|
<br />STATUS OF ENDANGERED COLORADO SQUAWFISH
<br />
<br />961
<br />
<br />TABLE 2.-Capture-recapture open models fit to the
<br />Colorado squawfish capture data. Also shown are the num-
<br />ber of estimable parameters in each model (K), AIC
<br />(Akaike's Information Criterion) values and AIC incre-
<br />ments (~AIC) compared to the AIC best model, sorted by
<br />AIC value, smallest (best model) to largest AIC (worst
<br />model).
<br />
<br />Model K AIC Ll.AIC
<br />{4>short ~ I; long' p} 2 723.776 0.000
<br />{4>,p} 2 723.810 0.034
<br />[4>, py} 3 724.976 1.200
<br />[4>, PT} 3 725.207 1.431
<br />(<Pshort; long' p) 3 725.772 1.996
<br />{ <Pseason, p} 4 726.061 2.285
<br />{ <Pshon ~ 1; long' Pyear} 5 726.400 2.624
<br />{<p, Pye.,} 5 726.439 2.663
<br />(<p,ptI 12 734.370 10.594
<br />{<p"p} 12 736.651 12.875
<br />[<p" Pye.,} 15 738.859 15.088
<br />{<P,,?,] 21 747.348 23.572
<br />
<br />.
<br />
<br />AIC (Table 2) had constant capture rate on each
<br />occasion (p = 0.107, SE = 0.014), survival rate
<br />set to 1 in the short periods among sampling oc-
<br />casions in the spring, and constant survival (~ =
<br />0.845, SE = 0.076) the remaining portion of the
<br />year. The next best model, with constant survival
<br />all year (~ = 0.860, SE = 0.069, 95% confidence
<br />interval [CI] = 0.662-0.950 based on logistic
<br />transformation) and constant capture rate, was
<br />more credible biologically because some natural
<br />mortality probably occurred during the late-April-
<br />mid-June sample periods.
<br />To assess whether the sampled population was
<br />stable, increasing, or decreasing during the four
<br />years of study, it was important to consider models
<br />with time trends in capture rates. The best of these
<br />models were those with constant survival and log-
<br />it-p having a linear trend over time (by sample
<br />occasion or by year). For model {</l,py}, ~ = 0.841
<br />(SE = 0.072) and py = 0.093 (SE = 0.020), 0.102
<br />(SE = 0.015),0.112 (SE = 0.015), and 0.123 (SE
<br />= 0.023) for 1991, 1992, 1993, and 1994, respec-
<br />tively, suggesting a 10% annual rate of increase
<br />in capture probabilities. This increase was not sta-
<br />tistically significant (z-test = 0.914, P = 0.3607,
<br />two-sided test), as corroborated by the small ,lAIC
<br />value (1.2). Still, in judging possible increases in
<br />the sampled population based on more fish caught
<br />in later years, it would be conservative to consider
<br />that the efficiency of capturing fish may have in-
<br />creased 10% per year.
<br />n seemed appropriate to use closed models to
<br />estimate population size separately by year be-
<br />cause estimated survival rates using model {c!>short:
<br />long' p} during the short time period between cap-
<br />
<br />TABLE 3.-Yearly population size estimates (Ni) for the
<br />upper reach, based on model Mo (constant within-year
<br />capture rates) from program CAPTUREa. Also shown are
<br />theoretical SEs (in parentheses) and 95% confidence in-
<br />tervals (profile likelihood) by year. For mean abundance,
<br />the SE is empirical and the confidence interval is based
<br />on a shifted log transform and a t3 distibution. Capture
<br />rate estimates (fJ) from CAPTURE are also shown.
<br />
<br /> 95% confidence
<br />Year i Ni (SE) interval p
<br />1991 205 (68) 124-520 0.106
<br />1992 311 (125) 179-1204 0.074
<br />1993 163 (29) 121-246 0.194
<br />1994 33:n90) 223-728 0.103
<br />Mean 253 (41) 161-440
<br />
<br />· White et al. (1992).
<br />
<br />tures were about 0.99. Table 3 gives the results
<br />from model Mo (constant within-year capture
<br />probabilities) computed by CAPTURE. The in-
<br />dependent annual population size estimates are too
<br />imprecise for assessing trends, and an average for
<br />the four years was calculated as N = 253 (SE =
<br />41, 95% CI = 161-440). The program RECAP
<br />was also used to estimate annual population size
<br />as well as annual survival and recruitment: N =
<br />263 (SE = 38, 95% CI = 186-333); ~ = 0.822
<br />(95% CI = 0.611-0.922); and iJ = 40 (SE = 10,
<br />95% CI = 24-59). These results, consistent with
<br />estimates from SURGE and CAPTURE, support
<br />the idea of a stable population during the four years
<br />of study while not excluding the possibility of pop-
<br />ulation increase during this time.
<br />Data summaries assessing potential trends in
<br />size of the upper reach subpopulation. during
<br />1991-1994 are presented in Tables 4 and 5. As-
<br />suming constant capture probabilities over sample
<br />occasions and years, total number of fish captured
<br />from 1991 to 1994 increased 17.1% (95% CI =
<br />8.5-24.6%) per year (linear regression, P = 0.001,
<br />two-sided t-test, df = 10).
<br />Mean CPUE of adults in the upper reach also
<br />steadily increased from 0.32 fish/net in 1991 to
<br />0.69 fish/net in 1994 (Table 5). When fish were
<br />partitioned by size, the increase in CPUE was ap-
<br />parently the result of increasing numbers of fish
<br />less than 550 mm, which steadily and significantly
<br />increased from 0.07 fish/net in 1991 to 0.35 fish/
<br />net in 1994 (P = 0.033). Older fish greater than
<br />550 mm increased, but not significantly (P = 0.59),
<br />from 0.25 to 0.33 fish/net. Colorado squawfish less
<br />than 550 mm made up approximately 22% of those
<br />netted in the upper reach during 1991 and ac-
<br />counted for 51 % by 1994.
<br />Because of this evidence of increasing popula-
<br />
|