Implementation and application of a nested numerical storm surge forecast model in the East China Sea

A nested numerical storm surge forecast model for the East China Sea is developed. Aone-way relaxing nest method is used to exchange the information between coarse grid and fine grid. In the inner boundary of the fine grid model a transition area is set up to relax the forecast variables. This ensures that the forecast variables of the coarse model may transit to those of fine grid gradually, which enhances the model stability. By using this model, a number of hindcasts and forecast are performed for six severe storm surges caused by tropical cyclones in the East China Sea. The results show good agreement with the observations.     

A numerical storm surge forecast model with Kalman filter

ImUcrIONThe deterministic storm stirge nurnrical fOrecast Tnedel has played an imPOrtant role inroutine storm surge real-time fOrecast. But somtimes the error of forecast is still large by usingdeterministic medels (Je1esnianshi et al., l992). The source of these errors mainly comesfrom (1 ) errors of wind stress and medel's open boundary, (2) non--optimized medel param-eter, (3) error of model equations, (4) error of medel's numrical methed, etc. The effec-ti ve methed to solve this probl…     

基于改进的神经网络方法的风暴潮灾害经济损失预测

风暴潮灾害一直以来对中国东南沿海地区的社会经济发展具有较为严重的负面影响, 是对中国造成危害最为严重的海洋灾害之一, 建立一个准确有效的损失评估模型进行风暴潮灾害损失预测, 对风暴潮灾害的预防具有重要的意义。本文在现有研究的基础上收集了2000—2018年中国东南沿海的琼、粤、闽、浙等省份记录较为完整的风暴潮灾害相关数据, 在综合考虑危险性、承灾体脆弱性、孕灾环境和防灾减灾能力的基础上, 建立起更为完整的风暴潮灾害损失的指标体系。相较于单一的BP神经网络, 本文在借鉴机器学习相关理论的基础上搭建了差分进化灰狼算法(DEGWO)优化的BP神经网络, 对样本进行训练和仿真测试。结果表明, 通过DEGWO算法优化后的模型误差更小, 数据的拟合程度更高, 对比而言, 提高了风暴潮灾害损失预测的精确性, 能够为风暴潮灾害损失预测的研究提供新的思路, 同时也为风暴潮灾害的防灾减灾管理提供了指导。     

Forecast of storm surge by means of artificial neural network

This study describes the construction and verification of a model of sea level changes during a storm surge, applying artificial neural network (ANN) methodology in hydrological forecasting in a tideless sea where the variation of water level is only wind generated. Some neural networks were tested to create the forecast model. The results of ANN were compared with observed sea-level values, and with the forecasts calculated by different routine methods. The results of verification show that the neural network methodology could be successfully applied in the routine, operational forecast service.     

一个东海嵌套网格台风暴潮数值预报模式的研制与应用

建立了一个覆盖东海的两重嵌套网格高分辨率台风暴潮数值预报模式.粗、细网格模式分辨率分别为6'和2'.两套网格的嵌套采用单向松弛套网格技术,即在细网格的内边界附近建立了一个“过渡区”,对预报的物理量进行松弛,使粗、细网格模式变量逐步过渡,避免了边界附近寄生波的产生,从而增加了模式的稳定性.利用该模式,对显着影响东中国海地区的6次风暴潮过程进行了后报和预报试验.与观测资料比较,数值结果令人满意.     

神经网络在珠江口风暴潮预报中的应用

当风暴潮沿河道上溯时,处于珠江口地区的灯笼山测站和黄埔测站的水位之间存在着非线性响应关系。文章利用BP人工神经网络,建立了两测站台风暴潮和天文潮的综合增水效应预报模型,对9903、9908和9910号台风期间黄埔站的综合增水进行了预报,并针对不同预报时段对计算结果和潮位极值的准确程度进行了相应的讨论。结果表明,该方法的预报精度较高。最后通过与纯风暴增水模型的对比,说明了综合增水模型的优越性。     

灰色系统灾变预报方法在风暴潮预报中的应用

本文把灰色系统灾变预报方法应用到风暴潮预报中，对逐年最大风暴潮增水资料确定一个阈值ξ，对于年最大风暴潮位资料大于阈值ξ的年份组成一个序列，用一阶线性模型GM（1，1）预报风暴潮灾的出现年份，结果表明：用GM（1，1）模型可较好地预报风暴潮灾的出现年份。     

Deterministic and ensemble storm surge prediction for Atlantic Canada with lead times of hours to ten days

Regional deterministic and ensemble surge prediction systems (RDSPS and RESPS respectively) are used to forecast sea levels off the east of Canada and northeast US. The surge models for the RDSPS and RESPS have grid spacings of 1/30° and 1/12° respectively. The models are driven by surface air pressure and 10 m winds generated by operational global deterministic and ensemble prediction systems that are run operationally by the Canadian Meteorological Centre. Surge forecasts are evaluated for the period 1 March, 2013 to 31 March 2014. Based on traditional statistics (e.g., standard deviation of the difference between observations and predictions) both systems are shown to have skill in forecasting surges six days into the future. It is shown however that skill exists beyond six days if allowance is made for errors in the timing of large surges. The usefulness of the RESPS is demonstrated for two positive surges (important for coastal flooding and erosion) and a negative surge (important for safe navigation in shallow water). It is shown that the RESPS can identify events not forecast by the RDSPS, and can also add useful additional information on the timing of the surge, an important consideration in tidally dominated waters. Several new types of display are used to illustrate the sort of information that can be generated by the RESPS to support the issuers of warnings of unusually high and low total water levels.     

基于多种神经网络的风暴潮增水预测方法的比较分析

简要介绍了利用BP神经网络、小波神经网络、递归神经网络进行风暴潮增水值预测的原理。选取广东省珠江口以南的阳江站2017年风暴潮增水数据进行测试。结果表明,三种神经网络方法针对阳江地区风暴潮增水的预测均具有可靠性和实用性。以当前增水值为输入量的单因子模型更能反映真实风暴潮增水趋势,而从增水极值预测的准确性来看,以台风风力、气压、风向等相关参数为输入量的多因子模型优于单因子模型。BP神经网络更适用于多因子长时间预测,小波神经网络在单因子短时间预测上准确性更高,递归神经网络预测值与实测值相关性更强。在工程运用中,需根据地域时空特点、数据资料的丰富度与预测值评估指标选择合适的方法。     

基于高低潮的风暴增水人工神经网络预报模型

针对只有高低潮数据的情况,利用人工神经网络建立起一种预报当前台风时刻后第一个高潮时增水的模型。该模型选取台风在当前时刻、前6 h、前12 h、前18 h的中心经度、纬度、最大风速、中心气压以及当前时刻前第一个高潮时刻的风暴增水为输入单元。台风当前时刻后第一个高潮时刻风暴增水为模型输出单元。利用历史资料形成的规范化后的模式对,对模型进行训练,训练成功后,结合台风因子预报模型,即可用于风暴增水的预报。经过长江口高桥站高低潮实测资料的检验,结果表明该模型提取到了风暴增水效应,说明该模型可用于风暴增水的预报。     

覆盖中国沿海地区的精细化台风风暴潮模型的研究及适用

精细化风暴潮预报是目前风暴潮预报重点发展方向之一,本文首次建立起了一个覆盖整个中国沿海地区的精细化台风风暴潮数值模型,克服了以往分区域数值模型的不足,该模型在中国沿海地区的分辨率达到300m左右。模型采用了并行计算,并对2012年和2013年灾害性台风风暴潮过程进行了数值检验,计算精度和计算所用时间都能够满足业务化运行的要求。本文同时还根据中国气象局、美国国家气象局等5家主要台风预报机构给出的24h台风预报,对2013年度灾害性台风风暴潮过程进行了24h数值预报检验,检验结果表明:根据中国气象局台风登陆前24h预报可以得到更准确的风暴潮预报结果,其预报结果优于其他各家预报结果。该结论可以为今后的台风风暴潮预报中台风路径的选取提供重要的参考。     

Artificial neural networks and their application to assessment of ultimate strength of plates with pitting corrosion

The potential for the structural capability degrading effects of both corrosion and fatigue induced cracks are of profound importance and must be both fully understood and reflected in vessel's inspection and maintenance programme. Corrosion has been studied and quantified by many researchers, however its effect on structural integrity is still subject to uncertainty, particularly with regards to localized corrosion. The present study is focused on assessing the effects of localized pitting corrosion on the ultimate strength of unstiffened plates. Over 265 non-linear finite-element analyses of panels with various locations and sizes of pitting corrosion have been carried out. The results indicate that the length, breadth and depth of pit corrosion have weakening effects on the ultimate strength of the plates while plate slenderness has only marginal effect on strength reduction. Transverse location of pit corrosion is also an important factor determining the amount of strength reduction. When corrosion spreads transversely on both edges, it has the most deteriorating effect on strength. In addition, artificial neural network (ANN) method is applied to derive a formula to predict ultimate strength reduction of locally corroded plates. It is found out that the proposed formulae can accurately predict the ultimate strength of locally corroded plates under uniaxial in-plane compression.     

Near-shore swell estimation from a global wind-wave model: Spectral process,linear, and artificial neural network models

Estimation of swell conditions in coastal regions is important for a variety of public, government, and research applications. Driving a model of the near-shore wave transformation from an offshore global swell model such as NOAA WaveWatch3 is an economical means to arrive at swell size estimates at particular locations of interest. Recently, some work (e.g. Browne et al. [Browne, M., Strauss, D., Castelle, B., Blumenstein, M., Tomlinson, R., 2006. Local swell estimation and prediction from a global wind-wave model. IEEE Geoscience and Remote Sensing Letters 3 (4), 462–466.]) has examined an artificial neural network (ANN) based, empirical approach to wave estimation. Here, we provide a comprehensive evaluation of two data driven approaches to estimating waves near-shore (linear and ANN), and also contrast these with a more traditional spectral wave simulation model (SWAN). Performance was assessed on data gathered from a total of 17 near-shore locations, with heterogenous geography and bathymetry, around the continent of Australia over a 7 month period. It was found that the ANNs out-performed SWAN and the non-linear architecture consistently out-performed the linear method. Variability in performance and differential performance with regard to geographical location could largely be explained in terms of the underlying complexity of the local wave transformation.     

Artificial neural network modeling of cross-shore profile on sand beaches: The coast of the province of Valencia (Spain)

The paper describes the training, validation, testing, and application of models of artificial neural networks (ANN) for computing the cross-shore beach profile of the sand beaches of the province of Valencia (Spain). Sixty ANN models were generated by modifying both the input variables as the number of neurons in the hidden layer. The input variables consist of wave data and sedimentological data. To select and evaluate the performance of the optimal model, the following parameters were used: R², absolute error, mean absolute percentage error, and percentage relative error. Finally, the results are compared with the numerical model proposed by Aragonés et al. (2016b Aragonés, L., Y. Villacampa, F. J. Navarro-González, and I. López. 2016b. Numerical modelling of the equilibrium profile in Valencia (Spain). Ocean Engineering, 123:164–73. doi:10.1016/j.oceaneng.2016.07.036[Crossref], [Web of Science ®] , [Google Scholar]) for the equilibrium profile in the study area. The results show a mean absolute error of 0.21?m compared to 0.33?m Aragones’ model, significantly improving the results of the numerical model in the bar area around de Valencia Port. In addition, when comparing the results with other methods currently used (Dean’s or Vellinga formulation), the errors of these compared to ANN are of the order of 167 and 1538% higher, respectively.