Sensitivity of the inverse problem to size-class selection

Abstract

Fisheries sonar systems typically operate at discrete frequencies between 38 and 420 kHz. Can length-frequency distributions of aggregated fish be accurately estimated using available frequencies and the inverse problem? The inverse problem requires measured fish lengths and realistic scattering models. The size distribution of threadfin shad (Dorosomapetenense) in Lake Nornman was estimated using 120-, 200-, and 420-kHz data, and a recently developed scattering model [Jech etal., J. Acoust. Soc. Am. (in press 1995)]. Size distribution estimates were compared to length frequency measurements from purse seine catches. Fits of probability density functions (PDF’s) using the inversion technique to those using length frequency measures were sensitive to the choice of fish size classes. Preliminary results indicate that estimation of length frequencies using multi- frequency data and the inverse problem appears dependent on the shape of measured length-frequency distributions. [Work supported by NOAA Coastal Ocean Program (NA16RGO 492-01) and NSF (OCE-9115740).]