9588

• UCREFRP Catalog Number: 9588
• Author: Bestgen, K. R. and e. al.
• Title: Population Status of Colorado Pikeminnow in the Green River Basin, Utah and Colorado.
• USFW Year: 2005.
• USFW - Doc Type: Fort Collins, CO.
• Copyright Material: NO

and the lower Green River were then correlated with abundance estimates for the same reaches.
No ISMP sampling was conducted in the Desolation-Gray Canyon reach of the Green River, so
no comparisons were available for that reach.
Robust design for capture-recapture studies.--Robust-design sampling and analysis
capitalizes on the strengths of closed and open population models used to estimate demographic
parameters (Pollock 1982; Pollock et al. 1990). Sampling occasions completed at closely-spaced
intervals (e.g. weeks) are used to estimate population size using closed population models. That
level of sampling completed in two or more consecutive years allows for estimation of
recruitment and survival rates between most years. In some reaches and years, we conducted a
fourth, relatively late sampling pass. We combined fourth pass data with third pass data because
using a fourth pass in the analysis created asymmetry of the capture histories and difficult model
fitting. Combining data was done only after analyses showed that only a couple offish («1%)
moved between sampling reaches between passes within a year. Thus, potential confounding due
to fish movement to spawning areas was minimized.
Statistical modeling.--The combined robust-design (Kendall 1999, Kendall et al. 1995,
1997) multistrata (Brownie et al. 1993, Hestbeck et al. 1991) model in Program MARK (White
and Burnham 1999) was used to estimate annual survival in year t (SJ, probability of transition
between reach i and j (?&,j), capture-recapture probabilities within reach i for each year t and
sampling occasion k (p«k), and Colorado pikeminnow abundance in each reach i for each year t
(N;,). A multi-strata (reach) model was justified because of the sampling design and because
probabilities of capture varied substantially among sampling occasions (passes) and reaches
within a given year, which precluded a model to estimate river-wide abundance directly. Instead,
river-wide abundance estimates were obtained by summing the separate reach estimates by year.
e
•
e
w
•
•
•
•
20