7960 - Laserfiche WebLink

Home Browse Search

114 Status ofRaxorback Sucker The general distribution of razorback sucker length frequency between 1980 and 1992 appeared relatively constant with the most frequent modes being in the in- terval of 505-515 mm (Fig. 2). Despite low average growth rates-1.66 mm/yr (SD = 4.72)-most individuals showed measurable growth through time (Fig. 3). Despite mea- surable growth, no dramatic shift in the length fre- quency occurred with time (Figs. 2 & 4). A regression test for linearity indicated no differences in average length among years, F1,13 = 0.03 (p = 0.858). The mea- surement of negative growth in length measurements through time (Fig. 3) probably represent variability rela- tive to measurement error. Population Parameter Estimates The program RELEASE, testing the assumptions of the CJS model, provided a goodness of fit chi-square value of 29.96 with 26 degrees of freedom indicating that the model fit the data. With the program SURGE the model {~, pt} was selected, a model with time-varying capture probabilities, but constant annual survival rate. In this model to was estimated to be 0.708, with a standard er- ror of 0.0246. The confidence interval for the annual sur- vival rate, based on a logistic transformation on ~, was 0.658 to 0.754. Annual population estimates using the Lincoln-Petersen method (Table 1) showed considerable variation, particularly in the initial estimates, but this variation does not exceed that expected given the sam- pling standard errors and the N. The unweighted aver- age estimate of 524 has an empirical standard error of 85, on 8 degrees of freedom. The estimated theoretical sampling standard error based on individual standard er- rors is 88. The estimate of average annual abundance and recruitment using program RECAPCO was Table 1. Lincoln-Petersen estimates of razorback sucker population from the middle Green River between 1980 and 1992 Year Nt SE (NI) cvE Log (NZ) 1980 - -- - 1981 -- - 1982 1051 539 0.51284 6.95750 1983 - - - 1984 - - - 1985 282 99 0.35106 5.64191 1986 630 358 0.56825 6.44572 1987 736 268 0.336413 6.60123 1988 255 60 0.23529 5.54126 1989 477 144 0.30189 6.16752 1990 521 203 0.38964 6.25575 1991 307 132 0.42997 5.72685 1992 453 217 0.47903 6.11589 Mean: 524 224 0.40360 95% C.I. N: 351 to 696 Modde et al. N = 508, SE (N) = 106 and B = 108, SE (B) = 36. A high estimate of recruits reflects the addition of indi- viduals that have lost tags to those individuals recruited annually into the population. It seems reassuring that two very different methods (open versus closed models to estimate N) provided an average annual estimate near 500 with similar standard errors. Note, however, that the method based on closed models was actually estimating N/~; hence, a better abundance estimate from that method is 524 X 0.7, or about 367 (SE = 61). Given the uncertainties of the methods, we think it reasonable to believe the popula- tion size during 1980-1992 was, on average, between 300 and 600 fish. The hypothesis of a trend in the population size of the razorback sucker was tested with linear regression of log (N=) on year i. The least conservative test is an un- weighted regression using all Na. That analysis gives r = -0.0694 (SE = 0.0498, 95% C.I. _ -0.187 to 0.048), with none-sided p value equal to 0.1031. The first NZ = 1051 (SE = 539) has undo influence (leverage) on this result. If the regression of log (N~ on time is weighted by 1/(cvZ )2, then r = -0.0456 (SE = 0.0668), p = 0.2584. If the first N= is not used in the regression, then r = -0.0082 (SE = 0.0644), p = 0.4516. Because none of these tests were significant, the data do not support a decline in population number through time. Power calculations were done to get approximate power of a test for population decrease for our data. We assumed a constant annual decrease (r), a one-sided test (alpha = 0.05) based on log (N), and a constant coeffi- cient of variation of 0.4 for each N. Results for a few val- ues of r follow: r Power -0.025 0.14 -0.050 0.30 -0.075 0.51 -0.100 0.73 -0.125 0.89 -0.150 0.97 These results and the confidence intervals on r show that we do not have power enough to detect small de- creases in population abundance. Yet, if there was no re- cruitment, then we would have r = 1 - S, so unless an- nual survival is 0.875 or more we might expect to detect population declines in the absence of any recruitment. Population and Habitat Correlations Discharge of the Green River was significantly regressed only with the abundance of fish within the length inter- 1 Conservation Biology Volume 10, No. 1, February 1996