Abstract
Data for the Canadian fisheries system in Lake Superior were organized into monthly time series of catch and effort from January 1963 through December 1976 for six fish species. Multivariate, autoregressive (ARMA) models were identified for the system based on data for the first 140 mo. Forecasts were compared to data for the last 28 mo. The structure of the models indicate that (1) within the system, AR processes, as opposed to MA processes, were of overriding importance, (2) intraspecific interactions (inferred from data on catch-per-unit-effort, CPUE) were more prevalent than interspecific interactions, (3) interactions within the system occurred with lags of 1, 4, 12, 24, 25, 28, and 36 mo, (4) some of the trophic relationships among the fish species were revealed by the models, and (5) CPUE time series of lake trout (Salvelinus namaycush) affected, but was not affected by, the CPUE time series of other species. The models were used to forecast catch and CPUE for the last 28 mo, and the data were generally within one standard error of the forecasts. The models may help policy decision makers to explore the effects of inputs (e.g. quota regulations) and feedbacks within the fisheries' system on outputs (e.g. production, CPUE).

This publication has 15 references indexed in Scilit: