Learning Something from Nothing - Quantitative Modeling of Salmon Runs and Spine Growth Rings

Dr. Lynn Waterhouse

Assistant Unit Leader, USGS Minnesota Cooperative Fish and Wildlife Research Unit

Abstract

Missing data is a problem for many observational studies. Bayesian methods are one possible way of dealing with these data gaps. Here I highlight two examples of applied Bayesian models. The first is a Bayesian nested patch occupancy model applied to PIT (passive integrated transponder) tag data from steelhead returns in the Upper Columbia River basin. This model structure provides probabilistic estimates of up-stream migration routes for each tagged individual based on a set of underlying nested state variables. These movement estimates can be converted into abundance estimates when an estimate of abundance is available for a location within the river network. The second example is a Bayesian like model for estimating ages from spine samples when vascularization has led to obliteration of the center core of the spine - with an application to Skipjack data. Both cases highlight ways Bayesian methods can be used to impute missing data, providing valuable information for conservation studies.