Automated Analysis of Australian Tropical Cyclones with Regression, Clustering and Convolutional Neural Network
Round 1
Reviewer 1 Report
This is interesting and cutting-edge research. The authors used different methods to make predictions. It is suggested to publish after minor revision. A few comments for reference:
(1) There are many methods of data prediction. Why the author chooses the present ones needs further explanation. At the same time, different methods have different applicable conditions, so it is suggested that the author make an in-depth comparative analysis of the results of different methods.
(2) What implications does the author's research have for other regions or countries facing the threat of natural disasters? Can the present method be applied to natural disasters other than hot cyclones? If other countries adopt the author's approach to natural disaster prediction, what should they pay attention to? The author is suggested to do some in-depth and necessary discussion.
Author Response
This is interesting and cutting-edge research. The authors used different methods to make predictions. It is suggested to publish after minor revision.
We greatly appreciated the honorable reviewers’ efforts in reading our paper and finding it interesting. We have gone through all the comments / suggestions of the reviewer and updated our manuscript accordingly. As a result, the updated manuscript now contains 6 additional references, one additional table (i.e., Table 4), and about 800 additional words.
A few comments for reference:
(1) There are many methods of data prediction. Why the author chooses the present ones needs further explanation. At the same time, different methods have different applicable conditions, so it is suggested that the author make an in-depth comparative analysis of the results of different methods.
We agree that there are many methods of data prediction. To highlight these methods, we have added the following new references:
1. M. B. Richman, L. M. Leslie, H. A. Ramsay and P. J. Klotzbach, "Reducing Tropical Cyclone Prediction Errors Using Machine Learning Approaches," Procedia Computer Science, vol. 114, pp. 314-323, 2017. |
2. Z. Wang, J. Zhao, H. Huang and X. Wang, "A Review on the Application of Machine Learning Methods in Tropical Cyclone Forecasting," Front. Earth Sci., vol. 10, no. 902596, pp. 1-17, 2022. |
In this paper we focused on two cyclone related parameters: central pressure and maximum air speed to find out what other cyclone related parameters influences these two main parameters using regression, clustering and CNN based anomaly detection. To describe this, we have added the following paragraph within the conclusion section:
“It should be noted that this study predominantly analyzed only 2 cyclone-related parameters like central pressure and maximum air speed (out of 80 cyclone related parameters available in the source data [13] [14]). While regression and clustering algorithms were used on analyze central pressure dynamics, CNN-based anomaly detection was used to analyze central pressure of Australian cyclones. There are many more critical parameters like longitude, latitude, imagery etc. corresponding to geospatial analysis that were outside of the scope of this study (as demonstrated in [19] [36]). Moreover, important cyclone related analyses like cyclone prediction, rainfall prediction, landfall prediction, detecting cyclones of unprecedented scale and cyclone damage analysis were not investigated in this study (as demonstrated in [17] [18] [37] [38] [39]).”
In regard to the pre-condition or applicable conditions of different algorithms, we have added a new table called Table 4.
“… the researchers need to understand, which algorithms is to be used for solving what problem and use these algorithms in appropriate manner suiting various pre-conditions as shown in Table 4. “
Table 4. Preconditions and purpose of using regression, clustering, and CNN based anomaly detection in another research
Name of Algorithm |
Applicability |
Pre-condition |
Linear Regression |
Finding out the factors that drive the metric being analyzed. It analyzes the dataset, ranks the factors that matter, and displays them as key influencers. |
Supports only numeric values |
Logistic Regression |
Finding out the factors that drive the metric being analyzed. It analyzes the dataset, ranks the factors that matter, and displays them as key influencers. |
Supports only categorical values |
Clustering |
Finding out similarities or dissimilarities among categorical values by grouping them |
Supports only 1 dimension/categorical value and up to 15 measures/numeric values |
CNN based anomaly detection |
Automatically detects anomalies in time series data. It also provides explanations for the anomalies to help with root cause analysis. |
Supports only time-series data |
(2) What implications does the author's research have for other regions or countries facing the threat of natural disasters? Can the present method be applied to natural disasters other than hot cyclones? If other countries adopt the author's approach to natural disaster prediction, what should they pay attention to? The author is suggested to do some in-depth and necessary discussion.
We appreciated this valuable feedback. Accordingly, we have added the following paragraph within discussion section:
“Regression, Clustering & CNN algorithm configured, parameterized, and described in this paper for analyzing Australian cyclones could also be used in analyzing various other disasters as demonstrated in our recent research [6] [7] [8] [9] [10] [11] [12] [22] [23] [33] [34]. For example, we analyzed landslides in [6] [7] [10], tornadoes in [8] [9], global events in [22] [23] [24], and even COVID-19 in [33] [35]. In our previous studies as demonstrated [11] [12], natural disasters like Earthquakes, Bushfires, Floods, Volcano, Drought, Tsunami etc. were successfully analyzed with higher levels of accuracies. These AI-based algorithms could seamlessly be integrated on any datasets be it global (as demonstrated in [6] [7]) or local (as demonstrated in [8] [9] [10]).”
Reviewer 2 Report
Thank you for the opportunity of reviewing the interesting paper. It covers the important issue of analysis of tropical cyclones. I have the following comments:
Line310-315, Figures 6&7, Please spell out the words of “MAX WIND SPD”, “MAC WIND GUST”, etc.
Discussion
The paper compares the results with Ai related papers. Please compare the results with meteorological papers and discuss the findings from the meteorological perspective.
Please discuss limitation of the research. For example, can these findings be applied to cyclones of unprecedented scale?
Author Response
Thank you for the opportunity of reviewing the interesting paper. It covers the important issue of analysis of tropical cyclones. I have the following comments:
We greatly appreciated the honorable reviewers’ efforts in reading our paper and finding it interesting. We have gone through all the comments / suggestions of the reviewer and updated our manuscript accordingly. As a result, the updated manuscript now contains 6 additional references, one additional table (i.e., Table 4), and about 800 additional words.
- Line310-315, Figures 6&7, Please spell out the words of “MAX WIND SPD”, “MAC WIND GUST”, etc.
We acknowledge this feedback and accordingly we have updated the manuscript. Line 310 to 315, now says the following:
“For the case item 21 of Table 1, the relationship Maximum wind speed and Central pressure was explained by NLP as “When Maximum wind speed goes up 11.71 is more than 167, the average of Central Pressure decreases by 11.48”. On the other hand, for item 33 of Table 1, the relationship between Maximum wind gust and Central pressure was explained by our system as “When Maximum wind gust goes down 14.98, the average of Central Pressure decreases by 0.61” (as shown in Fig. 7).”
Figure 6 and 7 have also been updated as per the valuable suggestion.
Discussion
- The paper compares the results with Ai related papers. Please compare the results with meteorological papers and discuss the findings from the meteorological perspective.
We greatly appreciated this valuable suggestion and as a result we have added following new references and discussed about the findings within Material & Methods, Discussion and Conclusion sections:
1. K. Dube, L. Chapungu and J. Fitchett, "Meteorological and Climatic Aspects of Cyclone Idai and Kenneth," in Cyclones in Southern Africa. Sustainable Development Goals Series, Cham, Springer, 2021. |
2. M. Paliwal and A. Patwardhan, "Analysis of Cyclone Tracks of North Indian Ocean Using Cluster Analysis," in Monitoring and Prediction of Tropical Cyclones in the Indian Ocean and Climate Change, Dordrecht, Springer, 2014, pp. 89-96. |
3. M. B. Richman, L. M. Leslie, H. A. Ramsay and P. J. Klotzbach, "Reducing Tropical Cyclone Prediction Errors Using Machine Learning Approaches," Procedia Computer Science, vol. 114, pp. 314-323, 2017. |
4. E. C. Joint Research Centre, "Tropical Cyclone IDAI: analysis of the wind, rainfall and storm surge impact," 9 Apr 2019. [Online]. Available: https://www.humanitarianresponse.info/sites/www.humanitarianresponse.info/files/documents/files/joint_research_centre_analysis_of_wind_rainfall_and_storm_surge_impact_09_april_2019.pdf. [Accessed 1 08 2022]. |
5. Z. Wang, J. Zhao, H. Huang and X. Wang, "A Review on the Application of Machine Learning Methods in Tropical Cyclone Forecasting," Front. Earth Sci., vol. 10, no. 902596, pp. 1-17, 2022. |
For example, section “2.3. Algorithms for Analysis” now clearly states the following:
“For this study, we specifically focused on analyzing central pressure and maximum wind speed because of their importance from meteorological perspective [16] [15] [25].”
- Please discuss limitation of the research. For example, can these findings be applied to cyclones of unprecedented scale?
We greatly appreciated this valuable feedback and accordingly, we have added the following paragraph highlighting the limitation of this study within conclusion section:
“It should be noted that this study predominantly analyzed only 2 cyclone-related parameters like central pressure and maximum air speed (out of 80 cyclone related parameters available in the source data [13] [14]). While regression and clustering algorithms were used on analyze central pressure dynamics, CNN-based anomaly detection was used to analyze central pressure of Australian cyclones. There are many more critical parameters like longitude, latitude, imagery etc. corresponding to geospatial analysis that were outside of the scope of this study (as demonstrated in [19] [36]). Moreover, important cyclone related analyses like cyclone prediction, rainfall prediction, landfall prediction, detecting cyclones of unprecedented scale and cyclone damage analysis were not investigated in this study (as demonstrated in [17] [18] [37] [38] [39]).”
The algorithm used within this algorithm can be used on any dataset on Tornadoes, Cyclones, Tsunamis, Floods, Landslides, Bushfires, Earthquakes, or any other disasters. Regardless of the dataset chosen CNN-based anomaly detection can identify, highlight, and explain natural disasters on an unprecedented scale by analyzing time-frequency data of disaster event. To further explain this concept, we have added the following paragraph within discussion section:
“Regression, Clustering & CNN algorithm configured, parameterized, and described in this paper for analyzing Australian cyclones could also be used in analyzing various other disasters as demonstrated in our recent research [6] [7] [8] [9] [10] [11] [12] [22] [23] [33] [34]. For example, we analyzed landslides in [6] [7] [10], tornadoes in [8] [9], global events in [22] [23] [24], and even COVID-19 in [33] [35]. In our previous studies as demonstrated [11] [12], natural disasters like Earthquakes, Bushfires, Floods, Volcano, Drought, Tsunami etc. were successfully analyzed with higher levels of accuracies. These AI-based algorithms could seamlessly be integrated on any datasets be it global (as demonstrated in [6] [7]) or local (as demonstrated in [8] [9] [10]). However, the researchers need to understand, which algorithms is to be used for solving what problem and use these algorithms in appropriate manner suiting various pre-conditions as shown in Table 4. “
Table 4. Preconditions and purpose of using regression, clustering, and CNN based anomaly detection in another research
Name of Algorithm |
Applicability |
Pre-condition |
Linear Regression |
Finding out the factors that drive the metric being analyzed. It analyzes the dataset, ranks the factors that matter, and displays them as key influencers. |
Supports only numeric values |
Logistic Regression |
Finding out the factors that drive the metric being analyzed. It analyzes the dataset, ranks the factors that matter, and displays them as key influencers. |
Supports only categorical values |
Clustering |
Finding out similarities or dissimilarities among categorical values by grouping them |
Supports only 1 dimension/categorical value and up to 15 measures/numeric values |
CNN based anomaly detection |
Automatically detects anomalies in time series data. It also provides explanations for the anomalies to help with root cause analysis. |
Supports only time-series data |
Round 2
Reviewer 2 Report
my comments have been incorporated into the revised draft.