A Systematic Study on Berthing Capacity Assessment of Sanya Yazhou Fishing Port by Typhoon Prediction Model
Abstract
:1. Introduction
2. Materials and Methods
2.1. Wind Field Model
2.2. Wave Models
2.2.1. MIKE 21 SW Model
2.2.2. MIKE 21 BW Model
2.3. Assessment of the Berthing Quantity of the Fishing Port
2.3.1. Single Anchor Mooring
2.3.2. Multiple Vessels Abreast with a Single Anchor
3. Results and Discussion
3.1. Wind Field Model
3.2. Verification
3.3. MIKE 21 BW Model
3.3.1. Setting of the Model
3.3.2. Maximum Significant Wave Heights and Peak Wave Periods
3.3.3. Wave Distribution in SYFP
3.4. Berthing Quantity of SYFP
4. Conclusions
Author Contributions
Funding
Conflicts of Interest
References
- Chang, C.-H.; Shih, H.-J.; Chen, W.-B.; Su, W.-R.; Lin, L.-Y.; Yu, Y.-C.; Jang, J.-H. Hazard Assessment of Typhoon-Driven Storm Waves in the Nearshore Waters of Taiwan. Water 2018, 10, 926. [Google Scholar] [CrossRef] [Green Version]
- Yan, Z.; Liang, B.; Wu, G.; Wang, S.; Li, P. Ultra-long return level estimation of extreme wind speed based on the deductive method. Ocean Eng. 2020, 197, 106900. [Google Scholar] [CrossRef]
- Gopikrishna, B.; Deo, M. Changes in the shoreline at Paradip Port, India in response to climate change. Geomorphology 2018, 303, 243–255. [Google Scholar] [CrossRef]
- Doong, D.-J.; Chuang, H.-C.; Shieh, C.-L.; Hu, J.-H. Quantity, distribution, and impacts of coastal driftwood triggered by a typhoon. Mar. Pollut. Bull. 2011, 62, 1446–1454. [Google Scholar] [CrossRef]
- Liang, B.; Ge, H.; Zhang, L.; Liu, Y. Wave resonant scattering mechanism of sinusoidal seabed elucidated by Mathieu Instability theorem. Ocean Eng. 2020, 218, 108238. [Google Scholar] [CrossRef]
- Chen, W.-B.; Chen, H.; Hsiao, S.-C.; Chang, C.-H.; Lin, L.-Y. Wind forcing effect on hindcasting of typhoon-driven extreme waves. Ocean Eng. 2019, 188, 1–17. [Google Scholar] [CrossRef]
- Yan, Z.; Pang, L.; Dong, S. Analysis of Extreme Wind Speed Estimates in the Northern South China Sea. J. Appl. Meteorol. Clim. 2020, 59, 1625–1635. [Google Scholar] [CrossRef]
- Yan, Z.; Wu, G.; Liang, B.; Li, P. A Stochastic Tropical Cyclone Model for the Northwestern Pacific Ocean with Improved Track and Intensity Representations. Appl. Ocean Res. 2020, 105, 102423. [Google Scholar] [CrossRef]
- Kong, C.; Lu, Y.; Han, T. Assessment of typhoon prevention level at fishing port. In IOP Conference Series: Earth and Environmental Science; IOP Publishing: Bristol, UK, 2021. [Google Scholar]
- Sun, Z.; Wang, C.; Zhong, S.; Ji, H. Numerical simulation of storm surge coupled with wave in zhoushan fshing port. Mar. Sci. Bull. 2019, 38, 150–158. [Google Scholar]
- Wang, Y.; Shangguan, Z.; Dong, W.; Qi, Q. Structural reliability of fshing chains at a single anchor in fshing port under typhoon calamity. J. Dalian Ocean. Univ. 2016, 31, 692–695. [Google Scholar]
- Fan, K.; Wang, H. A New Approach to Forecasting Typhoon Frequency over the Western North Pacific. Weather. Forecast. 2009, 24, 974–986. [Google Scholar] [CrossRef]
- Remya, P.G.; Kumar, R.; Basu, S. An assessment of wind forcing impact on a spectral wave model for the Indian Ocean. J. Earth Syst. Sci. 2014, 123, 1075–1087. [Google Scholar] [CrossRef] [Green Version]
- Hu, Q.K.; Chen, Q.; Fitzpatrick, P. Assessment of a Parametric Hurricane Surface Wind Model for Tropical Cyclones in the Gulf of Mexico; InTch: London, UK, 2012. [Google Scholar]
- Holland, G.J. An Analytic Model of the Wind and Pressure Profiles in Hurricanes. Mon. Weather. Rev. 1980, 108, 1212–1218. [Google Scholar] [CrossRef]
- Holland, G.J.; Belanger, J.I.; Fritz, A. A Revised Model for Radial Profiles of Hurricane Winds. Mon. Weather. Rev. 2010, 138, 4393–4401. [Google Scholar] [CrossRef]
- Ge, H.; Liu, H.; Zhang, L. Accurate Depth Inversion Method for Coastal Bathymetry: Introduction of Water Wave High-Order Dispersion Relation. J. Mar. Sci. Eng. 2020, 8, 153. [Google Scholar] [CrossRef] [Green Version]
- Hou, W.; Zhang, R.; Zhang, P.; Xi, Y.; Ma, Q. Wave Characteristics and Berthing Capacity Evaluation of the Offshore Fishing Port under the Influence of Typhoons. Appl. Ocean Res. 2020, 106, 102447. [Google Scholar] [CrossRef]
- Lee, H.S.; Kim, K.O. Storm Surge and Storm Waves Modelling Due to Typhoon Haiyan in November 2013 with Improved Dynamic Meteorological Conditions. Procedia Eng. 2015, 116, 699–706. [Google Scholar] [CrossRef] [Green Version]
- Hsiao, S.-C.; Chen, H.; Chen, W.-B.; Chang, C.-H.; Lin, L.-Y. Quantifying the contribution of nonlinear interactions to storm tide simulations during a super typhoon event. Ocean Eng. 2019, 194, 106661. [Google Scholar] [CrossRef]
- Madsen, P.A.; Murray, R.; Sørensen, O.R. A new form of the Boussinesq equations with improved linear dispersion characteristics. Coast. Eng. 1991, 15, 371–388. [Google Scholar] [CrossRef]
- Madsen, P.A.; Murray, R.; Sørensen, O.R. A new form of the Boussinesq equations with improved linear dispersion characteristics. Part 2. A Slowly-varying bathymetry. Coast. Eng. 1992, 18, 183–204. [Google Scholar] [CrossRef]
- Wang, S.P.; Rui-Jie, L.I.; Ming, J.I.; Dong, X.T.; Zhu, W.J. Combined application of swan and cgwave model in calculation of design wave parameters. J. Waterw. Harb. 2015, 4, 308–312. [Google Scholar]
- Gou, H.; Luo, F.; Li, R.; Dong, X.; Zhang, Y. Modeling Study on the Hydrodynamic Environmental Impact Caused by the Sea for Regional Construction near the Yanwo Island in Zhoushan, China. Water 2019, 11, 1674. [Google Scholar] [CrossRef] [Green Version]
- Ying, M.; Zhang, W.; Yu, H.; Lu, X.; Feng, J.; Fan, Y.; Zhu, Y.; Chen, D. An Overview of the China Meteorological Administration Tropical Cyclone Database. J. Atmos. Ocean. Technol. 2014, 31, 287–301. [Google Scholar] [CrossRef] [Green Version]
- Graham, H.E. Meteorological Considerations Pertinent to Standard Project Hurricane; National Hurricane Research Project Report; US Department of Commerce, Weather Bureau: Washington, DC, USA, 1959.
- Olfateh, M.; Callaghan, D.P.; Nielsen, P.; Baldock, T.E. Tropical cyclone wind field asymmetry-Development and evaluation of a new parametric model. J. Geophys. Res. Oceans 2017, 122, 458–469. [Google Scholar] [CrossRef]
- Ministry of Agriculture and Rural Affairs of the People’s Republic of China. National Standards of People’s Republic of China: Master Design Code of Fishery Port; The Ministry of Agriculture of the People’s Republic of China Beijing: Beijing, China, 2000.
- John, F. Nonparametric Regression. Appendix to an R and S-PLUS Companion to Applied Regression; Sage Publications: Beverly Hills, CA, USA, 2002; pp. 67–82. [Google Scholar]
- PRC MOAT. Grade of Tropical Cyclones (GB/T19201-2006); China Meteorological Administration: Beijing, China, 2006. [Google Scholar]
- Hsiao, S.-C.; Chen, H.; Wu, H.-L.; Chen, W.-B.; Chang, C.-H.; Guo, W.-D.; Chen, Y.-M.; Lin, L.-Y. Numerical Simulation of Large Wave Heights from Super Typhoon Nepartak (2016) in the Eastern Waters of Taiwan. J. Mar. Sci. Eng. 2020, 8, 217. [Google Scholar] [CrossRef] [Green Version]
- Feng, W.B.; Huang, Y.N..; Li, J.H. Application of wave feild simulation in sheltered water based on BW model. Port Waterw. Eng. 2015, 11, 8–13. [Google Scholar]
- Zhao, J.; Yan, Z.; Chen, X.; Pang, L.; Lu, Y. The Calculation of Extreme Wind Speed Based on Bayesian Method. J. Coast. Res. 2020, 99, 105–114. [Google Scholar] [CrossRef]
- Huang, C.C.; Fang, H.T.; Ho, H.C.; Jhong, B.C. Interdisciplinary application of numerical and manchine-learing-based models to predict half-hourly suspended sediment concentrations during typhoons. J. Hydrol. 2019, 573, 661–675. [Google Scholar] [CrossRef]
Level of Typhoon | Significant Wave Height (m) | T (s) |
---|---|---|
12 | 2.21 | 10.08 |
13 | 2.39 | 10.38 |
14 | 2.52 | 10.67 |
15 | 2.63 | 10.50 |
16 | 2.74 | 10.68 |
17 | 2.82 | 11.33 |
Anchoring Way | Type of Vessel | Mooring Area of Single/Single Group Fishing Vessel (m2) | Number of Vessels |
---|---|---|---|
Single anchor mooring | Small | 13,747 | 33 |
Large | 26,101 | 17 | |
Multiple vessels mooring side-by-side with a single anchor | Small (6 vessels one set) | 3755 | 735 |
Large (2 vessels one set) | 5812 | 158 |
Direction of Typhoon | Level of Typhoon | Number of Small Vessels | Anchoring Area (m2) of Small Vessels | Number of Large Vessels | Anchoring Area (m2) of Large Vessels |
---|---|---|---|---|---|
12 | 615 | 385,000 | 156 | 450,000 | |
13 | 599 | 375,000 | 156 | 450,000 | |
14 | 591 | 370,000 | 156 | 450,000 | |
15 | 583 | 365,000 | 156 | 450,000 | |
16 | 559 | 350,000 | 156 | 450,000 | |
17 | 503 | 315,000 | 156 | 450,000 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Ge, H.; Wang, Z.; Liang, B.; Zhang, Z.; Yan, Z.; Li, Z. A Systematic Study on Berthing Capacity Assessment of Sanya Yazhou Fishing Port by Typhoon Prediction Model. J. Mar. Sci. Eng. 2021, 9, 1380. https://doi.org/10.3390/jmse9121380
Ge H, Wang Z, Liang B, Zhang Z, Yan Z, Li Z. A Systematic Study on Berthing Capacity Assessment of Sanya Yazhou Fishing Port by Typhoon Prediction Model. Journal of Marine Science and Engineering. 2021; 9(12):1380. https://doi.org/10.3390/jmse9121380
Chicago/Turabian StyleGe, Hongli, Zhenlu Wang, Bingchen Liang, Zhaozi Zhang, Zhiduo Yan, and Ziwang Li. 2021. "A Systematic Study on Berthing Capacity Assessment of Sanya Yazhou Fishing Port by Typhoon Prediction Model" Journal of Marine Science and Engineering 9, no. 12: 1380. https://doi.org/10.3390/jmse9121380
APA StyleGe, H., Wang, Z., Liang, B., Zhang, Z., Yan, Z., & Li, Z. (2021). A Systematic Study on Berthing Capacity Assessment of Sanya Yazhou Fishing Port by Typhoon Prediction Model. Journal of Marine Science and Engineering, 9(12), 1380. https://doi.org/10.3390/jmse9121380