A new in-season forecast density of anadromous fish return abundance

Abstract

We developed a new probability density function (pdf) for an in-season forecast of anadromous fish return abundance, given two kinds of data: (i) historical run sizes by day and (ii) daily in-season run sizes. The new pdf, which we call “inverted-beta”, can be generally applied to an in-season forecast of anadromous fish return abundance regardless of population and setting, and can be part of an integrated stock- and age-structured model for an in-season abundance forecast made with all available information that includes fishery catches, escapements, and stock/age compositions in catches and escapements. The merit of the inverted-beta forecast model lied in the reflection of the annual differences in weightings of historical run proportions at day. Using actual data on sockeye salmon (Oncorhynchus nerka) runs to a district in Bristol Bay, Alaska from 1996 to 2020, we made hind-casting forecasts to examine the forecast performance. The inverted-beta forecast model outperformed the other candidate models in both the bias and the precision of forecasts especially during the early part of the run season when forecasting was challenging because of a lack of early in-season data. Also, the performance of the inverted-beta model was less sensitive to different time series of historical data in the hind-casts than the other models.