Processes of migration, growth, and survival of Pacific salmon in off-shore marine environments are poorly understood. Because of logistical difficulties in sampling and experimentation on the high seas, very few directed field efforts have been conducted to try to gain insight into biological mechanisms regulating growth and survival of salmon in this environment. Many of the measures of final ocean weight of salmon stocks from British Columbia suggest a long-term reduction in mean size, perhaps related to climate change (Hinch et al. 1996, Cox and Hinch in review). Many questions of critical importance both to basic science, as well as to developing better practical management policies, need to be addressed to gain a more comprehensive understanding of growth and survival of these and other important stocks during their high seas life history stage.
We contend that much can be gained through reanalysis of past data bases involving extensive biological sampling, coupled with the application of state-of-the-art approaches in statistics, modeling, and data visualization that can offer fresh insights into spatial and temporal patterns in the data. In this paper we describe a series of data bases that have been developed over the past 40+ years and some preliminary results from analyses. We also describe the development of an interactive software shell that allows users to test hypotheses concerning migration and growth processes of salmon on the high seas. We apply the results of the data analyses and modeling exercises to test the following hypotheses: 1) British Columbia sockeye salmon (Oncorhynchus nerka) complete two loops around the Alaskan Gyre as a result of simple behavioral rules of compass orientation and swim speed and estimated off-shore current fields, 2) meso- (100-300 km) and gyre-scale (>1000 km) patterns exist in zooplankton biomass and salmon feeding and growth that arise from physical oceanographic processes, 3) the sharply defined southern limits of salmon biomass is a result of a reduction in growth-rate potential across latitude, and 4) during spring of 1962, the area of highest growth rate potential for salmon can be found in the center of the gyre, resulting from an optimal mix of prey abundance and thermal properties.
We compiled data on sea surface temperatures (SST, COADS database), off-shore current fields (Thomson et al. 1994), zooplankton biomass (LeBrasseur 1965a, Brodeur and Ware 1992), and salmon diets (LeBrasseur 1965b). The latter two data sets were based on archived data records from the Pacific Biological Station at Nanaimo, British Columbia. The period of biological sampling occurred during 1956-64. The sampling was extensive in space (lat. 40-60°N, long. 120-160°W) and time (1957-64, all seasons). We focused data analysis on the biological measures of prey (zooplankton) and predator (salmon) and tested for significant spatial patterns and responses along latitudinal environmental gradients. We expressed the diet data in terms of a measure of stomach fullness (proportion of maximum daily ration), which accounts for body size-allometric effects on feeding rates (Hewett and Kraft 1993). We applied Mantel's tests (Mantel 1967) to detect spatial dependence in the zooplankton and salmon gut fullness data, and developed correlograms using distance class intervals to gain insight into the nature of the spatial patch structure. We also explored the relationship between SST and latitude and these biological measures.
We found evidence of spatial patterns (i.e., data were spatially autocorrelated) in a majority of the zooplankton biomass data sets. There appear to be two consistent scales at which spatial dependence is evident. We detected meso- (100-300 km) and gyre-scale (>1000 km) spatial dependence (Fig. 1). These results are consistent with the hypothesis that meso-scale physical structure (e.g., eddies) may lead to the formation of meso-scale food "patches," while dominant offshore current fields within the Gulf of Alaska may provide a physical template for the distribution of zooplankton at much larger scales of observation.
Analyses of pink (O. gorbuscha) and sockeye salmon stomach fullness during spring and summer of 1962 indicated spatial similarities at a scale of approximately 1,200 km (Fig. 1). The presence of coherence in the data at this scale suggests that predators are responding to the general gyre-scale patterns detected in the prey "landscape." We also found a four- to five-fold reduction in stomach fullness measures between approximately 9°C and 10°C SST for sockeye salmon and between 10°C and 12°C SST for pink salmon. We also discovered a sharp reduction in stomach contents below lat. 50°N, which was coincident with a reduction in zooplankton biomass. We used the results of these analyses to develop temperature- and prey-dependent foraging functions that were applied in the models described below.
The model tracks individual migration and growth trajectories of salmon beginning on 15 August in the vicinity of the Queen Charlotte Islands and is terminated 660 days later (15 June) at predicted endpoints on the high seas. The model is individual-based (IBM) and operates on a 5-day time step at a spatial resolution of 1° lat. by 1° long. The structure of the migration submodel has been described elsewhere (Walter et al. in review). Algorithms to estimate swim speed and bioenergetic losses can be found in a number of publications (e.g., Beauchamp et al. 1989). The model predicts ration and growth based on occupied temperature and prey biomass within each cell in the spatial matrix (Fig. 2). We express growth rates in two different ways. The first is the observed mean and range of projected weights across all individuals in an IBM simulation. The second is the growth rate potential (e.g., Brandt et al. 1992) for a representative 500-g sockeye salmon within each cell in the ocean matrix given the spatial variability in SST and zooplankton biomass, which provides a visual "snapshot" of growth conditions across the Northeast Pacific Ocean.
We tracked the migration trajectory, feeding, bioenergetics, and growth of 200 individual sockeye salmon over 660 days in the coast and high seas environments of the Northeast Pacific Ocean. The model results suggest that salmon are more likely to complete only one loop around the Gulf of Alaska prior to their spawning migration (Fig. 3). In order for individuals in the model to complete two loops around the gyre, it was necessary to invoke directed swimming behaviors that moved salmon in a direction parallel to the dominant gyre currents. We report here the results of the IBM using the former, simpler migration model. In the IBM simulation, variability in occupied temperatures was relatively low, while variability in encounters with prey was highly variable (Fig. 4), resulting from the prey patch structure defined above. Growth rate in the model appeared to conform to a seasonal cycle, with highest rates of growth observed during the spring and summer, and lowest growth rates during fall and winter.
We computed spatially explicit growth-rate potential for a 500-g sockeye salmon in the Northeast Pacific Ocean during four seasons. Growth-rate potential was clearly highest and most spatially heterogeneous during the spring (Fig. 5). The greatest potential for growth during this season appeared to be centered in the mid- to lower-region of the Alaskan Gyre. This central region appeared to correspond to an optimal mix of sufficient levels of prey abundance and preferable thermal properties. A very sharp reduction in growth rate potential is evident in the southern latitudes, where growth rate declines from a high of nearly 2% body weight per day to -0.5% body weight per day along a transect of approximately 300 km. This feature arises from the interaction between levels of zooplankton biomass (and related feeding rates) and temperature-dependent metabolic demands. This may provide a plausible hypothesis to explain the sharp "thermal" limits in the oceanic distribution of salmon as described by Welch et al. (1995). One can also discern an area in the center of the gyre that supports relatively high growth-rate potential for salmon, surrounded by concentric "rings" of lower potential (Fig. 5). These general patterns in growth-rate potential emerge from the results described above identifying the importance of prey abundance and temperature on salmon feeding.
The present model represents the product of collaborative efforts by physical, biological, and fisheries oceanographers, along with expertise in computer software design and data visualization. This effort represents a rare, successful marriage between biology and technology, and will allow new insights into the ecology of Pacific salmon in coastal and high seas environments. These efforts will culminate in explicit predictions in spatial distribution and growth rates that can be rigorously tested with innovative field programs employing new sampling technologies such as acoustics and smart tags (e.g., MacLennan and Simmonds 1992, DeLong et al. 1992). Although much of what was described above addresses issues related to growth and migration, we intend to apply the model to gain insights into mechanisms responsible for regulating mortality, which is the explicit focus of this scientific conference. We have begun to explore the trade-off between growth and mortality in Pacific salmon using techniques of dynamic programming (Scandol et al. in review). If indeed the critical period in the Pacific salmon life history is during the first month or months at sea, we could gain insights into mechanisms of this process by hindcasting growth conditions in past years. There is a substantial literature that supports the notion that mortality is strongly size dependent, and hence growth conditions experienced by juveniles during this period may serve as an accurate predictor of year-class strength. We are presently working with Canada's Department of Fisheries and Oceans and the Pacific Salmon Commission to compile data sets on smolt migration timing, size distribution, abundance, and subsequent survival rates for select salmon stocks. This information, coupled with our simulation model, may allow us to better explain the observed variance in smolt-to-adult survival.
One element shared by most conference presentations was a deep appreciation for the variability that salmon survival exhibits in both space and time. This became clear through comparisons of salmon survival across, for example, northern vs. southern salmon stocks, pre- vs. post-1977 oceanographic conditions, and El Niño vs. La Niña climate scenarios. This argues for future efforts that explicitly integrate effects of both space and time in the analyses, to help explain changes in growth and survival of these important fish populations. We hope the present effort can be successful at accomplishing this goal to help improve conservation efforts and develop better policies for stock management.
Citations
Beauchamp, D. A., D. J. Stewart, and G. L. Thomas. 1989. Corroboration for a bioenergetics model for sockeye salmon. Trans. Am. Fish. Soc. 118:597-607.
Brandt, S. B., D. M. Mason, E. V. Patrick. 1992. Spatially-explicit models of fish growth rate. Fisheries (Bethesda) 17(2):23-31, 34-35.
Brodeur, R. D., and D. M. Ware. 1992. Long-term variability in zooplankton biomass in the subarctic Pacific Ocean. Fish. Oceanogr. 1:32-38.
Cox, S. P., and S. G. Hinch. In review. Size at maturity of Fraser River sockeye salmon: Interannual trends and the potential effects of temperature. Can. J. Fish. Aquat. Sci.
DeLong, R. L., B. S. Stewart, and R. D. Hill. 1992. Documenting migrations of northern elephant seals using day length. Mar. Mammal Sci. 8:155-159.
Hewett, S. W., and C. E. Kraft. 1993. The relationship between growth and consumption: Comparisons across fish populations. Trans. Am. Fish. Soc. 122:814-821.
Hinch, S. G., M. C. Healey, R. E. Diewert, K. A. Thomson, R. Hourston, M. A. Henderson, and F. Juanes. 1996. Potential effects of climate change on marine growth and survival of Fraser River sockeye salmon. Can. J. Fish. Aquat. Sci. 52:2651-2659.
LeBrasseur, R. J. 1965a. Biomass atlas of net zooplankton in the northeastern Pacific Ocean, 1956-1964. Fish. Res. Board Can. Manuscr. Rep. Ser. (Oceanogr. Limnol.) 201:1-260.
LeBrasseur, R. J. 1965b. Data Record. Stomach contents of salmon from the northeastern Pacific Ocean, 1958-64. Circular (Statistical Series), Nos. 15-23, Fish. Res. Board Canada, Biological Station, Nanaimo, BC.
MacLennan, D. N., and E. J. Simmonds. 1992. Fisheries Acoustics. Chapman & Hall Fish and Fisheries Series 5:1-325.
Mantel, N. 1967. The detection of disease clustering and a generalized regression approach. Cancer Res. 27:209-220.
Scandol, J. P., P. S. Rand, F. Juanes, M. C. Healey, and L. Huato. In review. Examining sockeye salmon growth and survival using dynamic optimisation to solve for bioenergetic activity. Can. J. Fish. Aquat. Sci.
Thomson, K. A., W. J. Ingraham Jr., M. C. Healey, P. H. LeBlond, C. Groot, and C. G. Healey. 1994. Computer simulations of the influence of ocean currents on Fraser River sockeye salmon (Oncorhynchus nerka) return times. Can. J. Fish. Aquat. Sci. 51:441-449.
Walter, E. E., J. P. Scandol, and M. C. Healey. In review. Ocean migration patterns of Fraser River sockeye salmon: A reanalysis of accepted patterns. Can. J. Fish. Aquat. Sci.
Welch, D. W., A. I. Chigirinsky, and Y. Ishida. 1995. Upper thermal limits on the oceanic distribution of Pacific salmon (Oncorhychus spp.) in the spring. Can. J. Fish. Aquat. Sci. 52:489-503.