Investigators have spent considerable effort exploring the basis of chinook salmon (Oncorhynchus tshawytscha) return variability. All studies (e.g. Sharma et al. [1], Miller et al. [2], Duffy [3], Scheuerell and Williams [4], Woodson et al. [5]) have implicitly focussed on the effects of food availability on size, with mortality resulting from predation (mis-mismatch hypothesis, see [6]; critical period hypothesis, see Cushing [7]; bigger is better hypothesis [8]) or the inability to store adequate energy reserves to survive the first winter at sea (critical size hypothesis, see [9]). These investigations used a variety of physical oceanographic measures as proxies for food. In addition, there are instances where size at ocean entry, influenced by freshwater or hatchery rearing, are also considered. The results suggest that the mortality during the early marine life history determines return variability.

The results of these studies are of concern to us for two reasons. The first one is that there is no empirical basis for relating physical oceanographic conditions directly to food availability. There are correlative relationships between physical oceanographic measures and zooplankton biomass (see Peterson et al. [10] for examples) but Tanasichuk and Routledge [11] and Tanasichuk et al. [12] found that juvenile salmon can select very specific parts of the zooplankton community as prey and correlative relationships at such fine taxonomic size scales have not been calculated. Wal- ters (cited in Coronado and Hilborn 1998 [13]) warned that the correlations between ocean condition and Oregon Production Index Area (OPI)coho (O. kisutch) production were spurious and consequently of no long term predictive value; the relationship between upwelling and OPI coho return described by Nicholson [14] has broken down. Myers [15] revisited 48 analyses that described correlative relationships between measures of aquatic conditions and recruitment and found that only one, for a fish population in a lake in Ontario, persisted. Scheuerell and Williams [4] noted that the correlations between North Pacific salmon production and oceanographic variables rarely lead to accurate predictions because the correlations degrade over time. Our second concern is about the statistical methodology. In all instances, there are a number of factors being tested and Bonferroni adjusted probabilities (see [16]) are not used to minimize the likelihood of committing a Type I experimental error. Tovey [17] is exceptional by her use of Bonferroni adjusted probabilities.

Fig. (1). Alberni Inlet and Barkley Sound showing fishing locations used to learn about hatchery chinook migration timing, distribution and diet, and zooplankton sampling stations used to monitor prey availability. A - Amphitrite Point lighthouse. Robertson Creek Hatchery is 2 km downstream of the outlet of Great Central Lake.

We took advantage of multi decadal time series of information about rearing and release practices for chinook salmon from the Robertson Creek Hatchery, results of an investigation of the early marine life history of chinook from the Hatchery [12], and a multi decadal time series of prey availability and predator biomass/abundance during the early marine life history (the first marine summer), to test a number of null hypotheses regarding release specific return variability for the commonly returning ages of fish. We chose survival rate, the proportion of fish released that return or are caught at a given age, as the response variable because it is the product of the actual survival rate and the maturation rate; this survival rate is an apparent one. Variation in age at maturity may contribute to age specific return variability. Hankin et al. [18] found that age at maturity was heritable for males and females but suggested some plasticity as a consequence of growth variation. Wells et al. [19] reported that growth during the year before the return year affected the rate at which age 4 chinook from northern California matured. The apparent survival rate we used gave us the opportunity to consider survival and maturation simultaneously. The null hypotheses we tested were that there were no fish size at release, rearing density, the timing of release, the number of fish in a given release, the total number of fish released, and prey biomass and predator biomass/abundance during the early marine life history on the age specific survival rate of ages 2 through 5 chinook salmon from releases from the Robertson Creek Hatchery. We also took advantage of the opportunity to test the efficacy of physical oceanography measurements to explain variation in apparent age specific survival rate.

We found that Pacific mackerel and sea lion abundance variability accounted for all of the explained variation in the apparent age and pond specific survival rate of ages 2 through 5 chinook salmon released as age 0+ smolts from the Robertson Creek Hatchery. The VIF’s indicated that the independent variables were not correlated. There was one outlier each for the initial regression tests of the effects of Pacific mackerel and sea lion abundances on logit survival rate of ages 4 or 5 fish; these were excluded before the solutions were calculated. The statistics for the regression analyses are presented in Table 1. Sea lion abundance at the NWCVI haul out sites accounted for 65, 74, 79, and 71% respectively of the explained variation in apparent survival rate for ages 2 through 5 Robertson Creek Hatchery chinook. Plots of predicted age-specific survival rates against observed pond specific rates are shown in Fig. (3).

Fig. (3). Observed (open circles) and predicted (closed circles) survival rates.

There were no statistically significant solutions when we repeated the regression analyses using physical oceanographic measurements as explanatory variables.

Our tests of hatchery practice, prey and predator abundance/biomass effects suggested that return variability for ages 2 through 5 chinook from the Robertson Creek Hatchery is a consequence of predation with no effect of hatchery rearing or release practices, or prey variability. We note that, based on the results of a cluster analysis of survival rate (see [13]), Robertson Creek Hatchery chinook survival rate trends differ from those of Washington and Oregon populations that have been studied; there are no previous studies of survival variation for other Canadian chinook populations. The inability to explain survival rate variability using the physical oceanographic variables suggests that these variables are not accurate proxies, or reflect different processes for other chinook populations. The recurring conclusion of other studies (e.g. [5]) is that there is a relationship between local SST and survival for Pacific Northwest (Oregon to Alaska) chinook populations. It would be interesting to revisit previous explanations of return variability which are based on physical oceanographic variables to see if the results persist or fail as described in [15].

This study is the most comprehensive analysis of hatchery chinook return variability to date because it considers hatchery and early marine biological effects explicitly and simultaneously. Quiñones et al. [29] investigated the effects of hatchery release practices and some freshwater and oceanographic conditions on the return of salmonids to the Klamath River Basin, California. They reported that fall chinook returns were affected mostly by hatchery practices and that the spring and fall chinook populations were also affected by climatic effects (upwelling, summer river flows). Somewhat related, Haeseker et al. [30] noted that freshwater and marine survival rates of Snake River chinook were correlated.

We are intrigued by finding no effect of hatchery rearing or release practices because it contributes to the ongoing discussion about whether the hatchery affects return; as shown in Fig. (2), rearing and release conditions varied substantially during our study period. Results of previous experiments of the effect of time and size at-release (e.g. [31, 32]) showed that survival was affected more by release date than release size. However, Irvine et al. [33], based on 25-27 years of data, concluded more recently that the declining trend in survival of Strait of Georgia coho could not be reversed by manipulating release time and/or size as suggested by the results of the aforementioned short-term experiments. These authors also reported mixed effects of time and size at release among the five hatchery populations they examined. The results presented in [33] are unique because studies of hatchery effects on return are generally short term ones. Mathews and Ishida [34] using experiments conducted at a Columbia River hatchery for one brood year and a southern Oregon hatchery in the next brood year, found that the effect release size on coho survival were mixed. Burgess and Skalski [35] explored the basis of survival variability for fall chinook salmon released from 10 hatcheries along the Washington or Oregon coasts. They used coded wire tag release and return data for between 11 and 20 years. The authors reported mixed results for the effect of weight at release on the survival and that survival rate increased with date of release until September. Chittenden et al. [36] found that smolt to adult survival of coho released from the Quinsam River Hatchery, British Columbia peaked when fish were released at the time of the seasonal peak of zooplankton abundance. Duffy [3] reported that marine survival of coded-wire tagged release groups of juvenile chinook in Puget Sound was dependent on size in July of the first marine year so declines in marine survival since the 1980’s could be a result of reduced food availability in May and June; she found that release size and release date had a minor positive effect on survival. Duffy noted that the lack of an effect of fish size at release may be due to the small range (5-11 g) of mass at release. Quinn et al. [37] reported that, based on release sizes ranging between 3 and 20 g, intra-annual variations in marine survival were greater for heavier fish but fish size did not explain inter-annual survival variations. Tomaro [38] found, over an eight year study, that the return of adult mid upper Columbia River spring chinook was best explained by size of yearling smolts at ocean entry. Banks and LaMotte [39] found that rearing density had a minor, inconsistent effect on smolt to adult survival. Olsen and Paiya [40] reported mixed effects of lower density on adult survival; however they thought their low, medium and high experimental densities would all be considered to be low rearing densities. They, and [41], suggested that lower rearing densities enhanced survival when ocean conditions were poor. Bilton [32] reported a strong and positive curvilinear relationship between mass at release and return for chinook reared at the Big Qualicum Hatchery in the Strait of Georgia.

We found that no hatchery practice we tested explained variation in age at return. Ricker [42] reported that age at maturity of chinook is highly heritable and that age at maturity of males and females is, to some extent, determined by the sex of the parent. Hankin et al. [18] found that, in wild chinook experiments at the Elk River Hatchery, Oregon, heritability in age at maturity was relatively high, age at maturity of females was independent of age of the male parent, faster growing progeny for a given parental age matured at younger ages, and that size-at-age of progeny from older parents is not smaller. They concluded that age at maturity is not a function of growth rate. They proposed that there are heritable minimum threshold lengths that differentially trigger maturation as a consequence of parental age and sex. Wells et al. [19] explored environmental effects on at maturity of age 4 Northern California chinook. They found that accelerated growth in the third rather than the fourth year of life, as a consequence of physical oceanographic factors (negatively with spring sea level height, SST and scalar winds, and positively to curl, upwelling and northerly wind stresses), increased the likelihood of fish returning as 4 year olds. These authors stressed that their results do not imply cause and effect. Sharma [43] reported no link between age at maturity and ocean conditions for 23 populations of Pacific Northwest chinook salmon.

We conclude that predation explained the variation in return variability of ages 2 through 5 chinook salmon to the Robertson Creek Hatchery and there was no effect of the aspects of hatchery practice we tested on survival rate or age at maturity. Physical oceanographic variables did not provide an explanation for survival variation. We suggest that return variation of Robertson Creek Hatchery chinook is under biological control and hatchery practice modifications would have no influence on fish production.