Ocean Acoustic Inversion

海洋对电磁辐射不透明，但对声音透明；因此，科学家们设计了许多方法来利用声学来代替光、无线电或微波。海底特性/过程或声源（如潜艇、海洋哺乳动物）位置的遥感是地质、岩土、环境和国防应用方面的重要问题。然而，声学遥感需要解决一个非线性逆问题，这是一个固有的非唯一的不确定性和分辨率往往不是很好地理解。本研究是建立在我们在海洋声学遥感定量贝叶斯（概率）反演及相关地震问题的理论和应用基础上的。(1)海底声散射参数与过程反演。我们的初步工作表明，声波后向散射数据的反演可以解析海底粗糙度和沉积物非均质谱（一种替代费力的直接测量的遥感方法）。我们建议发展定量模型选择方法，以确定各种物理散射机制的相对重要性，并估计不同海底类型的相关散射参数和不确定性，从而更好地了解海底散射的物理特性，并完善地声数据库和声呐模型。(2)为海军数据库或电缆铺设等岩土工程应用估算横向变化的海底散射和地声参数。该方法基于连续贝叶斯反演新声学数据集，该数据集使用由自主水下航行器拖曳海床附近的声源和接收器阵列组成的移动系统沿轨道收集。(3)利用固定接收阵列记录的全场声波数据进行海底地声特性层析（2D）反演。二维定量非线性反演的一个主要组成部分涉及开发新的自适应稀疏/不规则数值网格，以与数据的分辨率一致的方式参数化环境。(4)声源定位与不确定性估计，包括不确定性对海洋环境特性的影响。我们最初的工作是在海军（被动声纳）应用，但我们正在与加拿大领先的海洋技术公司JASCO研究有限公司合作进行海洋哺乳动物研究。新的方法包括使用非同步阵列系统/子系统进行匹配场定位的新方法，以及在单个接收器上通过模态色散（时频分析）进行源测距。(5)利用长期声学监测数据跟踪未知数量的发声海洋哺乳动物。这需要开发一种贝叶斯跟踪算法，在游泳速度的先验信息约束下，对哺乳动物的数量及其轨迹进行采样，并提供一个概率解决方案，从中可以提取种群和行为的信息。(6)与加拿大地质调查局合作，对加拿大地震/海啸高发地区Cascadia的地壳/构造结构进行被动地震反演。这涉及开发远震数据（接收函数）和环境地震噪声（表面波色散）的联合层析反演，以解决与历史地震震源相关的二维结构。考虑到2012年的7.8级地震和当地的海啸，夏洛特女王边缘从俯冲到转换断层的转变特别令人感兴趣，来自海王星系统的海底地震仪为横跨胡安德富卡板块和卡斯卡迪亚俯冲带的一段提供了新的中板块数据。

Oceans are opaque to electromagnetic radiation but transparent to sound; hence, scientists have devised many methods to use acoustics underwater instead of light, radio or microwaves. The remote acoustic sensing of seabed properties/processes or the locations of sound sources (e.g., submarines, marine mammals) are important problems with geological, geotechnical, environmental and defence applications. However, acoustic remote sensing requires solving a nonlinear inverse problem which is inherently non-unique with uncertainties and resolution that are often not well understood. The proposed research builds on our work in theory and applications of quantitative Bayesian (probabilistic) inversion for ocean-acoustic remote sensing and related seismic problems with several projects. (1) Inversion for seabed acoustic scattering parameters and processes. Our initial work shows that inversion of acoustic backscatter data can resolve in-situ seafloor roughness and sediment heterogeneity spectra (a remote-sensing alternative to laborious direct measurements). We propose to develop quantitative model-selection approaches to determine the relative importance of various physical scattering mechanisms and to estimate relevant scattering parameters and uncertainties for different seabed types, to better understand the physics of seabed scattering and advance geoacoustic databases and sonar models. (2) Estimation of laterally-variant seabed scattering and geoacoustic parameters for naval databases or geotechnical applications such as cable-laying. This approach is based on sequential Bayesian inversion of novel acoustic data sets collected along a track using a mobile system comprising an acoustic source and receiver array towed near the seabed by an autonomous underwater vehicle. (3) Tomographic (2D) inversion for seabed geoacoustic properties using full-field acoustic data recorded at a fixed receiver array. A major component of quantitative nonlinear inversion in 2D involves developing new self-adaptive sparse/irregular numerical grids to parameterize the environment in a manner consistent with the resolving power of the data. (4) Acoustic source localization and uncertainty estimation, including the effects of uncertainty in ocean environmental properties. Our initial work was in naval (passive sonar) applications, but we are moving into marine-mammal studies in collaboration with JASCO Research Ltd, a leading Canadian marine-technology company. New methods include a novel approach to matched-field localization with non-synchronized array systems/sub-systems, and source ranging via modal dispersion (time-frequency analysis) at a single receiver. (5) Tracking an unknown number of vocalizing marine mammals using long-term acoustic monitoring data. This requires developing a Bayesian tracking algorithm to sample over the number of mammals and their tracks, constrained by prior information on swim speed, and provide a probabilistic solution from which information on population and behaviour can be extracted. (6) Passive seismic inversion for crustal/tectonic structure of Cascadia (Canada’s highest earthquake/tsunami hazard), in collaboration with the Geological Survey of Canada. This involves developing a joint tomographic inversion of teleseismic data (receiver functions) and ambient seismic noise (surface-wave dispersion) to resolve 2D structure which can be correlated with historical earthquake foci. The transition from subduction to transform faulting at the Queen Charlotte margin is of particular interest given the 2012 magnitude 7.8 earthquake and local tsunami, and ocean-bottom seismometers from the NEPTUNE system provide new mid-plate data for a section across the Juan de Fuca plate and Cascadia subduction zone.