Multi-step ahead short-term predictions of storm surge level using CNN and LSTM network

Storm surges pose significant danger and havoc to the coastal residents' safety, property, and lives, particularly at offshore locations with shallow water levels. Predictions of storm surges with hours of warning time are important for evacuation measures in low-lying regions and coastal management plans. In addition to experienced predictions and numerical models, artificial intelligence (AI) techniques are also being used widely for short-term storm surge prediction owing to their merits in good level of prediction accuracy and rapid computations. Convolutional neural network (CNN) and long short-term memory (LSTM) are two of the most important models among AI techniques. However, they have been scarcely utilised for surge level (SL) forecasting, and combinations of the two models are even rarer. This study applied CNN and LSTM both individually and in combination towards multi-step ahead short-term storm surge level prediction using observed SL and wind information. The architectures of the CNN, LSTM, and two sequential techniques of combining the models (LSTM–CNN and CNN–LSTM) were constructed via a trial-and-error approach and knowledge obtained from previous studies. As a case study, 11 a of hourly observed SL and wind data of the Xiuying Station, Hainan Province, China, were organised as inputs for training to verify the feasibility and superiority of the proposed models. The results show that CNN and LSTM had evident advantages over support vector regression (SVR) and multilayer perceptron (MLP), and the combined models outperformed the individual models (CNN and LSTM), mostly by 4％–6％. However, on comparing the model computed predictions during two severe typhoons that resulted in extreme storm surges, the accuracy was found to improve by over 10％ at all forecasting steps.     

基于KPCA-RBF模型的风暴潮灾害经济损失预测

作为破坏性最强的海洋灾害,风暴潮灾害每年都给我国沿海地区造成了巨大的经济损失,运用科学的方法模型合理预测风暴潮灾害经济损失对指导沿海地区的防灾减灾工作意义深远。本文基于风暴潮灾害的成灾特点建立了风暴潮灾害直接经济损失预评估指标体系,由于评估指标数据高度非线性,采用核主成分分析（KPCA）对高维非线性数据进行降维优化,并利用径向基函数（RBF）神经网络对降维后的数据进行训练,从而实现对风暴潮灾害直接经济损失的预测。选取广东省1996—2018年的32个风暴潮灾害损失样本对模型进行仿真测试,结果表明,KPCA-RBF预测模型集成了核主成分分析和径向基函数神经网络的优势,预测结果精度高,学习收敛速度快,对风暴潮灾害数据序列有较好的非线性拟合能力。     

基于潮汐表数据同化的天文潮数值预报模型及其模拟预报效果

潮汐表是利用长期潮汐观测结果经调和分析实现的主要港湾潮汐预报结果,具有较高的预报精度,而通常的天文潮数值预报目前还难以达到潮汐表的预报精度.本研究在建立常规天文潮数值预报模型的基础上,建立了基于潮汐表数据同化的天文潮数值预报模型,并分别采用这2种模型预报福建沿岸海域的天文潮.其结果表明同化模型的预报结果无论是在潮时还是在潮高均明显优于常规模型;同化模型能显著地改善所研究的沿岸海域90个水位点中至少45个水位点的潮汐预报结果,而其他水位点的预报结果也有不同程度地改善.     

山东邻海长周期分潮对深度基准面的影响分析

利用潮汐模型NAO.99Jb和FES2014确定了山东邻海的深度基准面模型并对其精度进行了评估,结果表明,NAO.99Jb模型确定的深度基准值L10的中误差为23.28 cm,FES2014模型确定的深度基准值L13的中误差为34.37 cm,长周期分潮的相对误差过大导致加入长周期分潮改正项后深度基准值中误差分别增大了11.04 cm和12.38 cm,较其他分潮对深度基准值精度的影响更明显,所以基于潮汐模型构建深度基准面模型时,长周期分潮部分必须加入实测数据改正。进一步采用山东邻海13个长期验潮站实测数据,定量地分析了长周期分潮对深度基准面确定的影响,结果表明,长周期分潮改正项的量值介于13.89~22.39 cm,平均改正值为18.03 cm,在深度基准值中占比达到15.15%。因此,长周期分潮改正对深度基准面的精确确定研究贡献较大,准确的长周期分潮模型是构建高精度深度基准面模型的基础。     

人工神经网络在潮汐数值预报中的应用

潮汐数值预报经过了几十年的发展,但是其预报精度并不能让人十分满意,本文试图将传统的潮汐数值预报模式与近年来发展迅速的人工神经网络相结合并改进潮汐数值预报的精度。文章建立了一个神经网络系统,采用潮汐数值模式的输出结果作为网络输入,潮位观测资料作为输出,用建立的神经网络进行训练,结果表明人工神经网络可以明显地改进潮汐数值预报的精度。     

用混沌理论提高潮汐预报的准确度

为提高潮位预报的准确性,在具有较长潮汐观测数据的站点,基于混沌理论,对观测值与潮汐模型预测值之差所构成的余水位序列(即误差序列),采用局域线性模型的分析方法,给出可能误差预测,修正模型的预报结果,提高潮汐预报的准确性。所给例子,对预测跨度T=2 h,经局域法修正后,崇武站2007年12月份1个月预测水位统计的RMSE值减少74.7%,厦门站减少60.5%;对T=24 h,崇武、厦门两站RMSE值减小都在50%左右。     

一种基于小波分解和模糊推理神经网络的潮位预测方法

潮位预测严重影响沿海区域，尤其是近海浅水沿岸地区居民的生产生活和涉海活动。谐波分析是长周期潮位预测的传统方法，但无法预测非周期性气象过程发生时的水位变化。与数据处理方法相结合，人工智能的方法通过拟合输入与输出数据的历史数值关系，能够有效预测高度非线性和非平稳的流模式，因而在时间序列数据预测领域得到了广泛的应用。本文结合自适应模糊推理系统（Adaptive Neuro-Fuzzy Inference System, ANFIS）和小波分解方法，利用水位异常和风切变分量作为输入数据，实现了一种综合的多时效潮位预测方法。文中测试了多种输入变量组合和小波-ANFIS（WANFIS）模型，并与人工神经网络（Artificial Neural Network, ANN）、小波-ANN（WANN）和ANFIS模型进行了预测结果对比。通过不同指数的误差分析来看，相比ANN模型，ANFIS模型能够更准确的预测潮位变化，小波分解对ANFIS预测精度有一定的提高，且模型中水位异常和风切变分量数据的加入比单一的潮位数据输入能取得更好的预测结果。     

下一代大气模式中的数值方法综述

数值预报是逐日天气预报、气候预测和气象防灾减灾的核心科技支撑。为进一步提高预报预测的准确度和延长预见期,甚高分辨率、多圈层耦合、多尺度嵌套、多尺度集合、数值地球系统模拟技术等是下一代数值预报的重要发展方向。异构众核高性能计算机和E级计算的高速发展为这一发展提供了契机,但也对现有业务数值预报中采用的数值方法提出了挑战。此文仅对国内外下一代大气模式涉及到的数值方法进行综述,着重于数值算法、准均匀球面网格和时间积分方案等3个方面,期望为相关研究者提供参考。     

Wavelet and artificial neural network analyses of tide forecasting and supplement of tides around Taiwan and South China Sea

In multi-resolution analysis (MRA) by wavelet function Daubechies (db), we decompose the signal in two parts, the low and high-frequency contents. We remove the high-frequency content and reconstruct a new “de-noise” signal by using inverse wavelet transform. The calculation of tidal constituent phase-lags was made to determine the input and output data patterns used in building network structure of Artificial Neuron-Network (ANN) model. The “de-noise” signal was, then, used as the input data to improve the forecasting accuracy of the ANN model. The wavelet spectrum, conventional energy spectrum (fast Fourier transform, FFT), and harmonic analysis were used to analyze the characteristics of tidal data.Using only a very short-period data as a training data set in Artificial Neuron-Network Back-Propagate (ANN-BP) model, the developed ANN+Wavelet model can accurately predict or supply the missing tide data for a long period (1–5 years). The results also show that the concept of tidal constituent phase-lags can improve ANN model of tidal forecasting and data supplement. The addition of the wavelet analysis to ANN method can prominently improve the prediction quality.     

数值模式与统计模型相耦合的近岸海浪预报方法

针对数值模式和统计模型预报近岸海浪存在的局限性,构建了数值模式和统计模型相耦合的近岸海浪预报框架,在模式计算格点和近岸预报目标点之间定义一个海浪能量密度谱传递系数,通过经验正交函数分解和卡尔曼滤波方法建立传递系数的统计预报模型并与数值模式进行耦合。经过对近岸波浪观测站1a的预报试验表明:该方法能够提高近岸海浪有效波高预报精度,有效波高的均方根误差降低了约0.16m,平均相对误差降低约9%。进一步试验和分析发现,该方法的预报有效时间小于24h,将海浪能量密度谱经过分解后得到的基本模态反映了近岸波侯的主要特征,海浪能量密度谱传递系数的变化体现了波侯的季节变化特点。     

Application of artificial neural networks in tide-forecasting

An accurate tidal forecast is an important task in determining constructions and human activities in ocean environments. Conventional tidal forecasting has been based on harmonic analysis using the least squares method to determine harmonic parameters. However, a large number of parameters are required for the prediction of a long-term tidal level with harmonic analysis. Unlike conventional harmonic analysis, this paper presents an artificial neural network (ANN) model for forecasting the tidal-level using the short term measuring data. The ANN model can easily decide the unknown parameters by learning the input–output interrelation of the short-term tidal records. Three field data with three types of tides will be used to test the performance of the proposed ANN model. The numerical results indicate that the hourly tidal levels over a long duration can be predicted using a short-term hourly tidal record.     

Uncertainty Analysis of Wind-Wave Predictions in Lake Michigan

With all the improvement in wave and hydrodynamics numerical models, the question rises in our mind that how the accuracy of the forcing functions and their input can affect the results. In this paper, a commonly used numerical third-generation wave model, SWAN is applied to predict waves in Lake Michigan. Wind data are analyzed to determine wind variation frequency over Lake Michigan. Wave predictions uncertainty due to wind local effects are compared during a period where wind has a fairly constant speed and direction over the northern and southern basins. The study shows that despite model calibration in Lake Michigan area, the model deficiency arises from ignoring wind effects in small scales. Wave prediction also emphasizes that small scale turbulence in meteorological forces can increase prediction errors by 38%. Wave frequency and coherence analysis show that both models can predict the wave variation time scale with the same accuracy. Insufficient number of meteorological stations can result in neglecting local wind effects and discrepancies in current predictions. The uncertainty of wave numerical models due to input uncertainties and model principals should be taken into account for design risk factors.     

Local model approximation in the real time wave forecasting

The notion of data assimilation is common in most wave predictions. This typically means nudging of wave observations into numerical predictions so as to drive the predictions towards the observations. In this approach, the predicted wave climate is corrected at each time of the observation. However, the corrections would diminish soon in the absence of future observations. To drive the model state predictions towards real time climatology, the updating has to be carried out in the forecasting horizon too. This could be achieved if the wave forecasting at the observational network is made available. The present study addresses a wave forecasting technique for a discrete observation station using local models. Embedding theorem based on the time-lagged embedded vector is the basis for the local model. It is a powerful tool for time series forecasting. The efficiency of the forecasting model as an error correction tool (by combining the model predictions with the measurements) has been brought up in a forecasting horizon from few hours to 24 h. The parameters driving the local model are optimised using evolutionary algorithms.     

基于调和分析法与ANFIS系统的综合潮汐预报模型

港口沿岸地区以及河流入海口等地区的精确潮汐预报对于各种海洋工程作业有着非常重要的意义。潮汐水位的变化受到众多复杂因素的影响,而且这些复杂的因素往往有着较强的实变性和非线性。为了进一步提高沿岸港口码头等水域的潮汐水位的预测精度,本文提出了一种基于调和分析模型与自适应神经模糊推理系统相结合的模块化潮汐水位预测模型;并采用相关分析确定整个预测模型的输入维数;模块化将潮汐分解为两部分：由天体引潮力形成的天文潮部分和由各种天气以及环境因素引起非天文潮部分。其中调和分析法用于天文潮部分的预测,ANFIS用于预测具有较强非线性的非文潮部分。模块化综合了两种方法的优势,即调和分析法能够实现长期、稳定的天文潮预报,ANFIS能够以较高的精度实现潮汐非线性拟合与预测。模型使用ANFIS模型和调和分析模型分别对潮汐的非天文潮和天文潮部分进行仿真预测,然后将两部分的预测结果综合形成最终的潮汐预测值。此外,本文选用三种不同的模糊规则生成方法（grid partition （GP）,fuzzy c-means （FCM） and sub-clustering （SC））生成完整的ANFIS系统,并使用实测数据进行验证用以选取最优的ANFIS预测模型。最后将最优的ANFIS模型与调和分析模型相结合进行潮汐水位的最终预报。仿真实验选用Fort Pulaski潮汐观测站的实测潮汐值数据进行预报的仿真实验,仿真结果验证了该模型的可行性与有效性并取得了良好的效果,具有较高的预报精度。     

Effects of DEM Resolution on Modeling Coastal Flood Vulnerability

Abstract

This article examines whether Digital Elevation Model (DEM) resolution affects the accuracy of predicted coastal inundation extent using LISFLOOD-FP, with application to a sandy coastline in New Jersey. DEMs with resolution ranging from 10 to 100 m were created using coastal elevation data from NOAA, using the North American Vertical Datum of 1988. A two-dimensional hydrodynamic flood model was developed in LISFLOOD-FP using each DEM, all of which were calibrated and validated against an observed 24-h tidal cycle and used to simulate a 1.5 m storm surge. While differences in predicted inundated area from all models were within 1.0%, model performance and computational time worsened and decreased with coarser DEM resolution, respectively. This implied that using a structured grid model for modeling coastal flood vulnerability is based on two trade-offs: high DEM resolution coupled with computational intensity, but higher precision in model predictions, and vice versa. Furthermore, water depth predictions from all DEMs were consistent. Using an integrated numerical modeling and GIS approach, a two-scale modeling strategy, where a coarse DEM is used to predict water levels for projection onto a fine DEM was found to be an effective, and computationally efficient approach for obtaining reliable estimates of coastal inundation extent.     

三种基于均生函数的统计模型在厄尔尼诺预测中的应用

为了探究厄尔尼诺的预测方法,提高其预报准确度与预报时效,本文基于均生函数的基本原理,在普通均生函数模型中分别加入气候序列的本征模函数和预报因子变量,构建了基于EEMD(Ensemble EmpiricalModeDecomposition)的均生函数模型和多变量均生函数模型;并应用三种方案对NINO3区海温指数进行了预报试验。结果表明,两种改进模型对厄尔尼诺的预报效果好于普通均生函数模型,是提高预报准确度的有效手段;同时,利用统计的预报模型,可以在一定程度上有效延长厄尔尼诺的预报时效,具有一定实践意义。     

远海航渡式水深测量水位改正方法研究

针对远海航渡式水深测量作业中的潮汐改正难题,基于全球潮汐场DTU10模型及GPS无验潮测深两种改正模式,通过潮汐场预报精度评估、验潮站实测数据比对分析以及GPS大地高计算潮汐值等多种手段,开展了大范围、长时段、单测线情况下水深测量水位改正研究,形成了一套适用性强的航渡水深测量水位改正方法与流程,为面向全球的海洋水深测量资料处理提供了潮汐、垂直基准和水位归算的方法和技术支持。     

海道测量水位改正通用模式研究

为满足海道测量作业发展实际需求,进行了海道测量水位改正通用模式研究.在单波束测深逐点逐时分区改正基础上提出了适于单、多波束测深的海量数据虚拟单验潮站改正模式,基于时差法、最小二乘拟合法数学模型以及海洋潮汐数值预报模型,研制了适于沿岸、近海水深测量的水位改正软件.     

一种利用浮标站资料改进海浪模式有效波高预报的方法

基于浮标站海浪历史数据,利用回归分析方法建立了海浪数值模式有效波高预报产品的一元二次回归方程订正统计模型。通过2017年7月1日-2018年10月10日期间业务试运行结果发现:订正方程能有效改善有效波高数值预报产品的预报精度,且预报时效越短订正效果越显著。其中,第6~11 h预报时效内的订正前后平均绝对误差值减小0.17~0. 241 m,第6~18 h预报时效内订正前后均方根误差减小幅度为0.103~0. 28 m。这说明应用订正统计模型对海浪模式输出产品进行订正,也是改进海浪模式预报准确率的一种有效途径。     

Prediction models for tidal level including strong meteorologic effects using a neural network

Accurate prediction of tidal level including strong meteorologic effects is very important for human activities in oceanic and coastal areas. The contribution of non-astronomical components to tidal level may be as significant as that of astronomical components under the weather, such as typhoon and storm surge. The traditional harmonic analysis method and other models based on the analysis of astronomical components do not work well in these situations. This paper describes the Back-Propagation Neural Network (BPNN) approach, and proposes a method of iterative multi-step prediction and the concept of periodical analysis. The prediction among stations shows that the BPNN model can predict the tidal level with great precision regardless of different tide types in different regions. Based on the non-stationary characteristic of hourly tidal record including strong meteorologic effects, three Back-Propagation Neural Network models were developed in order to improve the accuracy of prediction and supplement of tidal records: (1) Difference Neural Network model (DNN) for the supplementing of tidal record; (2) Minus-Mean-Value Neural Network model (MMVNN) for the corresponding prediction between tidal gauge stations; (3) Weather-Data-based Neural Networks model (WDNN) for set up and set down.The results show that the above models perform well in the prediction of tidal level or supplement of tidal record including strong meteorologic effects.