Life, Death, and Resurrection: Accounting for State Uncertainty in Survival Estimation from Tagged Grass Carp

Abstract

Information about Grass Carp Ctenopharyngodon idella survival would be useful for improving the management of fish used for aquatic weed control. Reliable methods for estimating annual poststocking survival of Grass Carp from radiotelemetry data do not exist because the fish remain sedentary for prolonged periods between movements, giving the false impression of death, only to be observed alive (i.e., “resurrected”) at a later date. We constructed a state-space, multistate mark–recapture survival model accounting for uncertainty in the live/dead states of tagged Grass Carp in a large (8,500 ha) reservoir, and we estimated monthly and annual survival. Model results were compared with life history-based methods for estimating survival, and survival estimates that were corrected for state misclassification were compared with uncorrected estimates. Corrected estimates of annual survival (mean = 0.23; 95% credible interval [CRI] = 0.15–0.41) contained less bias than uncorrected estimates (0.12; 95% CRI = 0.08–0.18). However, both corrected and uncorrected estimates were substantially lower than the survival expected based on life history theory (mean = 0.69; 95% confidence interval = 0.52–0.78), suggesting that mark–recapture survival estimates for Grass Carp might be negatively biased due to tag shedding, tag-related mortality, or both. Our model effectively reduced bias in monthly and annual survival estimates due to state misclassification, illustrating the potential for application of existing mark–recapture frameworks to estimate Grass Carp survival with telemetry data, despite the behavioral idiosyncrasies of the species. Furthermore, these methods may have application for studies of other animals that undergo periodic quiescence between movements, such as salmonids, ictalurids, and reef fishes. To account for bias resulting from tag loss, future mark–recapture studies of Grass Carp could incorporate tag shedding rates within the framework developed here.

Received July 19, 2014; accepted December 3, 2014

ACKNOWLEDGMENTS

We thank Vic Dicenzo and Emmanuel Frimpong (Department of Fish and Wildlife Conservation, Virginia Polytechnic Institute and State University) and William Halteman (Department of Mathematics and Statistics, University of Maine) for thoughtful reviews, guidance, and constructive discussions about early versions of this paper. We thank Bill Kendall and an anonymous reviewer for constructive reviews that greatly improved the quality of the manuscript. We acknowledge considerable reliance upon R and JAGS code written by Marc Kéry and Michael Schaub (available: www.vogelwarte.com/bpa.html). As such, the model code used in the present study will be made available upon request. This research was primarily funded by the Lake Gaston Weed Control Council and the Acorn Alcinda Foundation. Additional funding for this work was provided by the Virginia Agricultural Experiment Station and by Hatch funds from the U.S. Department of Agriculture's National Institute of Food and Agriculture.