SHORT-TERM FORECASTING OF THE STRIPED BASS (MORONE-SAXATILIS) COMMERCIAL HARVEST IN THE MARYLAND PORTION OF CHESAPEAKE BAY

Using the discrete time series of the Maryland striped bass juvenile indices and commercial harvests, 1954-1984, the classic regression model, the time series regression model, the univariate ARIMA model and the transfer function noise model were evaluated for forecasting the commercial harvests of striped bass in the Maryland portion of Chesapeake Bay. The residual terms of the classic regression model were autocorrelated and therefore the model was statistically inappropriate as a forecasting tool. The transfer function noise model and the time series regression model had a better simulation fit (smaller root mean square error) than the univariate ARIMA model. They can both be used for forecasting, but the forecast error can be expected to be high, averages of 84.56% and 80.74%, respectively, in terms of the mean absolute percent error. Improving the precision of the juvenile index estimates and reliable fishing effort data are the keys to forecasting success.