Predicting Offshore Oil Slick Formation: A Machine Learning Approach Integrating Meteoceanographic Variables

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear authors

 

The study “Prediction of oil films on the ocean: an analysis of the impact of mete oceanographic variables and total oil and grease during primary petroleum processing’’.  The article focused on monitoring and mitigating the environmental impacts of offshore oil processing operations. The article presentation is obvious and the results discussion was an in-depth explanation and conclusion.  

Best regards

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper of 21 page, with half of a page Discussion and Conclusion does not look convincing. However, the paper can be improved.

In Introduction more attention should be given to ocean currents, which are mentioned too briefly.

In Methodology chapter there should be presented at least one satellite image with the  results, all the data for that image, so that reader can imagine what are the data you are dealing with.

Also, it is necessary to mention how meteorological and oceanographic parameters were measured. For example, how ocean currents were determined. What is mentioned from a platform and what from satellite images?

The chapter 5. Discussion and Conclusion must be changed.

I suggest that chapter 4 be called Results and Discussion. Here more explanations should be given about the advantage of each of the methods.

Separate chapter 5. Conclusion should be rewritten. In this chapter do not use acronyms. It should contain in brief the main results, evaluation of the methods and results.  It can contain remarks in relation to the methods used elsewhere.

In Conclusions people sometimes write what are they going to do next, so if a method is proven useful, the authors for example may think of establishing automatic monitoring system based on the methods. There may be some ideas how to improve the methods.

Line 53   "-0.005"   is it >0.005?

Legends of figures and tables should be more informative.

Figure 1 should be of a size of a full page. The legend should contain expanded acronyms and units.

Table 3. Correlation coefficients and significance levels between parameters:....  Note what is in the first and what in the second row. Expand acronyms in the legend.

line 411    change "can viewed"   with  " can be seen"

In equation (14)  what is the meaning of five decimal digits? Can it be shortened to two digits?

Figure 7. Explain what is in every row. This figure should be at one page. Describe the meaning of this figure in the legend.

Author Response

Please see the attachment

 

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Review of “Prediction of oil films on the ocean: an analysis of the impact of meteoceanographic variables and total oil and grease during primary petroleum processing”

 

This works seeks to provide a quantitative assessment of how metocean variables and oil layer thickness impact slick size and persistence on the ocean surface. Unfortunately, the bulk of the text explores the performance of different machine-learning techniques and very little effort is made to provide results that are useful. Virtually no effort is made to review relevant literature with regards to how the six metocean parameters (wind speed/direction, current speed/direction, wave period/direction) or oil layer thickness impact slick persistence or spreading.

The introduction is emblematic of the lack of balance in the paper. The first paragraph of the introduction is completely irrelevant and should be deleted. The second paragraph implies that grease might be released from an offshore separator – which is nonsense since grease is a refined product – only crude oil is potentially released along with produced water. The third paragraph review NOAA’s oil layer thickness classification but this is not used in the subsequent analysis and essentially irrelevant. The real content of the paper isn’t addressed until the fourth paragraph of the introduction and receives only 6 lines out of 87 of the introduction, when it should probably be closer to 60 lines out of 87.

The background and literature review section completely ignores the topic of the paper focusing only on the analytical methods for working with the data. If the paper is supposed to tell the reader something about the impact of metocean conditions on oil on the water surface, then the current understanding of that topic should be an essential part of the background. Most of what is included in this section should be relegated to an appendix.

The presentation of the data used is surficial and entirely inadequate. The observations are divided into those where the oil extends less than 500m from the platform and those where it extends more (classes 0 and 1). My reading of this is that oil is present in all 300 images and the distinction is it displacement from the platform. However subsequently the class 0 observations seem to be treated as null observations (in section 3 of the discussion). The oil detection is done via satellite but is it SAR detection or multispectral (or both)? Are all the data from the same satellite using the same detection mode? How are the limitations of the satellite detection capabilities imposed by the metocean conditions dealt with? The detection method itself imposes a bias on the data since neither type of satellite can detect oil on the water surface under all conditions.

The source of the metocean data used seems to be from a large-scale model via the Center for Weather Forecasting and Climate Studies (Centro de Pre-visão de Tempo e Estudos Climáticos - CPTEC). This is likely at 1/8^th of a degree resolution at best. It has been established in the slick literature that this type of data is generally inadequate for local modeling of individual surface slicks. Even in the Gulf of Mexico where the models have 1/25^th degree resolutions the models are often inadequate. Local wind and current data from the platform should used if possible. If local data is not available, the authors need to explore the limitations of the data they are using more thoroughly in the text.

The metocean parameters used have complex interdependencies that are not discussed. When the bearing of the current and wind are nearly opposite, they work against each other leading to less spread of the oil. It doesn’t matter what the actual bearing is for either parameter just that they are counterposed. When the wind is rotated at close to a right angle to the current it promotes lateral spreading of oil creating a wider slick (more area covered) but can limit the along current spread of the oil depending on the oil-layer thickness. Again, the actual orientations don’t matter it’s the relative orientations. If the wind and current directions are aligned the slick elongation is favored, again independent of the bearing. The parameter the authors should be using is the angular mismatch of the current and the wind.

The wind speed impacts the oil spread not simply by moving the oil across the water but also by promoting the formation of breaking waves which will destroy thinner surface slicks by forcing the water into the upper mixed layer in droplets that are too small to resurface. At lower wind speeds the wind can promote oil spreading and slick elongation (depending on its alignment with the current and their relative speeds) but at higher wind speeds the oil layer is destroyed leading to shorter/smaller slicks.

The wind wave directions and peak period are lagging indicators of the wind speed and direction and so could be useful for understanding if the current wind conditions are mostly in equilibrium with the current sea state or if the system is in significant flux. During periods of waning wind speed the wind driven waves will be larger than predicted by the wind speed and the dispersion of surface slicks greater than expected based on the wind speed. During periods of waxing wind speeds the situation is reversed. As the wind and current bearings, the wave parameters need to be convolved with the wind parameters to generate a parameter that should be predictive.

The oil layer thickness (I suggest the authors drop the term TOG and since grease would not spread on the surface of water) is determined colorimeterically. However, no conversion is provided between this measurement and an actual thickness in microns or millimeters. So as a reader I have no idea if the oil releases are more similar to: thin natural seepage slicks, moderate bilge-dump slicks, or thick pollution-spill slicks. Since the physics of oil spreading on water are strongly dependent on layer thickness the actual thickness needs to be provided in the text.

In the end it is concluded that wind speed is the most important parameter and that all the other parameters are of subequal importance. I agree that wind speed is likely the most important parameter but I think that by treating the other parameters as independent the authors have failed to capture their impact. For example, during times with low wind speeds and high current velocities substantial surface spreading of oil will take place.

I suggest that the authors refocus the paper on the physical processes they want to model. They need to review the relevant literature in a much more complete manner and decide how to convolve the parameters they ultimately use. Knowledge of the layer thickness of oil should drive them to either consider the slicks as being more like thin natural-seepage slicks or thicker pollution slicks. This will in turn point them toward the best way to convolve their parameters.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

This paper investigates the impact of wind direction (WD), wind speed (WS), current direction (CD), current speed (CS), wind wave direction (WWD), and peak period (PP)  factors on the occur-rence, detection, and extent of oil films using various statistical classifiers. Key findings reveal that Random Forest outperformed other classifiers, achieving an area under the ROC curve of 0.93.  Higher values of WS, WD, and CS were associated with a lower likelihood of oil film occurrence and detection, whereas higher TOG, PP, WWD, and CD values increased this probability. CS and TOG positively contributed to the extent of oil films, while high WS values reduced it. These results provide a robust decision-support framework for moni- toring and mitigating the environmental impacts of offshore oil processing operations.

This is a very meaningful work. However, the content description is lacking and it is not easy to read.

For example, how does paragraph 2 of the §1introduction page 2 support the full text? In §2, the author spends a lot of space on 6 methods, which are mature methods.Should you introduce people who have used these methods to do oil spill research, and help you to use these methods to innovate.

 

Here are some suggestions to consider：

1. What’s the meaning of ROC and TOG (The definition of TOG in the keywords is inconsistent with that in the abstract) in the abstract?
2. Abbreviations like KMO in Figure 2 are not explained.
3. What do Acc, Sn and Sp mean in Table 1?
4. What do Ac and tualmean in Table 2?
5. What do Effect,Coeff,SE Coeff ,T-Value, P-Value and VIF mean in Table 5?
6. What doF1-F6 mean in Table 4, Fig7, Formula15-21 ?

 

Comments on the Quality of English Language

Hope to explain clearly the variables in the chart, table and formula of the article.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 5 Report

Comments and Suggestions for Authors

The manuscript analyzes the impact of several meteorological and oceanographic variables on the formation and detection of oil films and compares the accuracy of five classifiers. The content is substantial, but there are still some shortcomings.

1. The abstract could be written more clearly to show that the authors have completed three parts of the work, highlighting which part is the key focus, and clarifying the internal connections between the three parts—what was done first and what followed.
2. Since the classification methods used in this manuscript are all well-known and commonly used techniques, it is recommended to streamline the introduction of classification methods in Chapter 2.
3. The literature review in the manuscript is not sufficiently comprehensive; it is advisable to include a summary of research findings related to the study's content.
4. The figure legends for lines 383 and 425 are incomplete; the legend colors in Figure 1 do not match the colors in the figure. It is suggested to add necessary legends for Figure 3.
5. It is recommended to provide corresponding explanations for the formulas used in Section 2.4, particularly for Formula 5 and Formula 6.
6. The discussion section of the manuscript is not thorough enough; it is suggested to incorporate findings from peers' studies and conduct
   a more in-depth analysis in conjunction with the results of this manuscript.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

English can be improved to express research more clearly.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

Comments and Suggestions for Authors

The article has made some adjustments, and the structure is higher than before; Here are some suggestions to consider.

1. What’s the meaning about “The wind-driven transport of oil occurs at approximately 3-4% of the wind speed"(The wind speed should be m/s or km/h. What does % mean?)

 

2. You have two FP's in the denominatorof Formal(1), right?
   3. What do the different colors in Figure 3 represent?

4. If the additional materials do not have their own innovation, just the results of predecessors, I think there is no comparison of additional, only in the research method can be mentioned. Because you're using the proven results of those who came before you.

Comments on the Quality of English Language

The article has made some adjustments, and the structure is higher than before; Here are some suggestions to consider.(The article has room for improvement)

1. What’s the meaning about “The wind-driven transport of oil occurs at approximately 3-4% of the wind speed"(The wind speed should be m/s or km/h. What does % mean?)

 

2. You have two FP's in the denominatorof Formal(1), right?
   3. What do the different colors in Figure 3 represent?

4. If the additional materials do not have their own innovation, just the results of predecessors, I think there is no comparison of additional, only in the research method can be mentioned. Because you're using the proven results of those who came before you.

Author Response

Comments 1: What’s the meaning about “The wind-driven transport of oil occurs at approximately 3-4% of the wind speed"(The wind speed should be m/s or km/h. What does % mean?)

Response 1: "The statement means that the velocity at which an oil slick moves due to wind is approximately 3-4% of the wind speed. If the wind speed is given in meters per second (m/s) or kilometers per hour (km/h), the oil slick's movement speed is calculated as a fraction of that value.

For example:

• If the wind speed is 10 m/s, the oil slick is expected to move at 0.3 to 0.4 m/s.
• If the wind speed is 36 km/h, the oil slick moves at 1.08 to 1.44 km/h.

The percentage represents a proportional factor used in oil spill modeling to estimate the effect of wind on oil transport. The exact value may vary depending on environmental conditions, such as water currents and oil properties."

We made this clear at line 148-152 

 

Comments 2: You have two FP's in the denominatorof Formal(1), right?

Response 2: Thank you so much for observing the error.

See line 305-306

 

 

Comments 3: What do the different colors in Figure 3 represent?

Response 3: Thanks for noticing that. The colors of the figures are related to several methods. Since we are using at the end only Random Forest, we updated the figure, explaining it

See lines 354-364

 

Comments 4: If the additional materials do not have their own innovation, just the results of predecessors, I think there is no comparison of additional, only in the research method can be mentioned. Because you're using the proven results of those who came before you.

Response 4: Thank you for your comment. The additional materials are included to provide context and support for the methodology used in our study. While they may not introduce new innovations themselves, they serve as essential references to validate our approach and ensure consistency with established research. Their inclusion offers a more comprehensive understanding of the background and enhances the reliability of our findings. However, we acknowledge your point and have ensured that the discussion focuses on how our research builds upon and applies these proven methods. As advised by some reviewers, we have included them at a minimum in the appendix.

Reviewer 5 Report

Comments and Suggestions for Authors

The author has revised as requested,good luck!

Author Response

Thanks!