III.4.1 - Fall Chinook survival

Fig. 70 Release () and recovery sites () of data for fall chinook validation. Juvenile habitat indicates where Snake River smolts were tagged prior to migration.

Snake River fall chinook

The evaluations of fall chinook survival predictions were made using PIT tag data from 1991-1994. Fish were collected in beach seines in their juvenile habitat in the Snake River reach above the confluence with the Clearwater (Fig. 70). Juvenile pre-smolt fish were tagged late May through mid-July in 1991, mid-April through mid-June in 1992, late April through mid-July in 1993, and mid-April through mid-June in 1994. The analysis used fish observed at Lower Granite, Little Goose, and McNary dams for 1991-1994, and at Lower Monumental in 1994 only.

Fitting the model to these data sets was problematic because of the variable nature of fall chinook behavior within a year and from year to year. In addition, fish were tagged prior to their active migration so they experienced an unknown amount of mortality prior to migration from the juvenile habitat. The variability in fish behavior between tagging and arrival at the first dam, Lower Granite, was considerable and using fixed model parameters provided poor fits to the arrival time and survival. In this situation strategy was to validate travel time and survival of fish (Table 66) once they had passed Lower Granite Dam. For this comparison smoltification onset (Table 64) and predator density (Table 66) in the juvenile habitat were adjusted so the model generated the observed arrival date and collection numbers at Lower Granite Dam. Since the collection numbers were determined with an assumed FGE as well as travel time and mortality rate the comparisons to observations is not a verification of survival parameters but of the overall predictions on fish collections numbers.

CRiSP predictions of are generally within a 2x factor of the observed collections for Snake River fall chinook (Table 66). For 1991, 1992 and 1994, the fit of model-predicted numbers to observed numbers is excellent. For 1993, the model consistently overpredicts the numbers of fish at each downstream site. This is partly due to a very high detection rate at Lower Granite as compared to the other three years but average detection rates (compared to the other three years) at the lower three dams.

Mid-Columbia fall chinook survival

Survival of Priest Rapids Hatchery fall chinook was evaluated for the outmigration years 1977 through 1987 by Hilborn et al. (1993). The study reported survival from hatchery release to entry into the ocean fishery. The CRiSP model predicts survival from hatchery release to the estuary so the difference in the observed and modeled survival is an estimate of the early ocean survival for each year.

To generate river survivals the Priest Rapids hatchery-derived travel time parameters were applied with the historical flows, spills, and temperatures, and estimates of predator activity derived from the John Day Pool study. Model release dates were set from Priest Rapids hatchery release information and were tracked to Jones Beach, near the site of the coded wire tag releases used by Hilborn et al. for comparison.

In the Hilborn et al. study recovery rates of coded wire tags (CWT) from Priest Rapids brand groups and branded fish from stocks below Bonneville Dam were used in a generalized linear model to estimate survival of the in-river and early ocean life stages.

CRiSP generates higher estimates of survival (23% average for CRiSP vs. 11% average for the Hilborn study) because the CRiSP model defines survival through the river and not to entry in to the ocean fishery. The difference in the Hilborn estimate is the inclusion of ocean mortality prior to entry into the fishery. The relationship between the two survivals is

(170)

The ocean survival can be estimated by a regression of S_crisp. against S_hilborn in which the regression is constrained through zero (Fig. 71). The slope of the regression is 1/S_ocean where the ocean survival is the average survival over the years of observations. The survival depends on the early ocean mortality processes and any estuary mortality processes not accounted for in CRiSP. The regression in Fig. 71-A gave a slope of about 1.66 which implies an early ocean survival around 60%.

The Hilborn analysis produced a relationship between river flow and estimated survival. This flow relationship was not observed in CRiSP. The difference in the relationships could result from 1) CRiSP not correctly representing a fall chinook flow vs. in-river survival relationship, or 2) a flow-estuary survival relationship exists that is not captured by CRiSP.1.

The temperature predator activity relationship could account for the difference in survivals. If this were the case the fractional deviation between the two estimates could vary with temperature. This relationship is evident and can be illustrated by defining the fractional difference in the two survivals according to the equation

(171)

Fig. 71-B illustrates that E has a relationship with temperature during migration. This suggests that to improve the modeling of fall chinook better temperature modeling is required.

Fig. 71 -A. Priest Rapids Hatchery fall chinook survival. CRiSP-estimated survival to Jones Beach vs. Hilborn et al. (in revision) estimated survival to age 1 fishery. With forced zero intercept slope is 1.66. -B. Difference (E) between CRiSP and Hilborn et al. survival (eq(171)) as a function of temperature at release.