Ongoing modeling studies examine potential impact of interannual
changes in circulation on survival of larvae in the western Gulf of
Alaska. Physical factors that pose challenges include complex
bathymetry with many islands, mesoscale (
20 km) meanders and
eddies, strong vertical shear (estuarine-like flow), and strong
forcing by winds and freshwater runoff. The circulation model we use
is based on the Semispectral Primitive Equation Model (SPEM) of
Haidvogel et al. [1991], modified for this region [ Stabeno et
al., 1995]. A total of 
250,000 gridpoints span the model
domain, and a typical simulation of spring and summer months entails
10
time steps. The model code is presently designed to
take advantage of vector-processing computing architectures.
Thus far SPEM has reproduced the observed general spatial features of circulation [ Stabeno et al., 1995]. A comparison between model output and measured currents yielded reasonable agreement. The model also generates eddies with similar spatial scales to those observed. Results from SPEM show that during 1978 (the strongest year class) larvae were more likely transported into coastal waters along the Alaska Peninsula, while in 1990 (a below average year class) they remained in the sea valley where currents then result in transport offshore [ Stabeno et al., 1995]. This latter scenario implies loss of recruits. These results support the original transport hypothesis.
SPEM is being coupled to a spatially explicit, individual-based, probabilities model (IBM) of egg and larval development. The IBM has distinct advantages over more traditional approaches that consider only the ``mean'' individual [ Huston et al., 1988]; since it follows the unique life history of each fish, the IBM approach yields specific information about survivors. The model employs a spatial tracking algorithm for each individual, that includes vertical migration according to life stage. Horizontal transport, growth, and behavior are governed by velocity, salinity and temperature fields generated by SPEM. Low-pass filtered velocity and scalar fields from SPEM are stored once per model day, then used as input for multiple runs of the biological model. The model-generated spatial distributions qualitatively compare favorably with observed distributions of larvae and juveniles. Interannual differences in wind and freshwater runoff lead to differences in the modeled spatial paths of individuals, and in the distributions of population attributes (e.g., growth).