improved in terms of written English. Some parts are not clear and sometimes figures are not connected with the text. In the following, I point out some important details that I found in the text:

Abstract: At least mention once that COD is
the acronym of chemical oxygen demand, then use COD in the rest
of the document

Introduction: In my opinion the introduction must be focused to the topics relevant to the current research (water quality). Sometimes the authors refers to research not related to UAV or remote sensing applications to assess water quality

Line 71: What is the point of these
examples?is remarking the integration of ground observations and
satellite estimates? suddenly changes mention in the text that these are examples of ground-RS integration

LINE 89: I do not think this is necessary. The topic is water quality not water or nutrient stress in plants. It is better an example of spectral response in water bodies such as chlorophyll concentration in water bodies

Methods: Calibration cloth or calibration plate?Calibration panel? Please include a brief description of the calibration cloth (material, dimensions etc)

Methods: How is the spectral information from the UAV si retrieved? as image? as spectral data only? what format? What do you mean by data strip?
you mentioned that data strip had a spatial resolution of 0.075 m then I supposed is a raster format.  What software did you use to perform the geometric correction? What software did you use to process the multi-spectral data?

Methods: How is the water quality data cloud service platform work?
I imagine is a device or system to transfer data remotely to a cloud
service where data can be stored. If this is the case, how the data is transferred remotely? (example raspberry Pi via GSM or common telemetry)
Data transmission is not clear. In figure caption is mentioned that dat is transmitted to the cloud via 4/5G network, therefore is sent via GSM module or similar. This is not mentioned in the 1.2.2 section

Line 233: For consistency, use COD with capital letters (like in the entire
document)

Results: burrs = spectral noise? clarify what you mean by burrs. It is not clear the idea of line 420 is it referring to a particular figure?

Line 421: Figure 7 does not reflect the changes in correlation coefficient

Line 425: To my knowledge and according to the methods, correlation
coefficient ranges from -1 to 1, therefore, I think these values are band matching positions.

Results is basically a description of the main findings but the discussion is very poor and need to be improved.

 

Reviewer 2 Report

In general, it is a good research work for water quality, but there are a few points to be clear as below:

1) what is the difference between RMS (page 8) and RMSE (page 11)?

2) In results, Figure 12 in page 17 should use line rather than  curve.

3) Discussion should be a section, separating from the results section.

Please make some revision as mentioned above.

Reviewer 3 Report

The authors have developed method to indirectly measure water quality parameters using images acquired from UAV-borne hyperspectral sensors. These methods include a technological part of buoys with sensors that upload continually data to a cloud server (where it is further processed) and methods to numerically determine the relevant hyperspectral bands to calibrate the method of contact-less water analysis.

The article deals with a challenging topic and some of its approaches are novel and would be worth publishing. However, it is so carelessly written (in particular, the introduction, which is apparently a collection of sentences about the conclusions in the cited literature, but without any kind of  continuity), and with so low quality English (many times it appears machine translation or translation by someone with no scientific background), that it requires a total rewrite before being considered for that publication; I encourage the authors to use a professional translation service for this.

In view of the previous, I must recommend it to be rejected in its present form.

However, I would like to include some comments of my revision that may probably help the authors with theirs:

L51-52: please, include only satellites and sensors that are still operational or, at least, indicate which of them you cite for historical reasons; please, do also differenciate between satellites and sensors on-board of those satellites.
L92-93: please, avoid excursions from the topic; the problematics of crop nitrogen or chlorophyll content is different from the water quality parameteres you will be interested in in the article; also focus on hyperspectral techniques, not multispectral ones (that will also highlight the novelty of your work).
L131-132: please, do not include in the introduction conclusions that cannot be drawn from the information presented in it or, otherwise, motivate that conclusion.
L169-172: please, include considerations about the public interest of the subject of your study in the description of the study area or, better, in the introduction and motivation of your article.
L175-179: please, provide details about the manufacturer of your sensors or other equipment, not just the country where they were developed; also include a description, not just the model ID (e.g. DJI M600 pro is a remote controlled quadcopter drone; the DR6000 is a UV-VIS spectrophotometer)
L218: Qu et al. 2008 is not in the references
L233: Ramoelo et al. 2011 did not use a UV-VIS spectrophotometer such as the DR6000 (actually, they used a VIS-NIR-SWIR spectrometer); please, cite properly.
L284: please, use a more clear notation for the important equation (2); are you calculating the autocorrelation globally or locally? is the range -20 to +20 referred to wavelengths? what are the units?; are you just cross-calibrating the wavelengths of the two analyzers or are you attempting to match specific features in the spectra measured by the two systems?
L291-296: please, explain what is this RMS measuring and what for is it being calculated: it does not appear in the results below.
L312-314 (and eq. 4): please, explain that reflectance is measured in percentage of reflected irradiance and not as a fraction of it; eq. 4 requires a reference to explain its origin/meaning: initially, if reflected radiation (actually, backscattered radiation) is what is not absorbed from incident radiation, then absorbance A_i would be just 1-R_i; please also write eq. 4 correctly: R_i/100 is not in the exponent and, furthermore, absolute value should not be required (use a minus sign, as the argument of the logarithm sould always be less than 1).
L317 (eq. 5): please, provide some rationale about this index; in its current form, the denominator may become zero whenever absorbance becomes equal to the average, and that is an unfortunate property.
L354-357: please, explain further the construction of megapixels from the pixel neighborhood of a given one; are the hierarchical clustering methods explained in the following paragraph (L366-377) used to compute these megapixels? (what would they be used otherwise?)
L379-388: this description is too general and not focused on the problem at hand; please, clarify how are linear regression, support vector machine or neural network methods trained and validated for the data.
Figure 8: please, explain graphs (c)-(f): they are not easy to interpret from the figure caption or the results description.
L447-448: please, explain the data validation criterion.
L453-454: spectroradiometer noise at wavelenghts on both limits of their spectral range is quite common (its cause is often low signal-to-noise ratio: low solar irradiation at those wavelengths combined with higher sensitivity of the detectors to operating temperature); this limitation could be introduced in the methodology, when the instrumentation is described (also the clarification in L482-483); please, include also the spectral resolution (2 nm bands?).
Figure 10 (b): please, keep the same color order in the plot and in the legend.
L536-540: more information is required: method, how scales were chosen, whether all scales were tested for all methods, etc. For example, why is it said that scales 16 and 24 are similar to scale 1? (provide some numerical arguments here)
L579-580: please, clarify this point of UAV data being acquired in two different days; how is it relevant to this discusion? were not the chemical analyses and the other spectral data acquired on those same two days?
L586-587 and Figures 13: these are not accuracy assessments, but correlation assessments; please, explain what data was used for the plots: it is intriguing how, after fitting the models with field data (obtained from chemical assays), these data appear to have trends so clearly different from the Y=X line.
Figure 14: please, include RGB composite images from suitable visible bands in order to assess visually thematic maps in Fig. 14 (b), and rule out shoreline shadow effects, among others.

Back to TopTop