Modelling salmon migration as a mixture problem

Abstract

Pulses of abundance in salmon migrations can arise from single populations arriving at different times, from multiple populations with different timing characteristics, or as a combination of these. Daily observations typically record an aggregate measure of abundance passing some location rather than the abundances of the individual components. An objective method is described that partitions a compound migration into its component parts by exploiting differences in the characteristics of each pulse. Simulated data were used to demonstrate when greater model complexity may be desirable. Three case studies of increasing complexity (Chilko Lake sockeye salmon smolts (Oncorhynchus nerka), large adult Columbia River Chinook salmon (Oncorhynchus tshawytscha), Fraser River salmon test fishery) demonstrate how the model can be applied in practice. Results indicated that Chilko Lake smolts rarely emigrate to sea as a single pulse, that the dates used to distinguish the spring run of Chinook salmon in the Columbia River may be overestimating its abundance, and that pulses of sockeye salmon abundance in a Fraser River ocean test fishery in 2014 may have arisen from some factor other than population composition.