story by Helen Hill
This month we look at work by Brian Dushaw of the Applied Physics Laboratory at the University of Washington. Brian has been working with Dimitris Menemenlis of JPL on using the MITgcm ECCO2 state-estimates to reconcile a long-standing mystery in ocean acoustics: How did sound travel from Perth to Bermuda during a 1960 ocean acoustic tomography test. Previous attempts based on, e.g., Levitus climatology, could not quite explain the observed propagation. In this new analysis the sharper, better resolved frontal gradients and eddy features in the state-estimate, together with higher fidelity bathymetry are shown to steer the signal along contiguous acoustic channels between Perth and Bermuda.
The colors in the animation show the acoustic mode-1 phase speed calculated from the output of the ECCO2 state estimate. The acoustic frequency employed for these calculations was 15 Hz, which is appropriate for the sound from the large explosive source. The state estimate used comes from the Estimating the Circulation and Climate of the Ocean, Phase II (ECCO2, http://ecco2.org/) project. This state estimate assimilates a wide variety of in situ and satellite data – altimetry, Argo profile data, hydrographic sections, etc., and produces estimates of initial condition and atmospheric surface boundary conditions, which are consistent with model physics and observations. These state estimates include the variables of sea-surface height and of temperature, salinity, and ocean currents at 50 vertical levels down to the full ocean depth. The ECCO2 solution is obtained on a eddy permitting cubed-sphere grid with 18-km horizontal grid spacing and it is subsequently sampled at monthly intervals on a 1/4-degree latitude-longitude grid to facilitate the acoustic analysis. The duration of the solution is 15 years: from 1992-2006. In the figure, the month is indicated by a small white label in the upper right of the bottom panel (snapshot number:
The two most dramatic aspects of ocean variability are the Agulhas Rings that form off Cape Agulhas and move into the South Atlantic and the sharp, filamental Antarctic Circumpolar Front. It is notable that many features of the ocean’s
variability are captured by acoustic mode phase speed. Both the Agulhas Rings and the circumpolar front are strong refractors of acoustic propagation. Acoustic propagation is governed by the gradients of sound speed. The Agulhas Rings are warm and salty and have sound speeds that are some 25 m/s faster than ambient sound speeds. Sound speed changes by 15-20 m/s as the circumpolar front is crossed.
The acoustic propagation interacts with the sea floor and continents at a few key locations. A shoaling sea floor causes the acoustic modes to be compressed, with a sharp increase in mode phase speed. Because the topographic details are fine-scale, modeling this interaction requires greater horizontal resolution than the 1/4 degree afforded by ECCO2. Therefore, the Smith-Sandwell 1-minute resolution topography (version 12.1; 1.8 km resolution) was employed within the 4 regions indicated by the white boxes. Within these boxes, the acoustic mode phase speeds were calculated at 1-minute resolution, using the ECCO2 estimate for the ocean and Smith-Sandwell for the ocean depth. Use of this detailed topographic data allows the shoaling effects of topography on mode phase speed to be accounted for at a more realistic horizontal scale than afforded by the ECCO2 state estimate.
The four areas indicated by the boxes are the Kerguelen Plateau, Crozet Islands, the southern tip of Africa and South America. Between them, the avenue for acoustic propagation from Perth to Bermuda is rather constricted. The mode phase speeds derived for each area reflect the nature of those areas: Kerguelen comprises a dense collection of seamounts, the southern tip of Africa is a broad continental shelf, and South America is a steep, narrow continental shelf. One may view Kerguelen as a near-perfect scatterer, Africa as a near-perfect refractor, and South America as a near-perfect reflector. Although the Kerguelen Plateau is rather large, indicated by the large white box, the region that influences mode-1 phase speed is confined to the areas around Kerguelen and Heard Islands. The region around Crozet Islands is a similarly seamount-dominated (acoustic scattering) region.
Acoustic ray tracing involves launching rays at progressively larger azimuthal angle and following the course of each ray until it gets near Bermuda. This calculation takes into account the horizontal refraction caused by the gradients in mode phase speed. The ocean variability lends a scintillating nature to the acoustic ray paths, with each ECCO2 snapshot giving a different set of rays. This scintillation also means that the rays interact with the topography completely differently in each snapshot. The influence of the ocean variability alone, particularly the sharp gradients apparent with the 1/4 degree resolution of ECCO2, is enough to shift the acoustic arrivals significantly closer to Bermuda than previous results based on smoothed ocean atlases. The ocean alone gives a direct arrival at Bermuda 3-4% of the time. In this movie, scattering from features near Kerguelen and the Crozets causes an essentially diffusive effect, giving successful paths to Bermuda most of the time. In many cases an additional path is obtained by reflection from South America. The African continental shelf causes repulsive bending, rather than attractive bending, of the ray paths, which works against getting rays to Bermuda – rays refracted by the African continental shelf head toward the West Indies.
The ray tracing does not include the effects of ocean currents. While normally small, the currents of the Antarctic circumpolar current may not be negligible.
It should be noted that the precise model employed for the sound speed within the sea floor is important for accurately modeling the acoustic refraction. Such details are poorly known for any given area at the present time, however. In many instances, sound speed actually decreases by a few percent just below the sea floor. If the African continental shelf were to have this sort of common sound speed phenomena in its sediments, it may well become attractive, resulting in a greater tendency for rays to reach Bermuda.
Overall the successful path to Bermuda seems to arise from several factors. The combined effects of the refraction by the circumpolar front, scattering by Kerguelen and the Crozets, and refraction by the Agulhas Rings all work to broaden the spread of rays at Bermuda and shift the ray group a little northward. Interestingly, successful paths are curiously near the great circle path between the end-points.