Does Temporal Stratification Improve the Accuracy of Mark–Recapture Estimates of Smolt Production? A Case Study Based on the Conne River, Newfoundland

Abstract

Based on an actual mark–recapture experiment with Atlantic salmon Salmo salar in the Conne River, a method was developed for simulating smolt migration when initial capture and tagging occur upstream and recapture occurs downstream at times that vary with the individual fish. The properties of Darroch's maximum-likelihood estimator of population size under time interval stratification were explored by Monte Carlo methods. In particular, the effects of population size and the width of the stratum interval on the bias and precision of the estimate of population size were examined under models of constant and variable probability of recapture. The viability of the asymptotic approximation of the variance of the estimate of population size was likewise examined. In addition to temporal strata, the case of no stratification was considered as well as a new estimator based on day-of-departure–day-of-recapture strata, from which daily estimates of the numbers leaving the upper site can be obtained. The study suggests that there is little advantage to be gained from temporal stratification for the situations considered. Standard errors computed from the asymptotic formula appear adequate except for small populations (and, thus, small numbers of recaptures), for very short stratification intervals, or when there is a variable probability of recapture that does not steadily increase or decrease, The study also confirms that previous standard error estimates of 3–4% for this population of approximately 70,000 fish are realistic. Although developed for a specific case, the methodology is modular and can be readily adapted to other scenarios.