Simulation Analyses of Integrated Tagging and Catch-at-Age Analysis Models and Application to Lake Erie Walley

Integrated tagging and catch-at-age analysis (ITCAAN) models incorporate tag-recovery data within statistical catch-at-age models for spatially-explicit assessment of fish stocks. I investigated two tag-recovery frameworks for ITCAAN models that assumed natal homing spawning behavior: release-conditioned and recovery-conditioned. In Chapter 1, I investigated the performance of a release-conditioned ITCAAN model under varying levels of parameter complexity, movement rates, data quality, and misspecification of natural mortality or reporting rates. The release-conditioned ITCAAN model simultaneously estimated movement rates, natural mortality, and tag reporting rates, though accuracy and precision of model estimates decreased with greater model complexity and fewer tags released. In Chapter 2, I investigated a recovery-conditioned ITCAAN model under a range of model complexities, different movement rates, misspecification of natural mortality or reporting rates, and spatially varying reporting rates. The recovery-conditioned ITCAAN model accurately estimated relative reporting rates at low intermixing rates. Biased estimates of individual population abundance resulted when intermixing was high and the reporting rate was assumed to be spatially constant, but the true reporting rates were not. I recommend recovery-conditioned ITCAAN model only be used for assessment of intermixed fish stocks when there is high certainty of spatially constant reporting rates. Both release- and recovery-conditioned ITCAAN models had difficulty estimating individual population abundances and reporting rates under high rates of movement and large differences in population sizes. For both ITCAAN frameworks, estimation of both natural mortality and reporting rates reduced the accuracy and precision of model estimates, but estimates were less biased than misspecifying one of these parameters by 50%. In all investigated simulations, parameter estimates from the release-conditioned ITCAAN model were more accurate and precise compared to the recovery-conditioned ITCAAN model. In Chapter 3, I examined the ability of a release-conditioned ITCAAN model to estimate time-varying natural mortality and reporting rate parameters and sensitivity to tag-shedding, high-reward tag-recovery data, and seasonal movement dynamics. Natural mortality and reporting rate estimates were most precise and accurate when estimated in 5-year time blocks. Estimation of natural mortality and reporting rates as temporally constant when reporting rates decreased linearly over time caused severe bias in abundance estimates, especially as the frequency of high-reward tag release events decreased. I recommend that high-reward tags be released annually to increase the precision and accuracy of ITCAAN model estimates. The release-conditioned ITCAAN model was sensitive to whether the operating model simulated tag-shedding and seasonal movement dynamics. In Chapter 4, a release-conditioned ITCAAN model was applied to Lakes Erie and Huron walleye (Sander vitreus) data and estimates were compared to current assessment models. Reasonable fits to all data sources were obtained for an ITCAAN model that estimated a single reporting rate for all recreational fisheries and assumed the effective sample size of the tag-recovery data was equal to the number of tags released divided by 10. Estimates of population abundance from the ITCAAN model were similar to the model currently used to manage Lake Huron walleye. Conversely, estimates of abundance and natural mortality for the western/central basin of Lake Erie were lower compared to the assessment model currently used to manage the stock. Estimates of natural mortality and abundance for the eastern basin of Lake Erie from the ITCAAN model were larger than estimates for the current assessment model. Based on results from Chapters 1 and 3 the estimates of abundance for the eastern basin of Lake Erie may be overestimated. Given the level of movement within and between Lakes Erie and Huron, ITCAAN models may be a beneficial assessment methodology for management. I recommend future tagging studies of walleye on Lakes Erie and Huron be designed such that high-reward tags and standard-reward tags be released annually so that ITCAAN models can provide accurate abundances, natural mortality, and reporting rates estimates. Additionally, I recommend creel surveys be conducted on the Huron-Erie corridor by the Michigan Department of Natural Resources and Ontario Ministry of Natural Resources and Forestry to measure the currently unaccounted for walleye harvest, which is likely a significant source of mortality.