These wild fish were beach-seined in their rearing grounds in the Snake River, PIT tagged, and released (details in Connor et al. (1996)). These fish are primarily subyearling fall chinook, but a small proportion may be yearling spring/summer chinook. Connor et al. (1996) estimated that for the 1994 migration year, approximately 80% of the fish tagged and detected that year at Lower Granite were fall chinook.
Analyzing these data for travel time presents several problems. First, since the fish are not active migrants when tagged, travel times from release site to first detection site represent a migration time plus a period of pre-migration. Thus, this information is not useful for predicting travel times in the downstream reservoirs. Second, sample sizes are small and seasonal passage distributions are protracted so the detections at observation sites do not necessarily characterize the passage distribution of the entire population. Methods of estimating travel time that involve comparing distributions of detections at an upstream to those at a downstream site produce unreasonable results. For the 1993 and 1994 data, for instance, the median detection date at McNary Dam (downstream site) was at least 4 days earlier than the median detection date at Lower Granite Dam (upstream site).
To overcome these problems, we only used travel times of individual fish that were detected at both Lower Granite Dam and one of the three downstream observations sites (Little Goose Dam, Lower Monumental Dam, and McNary Dam). Thus the data we analyzed were of direct observations of travel times of actively migrating fish. Unfortunately, double-detections were rare until 1995, so we only can analyze 2 years of data with this method.
To analyze these data, we grouped individuals into cohorts based on when they passed Lower Granite Dam. The Lower Granite passage dates extended over 5-7 days for each cohort with at least three individuals per cohort being detected at each of the downstream observation sites. Details are provided in Table 1.
We applied four nested models of increasing complexity to the data:
Model details are provided by Zabel, et al. (1997).
The population spread parameter (
2) was fit using the m.l.e.
approach outlined in Zabel
and Anderson (1997).
The plots below show observed average travel times versus modeled average travel times to each of the three observation sites for all four models.
corresponds to observations at
Little Goose Dam.
corresponds to observations at
Lower Monumental dam.
corresponds to observations at
McNary Dam.
Table 2 . Parameter estimates, standard errors, sum of
squares, and R2 for the four migration rate
models for all cohorts in the years 1995 - 1996. The units for
MIN and MAX are mi/day. FLOW, 
1, and 2 are non-dimensional.
TSEASN has units of Julian date, and 2 has units
mi2/day. For models 1-3, MIN in this table corresponds
to 0.
From model 1, the average migration rate for all the cohorts through all the reaches is 6.7 mi/ day. This is quite a bit faster than the average migration rate for subyearling chinook in the mid- Columbia River. With over 100 fish observed over two years, though, this result does not appear to be spurious.
The linear flow model (model 2) offers no improvement over the
simple constant migration rate model (model 1). In fact, model 2
reduces to model 1 because the value of the FLOW parameter is 0.0. We
constrained this parameter to be greater than or equal to 0 since
we know of no biological justification to have a negative
relationship between migration rate and flow. The absence of a
positive flow relationship is not surprising because in general,
migration rates increased during the migratory season while river
velocities decreased during this period in both 1995 and
1996.
Model 3, which incorporates a season/flow interaction, performed substantially better than model 2, with an R2 of 0.735; that is, model 3 described 73.5% of the variability present in model 1. With model 3, fish use more of the river flow for migration later in the season. Based on these results, the seasonal change in behavior is very pronounced in these fish.
Model 4 further improved performance over model 3 with the R2 increasing to 0.788. This model adds a flow-independent downstream acceleration term to model 3. With this model, fish migrated almost 9 mi/day faster in the lower reaches as compared to the uppermost reach.
For CRiSP Model runs: We recommend using the model 3 parameters for CRiSP model runs of Snake River subyearling chinook. Although model 4 conferred an improvement in fit, the standard errors also increased, indicating that model 4 is less stable than model 3 with these data.
Connor, W.P., H.L. Burge, R.D. Nelle, C. Eaton, and R. Waitt. 1996. Rearing and emigration of naturally produced juvenile Snake River fall chinook salmon. In Rondorf, D. W., and K. F. Tiffan, Identification of the spawning, rearing, and migratory requirements of fall chinook salmon in the Columbia River Basin. 1994 Annual Report to Bonneville Power Administra tion, Portland, OR. Project Number 91-029.
Zabel, R.W. and J.J.Anderson. 1997. A model of the travel time of migrating juvenile salmon, with an application to Snake River spring chinook salmon. North American Journal of Fisheries Management 17:93-100.
Zabel, R.W., J.J. Anderson, and P.A. Shaw. 1997. A multiple reach model describing the migratory behavior of Snake River yearling chinook salmon (Oncorhynchus tshawytscha). Canadian Journal of Fisheries and Aquatic Sciences. In press.