1. Exploring sensitivity of the relationship between salmon abundance indices with killer whale demographics Eric Ward, Chuck Parken, Robert Kope, John Clark, Antonio Velez- Espino, Larrie LaVoy, John Ford

2. Context Previous analyses (Ford/Wade/Ward) reached different conclusions (p 19 Sci Panel Rept) Workshop 1 results (FRAM) somewhat different New indices (Parken-Kope) presented – how do they compare?

3. Checklist CTC / FRAM comparison [high, p 15] Alternative approaches to FRAM [medium high, p 18]

4. Recap from workshop # 1 We evaluated data support for several salmon indices correlating with killer whale fecundity / survival FRAM (1983-2008) – Inland waters abundance – Coastal abundance – Inland kcals – Coastal kcals CTC indices (1979-2010) – Aggregate and stratified indices – e.g. inland, coastal, etc (E. Ward & L. La Voy)

5. Results from workshop # 1 FRAM inland abundance positively correlated with kw fecundity P(birth) = f(age) + f(salmon) FRAM model was slightly better than no salmon model, model averaging used (reduces correlation) But negative correlation with survival P(live) = f(stage) + f(pod) – L pod has lower survival

6. Recap of FRAM results FS

7. Recap of CTC results (WCVI & NBC) Stronger correlation for survival than fecundity (opposite of inland FRAM result) CTC WCVI CTC NBC SS FF

8. Objective for working group Link run reconstruction indices from workshop 1 (Parken/Kope) with kw demographic models – Alternative to CTC and FRAM indices – Less of a black box Counting fish: more empirical – Each index is composed of a terminal component and fishery impact component

10. Hypotheses to address How do new indices compare to CTC / FRAM? Is kw growth (survival, fecundity) more correlated with inland or ocean salmon abundance? – Inner v Outer, like FRAM coastal v FRAM inland Inland abundance (after ocean fishing occurs) – Only exclude ocean pre-terminal catch from inland migrating stocks (Fraser, Puget Sound) – Ocean fishing impacts from coastal stocks included Ocean abundance (terminal run size + fishing impacts)

11. Caveats We have to assume that the relative importance of a group of stocks is constant over time – More likely this is the case than if we use individual stocks Analyses with covariates are purely correlative – Sci panel report (p5): How do we determine if there is cause and effect? Without other data, we cant model the mechanism

12. What this comparison isnt This is not a comparison of fishing v no fishing – (next presentation: how much fishing impacts kw growth rate and ability to meet recovery criteria) – Fishing impacts sliced up into coastal / inland

13. Stratification 21 stocks, 2 stratification approaches considered: Ocean Distribution: north, central, California (south) Migration timing: spring, summer, summer/fall, fall

14. Overview of groupings for Parken-Kope time series For time period 1979-2010, we explored: 9 ocean distribution configurations 10 migration timing configurations KW Demographic models: SR fecundity ~ logistic regression (salmon = t-1) SR survival ~ logistic regression (salmon = t)

15. STOCKMIGRATION TIMING OCEAN DISTRIBUTION Col Lower R springSpringCentral Fraser Sp. 1.2SpringNorth/Central Fraser Sp. 1.3SpringNorth/Central SEAKSpringNorth Upper Col R springSpringCentral Areas 1-5SummerNorth Areas 6-10SummerNorth Col R summerSummerCentral Fraser Sm. 0.3SummerNorth Fraser Sm. 1.3SummerNorth Puget SoundSummer/FallCentral Upper Georgia StraitSummer/FallNorth Col R TuleFallCentral Col River bright fallFallNorth Fraser LateFallCentral Lower Georgia Strait H & NFallCentral Klamath fallFallSouth OR coastFallNorth Sacramento fallFallSouth WA coastFallNorth WCVIFallNorth

16. Not all fishery impacts included Some fishing on occurs in freshwater / near mouth of Columbia – E.g. exclude impacts for Columbia spring

17. Ocean Distribution Strata are north / central / California (south) Fraser spring 1.2 & 1.3 can be considered north OR central (we tried both) Inland pre-terminal fishery impacts (Fraser, Puget Sound) might be important or not (we tried both)

18. Outline for results Stratification by ocean distribution – 1. Fecundity (hidden slide) – 2. Survival Stratification by migration timing – 3. Fecundity (hidden slide) – 4. Survival

19. Results: SR fecundity, Chinook strata = ocean distribution Effect slightly lower when inland fishery impacts (Fraser) not included Fraser Sp = N, Ocean Fraser Sp = N, Inland Fraser Sp != N, Ocean Fraser Sp != N, Inland Fraser Sp = C, Ocean Fraser Sp = C, Inland Fraser Sp != C, Ocean Fraser Sp != C, Inland California

20. Results: SR survival, Chinook strata = ocean distribution Effect slightly higher when inland fishery impacts (Fraser) not included Fraser Sp = N, Ocean Fraser Sp = N, Inland Fraser Sp != N, Ocean Fraser Sp != N, Inland Fraser Sp = C, Ocean Fraser Sp = C, Inland Fraser Sp != C, Ocean Fraser Sp != C, Inland California

21. Migration timing Strata are spring, summer, summer/fall, and fall – Puget Sound / Upper St of Georgia = summer/fall Inland pre-terminal fishery impacts (Fraser, Puget Sound) might be important or not (we tried both)

22. Results: SR fecundity, Chinook strata = migration timing Ocean Inland Ocean Inland Ocean Inland Ocean (+ California) Inland (+ California) Ocean (- California)

23. Results: SR survival, Chinook strata = migration timing Ocean Inland Ocean Inland Ocean Inland Ocean (+ California) Inland (+ California) Ocean (- California) Inland (- California)

24. Summary of stratification SR fecundity: Mostly 0, except Summer/fall (+) SR survival: North / Central (+) Effect of all seasons mostly (+) Consistency with CTC (WCVI & NBC): Stronger correlation for survival than fecundity, positive correlation for both

25. What about model selection? How do new indices stack up to results from workshop 1? Caveat: FRAM is 1983-2008, so all time series need to be shortened Caveat # 2: AIC values are interpreted in log- likelihood units (lower is better). – Its difficult to discriminate models < 2 – Models within < 5 should still be considered good candidates – Models > 10 can probably be ignored or given little weight

26. Model selection results, 1983-2008 (SRKW) AIC values: 0 indicates best model, values greater than 0 indicate models that arent as good. Models with values > 10 can probably be ignored AIC values for fecundity / survival can be added together for total data support Group Include inland pre- terminal delta AIC (fecundity) delta AIC (survival) delta AIC (total) North (+Fraser Sp)Y3.43.82.1 North (+Fraser Sp)N5.11.5 North (-Fraser Sp)Y5.03.43.3 North (-Fraser Sp)Y5.10.9 Central (-Fraser Sp)Y0.06.41.3 Central (-Fraser Sp)N4.76.96.5 Central (+Fraser Sp)Y2.86.44.2 Central (+Fraser Sp)Y4.76.66.2 SouthNA5.421.722.0 SpringY4.611.811.2 SpringN4.711.711.3 SummerY5.321.421.7 SummerN5.018.718.6 Summer/fallY1.816.413.1 Summer/fallN2.916.714.5 Fall (+California)Y5.03.63.5 Fall (+California)N5.34.44.5 Fall (-California)Y4.71.30.8 Fall (-California)N5.10.0 FRAM #s inlandNA1.421.017.3 FRAM #s coastNA5.215.8 FRAM #s totalNA3.819.818.5 FRAM kcal inlandNA5.222.0 FRAM kcal coastNA5.410.711.0 FRAM kcal totalNA5.414.114.4 No salmonNA3.420.018.3 CTC WCVINA3.45.84.1 CTC NBCNA4.91.41.2

27. Model selection results 1979-2010 (SRKW) AIC values: 0 indicates best model, values greater than 0 indicate models that arent as good. Models with values > 10 can probably be ignored *California stocks and spring / summer stocks appear to be poor predictors of survival * North / fall migrating stocks are better predictors of Survival * NRKW results alone give more support to fall stocks (hidden slide) GroupInclude inland pre-terminal delta AIC (fecundity) delta AIC (survival) delta AIC (total) North (+Fraser Sp)Y1.72.40.6 North (+Fraser Sp)N3.52.1 North (-Fraser Sp)Y3.42.12.0 North (-Fraser Sp)N3.51.5 Central (-Fraser Sp)Y0.07.13.5 Central (-Fraser Sp)N3.27.87.4 Central (+Fraser Sp)Y2.77.46.6 Central (+Fraser Sp)N3.27.97.6 SouthNA3.617.317.4 SpringY2.612.611.7 SpringN2.813.012.2 SummerY3.517.0 SummerN3.716.216.4 Summer/fallY2.112.611.1 Summer/fallN1.912.310.6 Fall (+ California ) Y3.74.44.5 Fall (+ California ) N3.74.54.7 Fall (- California ) Y3.61.7 Fall (- California ) N3.70.00.1 No salmonNA1.715.413.6 CTC WCVINA2.54.73.7 CTC NBCNA3.00.60.0

28. Model selection results 1979-2010 (SRKW) AIC values: 0 indicates best model, values greater than 0 indicate models that arent as good. Models with values > 10 can probably be ignored *California stocks and spring / summer stocks appear to be poor predictors of survival * North / fall migrating stocks are better predictors of Survival * NRKW results alone give more support to fall stocks (hidden slide) GroupInclude inland pre-terminal delta AIC (fecundity) delta AIC (survival) delta AIC (total) North (+Fraser Sp)Y (= Ocean)0.30.7 North (+Fraser Sp)N (= Inland)0.71.6 North (-Fraser Sp)Y0.70.0 North (-Fraser Sp)N0.81.1 Central (-Fraser Sp)Y0.010.4 Central (-Fraser Sp)N1.510.4 Central (+Fraser Sp)Y1.810.4 Central (+Fraser Sp)N1.610.4 SouthNA2.111.0 SpringY0.411.7 SpringN0.711.7 SummerY2.211.6 SummerN2.211.7 Summer/fallY2.29.9 Summer/fallN2.29.8 Fall (+California)Y1.98.1 Fall (+California)N1.87.3 Fall (-California)Y1.34.9 Fall (-California)N1.12.0 No salmonNA0.39.7 CTC WCVINA1.16.3 CTC NBCNA0.32.2

29. Model selection results 1979-2010 (NRKW) AIC values: 0 indicates best model, values greater than 0 indicate models that arent as good. Models with values > 10 can probably be ignored GroupInclude inland pre-terminal delta AIC (fecundity) delta AIC (survival) delta AIC (total) North (+Fraser Sp)Y2.27.56.7 North (+Fraser Sp)N3.68.59.0 North (-Fraser Sp)Y3.16.5 North (-Fraser Sp)N3.67.88.3 Central (-Fraser Sp)Y4.010.911.9 Central (-Fraser Sp)N1.413.511.9 Central (+Fraser Sp)Y1.810.69.3 Central (+Fraser Sp)N1.513.612.0 SouthNA4.08.39.2 SpringY3.92.02.8 SpringN3.91.62.4 SummerY0.07.44.3 SummerN2.61.71.3 Summer/fallY3.25.25.4 Summer/fallN3.612.012.6 Fall (+California)Y2.99.18.9 Fall (+California)N2.910.610.4 Fall (-California)Y3.10.0 Fall (-California)N2.80.50.3 No salmonNA2.211.610.8 CTC WCVINA1.06.54.5 CTC NBCNA2.59.89.3

30. Model selection results 1979-2010 (SRKW & NRKW) AIC values: 0 indicates best model, values greater than 0 indicate models that arent as good. Models with values > 10 can probably be ignored *California stocks and spring / summer stocks appear to be poor predictors of survival * North / fall migrating stocks are better predictors of Survival GroupInclude inland pre-terminal delta AIC (fecundity) delta AIC (survival) delta AIC (total) North (+Fraser Sp)Y2.47.27.7 North (+Fraser Sp)N2.79.29.9 North (-Fraser Sp)Y2.15.45.6 North (-Fraser Sp)N2.87.98.7 Central (-Fraser Sp)Y3.218.519.8 Central (-Fraser Sp)N0.922.321.3 Central (+Fraser Sp)Y1.618.217.8 Central (+Fraser Sp)N1.022.221.3 SouthNA4.416.919.4 SpringY3.115.416.5 SpringN3.115.116.3 SummerY0.719.218.0 SummerN3.015.116.2 Summer/fallY3.512.814.4 Summer/fallN3.819.221.1 Fall (+ California ) Y2.714.615.4 Fall (+ California ) N2.615.716.4 Fall (- California ) Y2.42.22.6 Fall (- California ) N1.90.0 No salmonNA2.420.621.1 CTC WCVINA0.010.18.2 CTC NBCNA1.111.510.6

31. STOCKMIGRATION TIMINGGEOGRAPHY (Ocean) Col Lower R springSpringCentral Fraser Sp. 1.2SpringNorth/Central Fraser Sp. 1.3SpringNorth/Central SEAKSpringNorth Upper Col R springSpringCentral Areas 1-5SummerNorth Areas 6-10SummerNorth Col R summerSummerCentral Fraser Sm. 0.3SummerNorth Fraser Sm. 1.3SummerNorth Puget SoundSummer/FallCentral Upper Georgia StraitSummer/FallNorth Col R TuleFallCentral Col River bright fallFallNorth Fraser LateFallCentral Lower Georgia Strait H & NFallCentral Klamath fallFallCalifornia OR coastFallNorth Sacramento fallFallCalifornia WA coastFallNorth WCVIFallNorth

32. What about Puget Sound, Upper G. Strait (summer / fall)? Pre-terminal catch Puget Sound Georgia St Improvement in AIC Fall (-California)NNN-- NYN -0.09-0.91 Fall (-California)NNY -0.02-0.13 Fall (-California)NYY -0.12-0.98 More negative values indicate more improvement (reduction) in AIC value, better fit to model

33. Fall & Summer/Fall > 50%

34. Whats not included in fall group? California stocks (Sacramento, Klamath) Summer stocks Spring stocks – Fattier (e.g. Columbia spring) – Workshop 1 result: spring stocks may be most important (Wasser, Ayres et al.) – Spring stocks dont seem to be supported here

35. Summarizing Parken-Kope indices Results consistent with CTC results – strong positive correlations with survival, less so with fecundity FRAM inland #s seems to be somewhat of an outlier – Different pattern, especially since mid 1980s – Science panel report, p 12-13 (Figure 1)

36. Resolving inconsistencies with FRAM (coastal) FRAM coastal index is largely in agreement with the WCVI and More recent Parken-Kope time Series *FRAM coastal dominated By N. migrating stocks (upriver brights) Standardized salmon

37. Resolving inconsistencies with FRAM (inland) FRAM inland index Appears to be the most dissimilar index compared to CTC or Parken-Kope Standardized salmon