School of Earth and Ocean Sciences, University of Victoria, Victoria, British Columbia V8W 3P6, Canada.
J Acoust Soc Am. 2010 Jul;128(1):66-74. doi: 10.1121/1.3436530.
This paper applies Bayesian source tracking in an uncertain environment to Mediterranean Sea data, and investigates the resulting tracks and track uncertainties as a function of data information content (number of data time-segments, number of frequencies, and signal-to-noise ratio) and of prior information (environmental uncertainties and source-velocity constraints). To track low-level sources, acoustic data recorded for multiple time segments (corresponding to multiple source positions along the track) are inverted simultaneously. Environmental uncertainty is addressed by including unknown water-column and seabed properties as nuisance parameters in an augmented inversion. Two approaches are considered: Focalization-tracking maximizes the posterior probability density (PPD) over the unknown source and environmental parameters. Marginalization-tracking integrates the PPD over environmental parameters to obtain a sequence of joint marginal probability distributions over source coordinates, from which the most-probable track and track uncertainties can be extracted. Both approaches apply track constraints on the maximum allowable vertical and radial source velocity. The two approaches are applied for towed-source acoustic data recorded at a vertical line array at a shallow-water test site in the Mediterranean Sea where previous geoacoustic studies have been carried out.
本文将贝叶斯源跟踪应用于不确定环境中的地中海数据，研究了跟踪结果及其不确定性随数据信息量（数据时间段数量、频率数量和信噪比）和先验信息（环境不确定性和源速度约束）的变化。为了跟踪低水平源，同时对多个时间段（对应于跟踪轨迹上的多个源位置）记录的声数据进行反演。通过将未知的水柱和海底特性作为附加反演中的干扰参数，解决了环境不确定性问题。考虑了两种方法：聚焦跟踪通过最大化后验概率密度（PPD）来确定未知的源和环境参数。边缘化跟踪通过对环境参数进行积分，得到一系列关于源坐标的联合边际概率分布，从中可以提取最可能的轨迹和轨迹不确定性。这两种方法都对最大允许的垂直和径向源速度施加了轨迹约束。这两种方法应用于在地中海浅水区试验场的垂直线列阵上记录的拖曳源声数据，先前已经进行了地球声学研究。
