Data Assimilation Seminar

Prof. John Wells (16:00 - 17:30 March 31, 2025)

Affiliation Ritsumeikan University, Japan / RIKEN, R-CCS
Title Acoustic tomography in a stratified lake, with an initial attempt to estimate currents by a variational method
Abstract

For purposes of scientific investigation and operational monitoring of lakes, we have been testing the application of Coastal Acoustic Tomography ("CAT") to a deep subtropical lake, Lake Biwa in Japan. This talk will report on efforts to measure "differential travel time" (DTT) between transducer pairs, with a final target of assimilating these DTT into a 3D hydrodynamic model of the lake.
The largest lake in Japan, Biwa supplies water to 14 million residents of the Osaka-Kyoto region; in its main "Northern Basin" it has a mean depth of 41 m and a width of about 15 km. Starting in Nov. 2016 we have conducted the first successful tests of CAT at multikilometer ranges in any lake, with multiweek deployment of 3 (5) stations in Nov 2017 (2018). In 2021 we confirmed reciprocal transmission in spring through early summer between two stations at 7 km.
In autumn, we often observed two distinct groups of arrivals. Both ray-tracing and RAM simulations, based on sound-speed distributions from hydrodynamic simulations driven by meteorological observations, have reproduced the two arrival groups. Furthermore the ray-tracing simulations gave clear evidence for the existence of a "surface channel" that yields the earlier arrival times, and a "deep channel" producing the later arrivals.
In continuing work, we are targeting reliable acoustic measurement of path-averaged currents in Lake Biwa. We have confirmed that suitably processed phase differences between reciprocal acoustic transmissions varied smoothly between pings at 10 minute intervals at a range of 6.7 km, and believe that his phase information could facilitate reliable estimates of DTT under the difficult conditions of a stratified lake. This hypothesis leads us to investigate the applicability of the inversion algorithm proposed by Godin, Mikhin, and Mohkov (1996; Acoustical Physics 42 (4)), which to date has only been tested by numerical experiments. Those authors used phase difference as the basis for a cost function to be minimized when inverting for the values of parameters assumed to model the current field. Applying the RAM framework for acoustic simulation allows us to compute the required cross-correlation of theoretical and measured acoustic pressures and associated phase differences that are required to compute Godin et al's (1996) cost function. In this initial attempt, our model for the currents is simply a uniform horizontal current in the vertical plane between the transducers. For reciprocal sound pressure data recorded at a nearshore station and a deep station at R=6.7 km for three hours on Nov. 29 2018, the cost function is fairly consistently minimized by current values that cluster around 8 cm/s, flowing toward the transducer at the deep station. This represents the first report of Godin et al's (1996) technique being applied to field data. Work done in collaboration with Ernst Uzhansky and Naokazu Taniguchi.

Back to Seminar Series

Page Top