12.07.2015 Views

Fraser River Sockeye Fisheries and Fisheries Management - Cohen ...

Fraser River Sockeye Fisheries and Fisheries Management - Cohen ...

Fraser River Sockeye Fisheries and Fisheries Management - Cohen ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Forecasts for Bowron, Pitt, Chilko, <strong>and</strong> Stellako have been particularly poor, havingexplained only 8.7%, 0.4%, 9.1%, <strong>and</strong> 9.3% of return variation in the past 30 years. Thisis especially alarming for Chilko because this group contributes (on average) about 24%of the total <strong>Fraser</strong> return.The recent error in <strong>Fraser</strong> sockeye forecasts <strong>and</strong> recognized challenges with forecastingsalmon returns have led most managers to rely on in-season information to managesockeye fisheries. Fishermen <strong>and</strong> processors want reliable forecasts so they can makebetter business decisions, but most recognize that they should not make majorinvestments based on annual run size forecasts.Review of Pre-season Forecasting MethodsThe data sources <strong>and</strong> alternative methods (models) used to derive pre-season forecasts foreach of the 19 indicator stocks are fairly complex <strong>and</strong> described in detail in Cass et al.(2006). The suite of alternative models tested for each indicator stock depends on thetypes of data available for that group.Table 19 provides a list of the various models from Cass et al. (2006) that have been usedin the pre-season forecasting process from 2007 to 2010 (CSAS, 2006; 2007; 2009;2010). Most of the models use estimates of total returns, brood year spawners, or returnsper spawner available for each of the 19 indicator stocks; however, some models requiredata on fry or smolt abundance that are only available for a few stocks (e.g., Chilko,Cultus, Nadina, Weaver).The models range in complexity from simple, such as naïve models (i.e., simple timeseries models without explicitly modeled mechanisms), to the more complex, such asRicker stock-recruitment models that include environmental factors in their calculations.Cass et al. (2006) provides details on data sources, model equations, the retrospectiveanalysis approach, <strong>and</strong> measures used to assess the performance of the 14 models usedprior to 2008. Brief descriptions of the five additional models tested in 2009-2010 areprovided in (CSAS 2009, 2010). Most of the data required for these models are fromPSC databases.Four of the five models added in recent years use estimates of stock productivity, definedin terms of recruits per spawner (R/S), where recruits are the number of adults returning(catch plus escapement <strong>and</strong> “en-route losses”) for a specific stock <strong>and</strong> spawners are thenumber of “effective female spawners” in the brood year for that stock (i.e., female64

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!