|
<br />LI'ITLE COLORADO RIVER HUMPBACK CHUB
<br />
<br />239
<br />
<br />sampling effort was high, and lowest in the late 199Os,
<br />when sampling effort was limited. More intensive
<br />sampling effort since 2000 has resulted in increased
<br />capture probability from about 0.10 to 0.25 (Figure 4).
<br />
<br />Population Change
<br />
<br />The best TSM model (Table 2) was a fully time-
<br />dependent model that did not place time constraints on
<br />any model parameter. The estimated annual rates of
<br />population decline across all years were variable
<br />because of poor model fit and imprecise estimates in
<br />1996-1999, when sampling effort was low. Excluding
<br />these years, the geometric mean annual population
<br />change declined 14% annually (~= 0.86). In addition
<br />to using the fully time-dependent TSM model, we
<br />followed Franklin's (2001) suggestion to allow mor-
<br />tality and capture probability to remain time dependent
<br />and make population change time independent. The ~
<br />estimate for this model also indicates a negative trend
<br />in population growth of about 1 % annually (SE =
<br />0.006). Because humpback chub are a relatively large,
<br />long-lived species with few natural predators as adults,
<br />natural mortality should be relatively constant. To
<br />examine this, we fit a model with fixed sUlVival and
<br />time-dependent capture probability and population
<br />change. The geometric mean of population change
<br />from this model was also negative, indicating a decline
<br />of about 4% annually. When the low sampling years of
<br />1996-1999 are excluded, the overall findings for
<br />population growth indicate a decline of up to 14%
<br />annually.
<br />
<br />Recruitment
<br />
<br />As described in our companion paper (Coggins et al.
<br />2006), the ASMR method allows for estimation of
<br />recruitment preceding the onset of sampling activities
<br />by utilizing back-calculation methods from virtual
<br />population analysis. The recruitment reconstructions
<br />among the three ASMR formulations suggest a peak in
<br />recruitment in the late 1970s to early 1980s of 13,500-
<br />18,500 age-2 fish (Figure 5). After that peak, an overall
<br />decline was evident to the early 1990s, when annual
<br />recruitment stabilized at about 2,000 age-2 fish.
<br />Recruitment may have stabilized or increased in the
<br />late 1980s before resuming its decline and again
<br />stabilizing through the 1990s.
<br />
<br />Population Size
<br />
<br />The best (Le., biologically reasonable with a low
<br />AlCc value) Jolly-age model for estimating the
<br />population size of age-4 and older humpback chub
<br />allowed mortality to vary with age and capture
<br />probability to vary with time. This model gave a better
<br />fit than non-age-structured models, demonstrating the
<br />
<br />25,000
<br />~
<br />~ 20.000
<br />~ 15,000
<br />o
<br />~ 10.000
<br />,r,
<br />g 5,000
<br />z
<br />
<br />- ASMR I
<br />-...- ASMR 2
<br />.. ASMR 3
<br />
<br />
<br />.-- -.,;.,.(..-.-.
<br />
<br />o
<br />,0,'1" ~'1'1 ,0,'10, ,o,'b' ,0,"0'; ,0,"0" ,0,';;' ,o,'bo, ,0,0,' ,0,0,'; ,0,0," ,0,0,'1 ,,;po,
<br />
<br />Brood Year
<br />
<br />~ 16,000
<br />35 14,000
<br />:512.000
<br />'g 10,000
<br />~ 8.000
<br />if 6.000
<br />-'5 4,000
<br />':1 2,000
<br />o
<br />
<br />~<f' ,o,<f> ,0,0,' ,,;p'" ~o,'; ,o,~ ,0,0," ,0,0,'0 ,0,0,'1 ,o,o,'b ,0,0,0, ",# "'~'
<br />
<br />..
<br />
<br />- ASMR I
<br />-.-'" ASMR2
<br />....... ASMR 3
<br />-- Jolly-Age
<br />
<br />
<br />Year
<br />
<br />FIGURE 5.-Age-2 humpback chub recruitment by brood
<br />year estimated using the three formulations of the annual age-
<br />structured, open-population capture-recapture (ASMR) model
<br />(top panel) and adult (age-4 and older) humpback chub
<br />abundance estimates using the three formulations of the
<br />ASMR model and the age-structured Jolly-Seber model
<br />(bottom panel). The error bars on the ASMR estimates are
<br />95% credible intervals from Markov chain-Monte Carlo
<br />sampling of posterior parameter distributions; those on the
<br />Jolly-Seber estimates are 95% confidence intervals.
<br />
<br />strong affect of age on mortality (Table 2). Adult
<br />population estimates for this Jolly-Seber model ranged
<br />from about 14,500 in 1989 to about 2,400 in 2001
<br />(Figure 5). Adult population estimates from the ASMR
<br />methods were similar across methods and ranged from
<br />10,000-11,000 in 1989 to 3,100-4,400 in 2001
<br />(Figure 5). Population estimates from both the
<br />Jolly-age and ASMR models were similar in magni-
<br />tude and trend from the early 1990s through 2001.
<br />
<br />Movement Model Results
<br />
<br />Unfortunately, the combined movement and ASMR
<br />model is overparameterized and unable to provide
<br />reliable estimates of movement rates. For example,
<br />seasonal changes in the capture rates in the LCR can be
<br />explained equally well as the result of seasonal
<br />movement to the LCR inflow reach or seasonal
<br />changes in capture probability for fish still in the
<br />LCR. To complicate matters further, seasonal move-
<br />ment rates probably vary with age, younger fish being
<br />much less likely to migrate downstream after the
<br />spawning season (Gorman and Stone 1999).
<br />To avoid the overparameterization problem, we used
<br />capture probabilities and population size from the
<br />ASMR models. We used this information to examine
<br />a series of seasonal movement rate scenarios designed
<br />
|