FEBRUARY 2019 D U S H A W 183 Ocean Acoustic Tomography in the North Atlantic BRIAN D. DUSHAW Nansen Environmental and Remote Sensing Center, Bergen, Norway (Manuscript received 22 May 2018, in final form 9 October 2018) ABSTRACT An objective mapping exercise simulating observations of temperature in the North Atlantic Ocean was used to assess the resolution capabilities of ocean acoustic tomography in combination with Argo floats. A set of basis functions for a basinwide area was obtained from a singular value decomposition of a covariance derived from an ocean state estimate. As demonstrated by the formal uncertainty estimates from the objective maps, Argo and tomography are complementary measurements. In several examples, each separately ob- tained uncertainty for determining large-scale monthly average temperature of about 50% of prior (resolved 75% of variance), while when both data were employed, uncertainties were reduced to about 25% of prior (resolved 94% of variance). Possible tomography configurations range from arrays that span specific regions to line arrays that supplement existing observations to arrays that span the Atlantic basin. A basinwide array consisting of two acoustic sources and seven receivers can be used to significantly reduce the uncertainties of estimated broad-scale temperature. An optimal observing system study would comprise simulated mea- surements in combination with data assimilation techniques and numerical ocean modeling. This objective map study, however, showed that the addition of tomography to the existing observing system could sub- stantially reduce the uncertainties for estimated large-scale temperature. To the extent that tomography offers a 50% reduction in uncertainty at a fraction of the cost of the Argo program, it is a cost-effective contribution to the ocean observing system. 1. Introduction detection of coherent, mode-1, internal-tide radiation far into the interior of the North Pacific (Dushaw et al. 1995; Ocean acoustic tomography was introduced as an ap- Ray and Mitchum 1996) and 5) the determination that, proach to ocean observation in the late 1970s (Munk et al. much like the barotropic tides, such variability is pre- 1995; Munk and Wunsch 1982a; Munk 1986; Spiesberger dictable over most of the world’s oceans (Dushaw et al. and Metzger 1992). Several scientific programs in the 2011; Dushaw 2015). 6) An O(1000)-km-scale tomo- 1980s and 1990s developed and established this mea- graphic array deployed during 2000 in the equatorial surement as both unique and valuable. Examples of Pacific measured the mean relative vorticity, showing tomographic contributions to physical oceanography that it was positive during La Niña and negative during are many, including the following: 1) The Greenland the normal state (Nakano et al. 2001). 7) Basin-scale Sea Project made remarkable three-dimensional (3D) measurements of temperature variations of the layer of measurements of the evolution of deep-water forming Atlantic water in the Arctic were made by transbasin events during the winter of 1989–90 (Worcester et al. acoustic experiments in the 1990s, one of the first in- 1993; Morawitz et al. 1996a,b). 2) Highlighting the in- dications of Arctic warming (Mikhalevsky et al. 1999, herent averaging properties of tomography, precise 2015). 8) Measurements of temperature were made measurements of barotropic, open-ocean tidal cur- across the North Pacific over the 1996–2006 decade by rents, from the dominant M and S constituents to the 2 2 the Acoustic Thermometry of Ocean Climate (ATOC) small P and Q constituents, have been made in sev- 1 1 program (ATOC Consortium 1998; Worcester et al. eral places (Dushaw et al. 1997; Stammer et al. 2014). 2 1999; Dushaw et al. 2009), indicating that sometimes 3) The tiny tidal relative vorticity [O(1029)s 1]atM 2 rapid changes, comparable in size to the seasonal cycle, frequency caused by tidal variations in sea surface height in basin-scale temperature occur. 9) Data assimila- was also measured. 4) Tomography was central to the tion techniques used in combining tomography and other data with dynamical constraints in western bound- Corresponding author: Brian D. Dushaw, [email protected] ary current regions substantially reduces the uncertainties DOI: 10.1175/JTECH-D-18-0082.1 Ó 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses). Unauthenticated | Downloaded 09/25/21 09:31 PM UTC 184 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 36 of the state estimate (Gaillard 1992; Lebedev et al. 2003). purposes of this paper, at least, an OSSEs is defined as a In all of these measurements, tomography detected and study employing a numerical ocean model and data quantified oceanographic phenomena in ways not assimilation. While such numerical studies are the gold achievable by other means. Munk et al. (1995) describe standard of system assessment and design, in practice the technique and applications of acoustic tomography in they are technically challenging and time consuming, general. Updated reviews have been given by Worcester such that few OSSEs, examining tomography or any (2001), Dushaw et al. (2001),andDushaw (2014). other data type, in the context of general circulation In the course of these experimental programs, the re- models have been completed (Halliwell et al. 2017; markable stability of the acoustic environment over most Gasparin et al. 2018). of the world’s oceans was demonstrated (Munk et al. Considerable success has been achieved in obtaining 1995; Cornuelle et al. 1993; Worcester et al. 1999)such ocean state estimates by data assimilation of existing that in most open-ocean regions, quite accurate pre- data, however. Many such global and regional estimates dictions of the acoustic propagation and spatial sampling are now available for research or operational purposes. characteristics can be made (Jensen et al. 2011; Dushaw State estimates represent accurate ocean environments et al. 2013). Such predictions can use a climatological that can be used for observing system design, for exam- ocean such as the World Ocean Atlas (Antonov et al. ple, Mazloff et al. (2018). While not as definitive as a 2010; Locarnini et al. 2010) of the National Oceanic rigorous OSSE in quantifying the information contribu- and Atmospheric Administration (NOAA). The ocean- tion of data types and their configurations, analyses ographic signals are derived from the small deviations in based on realistic ocean environments offer a simple, but acoustic travel times from the basic climatological con- objective, way to determine information requirements ditions. Variations of travel times over a 5000-km range and to assess the relative resolution capabilities of dif- across an ocean basin can be measured to an accuracy of ferent data types. For the exercise here, simulated mea- about 0.010 s, out of a total travel time of over 3300.000 s. surements of temperature by ocean acoustic tomography This accuracy in travel time is roughly equivalent to a in conjunction with simulated float measurements in the range- and depth-averaged sound-speed (temperature) North Atlantic were assessed. A recent state estimate 2 change of 0.005 m s 1 (1 m 8C). from the Estimating the Circulation and Climate of Ocean observing systems (OOSs) began to be de- the Ocean (ECCO) consortium was employed. The ap- veloped in the 1990s, almost concurrent with the devel- proach used here is objective analysis, or objective opment of the acoustical measurements. Employment of mapping, which has been commonly used for data acoustical methods within OOSs has been elusive, how- mapping or observing array analysis (Bretherton et al. ever. International workshops focused on OOSs, such as 1976; Roemmich and Gilson 2009). Previous analyses the 1999 and 2009 OceanObs conferences, highlighted have often represented ocean variability using covariances acoustic tomography as a good potential contribution, derived directly from data. In the present case, the ocean suggesting the deployment of pilot systems in the North variability was represented by a covariance derived from Atlantic (Scientific Committee on Ocean Research 1994; the ECCO state estimate. A large set of mapping basis Dushaw et al. 2001, 2010; BOM 2001; Fischer et al. 2010). functions was obtained from a singular value decomposi- Despite this community consensus, there has been little tion (SVD) of the covariance. progress in developing such prototype sustained systems. The ECCO global ocean state estimate is described in One difficulty has been the challenge of completing section 2. Some examples of notional tomographic obser- modeling studies that would quantify the precise contri- vations for the North Atlantic are described in section 3, butions of tomography to the ocean observing system and and the basic, predicted acoustic characteristics asso- determine optimal configurations of the acoustical sys- ciated with those observations are described in section tems. In the context of the Global Ocean Data Assimilation 4. This study makes the simplifying, but justified, as- Experiment (GODAE; https://www.godae-oceanview.org/) sumption that a tomographic measurement represents and other such programs, observing system simulation a simple temperature measurement, averaged over experiments (OSSEs) that would bring numerical ocean depth and over range. The derivation of the basis modeling and data assimilation techniques to bear