Comparing the Relative Performance of Downstream Migrant Abundance Estimation Methods Through Simulation

Mark-recapture based methods are commonly employed to estimate the downstream migrant abundance of salmonid populations using rotary screw trap applications.  There are few comparative examples of the relative performance of these abundance estimation methods, their bias, and whether the confidence intervals generated accurately capture true abundance.  Such analyses are difficult given the lack of feasible methods to independently estimate migrant abundance under field conditions.  Recent advances to mark-recapture methods (e.g., Bayesian Time Stratified Population Analysis) may prove to be useful alternatives to traditional methods, though have not been widely applied in the field.  Our goal was to compare the relative performance of common and emerging abundance estimation methods under two field scenarios to provide trapping practitioners an understanding of: estimator performance; the consequences of violating assumptions; how increased mark-recapture effort affects performance; and the effect that daily variability in trap efficiencies may have on estimates.  We used a simulation based approach to explore abundance estimator performance in two rivers with different physical characteristics, migrant abundances, and trap efficiencies where migrant trapping programs are established.  Test migrant universes with known daily trap efficiency and outmigrants were generated based on existing rotary screw trap data.  Estimators compared include: Stratified Mark-Recapture (with several different methods for specifying strata); Modeled Trap Efficiency (based on flow); Bayesian Time Stratified Population Analysis; and Pooled Peterson.  Our results illustrate the consequences of not identifying or addressing where mark-recapture assumptions may be violated.  For many of the scenarios analyzed, estimators are biased and 95% confidence intervals do not capture the true abundance.  Mark-recapture assumptions can commonly be violated under fluctuating stream conditions.  In the field, the violations of assumptions may be evident or may go undetected with the trapping and trap efficiency estimation methods employed.  Substantial investments of increased effort, such as using additional fish to supplement trap efficiency trials to yield sufficient recaptures or doubling the number of trap efficiency trials, may not yield better abundance estimates where other violations of estimator assumptions exist and are not identified or addressed.