9697

• UCREFRP Catalog Number: 9697
• Author: Bestgen, K.R., D.W. Beyers, J.A. Rice and G.B. Haines.
• Title: Factors affecting recruitment of young Colorado pikeminnow
• USFW Year: 2006.
• USFW - Doc Type: synthesis of predation experiments, field studies, and individual-based modeling.
• Copyright Material: YES

1730

30

G
~ 25

OJ
....
;::l
~
~ 15
E
OJ
E-<

20

10

5
I-Joo 25-Joo 19-]ul 12-Aug 5-Sep29-Sep
Date

FIGURE 4.-Green River summer water temperature regimes
during cool (1983) and wann (1994) years, as recorded near
Iensen, Utah (U.S. Geological Survey gauge 09261000), and
used in individual-based model simulations.

temperatures is further justified because development
and growth of early life stages of Colorado pike-
minnow do not vary between fluctuating or constant
water temperature conditions, as long as a similar
number of degree-days are available (Bestgen and
Williams 1994; Bestgen 1996).
IBM simulations.-An individual-based modeling
approach was selected because we suspected that
individual differences in size and growth rates of
larvae may have a strong influence on recruibnent
patterns (DeAngelis and Rose 1992; Rice et al. 1993).
The IBM focused primarily on the interacting effects of
life history variables and temperature on larval growth
and survival and size-dependent predation by red
shiners. Each simulation started with a cohort of
10,000 larvae arriving in nursery habitat. Each fish was
assigned a hatching date (e.g., backwater arrival minus
8 d) so that simulation data could be compared with
field data Hatching date used in model simulations
varied from 1 June to 1 August, a range consistent with
dates of capture of newly hatched larvae in drift nets in
the lower Green and Yampa rivers (Bestgen et al.
1998). Initially all larvae were 9.0 mm TL which is a
typical size of Colorado pikeminnow larvae entering
backwaters (Bestgen et al. 1998). Each larva was then
assigned a baseline growth rate (mm/d) and the number
of attacks it encountered during the day was calculated
using equations (3) and (4). The estimated total number
of attacks per predator in each 6-h mesocosm trial was
multiplied as a function of pool area (1.48 to
standardize to 1 predatorfm2) and time (14-h day/6-h
trial = 2.33, where a 14-h day represented approxi-
mately the daylight period during summer in the
Colorado River basin) to provide mean number of
attacks per 14-h day on a larva For each attack, a
predator size was drawn from a predator size
distribution, and the probability of capture was
determined using equations (1) and (2). The probability

BESTGEN ET AL.

that a larva would survive all attacks on a given day
was the product of individual probabilities of surviving
each attack on that day. Whether the larva survived the
day was determined by drawing a number from a
uniform distribution between 0 and 1; if the number
was less than or equal to the larva's probability of
survival for the day, then it survived. Probability of
survival was set to 1 for larvae longer than 25 mm TL
because laboratory observations showed that red
shiners were not able to capture Colorado pikeminnow
of that size. At the end of a day, length of each
surviving larva was increased based on the larva's
initial baseline growth rate and the daily water
temperature via equation (4). This process was repeated
for each larva; then the time step was incremented, and
surviving larvae were exposed to predation the next
day. Fish that survived to 25-mm or 1 October were
considered to have left the window of predation
vulnerability and recruited to the age-O population in
autumn. The duration of the window of predation
vulnerability for Colorado pikeminnow is dependent
on temperature and food abundance (Bestgen 1996;
e.g., at maximum food ration a 9-mm Colorado
pikeminnow would take 80 d to grow to 25 mm at
180C but only 53 d at 260C). Each simulation was
replicated three times using different random number
seeds and results were averaged. Data describing
numbers, size distributions, and growth rate distribu-
tions of survivors and mortalities were output for
further analysis. .
Comparison of model and field patterns.-Simula-
tions were conducted with the IBM using hatching date
distributions derived from larvae captured in drift nets
in 1991 and 1992. This allowed us to compare
recruibnent patterns derived from IBM simulations to
those obtained from analysis of otoliths in wild fish.
The number of larvae input into the model on each day
was scaled directly as a function of the number of
larvae actually captured in drift nets multiplied by
10,000 (i.e., for each fish captured in a drift net, 10,000
were input to the model that day). On days when no
larvae were captured within the known spawning
season, a total of 10,000 larvae were added to the
simulation day to account for inability of our sampling
to detect low numbers of larvae. The magnitude of the
scaling factor was chosen for convenience of conduct-
ing simulations and did not influence the outcome of
the analysis because only relative survival indices were
compared. Baseline growth rates used in simulations
were based on otolith-estimated mean summer growth
rates of wild Colorado pikeminnow (Bestgen 1997).
Baseline growth rates were 0.41 mm/d TL for fish in
lI1e middle Green River in 1991 and 0.43 mm/d in
1992, and daily growth rates in the lower Green River