Abstract:
This dissertation presents a mathematical model of the overall fisheries acoustics problem posed by enumeration of fish populations using sonar. Emphasis is placed on three key components: a new geometric model for the target strength (TS) of Pacific salmon, a fish distribution for sockeye salmon, and generation of artificial sonar data. Results of the TS and fish distribution models show TS varies on height and breadth of fish as much as on fish length and TS from the air-filled swimbladder is the major contributor as reported by Foote [1985]. A fish roll factor within 45° leads to TS variations within 7 dB for normal incidence, side aspect and 2 dB for dorsal aspect. Also second order effects of ray propagation through fish flesh on TS from the swimbladder provide TS results up to 20 dB lower at high aspect angles. The geometric model predicts TS values that match extremely well with TS data collected on Pacific salmon and other species in river and ocean environments. By varying fish size and swimbladder parameters and considering the effect of fish flesh, the model covers the range of TS values that occur in the field, thus identifying and quantifying the uncertainty in the experimental data. The overall approach in this work is to construct a direct model providing artificial sonar data, then use an inverse model (echo integration algorithm) with that data or with experimental data to compare results. The echo integration results are not reliable at any fish rate for a fixed river cross-section. Estimated fish counts of 0--7 are obtained from 100 simulations for a known fish distribution of 3 fish (0.1 fish/sec). Similarly, at 0.5 fish/sec with 15 known fish, estimates of 0--30 were obtained; at 1 fish/sec with 30 known fish, estimates of 0--50; and at 5 fish/sec with 150 known fish, estimates of 0--100 fish. Fish counts ranged from 0--19 for 3 known fish when ping rate changed from 1--10 pings/sec and when pulse width varied from 0.1--1.0 ms.
