III.4.2 - Spring chinook survival

Data sets analyzed included PIT tag survival studies form the Snake River, brand release survival studies from the Snake to the lower Columbia (the Sims and Ossiander data), mid-Columbia brand release survival studies and radio tag survival studies below Bonneville Dam.

Snake River PIT tag survival studies

The Snake River PIT tag survival studies conducted by NMFS in 1993 and 1994 were used as validation date sets. The studies are described by Iwamoto et al. (1994) and Muir et al. (1995). Spring chinook were tagged with PIT tags and released at Nisqually John Landing 20 km upstream of Lower Granite Dam in 1993 and at the Snake/Clearwater confluence in 1994. Passage of individually tagged fish through Lower Granite, Little Goose and McNary Dams was monitored with PIT tag detectors in 1993. In 1994 Lower Monumental was added as a detector site (Fig. 72).

Two approaches were applied to compare the model predictions to the observed survivals. In both years of experiments the survival of fish from the release site to Lower Granite Dam were high, approaching 100%. These high survivals are either an artifact of the experiments or they reflect an actual high survival in Lower Granite Reservoir. Since there is insufficient information to determine if the survival experiments are representative of typical fish survival through Lower Granite Reservoir the model was fit in two ways: (A) as the model is calibrated with Lower Granite Reservoir predator densities, and (B) adjusting the Lower Granite Reservoir density so the model predicted survival through Lower Granite tailrace fits the PIT tagged estimates (Table 67). In the case (A) predicted survivals to the lower collection point are less than the observed estimates of survival but the survivals may better represent actual survival conditions in the river assuming that some mortality occurs in Lower Granite Reservoir. The second validation approach, (Case B) gives survivals closer to the observed levels. This implies that the model accurately predicted chinook survival once smolts passed Lower Granite Dam.

Fig. 72 Location of release and collection sites for the 1993 and 1994 survival studies. Collection sites are Lower Granite Dam (LGR), Little Goose Dam (LGO), Lower Monumental (LMO) and McNary Dam (MCN).

Spring chinook dam passage survival

The PIT tag study in 1993 provided estimates of total survival past two dams (Iwamoto et al. 1994). These were compared to the dam passage generated from CRiSP. The CRiSP predictions are within the standard error of the experimental estimates (Table 68).

Mid-Columbia spring chinook survival

Mark-recapture experiments conducted in the mid-Columbia in the 1980s released fish in the Methow River and below Priest Rapids Dam (Fig. 73). Both releases were recaptured at McNary Dam and an estimate of survival between the two release sites was obtained. The data are given in smolt monitoring program annual reports (1986, 1987). The resulting survival estimates obtained using CRiSP are in very close agreement with those obtained in the study. Travel time observations suggest that the released fish "held up" for some time before initiating migration; when releases are delayed in CRiSP an excellent fit to travel time estimates is obtained (Table 69).

Fig. 73 Release and recovery sites for mid-Columbia spring chinook survival studies.

The estimated survival and travel time, and CRiSP estimates for the same parameters, for both years of experiments are given in Table 69.

Snake River spring chinook survival 1966-1983

CRiSP was evaluated with survival estimates based on the brand release studies conducted between 1966 to 1983. These studies are informally knows as the Sims and Ossiander data after the report by Sims and Ossiander (1981) summarizing research between 1973 and 1979. The studies have been supplemented with information in Raymond (1979) and reports by the Coastal Zone and Estuaries Studies Division of the Northwest and Alaska Fisheries Center of NOAA for the years 1979 through 1983. These data cover a period prior to the construction of the Snake River dams above Ice Harbor Dam.

Because of the changes in the hydrosystem and differing research questions addressed over this period, release and recapture sites and the number of dams fish passed through changed from year to year. In the early years up through 1975 fish were released at Whitebird or Riggins on the lower Salmon River. These are represented by a Whitebird release in the model (Fig. 74). Recovery sites included the upper dam in the Lower Snake River, a mid dam, which was typically Ice Harbor, and a downstream site which was either the Dalles Dam or John Day Dam on the lower Columbia River. In 1966 Ice Harbor dam was the only project on the lower Snake River (Table 70).

Fig. 74 Release and recapture sites of studies used to validate spring chinook survival from Snake River to lower Columbia.

A number of factors altered the mortality in passing through turbines and bypass systems. Debris accumulated in the forebay of the Snake River dams causing significant descaling and mortality. This trash was not removed on a regular basis until 1980 (Raymond and Sims 1980, Williams and Mathews 1995). Experimental slotted bulkheads were installed in Little Goose and Lower Monumental dams in 1972 to reduce gas levels. The action lowered gas levels but mortality through the bulkheads was high (Raymond 1979). In the model validation passage mortalities were adjusted for years where studies or estimates of survival were available (see Table 54).

Fish passage was also affected when new dams were brought on line (Raymond 1979) (Table 71). In the first year of operation at Lower Monumental and Little Goose dams the turbines were not in operation and the river was passed in spill. Because of the spill, Snake River supersaturations reached 120 to 140% and juvenile fish mortality through gas bubble trauma was high (Raymond 1979).The survival studies covered 17 years and fish survivals were estimated in a variety of ways (Steward 1994). Between 1966 and 1979 survival was estimated from the number of fish passing Ice Harbor and the Dalles dams. Calculations were made from recoveries of juveniles that were marked and released in the Salmon River at hatcheries and in the forebays and tailraces of dams. In some years flow collection efficiency curves were used to estimate smolt passage at dams. In other years collection efficiency was estimated with forebay releases (Raymond 1979, Sims and Ossiander 1981). In years with transportation, accurate estimates of smolt survival were not possible because accurate estimates of the number transported were not available (Sims, Giorgi, Johnsen and Brege 1983).

To fit CRiSP to the survival estimates, the strategy was first to configure the model according to all relevant information on dam operations and passage conditions. Second, model dam passage parameters were adjusted to fit the reported dam passage information. Third, the model was fit to the survival information.

The river conditions, including daily river temperature and flows through the Snake and the Columbia, were set for all years. Hydrosystem operations for each year were also set; these included the number of dams, project hydraulic capacity (which reflected the number of turbines on line), reservoir elevation levels, spill on a daily basis, turbine and bypass mortalities reflecting projects, and the year with slotted bulkheads (Table 54).

Fish guidance efficiency estimates were obtained from a number of sources (see FGE Calibration section II.9.2). For transport dams the FGE was adjusted so the model predicted fraction of fish transported from the Snake River equaled the reported fraction (Table 72).

Spill was variable from year to year. During the construction of the Snake River dams spill was high because not all turbines were installed when the dams were completed. As a result, large spills were required in the early years of a dam's operation. The large spills generated high gas supersaturation. These were generally described by the model. The model-generated and reported ranges of supersaturation varied between 120% and 140% (Ebel Krcma and Raymond 1974; Ebel, Raymond, Monan, Farr and Tanonaka 1975; Raymond 1979). The effect of the gas depended on fish depth. The modal fish depth in the model was set at 12 ft. Observed chinook salmon depth varied between 5 and about 40 ft (Table 34 and Table 35).

Predator densities between the lower and upper dams were set by the predator densities studies (See Predator Density section II.6.1). For the years 1966 through 1968, prior to the construction of dams above Ice Harbor, predator density was set at 200 predators/sq km. Starting 1969 predator density above the upper most dam was set at the value observed for Lower Granite reservoir (440 predators/sq km).

References for validation of travel time and survival for the data from 1966 through 1983 is given in Table 73.

The observed median date of arrival of fish at dams and model fit are given in Table 74. For all years except 1973 and 1977, the two lowest flow years, a fit within a few days was obtained by adjusting the model parameter Tseasn. This parameter characterizes how quickly the flow dependent component of migration increases over season. A pattern emerged with Tseasn. Prior to hydrosystem development above Ice Harbor dam, 1966 through 1969, the Tseasn required to fit the migration data was early, Julian day 127 (May 7). In these years fish were predominately wild and their migration behavior was characteristically different from hatchery fish (Zabel 1994). Beginning in 1970, the hatchery contribution of fish migrating through the Snake River increased from less than 20% to over 40% (Raymond 1979) and the median arrival time at Whitebird trap was later by a week. This is coincident with a change in Tseasn: the model-fitted smoltification date was later, Tseasn = Julian day 160 days (June 9). From 1976 onward the median arrival data of the fish at the Whitebird trap was earlier by one to two weeks and the smoltification date again changed to Tseasn = Julian day 135 to 140 (May 15 to 20).

In 1973 and 1977 fish moved significantly slower than in all other years and the arrival time at dams could not be fit by adjusting Tseasn only. In both these years a good fit to the data was obtained by adjusting the maximum flow independent migration rate _max. In contrast, for all years except 1973 and 1977 _max = 20.2 miles/day. In 1973, to fit the observed arrival dates the parameter had to be reduced to 6 miles/day signifying a three fold decrease in the maximum flow independent component of migration. In 1977 _max was set to _min, which is the initial flow independent migration speed. In the model _min = 1.34 miles/day which infers that the minimum initial migration speed, independent of any contribution of current, would be 1.34 miles/day. This suggests that migration rate was constant and very slow in 1977. The analysis indicates that travel time under these two low flow years was abnormal.

Several biological reasons for slow migrations are possible. Fish in poor condition in 1973 and 1977, due to descaling at dams and high water temperatures, may have been less likely to migrate. Also, a reduced spring freshet in these years may have affected migration speed. An adequate environmental signal required to initiate migration might not have occurred in these two years. As a result, the spring chinook could have drifted through the river system much like subyearling chinook.

Table 74 Spring chinook median Julian day of arrival at dams, observed and model plus the model parameters adjusted to obtain the fit. References (Ref) are given in Table 73
Year	Rls date	Upper dam		IHR		Lower dam		max	Tseasn	Release site1 - Upper dam - Lower dam	Ref.
	Obs	Obs	Model	Obs	Model	Obs	Model
1966	110	124	124	-	-	-	-	20.2	127	WB-IHR-TDL	1
1967	110	125	125	-	-	-	-	20.2	127	WB-IHR-TDL	1
1968	110	126	125	-	-	-	-	20.2	127	WB-IHR-TDL	1
1969	No data							20.2	127	WB-LMO-TDL	1
1970	96 105 110 126 135	127 131 133 139 146	117 123 127 139 145	-	-	-	-	20.2	160	WB-LGS-TDL	1
1971	88 96 105 111 120	121 123 122 125 131	108 112 119 124 132	-	-	-	-	20.2	160	WB-LGS-TDL	1
1972	76 96 111 121 126	114 116 130 133 140	99 112 124 133 136	-	-	-	-	20.2	160	WB-LGS-TDL	1
1973	115	134	132	141	142	156	156	6	140	WB-LGS-TDL	3
1974	115	121	121	126	126	133	135	20.2	160	WB-LGS-TDL	3
1975	117	137	133	141	141	149	149	20.2	160	WB-LGR-TDL	3
1976	99	111	112	118	119	128	127	20.2	135	WB-LGR-JDA	3
1977	108	129	129	146	153	168	168	1.34	130	WB-LGR-JDA	0, 3
1978	104	119	118	125	125	132	133	20.2	135	WB-LGR-TDL	3
1979	108	124	124	131	131	139	139	20.2	140	WB-LGR-TDL	3
1980	105	119	119	-	-	132	133	20.2	135	RR.-LGR-JDA	4
1981	106	1182	121	-	-	134	133	20.2	135	WB-LGR-JDA	5
1982	106	119	120	1283	128	134	132	20.2	140	SC-LGR-JDA	6
1983	103	116	117	128b	127	130	131	20.2	135	SC-LGR-JDA	7

A graphical comparison of observed and fitted arrival times is illustrated in Fig. 75. The points falling below the on-to-one line are arrival data from 1970, 1971, and 1972, which have some early fish release groups. It is likely that for these years fish released prior to Julian day 110 were not ready to migrate and consequently their migration was delayed. For these years the release group in the middle of the migration season had predicted and observed arrival dates within 1 day (Table 74).

Fig. 75 Comparison of observed and model predicted median arrival days at upper dam (k), Ice harbor dam (o) and lower dam (+) for years 1966-1982. (See Table 70 for specific dam)

Survivals were estimated from the CRiSP model once travel times were fit. Model survivals were computed as reported in the corresponding survival studies (Table 73). The equation was

(172)

This equation does not correctly represent the effects of transportation on estimated survivals. Generally, the difference in total system survival would be upwards of 25% if transport were correctly estimated. For example, in 1982 between LGR and JDA dams an in-river survival of 41% using eq(172) becomes 50% when computed as the in-river survival without transportation. The difference is less when percent of fish transported is less. For comparing observed and model survivals eq(172) can be used as long as both observed and model survivals are reported using the same equation.

The resulting model survivals were close to the observed values (Table 75). For the years 1969 through 1983 no additional model parameters were altered other than; 1) fitting the travel time, 2) adjusting the model to the reported dam passage conditions discussed previously, 3) using the observed river temperature and flow. To fit survival between release site and the upper dam for 1966-1969 the predator density above Ice Harbor dam was set at 200 predator/sq km. This represents a density 50% lower than is estimated for the present levels in Lower Granite reservoir.

These survival estimates cover years with a wide variety of project operations and river conditions. Gas supersaturation levels in the 1970's reached 140% in the river and sometimes accounted for half of the fish mortality in passage (Raymond 1979). The model captures this effect.

Between 1966 and 1968 survival was high because the Snake River above Ice Harbor was free-flowing. Between 1969 and 1974 survival was low because of dam construction on the Snake River and the need to spill at the new dams that did not have their full complement of turbines. The year 1973 did not have spill problems but mortality was high because fish migration was anomalously slow so fish experienced high exposure to predators.

The relationship between modeled and observed survivals between release site and the lower dam is illustrated in Fig. 75. A linear regression of 43 data points allowing for an intercept has an r² = 0.806, a slope of 0.88 and an intercept of 6.4%. Confining the regression to a zero intercept the slope becomes 1.004 and the r² = 0.957.

These regressions suggests that the CRiSP model provides a good fit to survival data between the Salmon River and the lower Columbia. Of particular note are the low flow years 1973 and 1977. The observed and predicted survivals in these years were closely fit by adjusting the seasonal flow independent component of migration. This is an indication that the low survivals were a result of changes in fish migration behavior, which were independent of the flow itself. For example, in 1977 survival between Whitebird and John Day dam was 1.1%. If fish migratory behavior was not abnormally slow the survival would have been 7.3%. For comparison survival between Whitebird and John Day in 1983, a good flow year, was 25%. Thus in 1977, flow of its own accounts for a decrease in survival by a factor of 3 while fish behavior accounts for an additional decrease by a factor of 7.

Fig. 76 Comparison of observed and model predicted survival from release site to Ice Harbor dam (k), upper dam to middle dam (0), and upper dam to lower dam (+) for years 1966-1983. (See Table 70 for specific dams). The solid line is least squares regression with a forced zero intercept and the dotted line is a least squares regression allowing for an intercept.

Below Bonneville spring chinook survival

Data giving a minimum estimate of survival below Bonneville Dam was available to evaluate CRiSP in the lower river. Shreck et al. (1994) conducted a radio-tag study on spring chinook in this area in 1994. Fish were released at the Bonneville tailrace and were observed at a monitoring station 86 miles downstream (Fig. 77).

Fig. 77 Release and recapture locations for data used to validate lower river survival of spring chinook.

The survival estimates are "minimal" for several reasons. First, the fish were tagged at Lower Granite and then transported to the Bonneville tailrace. Upon release, some of the fish (2.5-10.0%) were never detected. It was not known whether this was due to tag failure, transportation mortality, or the inability to detect all the fish. Also, it is possible that some of the fish passed the downstream monitoring site without being detected. Also, travel time estimates may be biased downward because only larger fish (average fork length of 139 mm) were used in the study and the monitoring site was only operating for 4 days after release so some of the fish may have passed after monitoring concluded and would not be included as survivors.

The results and model comparison are given in Table 76. Model survivals are 12-18% higher than the "minimal" survivals, and model travel times were 1.2 days longer than reported by Schreck et al. (1994). These results are consistent with expectations, given the sources of bias indicated above.