Publications

Detailed Information

Generation Mechanism and Machine-Learning Forecasting Model of Sudden High Waves in the East Sea : 동해 돌연고파의 발생 메커니즘과 기계학습 예측모델

DC Field Value Language
dc.contributor.advisor서경덕-
dc.contributor.author오지희-
dc.date.accessioned2017-10-27T16:31:40Z-
dc.date.available2017-10-27T16:31:40Z-
dc.date.issued2017-08-
dc.identifier.other000000145784-
dc.identifier.urihttps://hdl.handle.net/10371/136693-
dc.description학위논문 (박사)-- 서울대학교 대학원 공과대학 건설환경공학부, 2017. 8. 서경덕.-
dc.description.abstractExceptional high waves have occurred repeatedly in the East Sea of Korea. These disastrous waves claimed the losses of life more than 50 people during the ten years between 2005 and 2015 in the east coast of Korea. Several researchers have examined the generation mechanism and characteristics of sudden high waves. However, the definition of the high waves is still vague and insufficient to explain the characteristics of sudden high waves. Also, occurrence of sudden high waves is only roughly forecasted in the daily weather forecast. In this study sudden high waves were defined using a new intensity parameter and the generation mechanism of the sudden high waves was investigated. Next, significant wave height and period were forecasted in the East Sea of Korea using machine-learning. Finally, sudden high waves were forecasted using the intensity parameter proposed and the forecasted significant waves in the East Sea of Korea.
In this study, the index of sudden high waves was suggested as and it was calculated using wave data measured in Gangneung and Wangdolcho in 2005–2012. The criteria of sudden high waves was set 80 m^3/hr, which corresponds to the top 20% of cumulative percentage of .
Next, to find the generation mechanism of sudden high waves, the evolution of spatial patterns of wind velocity and sea level pressures was presented during the sudden high wave events by CSEOF analysis and regression analysis. The wave data in Gangneung and Wangdolcho were used and the meteorological data were reanalysis data of National Centers for Environmental Prediction and National Center for Atmospheric Research (NCEP/NCAR). There are two peaks in the modes of all CSLV considered the physical process of sudden high waves. The patterns were categorized two groups. The first pattern was that the first peak was generated by low pressure moving to the north east part of the East Sea and easterly wind blowing for 1 day, whereas the second peak was caused by strong wind. The second pattern was that the first peak was affected by the wind speed in the east coast and the second peak was influenced by wind in the offshore area.
To forecast significant waves in multiple locations simultaneously, an EOFWNN model was developed by combining the EOF analysis and wavelet analysis with the neural network. The wave data used in this research were measured at eight wave observation stations in the East Sea and the meteorological data were the NCEP/NCAR reanalysis data. The results of the EOFWNN model for significant wave height were compared with those of a wavelet and neural network hybrid (WNN) model in Gangneung, Sakata and Aomori for several lead times. The EOFWNN model is better than the WNN model in that the former shows higher accuracy for longer lead times regardless of the wavelet decomposition level. Significant wave period series were also forecasted using the EOFWNN model. The results of significant wave period also show quite high accuracy. Also, the proposed model was employed to the numerical wave modeling data in the entire area of the East Sea. The results also show relatively high accuracy for one and three hour lead times.
Using the proposed intensity parameter of sudden high waves and the forecasted significant waves by the EOFWNN model, sudden high wave was detected and forecasted. From the forecasted wave data at 24 hour lead time, was calculated in Gangneung and Sakata. Although there is a slight deviation between the results of observed and forecasted wave data, sudden high wave was detected clearly.
-
dc.description.tableofcontentsCHAPTER 1 INTRODUCTION 1
1.1 Background 1
1.2 Research objectives 6

CHAPTER 2 THEORETICAL STUDY 8
2.1 Analysis methods of mechanism of sudden high waves 8
2.1.1 CSEOF analysis 9
2.1.2 Regreggion analysis 11
2.2 Wave forecast methods 13
2.2.1 EOF analysis 14
2.2.2 Wavelet analysis 16
2.2.3 Artificial Neural Network 19
2.2.4 EOF-Wavelet-ANN (EOFWNN) model 23

CHAPTER 3 CHARACTERISTICS OF SUDDEN HIGH WAVES 27
3.1 Data for anlaysis of characteristics of sudden high waves 27
3.1.1 Wave data 27
3.1.2 Meteorological data 27
3.2 Definition of sudden high waves 30
3.3 Mechanism of sudden high waves 43

CHAPTER 4 FORECASTING OF SUDDEN HIGH WAVES 62
4.1 Data for forecasting of sudden high waves 62
4.1.1 Observed wave data 62
4.1.2 Numerical wave modeling data 64
4.1.3 Meteorological data for forecasting 66
4.2 Forecasting of significant wave height and period using the observed data 67
4.3 Forecasting of significant wave height and peak period using the numerical modeling data 92
4.4 Detecting and forecasting of sudden high waves 103

CHAPTER 5 CONCLUSIONS 111
5.1 Summary and conclusions 111
5.2 Future study 114

REFERENCES 117

APPENDIX 121

국문초록 137
-
dc.formatapplication/pdf-
dc.format.extent3350901 bytes-
dc.format.mediumapplication/pdf-
dc.language.isoen-
dc.publisher서울대학교 대학원-
dc.subjectArtificial neural network-
dc.subjectEmpirical orthogonal function-
dc.subjectSignificant wave-
dc.subjectSudden high wave-
dc.subjectWave forecasting-
dc.subjectWavelet-
dc.subject.ddc624-
dc.titleGeneration Mechanism and Machine-Learning Forecasting Model of Sudden High Waves in the East Sea-
dc.title.alternative동해 돌연고파의 발생 메커니즘과 기계학습 예측모델-
dc.typeThesis-
dc.contributor.AlternativeAuthorJihee Oh-
dc.description.degreeDoctor-
dc.contributor.affiliation공과대학 건설환경공학부-
dc.date.awarded2017-08-
Appears in Collections:
Files in This Item:

Altmetrics

Item View & Download Count

  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Share