A hierarchical model for salmon run reconstruction and application to the Bristol Bay sockeye salmon (Oncorhynchus nerka) fishery

Abstract

The goal of spreading the annual catch of a Pacific salmon (Oncorhynchus spp.) run proportionally across all segments of the migration is rendered difficult or impossible because of the interannual variability in both run size and run timing. This problem is particularly acute in the case of the fishery for sockeye salmon (Oncorhynchus nerka) in Bristol Bay, Alaska, for which traditional run reconstruction models are not applicable because of the extreme temporal compression of the run. We develop a run reconstruction model appropriate for sockeye salmon in Bristol Bay by accounting for the hierarchical structure of the problem and by including process error. Our results indicate that the hierarchical structure is, in fact, not necessary, whereas the process error parameters are needed to fit the data. We suggest further model development without the hierarchical structure, including incorporating in-river test fishing data. The results of our method can be used to address questions regarding environmental or intrinsic drivers of run timing and the possibility of artificial selection on run timing.

Coho salmon escapement and trends in migration timing to a data-poor river: estimates from a Bayesian hierarchical model

“…In data-rich systems, it is possible to obtain constrained estimates of escapement model parameters from single year count data (Flynn et al 2006). However, in data-poor systems separate estimation of parameters for individual year's results in wide confidence intervals (Hilborn et al 1999).…”

Evaluating the Performance of Two Salmon Management Strategies using Run Reconstruction

“…The number of days the fish spent migrating from the fishing grounds to the tower was assumed to be constant and was estimated as 2 d for the Egegik River (ADFG estimate from Flynn et al 2006) and 9 d for the Togiak River (Brannian 1982). The average number of days the fish spent pooling on the fishing grounds was estimated to be 3.6 d for the Egegik fishery.…”

“…The average number of days the fish spent pooling on the fishing grounds was estimated to be 3.6 d for the Egegik fishery. Pooling time is thought to decrease throughout the season Flynn et al 2006). We initially used an average pooling time of 2 d for the Togiak fishery (Brannian 1982), but because a shorter pooling time resulted in a better model fit, we used a pooling time of 1.5 d for the Togiak River run reconstruction (see Results).…”

“…The resultant daily harvest rate equation chosen for the run reconstruction (Tables 1, 2; see Results) assumed a sequential harvest in which the drift gill-net fleet harvested fish first and the set gill-net fleet harvested from the remaining fraction. The daily harvest rate was calculated aŝ Starr andHilborn 1988 andFlynn et al 2006) used a maximum likelihood approach. All fish that arrived at the fishing area on a given day were treated as a single unit (cohort) that was separate from fish arriving on other days; we used a daily run reconstruction because daily to weekly run reconstructions have been shown to be more reliable than annual reconstructions (Starr and Hilborn 1988).…”

“…When the mean number of days spent pooling was not a whole number, a spread of 2 d was used to allow the average time spent pooling to be a fraction of a day (Flynn et al 2006). In this case, the proportion of fish leaving the fishing area was calculated aŝ…”