9724 - Laserfiche WebLink

Home Browse Search

LITTLE COLORADO RIVER HUMPBACK CHUB 241 The modeling approach for the age-structured Jolly- Seber and ASMR methods assumes that the initial age is correctly assigned based on size at fIrst capture. Errors in initial age assignment could lead to a negative bias in population point estimates and to the "smear- ing" of estimated recruitment among cohorts in the ASMR model (Coggins et al. 2006). However, the population trend estimates from the TSM models are not age dependent, so that incorrect age assignment does not bias the population trend estimates from these methods. The most critical assumption in the open-population models is that mortality rates of fish have been stable over time. Without this assumption, cumulative survivors of fIsh tagged in previous years cannot be safely used in calculating capture probabilities (and population estimates) for anyone year. If, for example, there had been a high mortality rate of all fIsh in any one year, we would have seen a subsequent drop in the recapture rates of tagged fIsh from all previous years. But because sampling has changed over time (so that capture probabilities cannot be assumed to have been stable), we would attribute that drop in recapture rates to changes in capture probability rather than to the change in survival and we would therefore over- estimate the number of surviving fIsh. We do not see trends in calculated monthly capture probabilities that would indicate such an error. However, the limited sampling effort in the late 1990s and the associated low catch rates make it difficult to evaluate the mortality and contribution of these year-classes to the overall population (new recruits to the population in the late 1990s were unlikely to be collected given the low sampling effort). Is it possible that the adult population was in fact stable from the early 1990s to the early 2000s and that the analyses indicating large declines over this period have been misleading? All of the tagging-based estimates, using both open- and closed-population estimates, indicate such a decline, as does the trammel net catch rate index for the LCR inflow reach. Only the hoop net catch rate index of larger (fL >200 mm) fIsh does not show a decline over the key decade of intensive monitoring (the 1990s). However, the hoop net index suggests a decline from the beginning of data collection (i.e., the late 1980s) to the present. For the tagging-based methods, there is no assump- tion that sampling methodology was consistent over time (i.e., capture probabilities are not assumed to have been stable). For the newer tagging estimates to be biased downward by 50% or more, these estimates would have to involve a doubling in recapture probability for marked fIsh relative to the probability of capturing an unmarked fish. In capture-recapture studies, such bias is typically due to restricting sampling to a reduced portion of the area occupied by the population (Seber 1982). There has been no such systematic reduction in area sampled within the LCR or the LCR inflow reach. There has been a restriction in the seasonal time coverage of sampling (from monthly sampling in the early 1990s to spring and fall events in the 2000s) that could have the same effect as a spatial restriction of sampling. This season- ality could cause downward bias in closed-population estimates based on monthly aggregated data, but it is difficult to see how it would affect the tagging methods based on annually aggregated data (annual ASMR, Jolly-Seber) where there is ample time (1 year) for marked and unmarked fIsh to mix before recaptures are used in estimation. Another possible cause of differentially high capture probabilities for marked fIsh is the use of baited hoop nets in the LCR in 2001-2002. If that practice caused marked fIsh to be more likely to reenter nets than unmarked fIsh were to enter them in the first place, then capture probabilities were overestimated for both the closed- and open-population estimation methods. This would have caused underestimates of the numbers of unmarked fIsh at the end of 2002, and in the ASMR methods this would in turn have caused the back- calculated (virtual population) numbers of unmarked fIsh to be too low for at least the 1998-2001 time period. However, when we artificially inflated the 2002 unmarked numbers by a factor of two, we found that the back-calculated overall population trend from ASMR methods still indicated a decline of at least 40% over the last decade, even though the calculated 2001 population size was increased by only 500-800 adult fIsh. When we simply excluded the likelihood terms for catches of unmarked fIsh during the baiting period from the calculation of maximum likelihood estimates, there was no noticeable effect on the parameter estimates. When we attempted to directly estimate the relative capture probability for marked versus unmarked fIsh during the baiting period, we obtained estimates of 0.96-D.98 for all the ASMR (annual, monthly, and with or without movement) formulations, suggesting that there is no differential capture probability between marked and unmarked fish. These statistical checks do not prove that there was no strong differential baiting effect, but they do indicate that it is not parsimonious to assume such an effect. Finally, the effects of random sampling variation on the ASMR estimates can be evaluated by comparing estimates of recruitment or adult population trends