Accounting for Uncertainty in Acoustic Estimates in the Great Lakes

Fish abundance and behavior can be usefully assessed using hydroacoustic remote sensing. However, the process of converting raw acoustic signals into estimates of fish abundance can be noisy and complicated. Are the methods we use to calculate mean abundances the best that are statistically achievable? If we are to put effort into increasing sample size or refining estimates, where should we focus our attention? Once we understand the uncertainty associated with a given step in the conversion process, how do we incorporate that uncertainty in characterizing the final estimates? In this paper, we develop a mechanism that is part Monte Carlo and part Bayesian to address known sources of uncertainty in the acoustic assessment process. Estimation steps are followed sequentially from Sv to abundance and the variances are carried along through each step.  The structure of the analysis facilitates examination of the effects of the various parts of the analysis on the quality of the final outputs. Acoustic scientists can use this approach to help prioritize their own research and assessment designs.