It's about time: A multistate semicontinuous time mark–recapture model to evaluate seasonal survival and movement rates of juvenile Coho Salmon in a small coastal watershed

Abstract Objective Many mark–recapture models assume that releases and recaptures are discrete events, and researchers often aggregate continuous recapture data (e.g., passive integrated transponder [PIT] detections) into coarse temporal scales to satisfy this assumption. This temporal discretization could result in parameter biases by ignoring the individual heterogeneity in the time susceptible to mortality after recapture and the conditions experienced (e.g., temperature and predation risk) before and after recapture. Our objectives were to (1) estimate the amount of bias in survival and emigration rates due to different temporal discretization durations when recapture events occur continuously and (2) apply this semicontinuous model to estimate rates of early emigration and overwinter survival for Coho Salmon Oncorhynchus kisutch in a coastal California watershed. Methods We developed a semicontinuous time multistate mark–recapture model to separately estimated emigration and survival rates throughout the year. We used weekly time-varying occasions paired with discrete spatial states and conducted extensive simulation trials to explore potential model bias. We then applied the model to an existing 4-year dataset of Coho Salmon PIT tag detections. Result Our simulations indicated that that the amount of bias in survival and movement rates decreased as the temporal discretization duration decreased. The confidence interval of the bias estimates included zero with a duration of 8 days, indicating that this duration was sufficiently short to model movement and survival. Results from our Coho Salmon analysis suggest that overwinter survival rate ranged from 0.72 to 0.83, which is higher than previous estimates for Coho Salmon in this region. We estimate that a substantial proportion of smaller juveniles (0.21–0.28 annually) move to downstream nonnatal rearing habitats before the spring smolt migration. Conclusion Our semicontinuous modeling approach can be implemented relatively easily and used to analyze continuous detection data to accurately estimate survival and movement rates. Our analysis of Coho Salmon PIT tag detections implies that previous low estimates of apparent overwinter survival of Coho Salmon were partially due to high movement rates to alternative rearing locations. This contrasts with conclusions from the previous research that suggested that overwinter survival was a major limiting factor for population recovery and implies that species recovery may be improved by considering multiple emigration patterns in the design of future research, monitoring, and restoration projects.