From data to forecast: methods for estimation of the abundance of chum salmon runs in Kamchatka Territory

This paper presents mathematical models used to forecast the abundance of chum salmon runs in Kamchatka Territory. The models are classified into two groups, each characterized by specific approaches to data analysis: time-series extrapolation and regression-based models. The first group includes two models: an exponential smoothing model (ETS) and an autoregressive moving average (ARIMA) model. The second group includes models based on an analysis of the relationship between chum salmon recruitment and the abundance of juvenile escapement and feeding underyearlings, pink salmon year-classes, and climate indices during the juvenile downstream migration from spawning rivers to the coastal part of the sea and the autumn migration to open sea waters, as well as the abundance of older age cohorts versus younger cohorts. The implementation algorithm for the applied models is presented as R scripts.