Fish movement is a key characteristic of fish populations and is essential to account for from both conservation and management perspectives. Movement behavior can impact how fish are distributed, and whether their populations persist in face of ecosystem changes, and how stocks are assessed. This dissertation seeks to advance our knowledge in understanding of fish movement patterns in time and space, how those are related to other environmental variables, and how to best harness information on such movement as part of fishery assessment and management, in the case of overlapping fish populations where distinct or partially distinct spawning stocks mix during the harvest season. In Chapter 1, I developed a Bayesian variable selection framework for analyzing how factors impact movement intensity in large water areas. Based on the tag-recovery results of lake whitefish populations in Lake Huron from 2003 to 2011, I evaluated how different predictors influenced lake whitefish net movement distance in Lake Huron. These net movement distances were calculated from tagging results. By using a data-driven Bayesian variable selection method, results suggest that lake whitefish with greater total length had longer net distance, and fish started their annual spawning runs earlier in warmer years after acquiring and processing energy needed for spawning. Results also show that when relative Diporeia spp. density was high near the tagging site, lake whitefish tended to stay closer to their tagging site. In Chapter 2 I explored the use of spawning origin information of catch as a means for improving the stock assessments for overlapping fish populations. I also evaluated the influence of including annual recruitment penalties. Results suggested that incorporating information on population-specific harvest age composition improved spawning stock biomass estimation throughout the years being assessed, and improved recruitment estimates only in the early assessment period. Including penalties on annual recruitment residuals improved recruitment estimates in terminal assessment years. In Chapter 3, I extended the spatial Brownie-Petersen tagging model for modeling multiyear tag-recovery data in a fishery context, and incorporated catch-at-age, and tag monitoring data jointly, for lake whitefish populations in Lake Huron. Previous studies of extending Brownie tagging models considering spatial structure were all based on a spatial assumption that fish start moving from where they were in the last time period, and did not recognize spawning site fidelity. We assumed 100% spawning site fidelity for lake whitefish in our model, and results suggested spawning populations in U.S. main basin has higher probability to overlap with other populations during fishing season compared to those in Canadian waters. In chapter 4, I extended the tagging model proposed in Chapter 3 to a more comprehensive framework that allowed for a continuum of spatial structures through modeling homing probability. Based on simulations, we explored how the degree of homing, the extent of spatial movements, and the types of data used, influenced estimability of parameters of interest. My results suggest that the model framework with only tag-recovery and fishing effort data had robust assessment performance in estimating movement rates, homing probability, natural mortality, and fully selectivity fishing mortality rates. With additional tag monitoring data, tag reporting rate can also be accurately estimated simultaneously. Including additional catch-at-age data did substantially improve the estimates of selectivity at age, slightly improved estimates of tag reporting and natural mortality rates, but the bias in estimating recruitment and spawning stock biomass can be high, especially for low productivity populations.
Read
