[论文解读] A survey on tidal analysis and forecasting methods for Tsunami detection

Sergio Consoli, Diego Reforgiato Recupero+1 more|arXiv (Cornell University)|Mar 1, 2014

Oceanographic and Atmospheric ProcessesEarth and Planetary Sciences参考文献 56被引用 19

一句话总结

本综述回顾了用于海啸探测的经典与先进潮汐分析及预报方法，重点强调了调和分析、滤波、神经网络和小波变换。它评估了这些方法在处理天文与非天文影响方面的有效性，对不同技术在数据可用性各异条件下的准确、及时海啸预警能力进行了全面比较。

ABSTRACT

Accurate analysis and forecasting of tidal level are very important tasks for human activities in oceanic and coastal areas. They can be crucial in catastrophic situations like occurrences of Tsunamis in order to provide a rapid alerting to the human population involved and to save lives. Conventional tidal forecasting methods are based on harmonic analysis using the least squares method to determine harmonic parameters. However, a large number of parameters and long-term measured data are required for precise tidal level predictions with harmonic analysis. Furthermore, traditional harmonic methods rely on models based on the analysis of astronomical components and they can be inadequate when the contribution of non-astronomical components, such as the weather, is significant. Other alternative approaches have been developed in the literature in order to deal with these situations and provide predictions with the desired accuracy, with respect also to the length of the available tidal record. These methods include standard high or band pass filtering techniques, although the relatively deterministic character and large amplitude of tidal signals make special techniques, like artificial neural networks and wavelets transform analysis methods, more effective. This paper is intended to provide the communities of both researchers and practitioners with a broadly applicable, up to date coverage of tidal analysis and forecasting methodologies that have proven to be successful in a variety of circumstances, and that hold particular promise for success in the future. Classical and novel methods are reviewed in a systematic and consistent way, outlining their main concepts and components, similarities and differences, advantages and disadvantages.

研究动机与目标

• 为适用于海啸探测的潮汐预报方法提供系统性综述。
• 评估传统调和分析在数据稀缺或非天文条件下的优缺点。
• 评估神经网络与小波变换等替代方法在短时或噪声潮汐记录中提升预报精度的潜力。
• 从计算复杂度、数据需求与预测性能角度，比较经典与新型技术。
• 为研究人员和实践者在实时海啸预警系统中选择最优方法提供指导。

提出的方法

• 利用最小二乘法进行调和分析，从长期海平面数据中估计潮汐 constituent 的振幅与相位。
• 应用高通与带通滤波，从噪声和非潮汐信号中分离出潮汐成分。
• 采用人工神经网络建模复杂非线性潮汐模式，无需明确的物理假设。
• 利用小波变换分析将潮汐信号分解为时频成分，以实现局部特征提取。
• 将气象与海洋强迫效应整合进潮汐模型，以提升极端事件期间的预报精度。
• 基于数据长度、预报精度与计算效率，系统比较各项方法。

实验结果

研究问题

• RQ1 当天气等非天文因素显著影响潮汐时，传统调和方法表现如何？
• RQ2 在数据有限条件下，神经网络与小波变换在潮汐预报中的优势与局限性是什么？
• RQ3 滤波技术与物理建模在从噪声信号中分离潮汐成分方面如何比较？
• RQ4 哪些方法在短时或不完整潮汐记录中仍能保持高精度？
• RQ5 应依据何种标准选择适用于实时海啸预警系统的最优潮汐预报方法？

主要发现

• 调和分析仍具有效性，但需要长期数据，且在应对如风暴潮等非天文影响时表现受限。
• 神经网络与小波变换在短数据场景及复杂非线性潮汐动力学中优于传统方法。
• 滤波技术在降噪方面有效，但缺乏对长期或极端事件预报的预测能力。
• 基于小波的方法提供更优的时频定位，可实现对与海啸相关的瞬态潮汐异常的检测。
• 将物理模型与数据驱动技术（如神经网络）结合的混合方法，在构建稳健、自适应海啸预警系统方面展现出前景。
• 方法的选择在很大程度上取决于数据可得性、精度要求与实时处理约束。