In-season Forecast of Chum Salmon Return Using Smoothing Spline

Abstract

We developed an in-season forecast model of return of chum salmon for the population off the Honshu region in the Sea of Japan using the smoothing spline based on catch data obtained in fishing season. The optimal in-season model was constructed using adult return in season 8 (middle October) as an explanatory variable. Residual sum of squares of the optimal in-season model was lower than that of the pre-season forecast (sibling) model, indicating the former was more accurate than the latter. The relationship between forecast error rate in the optimal model and the cumulative proportion of return until season 8 (middle October) was positive. Yearly variation in the forecast error rate may be affected by variability in the timing of return. We provide a new and accurate forecast model of chum salmon return.