9724 - Laserfiche WebLink

Home Browse Search

238 ~~tj 10 ~b\v I~ l' \A1'10 ~?~I~O vrJ COGGINS ET AL. 10,000 I .~ 8,000 j 6,000 ~ 4,000 o "- 2,000 o I I I I I I I I I I I ,# ,,f,'I' ~'" ,,f,'I'" ,,f,'I'" i' ,0,0,'0 ,0,0," ,,f,'I"o ,$' 0,,# o,,~' o,,~'" o,,~'" I I I Year ~ ~(}~~ev -t ~l~ .o(\.t ~ '1110 \_~~ 7 v \~V ~ 7.000 ~ 6,000 Vi 5,000 _2 4.000 ~ 3.000 g- 2.000 "- 1.000 o ,o,rf' ,0,,,,, ~",o" ~",,,, ~q> ,0,,,,,> ,<It ,,f,'I" ~","o ,$' 0,,# 'f'<>' ",~o" 'f''i;)'" Year FIGURE 3.-Abundance estimates for humpback chub from closed-population capture-recapture models in the Little Colorado River (top panel; all estimates for fish larger than ISO mm total length) and the Little Colorado River inflow reach of the Colorado River (bottom panel; all estimates for fish larger than 200 nun total length). In the top panel, the estimates depicted as diamonds are from the studies of Douglas and Marsh (1996) and those depicted as squares are from this article. In the bottom panel, the estimates depicted as diamonds are from the studies of Valdez and Ryel (1995) and that depicted as a square are from the studies of Trammell and Valdez (2003). The error bars in both panels indicate 95% confidence intervals. sampling coverage throughout the entire reach. Annual sample sizes in 1994-2000 and 2002--2003 were between 2% and 50% of the 1990--1993 average sample size, and in some years effort was focused near the LCR confluence. The lack of robust coverage and limited sampling in the mid to late 1990s suggest that comparing the 1990-1993 and 2001 data best depicts the overall trend of relative abundance. However, though both data sets suggest an overall decline in adult humpback chub abundance, simple linear re- gression analyses provide estimated slopes that are not significantly different from zero (P = 0.14 for both regressions; Figure 2). Closed-Population Models Population size estimates for fIsh (TL > 150 mm) from the Chapman-modified, Lincoln-Petersen closed- population model ranged from about 2,000 in 2001 to about 3,400 in 2003, a decline from the 4,300-5,400 individuals estimated in the 1990s (Figure 3). Recent closed-population abundance estimates in both the LCR and the LCR inflow reach (Figure 3) suggest smaller population sizes in the early 2000s than in the early 1990s, although the preclSlon for all Lin- coln-Petersen estimates is low. Open-Population Models The bootstrap goodness-of-fIt test in MARK was able to calculate a C to adjust for the overdispersion of the data. This estimate was used to evaluate the effects of overdispersion on AlCc model selection. No changes in model selection were found across the range of c values evaluated. The best (lowest AlC) Jolly-Seber model for estimating mortality and capture probability was the age-dependent mortality and time-dependent capture probability model (Table 2). Annual mortality decreased with age, from about 68% for age-2 fIsh to about 16% for fIsh over age 15 (Figure 4). Mortality estimates from each of the ASMR model formulations were similar to those of the Jolly-age methods; these estimates were age dependent and ranged from about 65% for age-2 fIsh to about 18% for fIsh over age 15 (Figure 4). Annual capture probabilities were consistent across all models. Annual capture probability for the Jolly- age model ranged from 0.04 to 0.51 (Figure 4). Capture probability from the three ASMR model formulations was similar, ranging from about 0.04 to 0.47 (Figure 4). In both the ASMR and Jolly-age models capture probability was highest in the early 1990s, when 0.8 ~O.7 ~0.6 '; 0.5 50.4 ~0.3 $ C 0.2 ..:10.1 o - ASMRI ---.- ASMR2 ...... ASMR3 _.~. Jolly-Ab'" 4 6 10 II 12 13 14 15 8 9 Age 0.6 &'0.5 :0 .8 0.4 o .t 0.3 ~o.2 Q. <30.1 0.0 ,0,"0'" ,o,rf' ~o,' ,,f,'Io,, ,0,0,'> ,,f,'I'" ,,,,,,,'> ,<>,0,'0 ~o," ,,,,0,"0 ,,,,,f,'I o,,~<> o,,~' - ASMR I ---ASMR2 ASMR 3 --~ Jolly-Age Year FIGURE 4.-Humpback chub annual mortality rates across ages (top panel) and capture probabilities across years (bottom panel) from the three formulations of the annual age- structured, open-population capture-recapture (ASMR) model and the age-structured Jolly-Seber model. The error bars on the ASMR estimates are 95% credible intelVals from Markov chain-Monte Carlo sampling of posterior parameter distribu- tions; those on the Jolly--Seber estimates are 95 % confidence intelVals.