Monitoring and Forecasting Green Tide in the Yellow Sea Using Satellite Imagery

Round 1

Reviewer 1 Report

The manuscript entitled ‘Traceability and prediction of green tides in the Yellow Sea based on multi-source satellites’ proposed an method to extract green tied in the Yellow sea with multi-source satellite data. Remotely extract green tide, estimate its area and predict its spatiotemporal variation are important for protecting marine ecosystem. However, current manuscript has several drawbacks need to improve.

Introduction

It is necessary and important to point out the novelty of current work. The introduction section needs to reorganized due to the following facts: (1) several sentences are weird. For example, Line 73; (2) it is hard to find out the importance and necessity of current work.

Data introduction

Please given the preprocessing method of satellite data. In addition, it is necessary to validate the consistence between multi-source satellite data.

Method

Section 3.5: The proposed method needs to validate with several images that collected under different environment. In addition, it would be better to validate the method with high spatial resolution image.

Results and discussion

Line 411：Table 5 include Chinese characters.

Suggest to separate the results section and discussion section.

In addition, several contents belong to method, therefore, not suit to present in section 4.

Overall, current manuscript like a report, and the results need further validation.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Good study. Changings in marine ecosystems under climatic and anthropogenic impact are a subject of the many investigations. Satellite data give unique possibility for regular monitoring of the sea surface properties. Authors apply different satellite data for detection and analysis of green tide development  in the Yellow sea.  NDVI index was used for solving this problem.  Some statistical parameters of green tide are discussed. Spectral response of  U. prolifera green tide and Sargassum gold tide used for separate these algae.

Some remarks

What is the reason to give such accuracy

 1191.1294 km2?

Check all figures subscribes  add date, position and sensors info.

Pls add some geographical  names on fig 1 for better understanding, for example “Subei shoal along the Jiangsu”

“which are corrected earth observation” – what do You mean corrected?

“temporal resolution of 12 h” – night too?

Envelope line – what is it?

Extraction result – pls define term

The percentage of U. prolifera in Figure 7(b)  was lower than in Figure 7(a 、c 、d) which was mainly affected by Sargassum – not correct sentence

Accumulative cover area – pls define

Accumulative cover area depends from surface currents and number of images.

I recommend to take into account remarks.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Overview and general recommendation:

The paper “Traceability and prediction of green tides in the Yellow Sea based on multi-source satellites” by Shuwen Xu et al., proposes a semi-automatic method for extraction of green tide information by using multi-source satellite data. Remote sensing (RS) images from GF-1, Landsat 5 TM, Landsat8 OLI/TIRS, HJ-1A/B, HY-1C, and MODIS have been used for green tide detection in the time interval from 2008 to 2022. Model performance has been evaluated by using the machine learning metrics Precision, Recall, and F1-score. According to authors, the proposed method performed well with majority of RS images from all considered satellites. However, for application of these metrics, together with a set of predicted values obtained by using the model respective set of actual (true) values is needed. More detailed description shall be added how the actual “true” distribution of pixels classified as representing green tide has been determined or obtained. The proposed method did use normalized difference vegetation index (NDVI) for classification of pixels of RS images for presence or non-presence of the green tide. Selection of the NDVI threshold value separating seawater from green tide area is the key element of the model. In the article, a single threshold value is applied for the whole region of interest (ROI). However, reference [30] (Fig. 3 and 4 in [30]) is stating, “For any type of optical sensor, the NDVI will be more or less affected by atmospheric aerosols, muddy water, and shallow areas and changes in environmental factors on satellite images will lead to uncertainty in the extraction results.” All these factors are present in the ROI of the present study, and difference between outcomes if a unified or specific threshold value is used may be rather significant. It is very likely that due to this difference green tide detection uncertainty will be much larger than the high accuracy about 98 % stated by authors. Green tide detection in its initial phase likely is more difficult as the green tide patches are mostly evolving as thin strips. In this phase, likely, the capability of the method is close to its detection limit and the uncertainty is large. Source site statistics 2008–2022 (Fig. 6) shows that the initial growth area is rather large and varies from year to year due to large number of influencing factors. Therefore, performance of complicated prediction models cannot be easily evaluated, especially when there were no real data for verification. Value of calculations and predictions made by authors would be much higher if respective uncertainty estimates would be also presented.

General suggestions:

1.     Description shall be added how the set of actual values with presence of green tide is determined.

2.     Calculation shall be added with selection of ROI excluding shallow water and muddy water areas. For example, calculations should be repeated with the threshold selected at least for two ROI-s: (1) Fig. 4(a) left part of the ROI including shallow and muddy water areas, and (2) Fig. 4(a) right part of ROI – consisting of the deep water area; agreement between the regions shall be evaluated.

3.     Consider adding uncertainty estimate for the predicted cumulative green tide cover area.

4.     Suggestion for a Title: “Monitoring and Forecasting Green Tide in the Yellow Sea Using Satellite Imagery”.

I recommend that after revision by authors the article may be published in Remote Sensing.    

Some minor suggestions for authors:

Row 22: Precision, Recall, and F1-score.

Row 28: instead of “approximately 1191.1294 km2” should be “approximately 1190 km2, accounting for uncertainty of the presented value.

Row 103: 2 Description of the dataset.

Figure 3 legend: lower critical threshold.

Table 3: Excessive number of significant digits in numerical values shall be omitted.

Figure 6: Units and axes titles for the both charts of Fig. 6 shall be uniform.

Figure 7c: Rrs – remote-sensing reflectance, unit (sr^-1). To all four charts unit (sr^-1) shall be added. Multiplier 10-3 is not correct. The range of y-axes shall be the same for all four charts.

Table 5: symbol 代入 shall be replaced.

Figure 8:  The range of y- and x-axes shall be the same for all fourteen charts. Comparison of Green Tide effects of different years will be much easier. Title for x-axes: Time (d). Symbol for the unit day is d.

Figure 9: Units for y-axes?

Figure 10: Titles and units for both axes.

Figure 11: Titles and units for both axes.

Row 469: approximately 1191.1294 km2. Number of significant digits shall correspond to the uncertainty of the value. Likely two or three significant digit will be sufficient. Uncertainty estimate is highly advisable.

Figure 12b: Titles and units for both axes.

Figure 13: Titles and units for both axes: y-Duration time (d); x-Monitoring time (a). Symbol for the unit year is a.

Between Rows 499-500: 5 Conclusions.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Overall, the authors have performed a lot of work to revise the manuscript. However, there are several issues still need to further revise.

 

(1) The English expression in current manuscript suggest to polish.

 

(2) Identification of U. prolifera and Sargassum was by the spectral shape. Therefore, the atmospheric correction needs to clarified in the method section. In addition, it is necessary to validate the accuracy of atmospheric correction.

 

(3) Equation (4)-(9) and corresponding contents should belong to Method section. Current manuscript needs to reorganized.

 

(4) the first paragraph in Section 6 is not a conclusion.

 

(5) Climate factors, such as wind, rainfall and temperature, and human factors are also can impact the green tide. Therefore, suggest to forecast green tide with climate and human factors to improve the reliability of the results.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper “Monitoring and Forecasting Green Tide in the Yellow Sea Using Satellite Imagery” by Shuwen Xu et al., has been significantly modified. The authors have satisfactorily responded to a number of my comments and made changes to the manuscript. I suggest that after revisions and editing of English language and style the manuscript could be accepted for publication.

Minor comments:

Row 22: Precision

Row 27: and an uncertainty for the predicted growth curve was estimated.

Row 98 to 102: This work is organized as follows: Section 1 - Introduction; Section 2 - Description of the dataset; Section 3 - Method, the methodology of the study is explained in detail; Section 4 - Results, the results for different RS datasets are presented, the accuracy assessed, the green tide’s growth curves calculated, forecast curve for 2022 estimated, tide’s bloom times and accumulative/maximum cover area simulated; Section 5 - Discussion, statistics of source sites of the green tides in period from 2008 to 2022 is analyzed, and Section 6 - Conclusions.

Section 3.4, Fig 5: What is the minimum number of pixels needed for the assessment of accuracy? What is the optimal number of pixels?

Row 453: Equation (4) and Table 2 are closely connected; maybe (4) with explanations should be transferred to section 3 before the Table 2.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf