Research Document - 2003/033

Review of the Year 2002 Return of Barkley Sound Sockeye Salmon and Forecasts for 2003

By K. Hyatt, W. Luedke, P. Rankin, J. Till and D. Lewis

Abstract

Returns of Barkley Sound sockeye have exceeded the long term average return of 820,000 adults for the past two years (Fig. 2). Although variable, recent returns reflect the continuation of a pattern of predictable variations in ocean climate and survival conditions for juvenile sockeye salmon that have lead to repeated "crashes" (1978, 1985-86, 1989-90, 1994-95) followed within 1-4 years by recoveries (1979-81, 1987-88, 1991-93, 1996-98) of WCVI sockeye returns. The prolonged interval of sub-average returns between 1994 and 2000 ended as anticipated in 2001 as ocean climate signals suggested increased marine survival rates for Barkley Sound sockeye smolts migrating seaward during the 1998-2000 interval.

Over the past 15 years, four independent techniques have been tested for their utility in generating reliable pre-season forecasts of Barkley Sound sockeye returns for harvest managers. The four techniques are known as the Salinity Survival Method (SSM), the Survival Stanza Method (SStM), the Sibling Age Class Method (SACM), and the Salmonid Enhancement Program Biostandard Method (SEPB). Updates on the performance of three of these techniques in the year 2002 are as follows:

1. The SStM forecast exhibits the best overall performance by far with a mean absolute percent error (MAPE) value of 28% over the most recent 15 years of forecasting (Table 5). SStM forecasts exhibit a statistically significant association with returns among all years (returns in thousands = 1.14.SStM forecast in thousands + 55.19, R2 = 0.45, p<0.01). They also account for the majority of variations in returns if the extreme observation associated with the 1991 return year (when all forecasts techniques underestimated returns) is omitted from the analysis (returns in thousands = 1.07.SStM forecast in thousands + 21.71, R2 = 0.79, p<0.001).
2. The SEPB forecast exhibited the second best performance in predicting sockeye returns in 2002 (MAPE = 28% , Table 4). However, to achieve a statistically significant association with returns among years, the extreme observation associated with the 1991 return year must be omitted from the analysis (returns in thousands = 1.36.SEPB forecast in thousands - 462.57, R2 = 0.49, p< 0.01). Further, during the 1988-2002 testing interval, SEPB forecasts exhibited a much higher MAPE value (58%) than that displayed by SStM forecasts (28%).  Because large magnitude deviations between SEPB forecasts and actual returns tend to occur in consecutive years, it is viewed as having limited utility by both harvest managers and fishermen.
3. The SSM forecast rivalled the SStM forecast in performance between the 1988-1999 testing interval (Hyatt et al. 2000). However, SSM forecasts have overestimated total returns of Barkley Sound sockeye by 490% , 83% and 51% during the years 2000, 2001 and 2002 respectively (Table 4). Thus, SSM forecasts now exhibit a MAPE that averages 68%. SSM forecasts do not exhibit a statistically significant association with returns among all years but do achieve significance if the extreme observation associated with the 1991 return year (when all forecasts techniques underestimated returns) is omitted from the analysis (returns in thousands = 0.35 .SSM forecast in thousands + 317, R2 = 0.28, p<0.05).   Forecasts of Barkley Sound sockeye returns provided by different models vary greatly for 2003. Midpoint forecast estimates range from a low of 894,000 (SEPB model) to a high of 3,219,000 (SSM model) Barkley Sound sockeye of all ages (Table 13a). Comparative performance of the various forecast options, along with DFO's recent pursuit of a more risk averse approach to management recommends initial adoption of the SStM forecast range. Thus, 618,633 (75% probability) to 1,236,150 (25% probability) "adult" sockeye (i.e. excludes "jacks", Table 13b) constitute the preferred, pre-season forecast range for the year 2003. Supplementary information from coho leading indicator observations suggests that returns are likely to be closer to the upper than the lower end of this range.

Complete document (PDF, 185 KB)

Accessibility Notice

This document is available in PDF format. If the document is not accessible to you, please contact the Secretariat to obtain another appropriate format, such as regular print, large print, Braille or audio version.