WCVI Salmon Bulletin 2017 Area 23 Sockeye Forecast April 3, 2017 SUMMARY
There was an abundant return of about 1.1M Somass sockeye (Great Central and Sproat Lake stocks) in 2016. The majority of the abundance was 52 year olds produced from the record 2011 brood year escapement that experienced above-average marine survival for the associated 2013 sea entry year. The return of Henderson Lake sockeye in 2016 was moderate; estimated at about 34,000 adults. Similar to the Somass stocks, the majority of the return was associated with the 2011 brood year. For 2017, the recommended management forecast for Somass sockeye is the “critical” zone for harvest management; corresponding to an expected return of less than 200,000 adult fish. Key factors in the sharp decline of expected abundance relative to recent return years are i) very low observed smolt production and ii) relatively low marine survival rate for the 2014 and 2015 key sea-entry years associated with this year’s adult return. For 2017, the recommended management outlook for Henderson sockeye is in the “very low zone” for harvest management; corresponding to an expected return of less than 15,000. Similar to Somass sockeye, key factors in the low expected abundance for 2017 are the decline in the marine survival rate associated with recent sea-entry years and relatively low smolt production in 2015. BACKGROUND Three sockeye stocks return to Barkley Sound (Area 23) annually: the Great Central Lake, Sproat Lake and Henderson stocks. Status of each stock is assessed as a separate Conservation Unit (CU) for implementation of Canada’s Wild Salmon Policy. From 1980 to 2016, the average terminal returns (catch and escapement) of Great Central, Sproat and Henderson Lake sockeye are 400,000, 335,000 and 30,000, respectively. For fishery implementation and forecasting purposes, the Great Central and Sproat Lake stocks are aggregated as one stock management unit referred to as Somass sockeye. With the exception of Maanulth Treaty fisheries, Area 23 sockeye fisheries target Somass sockeye while managed to limit impacts on the Henderson stock. The pre-season management forecast for Somass sockeye guides early-season fishery planning and effort-limited harvest plans in June (Table 6). The run size forecasts are revised in late June based on the evaluation of in-season stock assessment results. The first in-season forecast revision is anticipated no earlier than June 29, 2017. Statistical forecast models for Henderson sockeye are not generated due to data limitations. A heuristic outlook is produced based on spawner and smolt abundance and indicators related to marine survival rate for the contributing brood years. This outlook determines the amount and timing of commercial gillnet openings in outer areas of Barkley Sound where the fishery is more likely to intercept Henderson sockeye (Table 7). FORECAST Methodology Four models have been used to forecast Somass sockeye returns for between 14 and 25 years with relative accuracy. These models include: the Survival Stanza Method (SStM), Surface Salinity Method (SSM), Salmonid Enhancement Program Biostandard Method (SEPB), and Coho Leading Indicator Method (CLI). Three of the four models (SStM, SSM, SEPB) use annual estimates of the numbers of smolts from Great Central and Sproat Lakes and correlates of early marine survival to predict returns. The CLI model uses marine survival estimates for coho salmon from corresponding sea-entry years to predict return from average returns per adult sockeye spawner. In general, the SStM and CLI forecasts have provided the most accurate forecasts over the long term. More recently, a “sibling” model has been developed that uses relationships between the returns at age from past brood years to predict future returns from earlier returning age classes. The forecasts generated from all methods are compared and based on their correspondence, relative accuracy at predicting past returns, and other relevant information (e.g. marine observations) a heuristic management forecast is produced to guide early season fisheries. This forecast sets pre-season expectations and guides early-season harvest planning. 2016 Return and Forecast Performance There was an abundant return of about 1.1M Somass sockeye (Great Central and Sproat Lake stocks) observed in 2016 (Table 1, Table 5, Figure 1). About 82% of this return resulted from the abundance of 5 year olds that smolted during the 2013 sea entry year. Ocean conditions for this sea entry year were favorable promoting an above average marine survival rate of about 10%. However, marine survival rate alone does not explain the abundance. The 5-year old return which contributed to 82% of the overall production in 2016 also results from the record high spawner abundance in the corresponding 2011 brood year. The Sibling and CLI forecast models were the best performing models, overestimated the return by 11 and 12%, respectively. All other models under-estimated the final run size (Table 2). The return of Henderson Lake sockeye in 2016 was moderate and estimated at about 34,000 (Table 1, Table 5 Figure 2). The pre-season outlook was for a very low return of less than 15,000 sockeye. Preseason expectations were based on the observation of relatively few 4-year olds in 2015 (brood year 2011). However, the majority of the return in 2016 was 5-year olds associated with this brood year. Therefore, the survival rate for the 2013 sea entry year was apparently higher than initially estimated, although still much lower than observed for the Somass stocks. 2017 Somass Sockeye Forecast The age of return for Somass sockeye ranges from 3 to 6 years with age 4 and 5 fish predominant. Sockeye produced from brood years 2011 to 2014 will return in 2017. Model forecasts for the Somass return range from a low of about 136,000 (SStM) to a high of about 511,000 (CLI Forecast Model, Table 3). The ‘sibling model’ forecast return for 2017 is 172,000. On average, this model has performed the best in recent years. The sharp decline of expected abundance relative to recent return years is explained by the following factors. One is the low level of smolt production observed in Sproat Lake in 2014 and the very low level observed for both stocks for 2015 (Figure 3). A second factor is an apparently low marine survival rate for those same sea-entry years (Figure 4). Correspondingly, relatively low numbers of “jacks” (or age 32) fish were observed from these broods so far (i.e. in both the 2015 and 2016 return years). The highest forecast prediction of 511,000 (CLI model) is discounted because that model does not incorporate observations of juvenile production1 . Therefore, for fishery management purposes, the recommended 1 The very low level of juvenile production observed in 2015 was apparently associated with higher than usual freshwater mortality during the over-winter period of lake rearing prior to sea-entry. Earlier fall surveys showed levels of juvenile abundances more commensurate with expectations given spawner abundances in the main associated brood year 2013 (GCL – 66K, SPL – 118K). management forecast is precautionary and in the “critical” for harvest management; corresponding to an expected return of less than 200,000 adult fish The expected Somass sockeye stock composition is about 37% Great Central and 63% Sproat Sockeye. The expected Somass sockeye age composition is about 50%, 37%, 7%, and 6% of age 42, 52, 53 and 63 adults, respectively. 2017 Henderson Sockeye Outlook The recommended management outlook for Henderson sockeye is the “very low” zone for harvest management, corresponding to an expected return of less than 15,000 (Table 4). The key factor influencing this outlook is the apparently low marine survival rate experienced by the two brood years (2012, 2013) that contribute to the 2017 return. Also, similar to Somass sockeye, the juvenile production from the 2013 (2015 sea entry year) was below average. SOURCES OF UNCERTAINTY The mean absolute percentage error (MAPE) for the forecast models used to predict Somass sockeye range from about 38 to over 100% for the years when the models have been applied, with best performing forecasts averaging about 38%. That is, on average, the observed return is about 38% higher or lower than the predicted return. Factors that contribute to forecast uncertainty include, but are not limited to: model structure, uncertainty associated with model inputs (i.e. source data), etc. For the Henderson sockeye outlook, there is considerable uncertainty due to lower quality assessment data relative to the Somass stocks. There are less complete age data, relatively high uncertainty in the estimates of spawner abundance, and also uncertainty in catch estimates. Catch estimates are particularly uncertain in recent years when the abundance of Henderson sockeye is very low relative to the Somass stocks. Under these circumstances, the probability of detection of Henderson sockeye in catch samples is lower therefore catch of Henderson sockeye may be underestimated.