8068

• UCREFRP Catalog Number: 8068
• Author: Bestgen, K. R., R. T. Muth and M. A. Trammell.
• Title: Downstream transport of Colorado squawfish larvae in the Green River drainage
• USFW Year: 1998.
• USFW - Doc Type: temporal and spatial variation in abundance and relationships with juvenile recruitment.
• Copyright Material: NO

1
1

collected instantaneously at U.S. Geological Survey stations because those records often did not
accurately reflect the average river conditions. Temperature values at first reproduction and at
the first peak of reproduction were the average of the 5 d-period centered on the estimated date of
interest. The 5-d mean temperature likely better reflected the average water temperature at
initiation and first peak of reproduction than would measurements taken on a single date because
of single day climatic anomalies and because of uncertainty (± 2.5 d) in estimates of hatch dates
of larvae that were aged with otoliths (Bestgen and Bundy 1998). Days post-peak discharge was
the number of days between the highest recorded daily average discharge during spring runoff
and first reproduction.
Diel and cross-channel abundance of larvae.--Sampling bias may explain differences in
seasonal or annual abundance of larvae captured in drift nets during summer. Mean abundance
of Colorado squawfish larvae in samples collected during combined diel and cross-channel (1992
only) and diel-only (1992-1996) sampling was analyzed by general linear models (PROC
GENMOD, SAS Institute, Inc. 1993). This analysis calculated maximum likelihood estimates of
model parameters and was used to evaluate sampling bias. The discrete nature of the data
(counts of larvae) and high occurrence of zeros or low capture values suggested a Poisson model.
Log transformation of the response variable (i.e., log-link) ensured that the mean number of
larvae predicted by the fitted model was positive (SAS Institute, Inc. 1993).
The independent variables used in the analyses of the diel and cross-channel sampling
data was sample date, net position, time, turbidity, and their interactions. Independent variables
for the diel-only sampling data were year of sampling, sample time, turbidity, and their
interactions. Turbidity was included as an analysis variable because levels may vary widely in
the Yampa River during summer and affect abundance of larvae in the drift. Sediment mobilized
by runoff from afternoon thunderstorms usually increased water turbidity in the Yampa River by
noon the following day. Therefore, samples collected on days when turbidity had increased due
to a storm event the previous day were classed as turbid; samples collected when turbidity levels
were stable or declining were classed as clear. The natural logarithm of the volume of water
filtered by each net (m3) was also included as a covariate (offset) in models to account for
differences in sampling effort. Volume of water filtered by each net was estimated by
multiplying sampling time by flow rate by area of the net frame. Model selection was by
Akaike's Information Criterion (AIC; Akaike 1981) adjusted for over-dispersed data (QAIC;
7