Page 34 - Fall2019

Computational Acoustics in Oceanography
With multiple synchronous receivers, one can use arrival time-difference analysis to geometrically locate a source. Tiemann et al. (2004) used propagation modeling to locate whales with great success. Unfortunately, spread out syn- chronous systems are usually too costly for bioacoustic studies. Alternatives include the use of directional sensors or propagation models and inverse methods (see Simula- tions for Inversion). Here, the aim is only to infer source location; the unknown environmental parameters are seen as a nuisance, and the fact that the source signal is usu- ally unknown is an extra difficulty. Nonetheless, when a synchronized vertical array is available, marine mam- mals can be localized with classical underwater acoustic inverse methods (e.g., Thode et al., 2000). Stepping down to a single sensor, a common arrangement for bioacoustic studies, causes further issues to arise. However, results are obtainable in this case when advanced signal-processing methods and propagation models are engaged. Toothed whales (high-frequency sources) can be localized by ana- lyzing single-hydrophone ray arrivals (Tiemann et al., 2006). On the other hand, baleen whales (low-frequency sources) can be localized by analyzing the mode arrivals (e.g., Bonnel et al., 2014).
Marine mammals are not the only underwater dwellers pro- ducing and using sounds. Adequate propagation models are required to study all species. Considering humans, global noise models are needed to study the impact of anthropo- genic activities on the entire marine ecosystem. Such noise models are discussed in Noise Modeling. On the other hand, many marine animals are studied in tanks and labs. Here, propagation models are required to (1) correctly understand the sound recorded in a reverberant tank and (2) predict sound properties in the sea from measurements performed in the lab. Of particular importance may be the use of prop- agation models for vector acoustics (water motion speed and direction; Heaney and Campbell, 2019) because fishes and crustaceans are highly sensitive to particle motion (Popper and Hawkins, 2018).
Simulations to Study Propagation Physics
As we have seen, ocean features control sound propaga- tion in complex ways. Sound fields are affected by multiple interactions whose effects cascade nonlinearly along the propagation path. As detailed in Acoustic Tomography and Thermometry and Simulations for Inversion, the resultant fields can be used to infer environment variables, assisted
by forward simulation. But the forward studies alone can be illuminating, particularly for nonlinearly chained events. The situation depicted in Figure 2 of a computation reveal-
ing a physics effect is not unique.
Some aspects to consider for studying chained propagation events is that the sound interactions with the environment are frequency dependent; higher frequencies are sensitive to smaller scale processes, whereas lower frequencies used over longer ranges will integrate the effects of more physi- cal constraints. Ocean internal and surface features are dynamic and time variable, whereas the bottom shape and sediments below can be taken as static. Note, however, that the locations and angles of bottom interactions can change over time due to changing water column conditions, illus- trating the chained nature.
Using these principles, sound simulation can be used as a tool to study and characterize complex sound/ocean- feature interaction (e.g., with internal waves, fronts, eddies and filaments, bottom vegetation, and coral reef rough- ness). After achieving that, a heady long-term goal would be the acoustic measurement of ocean phenomena with a synopticity that cannot be achieved by direct in situ mea- surements. The challenge is that sound signals arriving after traversing a volume are not each directly connected with a single ocean-state parameter but with many. This yields complex parameter-to-observation operators that require high-fidelity simulations to correlate the observed signal with the physical environment.
To illustrate, consider the acoustic effects of eddies, which have a huge range of parameters, and may control sound differently for varying source frequency and depth. Using an initial guess (ocean background), one can map the local features and identify the areas where alterations in sound speed (usually via altered temperature) are more likely to cause important sound field changes. The map in Figure 5, top right, shows an example with a lateral sound refraction metric based on the sound speed gradients that are based on US Navy Coastal Ocean Model (NCOM) ocean simulations. The maps in Figure 5, bottom, show simulations, from a 3-D ray/beam model (Porter, 2019), of the sound level estimate taken at two times for sound propagating outward from a point source. Sound rays beginning radially outward that strike eddy edges bend horizontally. At present, the quan- tification of the effects on sound fields of modeled eddies is
34 | Acoustics Today | Fall 2019