COMPARATIVE SURVIVAL STUDY (CSS) of PIT-tagged Spring/Summer Chinook and Summer Steelhead Annual Report. Prepared by

Transcription

1 COMPARATIVE SURVIVAL STUDY (CSS) of PIT-tagged Spring/Summer Chinook and Summer Steelhead 2006 Annual Report BPA Project # Prepared by Fish Passage Center and Comparative Survival Study Oversight Committee: Thomas Berggren and Peter McHugh, Fish Passage Center Paul Wilson and Howard Schaller, U.S. Fish and Wildlife Service Charlie Petrosky, Idaho Department of Fish and Game Earl Weber, Columbia River Inter-Tribal Fish Commission Ron Boyce, Oregon Department of Fish and Wildlife Project Leader: Michele DeHart, Fish Passage Center Final 11/30/2006

2 TABLE OF CONTENTS PAGE List of Tables... iii List of Figures...ix Acknowledgements...xiv Executive Summary...xvi Chapter 1 Introduction...1 Chapter 2 Methods...4 Chapter 3 SARs, T/C ratio, and D for Wild and Hatchery Chinook...8 Chapter 4 SARs, T/C ratio, and D for Wild and Hatchery Steelhead...27 Chapter 5 Relationships between wild and hatchery Chinook salmon smolt-to-adult survival and inriver, estuary/early ocean, and off-shore marine environmental conditions...41 Chapter 6 Associations between smolt outmigration experience and adult Chinook salmon Bonneville-to-Lower-Granite-Dam apparent survival rates...50 Chapter 7 Upstream-downstream comparisons: Differential mortality for upriver and downriver PIT-tagged wild and hatchery sp/su Chinook Chapter 8 Upstream-downstream comparisons: contrasting smolt life histories between Snake River and John Day River stream-type Chinook salmon populations...68 Chapter 8 Appendix Redd density estimation Chapter 9 Understanding the implications of smolt size detection probability relationships for CSS study-group comparisons...80 Chapter 10 Computer program to create simulated PIT tag input files for testing robustness of CJS survival estimates...88 References Appendix A Formula of parameters used in CSS analyses Appendix B Estimated number of smolts per study category with associated 90% confidence interval and number of returning adults per study category i

3 Appendix C Reach survival estimates with bootstrap 95% confidence intervals Appendix D Age distribution of returning adult Chinook and steelhead detected at Lower Granite Dam (or Bonneville Dam for downriver populations) Appendix E Number of PIT-tagged smolts transported at each collector dam (plus estimated number if tagged fish had been transported in same proportion as the untagged population) and site-specific SAR estimates Appendix F Data used in estimating the annual weighted SARs for wild and hatchery Chinook and steelhead Appendix G PIT-tagged hatchery Chinook release numbers and relation to production Appendix H Release sites of PIT-tagged wild Chinook and wild & hatchery steelhead (each release site is not present for every migration year) Appendix I Regional review comments and CSS Oversight Committee responses ii

4 LIST OF TABLES PAGE Table 1. Number of PIT-tagged wild Chinook parr/smolts from the four tributaries above Lower Granite Dam and Snake River trap used in the CSS analyses for migration years 1994 to Table 2. Estimated SAR LGR-to-LGR (%) for PIT-tagged wild Chinook in annual aggregate for each study category from 1994 to 2004 (with 90% confidence intervals) Table 3. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged wild Chinook for migration years 1994 to 2004 (with 90% confidence intervals) Table 4. Number of PIT-tagged hatchery Chinook parr/smolts from key hatcheries located above Lower Granite Dam used in the CSS analyses for migration years 1997 to Table 5. Estimated SAR LGR-to-LGR (%) for PIT-tagged spring Chinook from Rapid River Hatchery for each study category from 1997 to 2004 (with 90% confidence intervals) Table 6. Estimated SAR LGR-to-LGR (%) for PIT-tagged spring Chinook from Dworshak Hatchery for each study category from 1997 to 2004 (with 90% confidence intervals) Table 7. Estimated SAR LGR-to-LGR (%) for PIT-tagged spring Chinook from Catherine Creek AP for each study category from 2001 to 2004 (with 90% confidence intervals) Table 8. Estimated SAR LGR-to-LGR (%) for PIT-tagged summer Chinook from McCall Hatchery for each study category from 1997 to 2004 (with 90% confidence intervals) Table 9. Estimated SAR LGR-to-LGR (%) for PIT-tagged summer Chinook from Imnaha River AP for each study category from 1997 to 2004 (with 90% confidence intervals) Table 10. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged Rapid River Hatchery spring Chinook for 1997 to 2004 (with 90% confidence intervals) Table 11. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged Dworshak Hatchery spring Chinook for 1997 to 2004 (with 90% confidence intervals) iii

5 Table 12. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged Catherine Creek AP spring Chinook for 2001 to 2004 (with 90% confidence intervals) Table 13. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged McCall Hatchery summer Chinook for 1997 to 2004 (with 90% confidence intervals) Table 14. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged Imnaha AP summer Chinook for 1997 to 2004 (with 90% confidence intervals) Table 15. Number of PIT-tagged wild steelhead smolts from the four tributaries above Lower Granite Dam (plus Snake River trap) used in the CSS for migration years 1997 to Table 16. Estimated SAR LGR-to-LGR (%) for PIT-tagged wild steelhead in annual aggregate for each study category from 1997 to 2003 (with 90% confidence intervals) Table 17. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged wild steelhead for migration years 1997 to 2003 (with 90% confidence intervals) Table 18. Number of PIT-tagged hatchery steelhead smolts from the four tributaries above Lower Granite Dam (plus mainstem Snake River) used in the CSS for migration years 1997 to Table 19. Estimated SAR LGR-to-LGR (%) for PIT-tagged hatchery steelhead in annual aggregate for each study category from 1997 to 2003 (with 90% confidence intervals) Table 20. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged hatchery steelhead for migration years 1997 to 2003 (with 90% confidence intervals) Table 21. Model selection results for SAR environmental variable regression models fitted using MY PIT-tag-based annual weighted SARs. The bold-faced entry corresponds to the model with the lowest AIC c value (corrected for sample size) score. Q is discharge, WTT is water transit time, CUI-April is April upwelling, CUI-Oct is October upwelling, and PDO is Pacific decadal oscillation iv

6 Table 22. Model selection results for SAR environmental variable regression models fitted using and MY run-reconstruction-based SARs. The bold-faced entry corresponds to the model with the lowest AIC c value (corrected for sample size) score. See Table 1 description for variable definitions Table 23. Least-squares slope parameter estimates (+/- 95% CIs) for bivariate regressions between PIT-tag- ( Contemp. in table) and run-reconstruction-based ( Historic in table) SARs and environmental factors. Bold-faced cell entries correspond to those estimates differing significantly from zero Table 24. Counts of hatchery Chinook salmon adults that failed ( F ) or were successful ( S ) in surviving their BON-LGR migration in return years , grouped by migration year and outmigration experience (see Methods for group definitions). There was evidence for a significant association between transport history and migration success where sufficient observations-per-cell were available (see Table 26 for details) Table 25. Counts of wild Chinook salmon adults that failed ( F ) or were successful ( S ) in surviving their BON-LGR migration in return years , grouped by migration year and outmigration experience (see Methods for group definitions). There was evidence for a significant association between transport history and migration success where sufficient observationsper-cell were available (i.e., > 5; MY2002: χ 2 = 8.74, df = 2, P = 0.013; Combined: χ 2 = 7.94, df = 2, P = 0.019; MY2001, MY2003-4, not applicable) Table 26. Summary of MY-, RY-, and hatchery-specific χ 2 -tests for hatchery Chinook salmon. The P-values listed are not corrected for multiple tests. The success rate ranking corresponds to the ordering of % successful upstream migrants by juvenile outmigration history. The entry NA corresponds to table values that are not applicable because either a test was not performed due to low cell counts (i.e., RY2002) or the resulting test statistic was not significant (α = 0.05). df = 2 for all tests Table 27. Logistic regression model-selection results for CSS hatchery Chinook salmon. Note, Y = P(Success X), where X is the variable in question. The bold-faced model was the one most supported by the data, however those with a ΔAIC < 2 can be considered nearly equivalent. K is the number of estimated parameters (inclusive of variance) Table 28. Parameter estimates for the top logistic regression model describing BON-LGR migration success for CSS hatchery Chinook salmon returning in v

7 Table 29. Logistic regression model-selection results for CSS wild Chinook salmon. Note, Y = P(Success X), where X is the variable in question. The bold-faced model was the one most supported by the data, however those with a ΔAIC < 2 were viewed as equivalent. K is the number of estimated parameters (inclusive of variance) Table 30. Parameter estimates for the top logistic regression model describing BON-LGR migration success for CSS wild Chinook salmon returning from Table 31. Number of PIT-tagged Carson Hatchery Chinook released in the Wind River, estimated survival and resulting smolt population arriving Bonneville Dam in migration years 2000 to 2004 (with 90% confidence intervals) with detected adults at BOA Table 32. PIT-tag detections of returning adult Chinook (ages 2- and 3-salt) at Bonneville and Lower Granite dams with percentage of fish undetected at Bonneville Dam returns from smolts that outmigrated in 2001 to Table 33. Estimates of SAR from first dam encountered 1 as smolts to Bonneville Dam (BOA) as adults 2 for the upriver PIT-tagged wild Chinook aggregate and the downriver PIT-tagged John Day River wild Chinook that outmigrated in 2000 to Table 34. Number of PIT-tagged wild Chinook released in John Day River basin, estimated survival and resulting smolt population arriving John Day Dam in migration years 2000 to 2004 (with 90% confidence intervals) with detected adults at BOA Table 35. Estimates of SAR from first dam encountered 1 as smolts to Bonneville Dam (BOA) as adults 2 for the upriver PIT-tagged wild Chinook aggregate and the downriver PIT-tagged John Day River wild Chinook that outmigrated in 2000 to Table 36. Conversion of estimated upriver/downriver ratios to differential mortality rates for comparison to differential mortality rates computed by spawner-recruit analyses, 95% confidence intervals shown with each method Table 37. Summary statistics for wild Chinook salmon smolts captured, tagged, and released at CSS trap sites between March 15 and May 20 during migration years Table 38. Results from an ANCOVA-based comparison of smolt size across upstream and downstream release sites, using redd density as a covariate vi

8 Table 39. Results from an ANOVA evaluating smolt size variation across release sites and migration years Table 40. Dates of 50% passage (i.e., median emigration date) for Chinook salmon captured, tagged, and released at CSS-affiliated trap sites during MYs Table 41. Median estuary arrival (i.e., BON detection) dates for Chinook salmon smolts captured, tagged, and released at CSS-affiliated trap sites during MYs Table 42. Redd abundance and surveyed kilometers for production areas upstream of CSS trap sites used to contrast smolt size between upstream and downstream populations Table 43. Sample sizes for PIT-tagged release groups (sum of SNKTRP and CLWTRP releases between 11 April and 10 May) used in our estimation of P(det FL) relationships (1999, 2000, 2002, ) and comparison of size between detected and undetected study categories, by migration year (MY). Bold-faced values correspond to MYs included in our survival/detection probability modeling exercise Table 44. Candidate detection probability (p) models fitted for fish groups released in migration years , 2002, and For detection-probability model selection, the survival (φ) model structure was held constant based on the recommendations of Lebreton et al. (1992), in the most global form [i.e., φ(site FL, all), survival varies across sites as a site-specific function of length] Table 45. Model-selection results for wild Chinook salmon release groups with sufficient tags for survival and recapture probability estimation (i.e., >1,000), by migration year. ΔAIC values appear in cells. Top models (i.e., those with the lowest AIC value) are identified with bold-faced font and underlining; near-top models (i.e., those with a ΔAIC value < 2) also appear as underlined, but in italics. See Table 2 for description of survival and detection probability model structures Table 46. Maximum likelihood slope parameter estimates from detection probability fork length relationships for wild Chinook salmon captured, PITtagged, and released at the Snake River and Clearwater River smolt traps (rel_site = SNKTRP, CLWTRP). Bold-faced values correspond to those parameters with point estimates that were greater than twice the value of their standard errors (after Zabel et al. 2005). Estimates delineated by NOAA correspond to the values reported in Zabel et al., CSS corresponds to our upstream-of-lgr release analysis vii

9 Table 47. Summary statistics for detected and undetected wild Chinook salmon captured, tagged, and released from the Snake River and Clearwater River smolt traps during MYs Rows with bold-faced font are those MYs where a significant difference (α = 0.05) was detected between categories using a t-test Table 48. Comparison of simulated true survival rates S 2, S 3 and V C, and number of smolts in study categories T 0, C 0 and C 1 using default input parameters of Figures 42 to 48 and the resulting estimated parameter values obtained with the bootstrap program Table 49. Comparison of mean and 90% confidence intervals of parameters S 2, S 3, V C, T 0, C 0, and C 1 of simulated samples from common underlying population and the estimates obtained with the bootstrap program viii

10 LIST OF FIGURES PAGE Figure 1. Salmonid life cycle in the Snake River and lower Columbia River basins (source: Marmorek et al Figure 2. Trend in proportion of PIT-tagged wild Chinook transported at each Snake River collector dam, 1994 to Figure 3. Estimated SAR LGR-to-LGR for PIT-tagged wild Chinook aggregate in transport and inriver study categories for migration years 1994 to 2003 (only 2-salt adult returns for 2004) Figure 4. Trend in estimated annual SAR (with 90% confidence intervals for ) for wild Chinook based on PIT-tagged Chinook SARs in transport and inriver study categories weighted by estimated proportion of run-at-large in each study category for migration years 1994 to 2004 (only 2-salt adult returns for 2004) Figure 5. Trend in proportion of PIT-tagged hatchery Chinook transported at each Snake River collector dam, Figure 6. Trend in estimated transport and inriver SARs for Rapid River Hatchery spring Chinook for migration years 1997 to 2004 (only 2-salt adult returns for 2004) Figure 7. Estimated transport and inriver SARs for PIT-tagged Dworshak Hatchery spring Chinook for migration years 1997 to 2004 (only 2-salt adult returns for 2004) Figure 8. Estimated transport and inriver SARs for PIT-tagged Catherine Creek Acclimation Pond spring Chinook for migration years 2001 to 2004 (only 2-salt adult returns for 2004) Figure 9. Estimated transport and inriver SARs for PIT-tagged McCall Hatchery summer Chinook for migration years 1997 to 2004 (only 2-salt adult returns for 2004) Figure 10. Estimated transport and inriver SARs for PIT-tagged Imnaha River acclimation Pond summer Chinook for migration years 1997 to 2004 (only 2-salt adult returns for 2004) Figure 11. Trend in estimated annual SAR LGR-to-LGR for hatchery and wild sp/su Chinook based on PIT-tagged sp/su Chinook SARs in transport and inriver study categories weighted by estimated proportion of run-at-large in each study category for migration years 1994 to 2004 (only 2-salt adult returns for 2004) ix

11 Figure 12. Trend in in-river survival (V c ) for PIT-tagged Snake River wild and hatchery spring/summer Chinook in migrations years 1994 to Figure 13. Trend in ratio of SAR 2 (T 0 )/SAR(C 0 ) (log-transformed) for PITtagged Snake River hatchery and wild Chinook in migration years 1994 to Figure 14. Trend in D (log-transformed) for PIT-tagged Snake River hatchery and wild Chinook in migration years Figure 15. Trend in proportion of PIT-tagged wild steelhead transported at each Snake River collector dam, Figure 16. Estimated transport and inriver SARs (with 90% confidence intervals) for PIT-tagged wild steelhead aggregate for migration years 1997 to 2003 (incomplete returns for 2003) Figure 17. Estimated annual SAR for wild steelhead compared to wild Chinook based on PIT-tagged steelhead SARs in transport and inriver study categories weighted by estimated proportion of run-at-large in each study category for migration years 1997 to 2003 (incomplete returns for 2003) Figure 18. Trend in in-river survival (V c ) for PIT-tagged Snake River wild steelhead and wild Chinook for migration years 1997 to Figure 19. Trend in ratio of SAR 2 (T 0 )/SAR(C 0 ) (log-transformed) for PIT-tagged Snake River wild steelhead and wild Chinook in migration years 1997 to Figure 20. Trend in D (log-transformed) for PIT-tagged Snake River wild steelhead and wild Chinook in migration years Figure 21. Trend in proportion of PIT-tagged hatchery steelhead transported at each Snake River collector Dam, Figure 22. Estimated transport and inriver SARs for PIT-tagged hatchery steelhead aggregate for migration years 1997 to 2003 (incomplete returns for 2003) Figure 23. Trend in estimated annual SAR for hatchery and wild steelhead with associated 90% confidence intervals based on respective PIT-tagged steelhead SARs in transport and inriver study categories weighted by estimated proportion of run-at-large in each study category for migration years 1997 to 2003 (incomplete returns for 2003) x

12 Figure 24. Trend in in-river survival (V c ) for PIT-tagged Snake River hatchery and wild steelhead for migration years 1997 to Figure 25. Trend in ratio of SAR 2 (T 0 )/SAR(C 0 ) (log-transformed) for PIT-tagged Snake River hatchery and wild steelhead in migration years 1997 to Figure 26. Trend in D (log-transformed) for PIT-tagged Snake River hatchery and wild steelhead in migration years Figure 27. Time series plot of annual weighted PIT-tag SARs for wild ( Wild-PIT ) and hatchery ( Hatch-PIT, hatchery averaged) Chinook salmon across migration years Run-reconstruction-based SARs (from Scheuerell and Williams 2005; Wild-RR ) extending from and are presented for comparison. Note, SARs were interpolated between 1984 and 1991 for illustrative purposes only Figure 28. Time series plots of inriver (Water transit time; Columbia River discharge) and estuary/ocean environmental variables (PDO, CUI-April, CUI-October) across migration years 1964 to See text and Table 21 caption for variable definitions Figure 29. Scatter plots of hatchery and wild PIT-tag annual weighted SARs versus discharge during outmigration, April upwelling, and summer PDO Figure 30. Scatter plots of PIT-tag- ( Contemp. ) and run-reconstruction-based ( Historic ) bivariate SAR environmental variable (discharge and WTT during outmigration, April upwelling, and summer PDO) Figure 31. Bar chart of the percent of hatchery (left) and wild (right) Chinook salmon that were successful in migrating from BON to LGR for inriver, LGR, and LGS-down outmigration histories across return years Figure 32. Box-and-whisker plot of BON-LGR travel times for hatchery (left) and wild (right) Chinook salmon, by outmigration experience (pooled across RYs ) Figure 33. Trend in differential mortality ΔM=-ln(U/D) for hatchery Chinook (Snake River basin stocks [U] versus Carson NFH stock [D]) for smolt migration years 2000 to Figure 34. Differential mortality estimates from the Deriso et al. (2001) model updated through smolt year 2000 (Marmorek et al. 2004) compared to estimates based on SARs of wild Snake River and John Day River sp/su Chinook, smolt migration years xi

13 Figure 35. Trend in differential mortality ΔM=-ln(U/D) for wild Chinook (Snake River basin stocks [U] versus John Day stocks [D]) and hatchery Chinook (Snake River basin stocks [U] versus Carson NFH stock [D]) for smolt migration years 2000 to Figure 36. Wild Chinook salmon smolt size (mean fork length +/- 1 SD) for those fish captured, tagged, and released at CSS trap sites during migration years (between 15 March and 20 May). From left to right, release sites are: CLWTRP = Clearwater River Trap, GRNTRP = Grande Ronde River Trap, IMNTRP = Imnaha River Trap, JDAR1 = John Day River Trap, SALTRP = Salmon River Trap, SNKTRP = Snake River Trap. Note: there were no wild Chinook smolt size data available for CLWTRP prior to Figure 37. Scatter plot of mean fork length (mm) against redd density (redds / km) for wild Chinook salmon smolts collected, tagged, and released at CSS trap sites during migration years (between 15 March and 20 May). See Figure 36 caption for release site abbreviation definitions Figure year mean trap passage (i.e., emigration) distributions for JDAR1, SNKTRP, SALTRP, CLWTRP, IMNTRP, and GRNTRP release sites. Note: Julian date 75 is March 16th, 100 is April 10th, 125 is May 5th, and 150 is May 30th. See Figure 36 caption for release site abbreviation definitions Figure 39. Wild Chinook salmon smolt downstream migration rates (km / d, +/- 1 SD) for those fish captured, tagged, and released at CSS trap sites during migration years (between 15 March and 20 May). See Figure 36 caption for release site abbreviation definitions. Note, CLWTRP operations did not begin until 2002; also, too few tags were available for SNKTRP estimation in 2001, Figure 40. Scatter plot of first-to-third dam migration duration as a function of water travel time. Each dot reflects the mean value for a year-site combination. See Figure 36 caption for release site abbreviation definitions...76 Figure year mean estuary arrival (measured at BON) timing distributions for JDAR1, SNKTRP, SALTRP, CLWTRP, IMNTRP, and GRNTRP release sites. Note: Julian date 100 is April 10th, 125 is May 5th, 150 is May 30th, and 175 is June 24th. See Figure 36 caption for release site abbreviation definitions xii

14 Figure 42. Estimated fork length (mm) detection probability relationships for wild Chinook salmon at LGR for MYs 1999, 2000, 2002, 2005, and Figure 43. First input screen of simulator program initial settings including release number and survival to LGR, travel time related parameters, and assumed SAR levels Figure 44. Second simulator input screen arrival population characteristics, collection efficiency and removal rates at LGR, and smolt travel time and survival to LGS Figure 45. Third simulator input screen collection efficiency and removal rates at LGS, and smolt travel time and survival to LMN Figure 46. Fourth simulator input screen collection efficiency and removal rates at LMN, and smolt travel time and survival to MCN Figure 47. Fifth simulator input screen collection efficiency and removal rates at MCN, and smolt travel time and survival to JDA Figure 48. Sixth simulator input screen collection efficiency and removal rates at JDA, and smolt travel time and survival to BON Figure 49. Seventh simulator input screen collection efficiency and removal rates at BON, smolt travel time to trawl site, and trawl collection rate (joint survival-collection efficiency) Figure 50. Simulated arrival distribution of smolts at LGR and daily collection efficiency, based on default parameter inputs Figure 51. Passage timing of smolts at each dam and lower Columbia trawl site for fish returned-to-river at each site (upper plot) and transported (lower plot), based on default parameter inputs Figure 52. Reach survival rate from LGR to LGS for use with daily passing (undetected or detected and returned-to-river) smolts at LGR, based on default parameter inputs xiii

15 ACKNOWLEDGEMENTS The Comparative Survival Study (CSS) relies on cooperation of multiple agencies and individuals to ensure that the major tasks of marking, handling, releasing, and recovery of fish in three States and at hatcheries and dams are completed successfully. PIT-tag detection systems installed at mainstem Columbia and Snake River dams are maintained by PSMFC personnel and enable PIT-tag data to be collected and retrieved on a realtime basis. We commend Carter Stein and his staff for their diligence in repair of PIT-tag equipment in the juvenile and adult fish passage systems and providing the smooth transfer of PIT-tag data to the PIT Tag Operations Center in Portland, OR. We especially thank Dave Marvin of PSMFC, who conducts the routine programming for the separation-by-code operations at the dams with CSS study fish. We extend thanks to the Fish and Wildlife agencies and all hatchery managers and staff for their assistance in the planning, raising of, and recovery of study fish for the CSS at their hatcheries. The agencies include Idaho Department of Fish and Game (IDFG) for Rapid River and McCall hatcheries, U.S. Fish and Wildlife Service (USFWS) for Dworshak and Carson hatcheries, and Oregon Department of Fish and Wildlife (ODFW) for Lookingglass Hatchery s outplants into Imnaha and Catherine Creek acclimation ponds. These acclimation ponds are operated as a cooperative venture with Nez Perce and Umatilla tribes, respectively. We thank the field supervisors and crews for an excellent job in completing the PIT-tagging operations at these hatcheries. The USFWS Dworshak and Vancouver Fisheries Resource Office (FRO) personnel PIT tagged the fish at the USFWS hatcheries. PIT tagging at IDFG hatcheries was completed with supervision provided by the IDFG office in Lewiston, Idaho. Chinook at the Lookingglass complex were PIT-tagged by ODFW personnel from the Northeast District fisheries office in LaGrande, Oregon. In addition, we extend thanks to all crews PIT tagging wild Chinook and wild and hatchery steelhead in the region. These PIT-tagged fish have provided the opportunity for the CSS to expand our comparative evaluations to these salmonids also. We appreciate and thank the researchers at IDFG, ODFW, Confederated Tribes of Umatilla Indian Reservation (CTUIR), and Shoshone-Bannock Tribes (SHOBAN) who have allowed the CSS to route a proportion of their PIT-tagged smolts to transportation at the Snake River collector dams. The CSS commends the Nez Perce Tribal researchers for routing a portion of their PIT-tagged wild Chinook from the Imnaha and SF Salmon River sub-basins to transportation; this increases representation of those drainages in the wild Chinook PIT-tag aggregate population for SAR estimation. The Fish Passage Center s role in the implementation of this program was accomplished through coordination of PIT tagging and field logistics, performing database development, data compilation and preliminary analyses, and overseeing the budgetary aspects of this study. In addition to the coauthors, FPC staff members Michele DeHart and Dona Watson provided valuable contributions to the implementation of this program and Brandon Chockley provided assistance in graphics. A special thanks goes to former FPC staff Henry Franzoni, who programmed the bootstrap and simulator routines, Sergei Rassk for his contribution to updates and improvements to both the bootstrap and simulator programs, and Paul Wilson (USFWS) for his contribution to the planning and development of both the simulator and bootstrap programs. We also thank Nick Bouwes (EcoLogic, Logan UT) for his assistance in the planning phase of the bootstrap program. xiv

16 Bonneville Power Administration (BPA Project Number ) funded this project through the Northwest Power Planning Council Fish and Wildlife Program. BPA s Contract Officer s Technical Representative (COTR) for the CSS is Tracy Hauser. xv

17 EXECUTIVE SUMMARY The Comparative Survival Study (CSS) was initiated in 1996 as a multi-year program of the fishery agencies and tribes to estimate survival rates over different life stages for spring and summer Chinook salmon (hereafter, Chinook) produced in major hatcheries the Snake River basin and from selected hatcheries in the lower Columbia River. Much of the information evaluated in the CSS is derived from fish tagged with Passive Integrated Transponder (PIT) tags. A comparison of survival rates of Chinook marked in different regions (which differ in the number of dams Chinook have to migrate through) provides insight into the effects of the Federal Columbia River Power System (FCRPS, hereafter termed hydrosystem). The CSS compares the smolt-to-adult survival rates (SARs) for Snake River wild and hatchery Chinook that were transported versus those that migrated inriver to below Bonneville Dam. As in 2005, we also computed SARs for wild and hatchery summer steelhead PIT-tagged under other existing programs. These SAR estimates generate information reflecting the relative effects of the current management actions used to recover this listed species. Scientists and managers have recently emphasized the importance of delayed hydrosystem mortality to long-term management decisions. Delayed hydrosystem mortality may occur for both smolts that migrate inriver and smolts that are transported. The CSS PIT-tag information on inriver survival rates and SARs of both transported and inriver fish are relevant to the estimation of D, a parameter which partially describes delayed hydrosystem mortality. It is the differential survival rate of transported fish relative to fish that migrated inriver from below Bonneville Dam as smolts to adults returning to Lower Granite Dam. When D < 1, the transported smolts die at a greater rate after release below Bonneville Dam than smolts that have migrated inriver to below Bonneville Dam. Major objectives of the CSS include: (1) development of a long-term index of transport SAR to inriver SAR for Snake River hatchery and wild spring/summer Chinook smolts measured at Lower Granite Dam; (2) develop a long-term index of survival rates from release of smolts at Snake River hatcheries to return of adults to the hatcheries; (3) compute and compare the overall SARs for selected upriver and downriver spring and summer Chinook hatchery and wild stocks; and (4) begin a time series of SARs for use in hypothesis testing and in the regional long-term monitoring and evaluation. The CSS PIT tags and annually releases more than 200,000 smolts from Snake River hatcheries (e.g., Dworshak, McCall, Rapid River, Catherine Creek and Imnaha) and currently 15,000 smolts from a downriver hatchery (Carson NFH). In addition, the CSS provides 23,000 PIT-tags for wild Chinook and wild steelhead to augment various on-going trapping and tagging operations in the Snake River basin. These PIT-tagged smolts from the Snake River are detected in collection systems at Snake and Columbia River dams and diverted into transportation or bypassed to the river according to the annual study design. Beginning in 2002, the CSS increased releases of PIT-tagged wild Chinook in the Snake River basin and coordinated with other researchers to route more detected wild Chinook into transportation for subsequent use in the CSS. Because fewer PIT-tagged wild Chinook are available for study, the CSS continues to evaluate the extent to which the responses of hatchery Chinook to management actions can be used as a surrogate for wild Chinook. The PIT-tagged wild and hatchery Chinook and steelhead are assigned to study categories based on their route of passage (inriver vs transported) through the hydrosystem. The route of xvi

18 passage of individual fish is determined from its PIT-tag detection history through the hydrosystem. The inriver study groups include smolts that were never collected or bypassed at Snake River collector dams (Category C 0 ) and smolts that were collected and bypassed at one or more Snake River collector dams (Category C 1 ). The transport study group (Category T 0 ) includes smolts transported from a Snake River collector dam. Returning PIT-tagged adults detected at Lower Granite Dam are assigned to the appropriate study group. Then SARs, measured from smolts at Lower Granite to adult returns to Lower Granite, are were calculated for transport and inriver groups, along with ratios of transport SAR to inriver SAR (T/C ratios) and parameter D. Bootstrap confidence intervals are computed for all parameter estimates (Chinook: Chapter 3; Steelhead: Chapter 4). In addition to estimating relevant parameters, the CSS also performs upriver/downriver stock comparisons (Chapter 7). In an upriver/downriver evaluation, estimates of SARs for PITtagged wild Chinook from John Day River and hatchery Chinook from Carson NFH (both downriver stocks) are compared with Snake River stocks from the first dam encountered as smolts to Bonneville Dam as adults. In recent years, fisheries scientists have identified potential shortcomings of using an upriver/downriver comparison approach towards evaluating FCRPS effects. New to this year s report is a section addressing several of these concerns. We compare fish size and other smolt life history attributes between upriver and downriver populations in Chapter 8. Beyond reporting parameter estimates and summarizing upstream/downstream comparisons, we report on several related questions pursued as part of our analytical efforts. First, as in 2005, we evaluated dropout rates (combined effect of harvest, straying, and mortality) for returning adults between dams with adult PIT-tag monitors to see if differences occur between returning adults based on whether they migrated inriver or were transported as smolts. In addition to describing patterns, however, we provide a formal statistical comparison of dropout (i.e., apparent survival) rates between study groups in this year s report. Second, given that our wild and hatchery Chinook SAR datasets have reached 11- and 8-years duration, respectively, we provide our first evaluation of relationships between this performance response and large-scale environmental variables believed to influence salmon population productivity. That is, we evaluated associations between inriver, estuary/early ocean, and off-shore marine conditions and annual weighted SARs across study populations of Chinook (Chapter 5). We additionally provide an evaluation of the influence of the size collection efficiency relationships identified by NOAA-Fisheries on our transport vs. inriver comparative study approach (Chapter 9). Finally, we report on our simulation-based evaluation of the robustness of Cormack-Jolly- Seber (CJS) reach-survival estimates under various estimation methodologies (Chapter 10). Chapter 3 Findings: In Chapter 3, the estimates of SARs by study category, annual overall SAR, T/C ratio, and D, are presented for PIT-tagged wild and hatchery sp/su Chinook originating above LGR: 1. The annual SARs (indexed LGR smolts-to-lgr adults) for wild Snake River sp/su Chinook has been highly variable, rising from below 0.5% before 1997 to highs of 2.4% in 1999 before dropping each year to below 0.35 % in 2003 and 2004 (2-salt returns). xvii

19 Current overall annual SAR LGR-to-LGR estimates are far below the minimum 2% recommended in the NPCC Fish and Wildlife Program mainstem amendments (NPCC 2003), and estimated as needed for keeping the stocks stable (Marmorek et al. 1998). 2. Transportation provided little or no benefit to wild sp/su Chinook during the conditions experienced in most years during , except during the severe drought year The 10-year geometric mean (excluding 2001) SAR ratio transported to inriver migrants (T/C) was 0.98, while in 2001, the T/C was approximately 9-fold higher. The T/C ratio was significantly > 1 in only Delayed mortality of transported wild sp/su Chinook smolts was substantial most years relative to that of inriver migrants, based on a 10-yr geometric mean D estimate (excluding 2001) of 0.49, indicating transported smolts died at twice the rate as inriver migrants once they passed BON tailrace. In 2001, D was greater than 2, indicating inriver migrants died at twice the rate of transported smolts in the estuary and ocean. 4. The estimated inriver survival of wild sp/su Chinook from LGR tailrace to Bonneville Dam (BON) tailrace averaged 0.46 (geometric mean) for (excluding 2001, when estimated survival was 0.23). 5. During the 11-yr period 1994 to 2004, SAR(C 1 ) averaged approximately 32% lower than SAR(C 0 ) for wild sp/su Chinook. 6. SARs (LGR-to-LGR) for hatchery Snake River spring/summer Chinook have shown similar patterns as wild Chinook during , although the actual survival rates have differed among hatcheries and between spring and summer runs. For spring Chinook hatcheries, SARs for Rapid River Hatchery have exceeded those of Dworshak Hatchery, and SARs of hatchery summer Chinook (particularly from McCall) have exceeded those of hatchery spring Chinook. SARs of most hatchery Chinook (except Dworshak) have equaled or exceeded the SARs of wild Chinook in migration years In general, transportation provided benefits most years to Snake River hatchery sp/su Chinook , however benefits varied among hatcheries. Omitting 2001 (when all T/C ratios exceeded 5), the 7-year geometric mean T/C ranged from 1.08 at Dworshak, 1.46 at Rapid River, 1.50 at Imnaha and 1.54 at McCall hatcheries, indicating a higher return rate for the transported Chinook from these latter three hatcheries. Although having a shorter time series, annual T/C ratios at Catherine Creek AP hatchery Chinook have remained greater than Delayed mortality of transported hatchery spring and summer Chinook smolts was evident most years relative to that of inriver migrants, based on estimated values of D. Except for 2001 when all D values exceeded 1, the other seven years produced geometric mean D values of 0.62 at Dworshak, 0.78 at Imnaha, 0.81 at Rapid River, and 0.89 at McCall hatcheries. xviii

20 9. The 7-yr ( , ) geometric mean of the estimated inriver reach survival rate of hatchery sp/su Chinook from LGR tailrace to BON tailrace ranged from 0.49 to 0.54 across hatcheries. In 2001, the estimated reach survival rate ranged from 0.27 to 0.37 across hatcheries. 10. During the 8-yr period 1997 to 2004, SAR(C 1 ) has remained lower than lower than SAR(C 0 ) for Chinook from Rapid River, Dworshak, Imnaha, and McCall hatcheries. 11. While wild and hatchery populations demonstrated differences in magnitude for some parameters (T/C, D and SARs), the annual patterns of these parameters were highly correlated among wild and hatchery populations. Chapter 4 Findings: In Chapter 4, the estimates of SARs by study category, annual overall SAR, T/C ratio, and D, are presented for PIT-tagged wild and hatchery summer steelhead originating above LGR: 1. Wild steelhead from the Snake River basin had higher estimated annual SARs (indexed LGR to LGR) than hatchery steelhead in 6 of the 7 migration years (1997 to 2003). Wild steelhead had four years with annual SARs > 2%. 2. The pattern of decreasing estimated annual SARs for wild steelhead is following that of the wild Chinook, just not dropping as rapidly over the migration years 1999 to Transportation seems to provide benefit to wild and hatchery Snake River steelhead; the geometric mean T/C ratio ( , ) was 1.72 wild stocks and 1.46 for hatchery stocks. Migration year 2001 had very high, but imprecise T/C ratios, for both wild and hatchery steelhead. 4. Delayed mortality was evident with transported wild and hatchery steelhead relative to inriver migrants as the geometric mean D for (excluding 2001) was 0.80 for wild stocks and 0.64 for hatchery stocks. Migration year 2001 estimated Ds were >1 for wild and hatchery steelhead. Confidence intervals were wide due to small sample size. 5. Given small sample sizes and wide confidence intervals for both wild and hatchery steelhead, it is premature to conclude whether hatchery steelhead can serve as surrogates for wild steelhead. However, trends in Vc and T/C ratios were similar between wild and hatchery steelhead. xix

21 Chapter 5 Findings: In Chapter 5, we evaluated relationships between environmental conditions existing during outmigration, early ocean, and off-shore salmon life stages for hatchery and wild Chinook salmon using contemporary (based CSS PIT tags) and published historic (based on run reconstructions) data: 1. We found moderate-to-strong relationships between SARs and environmental variables using both contemporary PIT-tag and historic run-reconstruction information. Across populations and datasets examined, SARs varied in direct relation to large-scale marine and near-shore coastal climate indices and a single hydrological variable describing outmigration conditions. 2. SAR environmental variable relationships were generally convergent for both historic and contemporary datasets, with some minor exceptions. Specifically, given the larger domain existing for some environmental variables in the complete ( ) time series, more variance in run-reconstruction SARs could be explained by October upwelling and WTT than could be for recent ( ) PIT-tag SARs. 3. Measured SARs were highest for those MYs when fish emigrated during high-flow or under fast-moving WTT conditions, arrived at the coast during periods of increased upwelling, and completed their off-shore migration under cool-phase PDO conditions. 4. Future analyses will consider transport and inriver SARs independently and explore possible interactions between outmigration history and environmental effects on performance. Chapter 6 Findings: In Chapter 6, we quantified statistical associations between Chinook salmon outmigration experience and adult upstream-adult survival across several return years: 1. For both wild and hatchery Chinook salmon, we found a significant effect of outmigration experience on the upstream migration success or apparent survival of returning adults. This effect appeared most pronounced for fish that were transported from LGR as smolts, with these individuals surviving at an approximately 10% lower rate than those with either an inriver or an LGS or LMN transport history. 2. We found that outmigration experience does not affect the timing of adult return (based on BON detections) or the upstream travel times of those salmon surviving to LGR. xx

22 Chapter 7 Findings: In Chapter 7 the ratio of SARs for upriver and downriver wild Chinook are computed, plus the ratio of SARs for upriver and downriver hatchery Chinook. Additionally, trends in the estimated differential mortality rates are presented: 1. Differential mortality rates (between upstream and downstream populations) estimated from SAR data appear to correspond well with differential mortality rates estimated from recruit/spawner ratios for wild Chinook populations. 2. Differential mortality estimates based on SAR ratios of hatchery populations were generally less than those based on SAR ratios of wild populations. Chapter 8 Findings: Chapter 8 analyses address criticisms of an upriver/downriver comparison approach: 1. We observed little evidence indicating that a consistent and/or systematic difference in size-at-migration exists between upstream (Snake above LGR) and downstream (John Day) Chinook salmon smolt life histories. Both production areas yield smolts of similar, but variable (on an inter-annual basis) size. Further, we demonstrate that a portion of fork length variation can be attributed to density-dependent effects. 2. Our analysis of trap-passage timing distributions illustrates that both upstream and downstream populations depart from natal streams within a similar timeframe. 3. Across the years under consideration, we found that upstream-origin smolts migrated to the estuary at a faster rate (~ twice as fast) than those emigrating from the John Day system. 4. Upstream-origin smolts arrived at the estuary later (~7-10 days) than John Day River Chinook salmon smolts. Chapter 9 Findings: We evaluated the magnitude and likely influence of size collection efficiency (i.e., detection probability) relationships on CSS outcomes for wild Chinook salmon in Chapter 9: 1. For LGR, the bypass site where the majority of CSS Chinook are collected and assigned to their respective treatment groups estimated size collection efficiency relationships were weak to nonexistent. At LGS and LMN, relationships were quite variable across the 5-year record and of comparable magnitude to those estimated by NOAA-Fisheries. xxi

23 2. Based on a comparison of realized size distributions remaining inriver or going into a barge, there were no clear differences between detected and undetected fish, across projects and years. This was especially true at LGR, where sizes were virtually identical for the study groups. Chapter 10 Findings: Chapter 10 presents a description of the simulator program, including the input parameters, and provides preliminary results from runs of simulated data through the bootstrap program to compare estimated values for key parameters to the simulated true values: 1. The simulator software program was enhanced in 2006 with a user-friendly interface for inputting parameter values for different simulation runs. 2. Estimates of number of smolts in the CSS study categories in LGR equivalents and associated 90% confidence intervals computed with the bootstrap program closely agreed (within 0.5%) with default simulated true data sets. xxii

24 CHAPTER 1 Introduction Fisheries agencies and tribes have developed a multi-year program, the Comparative Survival Study (CSS), for the purpose of monitoring and evaluating the impacts of the mitigation measures and actions (e.g., flow augmentation, spill, and transportation) under the National Marine Fisheries Service (NMFS) Biological Opinion to recover listed stocks. This annual report covers smolt migration and adult return data for PIT-tagged spring/summer Chinook of wild (1994 to 2004) and hatchery (1997 to 2004) origin. New this year is coverage of PITtagged wild and hatchery summer steelhead (1997 to 2003). All study fish used in this report were uniquely identifiable based on a passive integrated transponder (PIT) tag implanted in the body cavity during the smolts life stage and retained through their return as adults. These tagged fish can then be detected as juvenile and adults at several locations of the Snake and Columbia Rivers. Reductions in the number of individuals detected as the tagged fish age provide estimates of survival. This allows comparisons of survival over different life stages between fish with different experiences in the hydrosystem (e.g. transportation vs. inriver migrants and migration through various numbers of dams) as illustrated in Figure 1. Harvest Management R/S Hatchery Wild S/S Eggs Lower Granite Little Goose Lower Monumental Ice Harbor SAR T:C McNary John Day The Dalles Bonneville Estuary D = λ t / λ n Direct survival through dams Freshwater Smolts/ spawner Direct survival of transported fish Ocean Mainstem Spawning / Rearing Habitat Actions Hydrosystem Actions Estuary Habitat Actions Figure 1. Salmonid life cycle in the Snake River and lower Columbia River basins (source: Marmorek et al. 2004). 1

25 The CSS has PIT-tagged large numbers of hatchery Chinook to obtain adequate sample sizes for these different comparisons. In addition, PIT-tagged wild Chinook, wild steelhead, and hatchery steelhead from other regional studies have also been used for survival estimation. Estimates and comparisons include: (i) survival of migrating smolts over different reaches of the hydro system; (ii) smolt-to-adult survival rates (SARs) from either Lower Granite Dam (LGR) back to LGR (i.e., SAR LGR-to-LGR ) or Bonneville Dam (BON) back to LGR (i.e., SAR BON-to-LGR ) for fish transported around dams and those migrating inriver; (iii) the ratio of SAR LGR-to-LGR of transported fish to SAR LGR-to-LGR of inriver migrants (T/Cs); and, (iv) the ratio of SAR BON-to-LGR of transported fish to SAR BON-to-LGR of inriver fish (Ds). The objectives of the CSS are as follows: 1. Develop a long-term index of transport to inriver smolt-to-adult survival rates (SARs) for Snake River hatchery and wild spring/summer Chinook and hatchery and wild summer steelhead. This includes computing annual ratios of transport SAR to inriver SAR (measured from LGR to LGR) with associated confidence interval. 2. Develop a long-term index of survival rates from release of yearling Chinook smolts at hatcheries to return of adults to hatcheries. This objective includes partitioning survival rates from (i) hatchery (smolts) to LGR (smolts), (ii) LGR (smolts) to back to LGR (adults), and (iii) LGR (adults) to the hatchery (adults). 3. Compute and compare overall SARs for selected upriver and down-river spring/summer Chinook hatchery and wild stocks. 4. Begin a time series of SARs for use in regional long-term monitoring and evaluation. One use of the SAR index will be for assessment of temporal changes in patterns of life cycle survival (e.g., recruit/spawner or R/S residuals; Schaller et al. 1999; Deriso et al. 2001). For Snake River wild spring/summer Chinook, changes in SAR explained most of the changes observed in life cycle survival following Columbia River Basin hydroelectric development and operation (Petrosky et al. 2001). A second application, in combination with SARs from downriver stocks, would be for assessing temporal and spatial changes in life cycle survival. Temporal and spatial R/S patterns indicated survival and productivity of Snake River stocks declined more than downriver stocks following hydrosystem development and operation (Schaller et al. 1999; Deriso et al. 2001; Marmorek et al. 2004). The upriver/downriver SAR comparison (Objective 3) will shed additional light on life stage survival patterns that drives lifecycle survival for Snake River populations. Continuing these assessments with PIT-tagged fish in the CSS will provide an independent measure to past R/S data of survival rates from smolt to adult, which incorporates variation in hydrosystem experiences and environmental conditions in the estuary and (early) ocean. Spatial and temporal contrasts of survival rates from different life stages (adult-to-adult, adult-to-smolt, and smolt-to-adult) provide valuable information to diagnose where in the salmon life cycle mortality rates have increased, and allow indirect inferences about alternative causes. The 2006 CSS Annual Report updates the status of the CSS with the addition of the wild and hatchery sp/su Chinook from the 2004 migration year and wild and hatchery steelhead from 2

26 the 2003 migration year. This annual report provides new analyses addressing the question raised by NOAA Fisheries in the review of last years annual report, specifically, whether size of smolts may be having an impact on the comparison of SARs between transported and in-river migrants. NOAA researchers have stated that the size distribution of fish collected for transportation tends to be smaller than that of the undetected fish, and assuming a higher survival for larger fish, this could partly explain the lower than expected SARs for transported fish. Additional analyses of have been conducted on the adult returns drop-out rate between Bonneville Dam and Lower Granite Dam. Plus evaluation of the adequacy of comparing wild Chinook from upriver stocks to the downriver stock in the John Day River has been undertaken in the report. These analytical methodologies will be applied to other questions of interest to fishery managers and interested public during the preparation of next year s 10-year CSS Summary Report recommended by the Council s Independent Scientific Advisory Board (ISAB). The use of Akcakaya (2002) method to remove sampling error from the overall variance estimated from a time series of survival rate data, as covered in the 2005 CSS Annual Report, will be deferred to the 10-year CSS Summary Report and covered in greater detail in that document. 3

27 CHAPTER 2 Methods Sources of study fish Fish utilized in the CSS are marked with a unique-coded passive integrated transponder (PIT) tag, which was evaluated for use on salmonids by NOAA (Prentice et al. 1986). The computer chips are encapsulated in glass with a 12-mm length and 0.05-mm width. PIT tags are cylindrical in shape and impermeable to water. Individual PIT tags are implanted into the fish s underbelly using a hand-held syringe with a 12-gauge veterinary needle (PTOC 1999 PIT-Tag Marking Procedures Manual). Tag loss and mortality of PIT-tagged fish are monitored, and the tagging files are transferred to Pacific State Marine Fisheries Commission s regional PTAGIS database in Portland, OR. The PIT-tagged wild steelhead, hatchery steelhead, and wild Chinook used in the CSS analyses as aggregate marked populations should be as representative of the untagged population as possible. For wild fish, the collection and tagging occurs over lengthy time periods from parr stages to smolt stages in each sub-basin located above Lower Granite Dam including the Clearwater, Grande Ronde, Salmon, and Imnaha rivers. These wild fish were PIT-tagged by various organizations over a 10 to 12-month period with varied sampling gear including inclineplane (scoop) traps, screw traps, electrofishing, hook and line, and beach seining. At the hatcheries, fish were obtained across as wide a set of ponds and raceways as possible to allow effective representation of production. Most hatchery steelhead releases have a small number of PIT-tagged fish, typically between 200 and 1000 fish per individual hatchery. The aggregate of these PIT-tag releases provided a fairly good cross-section of the hatchery production in each year, although it was not proportional to the magnitude of each hatchery production. Likewise, the number of wild fish PIT-tagged in each tributary is not expected to be proportional to the total population present; however, with PIT tagging occurring across a wide range of the total population, the resulting SARs of this aggregate PIT-tag population should be adequately reflective of the total population. The PIT-tagged wild Chinook, wild steelhead, and hatchery steelhead used in the CSS were initially PIT-tagged to satisfy the goals of several different research studies. At certain times of the year, multiple age classes of fish were being PIT-tagged. To ensure that smolts in our annual aggregate groups were actually migrating out in the respective year of interest, fish detected entirely outside the migratory year of interest were excluded. This was necessary since estimates of collection efficiency and survival must reflect a single year. For wild Chinook, we found that limiting the tagging season to a 10-month period from July 25 to May 20 each year reduced the instances of overlapping age classes. In this 10-month period, few additional fish were excluded due to being detected at the dams or trawl in a year outside the migration year; this was less than 0.1% in all years except 1994 when it was 0.18%. For wild steelhead, we found that size at tagging was a useful parameter for removing a high proportion of fish that reside an extra year or two in freshwater beyond the desired migration year of study (Berggren et al. 2005). Excluding wild steelhead below 130 mm and above 299 mm reduced the instances of multiple age classes and allowed the tagging season to be a full 12-months from July 1 to June 30 each year. 4

28 Detection of study fish PIT-tagged smolts were detected at six Snake and Columbia River dams, including Lower Granite (LGR), Little Goose (LGS), Lower Monumental (LMN), McNary (MCN), John Day (JDA), and Bonneville (BON). In addition, PIT-tag detections were obtained at the NOAA Fisheries trawl (TWX) operated in the lower Columbia River half-way between BON and the mouth of the Columbia River. The above juvenile fish detection site abbreviations will be used throughout this document. When PIT-tagged smolts enter the bypass/collection facility of a dam from which transportation occurs, there are four potential outcomes. The tagged fish may (1) be returned-toriver under the default routing option, (2) be routed to the raceways for transportation if requested by the researcher, (3) be routed to the sample room for anesthetization and handling prior to being routed to transportation, and (4) be seen only on the separator detector coils and therefore have an unknown disposition at that site. For PIT-tagged wild steelhead, hatchery steelhead, and wild Chinook originating above LGR, the number of tagged fish specifically routed to transportation has been very small in most prior years prior to 2002 (wild Chinook) and 2003 (wild steelhead and some hatchery steelhead releases). Since the default operation has been to return PIT-tagged fish to the river at collector dams, the only reason some PIT-tagged wild Chinook, wild steelhead, and hatchery steelhead were transported in the early years was because (1) the daily timed subsampling intervals of the Smolt Monitoring Program over-rides the default return-to-river operation for PIT-tagged fish (sampled fish are usually transported) and (2) the occurrence of periods when equipment malfunctions caused the collected PIT-tagged fish to go to the raceways. Based on the detection history of PIT-tagged smolts at the collector dams, we are able to determine to which CSS study category (defined below) these PIT-tagged fish belong. PIT-tagged returning adults were detected in the Lower Granite Dam adult fish ladder (GRA) in each year. Beginning in return year 2002, detectors were installed in all the adult fish ladders at Bonneville (BOA) and McNary (MCA) dams, allowing detection of returning PITtagged adults at these additional locations. In 2003, Ice Harbor Dam (IHA to 4/1/2005 and ICH thereafter) was fitted with a PIT tag detection system in its fish ladder. Lower Granite Dam has PIT tag detection coils located near the adult trapping facility and at the exit section of the adult fish ladder. As noted last year, the LGR adult PIT-tag detection efficiency is 98% (Berggren et al. 2005), so no adjustments to the number of detected adult PIT-tagged fish at LGR are necessary. The above adult fish detection site abbreviations will be used throughout this document. Holdovers within the hydrosystem below Lower Granite Dam In the estimation of inriver survival rates with the Cormack(1964) Jolly (1965) Seber (1965) method (hereafter termed CJS), it is assumed that all PIT-tagged smolts in a group are outmigrating together in a single migration year. Any PIT-tagged fish detected as a smolt only in a year later than the expected migration year was excluded from the release group. This exclusionary clause was necessary particularly for wild Chinook and wild steelhead, because at times when multiple age classes were being PIT tagged, our constraints of size on steelhead and tagging dates on Chinook were not enough to remove non-migratory fish for the year of interest. However, PIT-tagged fish detected at an upper dam and then holding over within the 5

29 hydrosystem with subsequent detections occurring the following year, were handled as follows. The capture history code for these fish showed detections at dams only during the year they initiated their outmigration. The detections in the following year were excluded during the estimation of CJS reach survivals and project collection efficiencies. Fortunately, few yearling Chinook and steelhead delayed in the hydrosystem until the following year except for steelhead that began their migration in 2001 (Berggren et al. 2005). No additional holdovers were observed for migration years 2003 (steelhead) and 2004 (Chinook). Annual SARs for each study category The population of PIT-tagged study fish arriving at LGR is partitioned into three categories of smolts related to the manner of subsequent passage through the hydro system. Fish have the opportunity to either (1) pass inriver through the Snake River collector dams in a nonbypass channel route (spillways or turbines), (2) pass inriver through the dam s bypass channel, or (3) pass in a truck or barge to below BON. These three ways of hydro system passage is used to define the three study categories, C 0, C 1 and T 0, respectively, of the CSS. Typically, study categories T 0 and C 0 are the most representative of the run-at-large untagged population (exception is 1997 when most fish collected, tagged and untagged, in April and May at LGS and LMN were bypassed to the river). See Appendix A for the formulas used to estimate the number of smolts in each study category. The SAR formulas for each study category (i.e., SAR 1 (T 0 ), SAR 2 (T 0 ), SAR(C 1 ), and SAR(C 0 ) are provided in Appendix A. As WDFW member of the CSS Oversight Committee, Ryding (2006) provided the concept and detailed analytical rationale (with examples) behind why the smolt numbers estimated for CSS study categories must be in LGR-equivalents so that the ratio of SARs forming the parameter T/C is unbiased in the CSS 2006 Design and Analysis Technical Report. The parameter T/C used throughout the 2006 CSS Annual Report is estimated by SAR 2 (T 0 )/SAR(C 0 ). The SAR 2 (T 0 ) estimate is used in most comparisons since it is the least affected by years in which too few PIT-tagged smolts are being transported at LGS and LMN to provide any returning adults for estimating the site-specific SARs used in computing SAR 1 (T 0 ). In years when the same proportion of collected PIT-tagged smolts are being routed to transportation at each of the three Snake River collector dams, the two estimators SAR 1 (T 0 ) and SAR 2 (T 0 ) are equivalent. Annual estimates of SAR LGR-to-LGR reflective of the run-at-large for wild steelhead, hatchery steelhead, wild Chinook, and hatchery Chinook are computed by weighting the estimated study category specific SARs of PIT-tagged fish by the estimated proportion of the run-at-large represented with each respective study category. Ninety-percent confidence interval for number of smolts in each study category, SARs for each study category, T/C ratio, and annual SARs are computed using nonparametric bootstrapping methods (Efron and Tibshirani 1993). Annual estimates of D The parameter D is the ratio of post-bon survival rate of transported fish to in-river fish. Basically, D is computed as {SAR 2 (T 0 )/V T }/{SAR(C 0 )/V C }. The parameter V C is the overall reach survival from LGR to BON of fish in Category C 0. The parameter V T is the overall in-river survival from LGR to the transportation sites and on barges or trucks till released below BON for 6

30 fish in Category T 0. Regardless of whether SAR 1 (T 0 ) or SAR 2 (T 0 ) is used in the computation of D, the estimate of V T should be computed as 0.98 (t 2 +t 3 + t 4 )/(t 2 + t 3 /S 2 + t 4 /S 2 S 3 ). In the 2005 CSS Annual Report (Berggren et al. 2005), it was noted that computed V T estimates have ranged between 88 and 98%. 7

31 Wild spring/summer Chinook CHAPTER 3 SARs, T/C ratio, and D for Wild and Hatchery Chinook The wild PIT-tagged juvenile Chinook used in the CSS analyses were obtained from all available marking efforts in the Snake River basin above Lower Granite Dam. Wild Chinook from each tributary (plus fish tagged at the Snake River trap near Lewiston) were represented in the PIT-tag aggregates for migration years 1994 to 2004 (Table 1). A list of the locations within the tributaries where the PIT-tagged wild Chinook were released is provided in Appendix H. Table 1. Number of PIT-tagged wild Chinook parr/smolts from the four tributaries above Lower Granite Dam and Snake River trap used in the CSS analyses for migration years 1994 to Migr. Year Number of PIT-tagged wild Chinook utilized in CSS by location of origin Total PIT Clearwater Snake River Grande Salmon Tags River trap 1 Ronde River River Imnaha River (Rkm 308) (Rkm 224) (Rkm 225) (Rkm 271) (Rkm 303) ,659 8,292 1,423 8,828 27,725 3, ,640 17,605 1,948 12,330 40,609 2, ,523 2, ,079 7,016 4, , None 3,870 3,543 1, ,836 4, ,644 11,179 8, ,493 13,695 3,051 11,240 43,323 10, ,841 9,921 1,526 7,706 39,609 9, ,775 3, ,354 23,107 14, ,286 14,060 1,077 9,715 36,051 6, ,012 15, ,057 60,261 13, ,743 17, ,104 56,153 13,731 Average % of total 16.3% 1.8% 15.5% 53.1% 13.3% 1 Snake River trap collects fish originating in Salmon, Imnaha, and Grande Ronde rivers. Estimated numbers of wild Chinook smolts in each study category are presented in Appendix B Table B-1 along with the estimated population of tagged fish arriving Lower Granite Dam. This appendix table provides a bootstrapped 90% confidence interval around each estimate, along with the number of returning adults in each study category. Most PIT-tagged wild Chinook are in the C 1 study category due to the default operation of routing most PITtagged fish back to the river at the Snake River collector dams. Until 2002, the number of PITtagged wild Chinook actually transported has been relatively small relative to the number of untagged wild Chinook transported (Figure 2 and Appendix E Table E-1). Beginning in 2002, the CSS coordinated with IDFG, ODFW, and CTRUIR research programs to purposely route 50% of the first-time detected PIT-tagged wild Chinook smolts at the Snake River transportation facilities to the raceways for transportation. This action has provided more PIT-tagged wild Chinook smolts in the transportation category in recent years. The individual reach survival estimates used to expand PIT-tag smolt counts in each study category to LGR equivalents are presented in Appendix C Table C-1 for each migration year. 8

32 Proportion of PIT-tagged smolts being transported relative to the untagged fish collected and transported prop(lgr) prop(lgs) prop(lmn) Proportion Migration year Figure 2. Trend in proportion of PIT-tagged wild Chinook transported at each Snake River collector Dam, 1994 to All SARs for wild Chinook are computed with only returning adults, age 2-salt and older. The full age composition of the returning jacks and adults for each migration year 1994 to 2004 is shown in Appendix D Table D-1. On average, only about 4.3% of the returning PIT-tagged wild Chinook detected at Lower Granite Dam have been jacks. The site-specific transportation SAR estimates [e.g., SAR(T LGR ), SAR(T LGS ), and SAR(T LMN )] used in estimating SAR 1 (T 0 ) for wild Chinook are presented in Appendix E Table E-2. Because of the low number of PIT-tagged smolts transported and small number of returning adults, this study s ability to detect potential differences in site-specific SARs will be limited. The 90% confidence intervals of the site-specific SARs are extremely wide and overlapping across all three dams in all years of study. However, this does not impact the conduct of this study since our goal is to create an overall multi-dam estimate of transportation SAR for comparison with the SARs of in-river migrants. The completion of the adult returns for migration year 2003 and addition of migration year 2004 with 2-salt returns has shown two sequential years with extremely low estimated SAR LGR-to-LGR (Table 2, Figure 3), not exceeding 0.35% in any study category. Wild sp/su Chinook appear to be back at the pre-1997 levels, which does not bode well for recovery efforts. Marmorek et al recommended levels above 2% to maintain a stable population and levels above 4% for recovery. SAR levels above 2% have recently been estimated in only a few years with specific study categories (e.g., transport T 0 Category in 1999 and inriver C 0 Category in 1997, 1999, and 2000). Only in migration year 2001 was the transport SAR 2 (T 0 ) significantly higher than that of the inriver migrants based on non-overlapping 90% confidence intervals. 9

33 Table 2. Estimated SAR LGR-to-LGR (%) for PIT-tagged wild Chinook in annual aggregate for each study category from 1994 to 2004 (with 90% confidence intervals). Mig. Year SAR 1 (T 0 ) SAR 2 (T 0 ) SAR(C 0 ) SAR(C 1 ) 1994 NA ( ) 0.28 ( ) 0.09 ( ) 1995 NA 0.35 ( ) 0.37 ( ) 0.25 ( ) 1996 NA 0.50 ( ) 0.26 ( ) 0.17 ( ) 1997 NA 1.74 ( ) 2.35 ( ) 0.93 ( ) ( ) 1.18 ( ) 1.36 ( ) 1.08 ( ) ( ) 2.44 ( ) 2.13 ( ) 1.90 ( ) ( ) 1.43 ( ) 2.39 ( ) 2.39 ( ) 2001 NA 1.28 ( ) Assume = SAR(C 1 ) 0.14 ( ) ( ) 0.80 ( ) 1.22 ( ) 0.99 ( ) ( ) 0.34 ( ) 0.33 ( ) 0.17 ( ) ( ) 0.30 ( ) 0.31 ( ) 0.18 ( ) 11-yr Avg. Std Error NA Not applicable since some sites have no adult returns for estimating a site-specific SAR 2 Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/ SAR Estimate Wild Spring/Summer Chinook sart0 sarc Migration Year Figure 3. Estimated SAR LGR-to-LGR for PIT-tagged wild Chinook aggregate in transport (sart0) and inriver (sarc0) study categories for migration years 1994 to 2004 (only 2-salt adult returns for 2004). The PIT-tagged wild Chinook smolts with a prior detection at a collector dam in the Snake River (C 1 Group) continue to have a lower SAR than those smolts undetected at these dams (C 0 Group). During the 11-yr period 1994 to 2004 (Table 2), SAR(C 1 ) averaged approximately 32% lower than SAR(C 0 ). The trend in annual estimated SAR LGR-to-LGR reflective of the wild Chinook run-at-large that outmigrated in 1994 to 2004 is shown in Figure 4 (data used to compute these annual weighted estimates is presented in Appendix F Table F-1). The trend in these estimates over the 11-yr period has been highly variable, rising from below 0.5% before 1997 to highs of 2.4% in 1999 before dropping each year to below 0.35 % in 2003 and 2004 (2-salt returns). Current 10

34 overall annual SAR LGR-to-LGR estimates are far below the minimum 2% recommended in Marmorek et al. (1998) for holding the wild Chinook stocks stable, and the 4% recommended for recovery. SARs are also less than the NPPC interim 2-6% SAR (average 4%) goal (NPCC 2003). Estimated SAR LGR-to-LGR for wild sp/su Chinook 0.03 SAR Survival Rate Migration Year Figure 4. Trend in estimated annual SAR (with 90% confidence intervals for ) for wild Chinook based on PIT-tagged Chinook SARs in transport and inriver study categories weighted by estimated proportion of run-at-large in each study category for migration years 1994 to 2004 (only 2-salt adult returns for 2004). The estimated transport SAR to inriver SAR (T/C) ratio for the PIT-tagged wild sp/su Chinook is presented in Table 3. With the addition of 2004 (2-salt returns), the T/C geometric mean (excluding 2001) remains at last year s level of The T/C ratio for 2001 was 9-fold higher than the geometric mean of other years. The lower limit of the 90% confidence interval for T/C exceeded a value of 1 only in 2001, indicating a significantly higher SAR for transported wild Chinook than in-river fish in that year. The estimated inriver survival from LGR tailrace to BON tailrace (V C ) and delayed mortality D for the PIT-tagged wild spring/summer Chinook aggregate group is also presented in Table 3 for migration years 1994 through The 10-yr geometric mean (excluding 2001) of V C was 0.46, while the 2001 V C value was half that average at With V C averaging under 50%, the geometric mean T/C ratios should have exceeded 2.0 if delayed mortality was no greater for transported wild Chinook smolts after release below BON than for inriver migrants, but that was not the case as was shown earlier. In the absence of this differential delayed mortality, D should average 1. However, for wild Chinook, the 10-yr geometric mean (excluding 2001) of D was 0.48, while the 2001 D estimate was slightly greater than 2. The 90% confidence intervals around the estimated D show relatively low precision in most of the years available, indicating the difficulty of estimating a precise D parameter with small sample sizes of PIT-tagged wild Chinook available. The individual reach survival estimates used to obtain Vc are presented in Appendix C Table C-1 for each migration year. 11

35 Table 3. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged wild Chinook for migration years 1994 to 2004 (with 90% confidence intervals). Mig. Year V C SAR 2 (T 0 )/SAR(C 0 ) D (77% expansion) A 1.62 ( ) 0.36 ( ) B (51% expansion) 0.95 ( ) 0.42 ( ) B (77% expansion) 1.92 ( ) 0.92 ( ) B (77% expansion) 0.74 ( ) 0.40 ( ) B (25% expansion) 0.87 ( ) 0.55 ( ) B ( ) 1.15 ( ) 0.72 ( ) B ( ) 0.60 ( ) 0.32 ( ) ( ) 0.65 ( ) 0.44 ( ) ( ) 1.05 ( ) 0.68 ( ) 2004 C 0.40 ( ) 0.97 ( ) 0.40 ( ) Geomean D 0.23 ( ) 8.96 ( ) 2.2 ( ) B A Expansion shows percent of reach with a constant per/mile survival rate applied. B Migration year 1994 to 1999 confidence intervals for D are shifted 0.01 to 0.06 higher than reported in 2005 CSS Annual Report to adjust for the corrected D presented in that report. C Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/2006. D For migration year 2001, the SAR(C 1 ) value is used in the denominator of the T/C ratio. 12

36 Hatchery Chinook Yearling spring and summer Chinook were PIT-tagged for the CSS at specific hatcheries within the four drainages above Lower Granite Dam including the Clearwater, Salmon, Imnaha, and Grande Ronde rivers. Hatcheries that accounted for a major portion of the Chinook production in their respective drainage were selected. Since study inception, the CSS has PITtagged juvenile Chinook at McCall, Rapid River, Dworshak, and Lookingglass hatcheries (Table 4). Chinook tagged at Lookingglass Hatchery included an Imnaha River stock released in the Imnaha River drainage and a Catherine Creek stock released in the Grande Ronde River drainage. This latter stock became available to the CSS in 2001 to replace the earlier releases made on-site at Lookingglass Hatchery. The proportion of Chinook production released with PIT-tags at each of the hatcheries listed in Table 4 is presented in Appendix G Table G-1 along with median length at time of tagging. Table 4. Number of PIT-tagged hatchery Chinook parr/smolts from key hatcheries located above Lower Granite Dam used in the CSS analyses for migration years 1997 to Migr. Rapid Dworshak Catherine McCall H Imnaha AP Year River H NFH Creek AP ,452 14, ,652 13, ,336 47, ,340 19, ,812 47, ,985 19, ,748 47, ,705 20, ,091 55,142 20,915 55,127 20, ,908 54,725 20,796 54,734 20, ,763 54,708 20,628 74,317 20, ,969 51,616 20,994 71,363 20,910 The estimated population numbers (with bootstrapped 90% confidence intervals) of PITtagged Chinook smolts arriving at LGR for each CSS hatchery group are presented in Appendix B Tables B-2 to B-4 for spring stocks and Appendix B Tables B-5 to B-6 for summer stocks. The appendix tables also provide the estimated number of smolts (with bootstrapped 90% confidence intervals) occurring in each CSS study category, T 0, C 0, and C 1, along with number of returning adults in each study category. Unlike their wild Chinook counterparts, the PITtagged hatchery Chinook populations arriving at LGR were fairly well split across the three study categories in all years except Few PIT-tagged smolts were in Category C 0 in 2001 due to the lack of spill at collector dams and subsequent high collection efficiency allowing for few fish to pass the three Snake River collector dams undetected that year. In the other years there were relatively large numbers in categories T 0 and C 0. Fish in categories T 0 and C 0 mimic the untagged population in each year except 1997, when approximately 40% of the inriver migrating hatchery Chinook smolts were of Category C 0 and the remaining 60% were of Category C 1 due to the bypass protocols implemented during portions of April and May at LGS and LMN that year. The individual reach survival estimates used to expand smolt counts per category to LGR equivalents are presented in Appendix C Tables C-2 to C-6 for each migration year and hatchery. A portion of the CSS PIT-tagged hatchery Chinook was purposely diverted into transportation at LGR in each of the years 1997 to 2004, but this was not the case at the other 13

37 two Snake River transportation facilities until 2000 (Appendix E Table E-3 and Figure 5). Since 2000, the proportion of first-time detected fish being diverted to transportation at the three Snake River collector dams was held constant within each year. But for parts of 1998 and 1999 (routing PIT-tagged fish to transportation ended on May 9 in 1998 and commenced on May 10 in 1999), PIT-tagged hatchery Chinook were routed to transportation at LGS for the CSS. The CSS did not route PIT-tagged hatchery Chinook to transportation at LMN until Again in 2002, the CSS did not route PIT-tagged hatchery Chinook to transportation at LMN because of the non-standard operations implemented that year to reduce the numbers of fish collected and transported in the absence of spill at that site. This non-standard operation included primary bypass without PIT tag detections during most of April and alternating 2-day transport and 1-day primary bypass without PIT tag detections during May and part of June at LMN. Springtime transportation at MCN did not occur in migration years 1997 to Proportion of PIT-tagged smolts being transported relative to the untagged fish collected and transported prop(lgr) prop(lgs) prop(lmn) 0.6 Proportion Migration year Figure 5. Trend in proportion of PIT-tagged hatchery Chinook transported at each Snake River collector Dam, 1997 to All SARs for hatchery Chinook are computed with only returning adults, age 2-salt and older. The full age composition of the returning jacks and adults for each migration year 1997 to 2004 is shown in Appendix D Table D-3. The average percentage of the total return that return as jacks was higher for the summer Chinook stocks than for the spring Chinook stocks, and was the highest for Chinook from Imnaha River AP. Throughout this report, we classify the Imnaha River Chinook as a summer stock (contrary to ODFW classification) due to its high return rate of jacks and later timing of its returning adults, which coincides with the summer stock from McCall Hatchery stock. This highly variable jack return rate among the hatcheries and the 14

38 extremely low jack return rate observed with the wild Chinook is one reason that SARs computed in the CSS report include 2-salt and 3-salt returning adults and no jacks. The site-specific transportation SAR estimates [e.g., SAR(T LGR ), SAR(T LGS ), and SAR(T LMN )] used in estimating SAR 1 (T 0 ) for hatchery Chinook are presented in Appendix E Table E-4. Because of the low number of PIT-tagged smolts transported from LGS prior to 2000 and from LMN in any year, and small number of returning adults from these site s transported fish, this study s ability to detect potential differences in site-specific SARs will be limited. The 90% confidence intervals of the site-specific SARs are extremely wide and overlapping across all three dams in all years of study. However, this does not impact the conduct of this study since our goal is to create an overall multi-dam estimate of transportation SAR for comparison with the SARs of in-river migrants. Estimated SARs for hatchery Chinook in study categories T 0, C 0, and C 1 are presented in Tables 5 to 7 for spring Chinook stocks and Tables 8 to 9 for summer Chinook stocks. When routing the same proportion of first-detected PIT-tagged smolts to transportation at each of the three collector dams as occurred in migration years 2000, 2001, 2003, and 2004, both estimators SAR 1 (T 0 ) and SAR 2 (T 0 ) produce the same result, illustrating the benefits of having selfweighting occur across the three dams. Migration year 2004 (2-salt returns) is producing a very low SAR LGR-to-LGR for the second consecutive year for Rapid River Hatchery spring Chinook with all study categories having SARs at or below 0.25% (Table 5). These current estimated SARs are far below the magnitudes of 1998 to 2000 (Figure 6). Relative to the 8-year average SAR(C 0 ) of Rapid River Hatchery Chinook that passed the three collector dams undetected, a 57% higher transportation average SAR 2 (T 0 ) and 24% lower bypass average SAR(C 1 ) was estimated. Table 5. Estimated SAR LGR-to-LGR (%) for PIT-tagged spring Chinook from Rapid River Hatchery for each study category from 1997 to 2004 (with 90% confidence intervals). Mig. Year SAR 1 (T 0 ) SAR 2 (T 0 ) SAR(C 0 ) SAR(C 1 ) 1997 NA ( ) 0.46 ( ) 0.53 ( ) ( ) 2.00 ( ) 1.20 ( ) 0.67 ( ) ( ) 3.05 ( ) 2.37 ( ) 1.63 ( ) ( ) 2.10 ( ) 1.59 ( ) 1.35 ( ) ( ) 1.09 ( ) {Assume =SAR(C 1 )} 0.05 ( ) ( ) 1.01 ( ) 0.67 ( ) 0.63 ( ) ( ) 0.25 ( ) 0.23 ( ) 0.16 ( ) ( ) 0.26 ( ) 0.14 ( ) 0.09 ( ) 8-yr Avg. Std_error Not applicable since some sites have no adult returns for estimating a site-specific SAR. 2 Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/

39 SAR Estimate Rapid River Hatchery Spring Chinook sart0 sarc MIgration Year Figure 6. Trend in estimated transport and inriver SARs for Rapid River Hatchery spring Chinook for migration years 1997 to 2004 (latter with 2-salt adult returns). Migration year 2004 (2-salt returns) is producing a very low SAR LGR-to-LGR for the second consecutive year for Dworshak Hatchery spring Chinook with all study categories having SARs at or below 0.22% (Table 6). These current estimated SARs are far below the magnitudes of 1998 to 2000 (Figure 7). Relative to the 8-year average SAR(C 0 ) of Dworshak Hatchery Chinook that passed the three collector dams undetected, a 10% higher transportation average SAR 2 (T 0 ) and 20% lower bypass average SAR(C 1 ) was estimated. Table 6. Estimated SAR LGR-to-LGR (%) for PIT-tagged spring Chinook from Dworshak Hatchery for each study category from 1997 to 2004 (with 90% confidence intervals). Mig. Year SAR 1 (T 0 ) SAR 2 (T 0 ) SAR(C 0 ) SAR(C 1 ) 1997 NA ( ) 0.47 ( ) 0.36 ( ) 1998 NA 0.90 ( ) 1.25 ( ) 0.91 ( ) ( ) 1.18 ( ) 1.19 ( ) 0.95 ( ) ( ) 1.00 ( ) 1.01 ( ) 0.85 ( ) ( ) 0.36 ( ) {Assume =SAR(C 1 )} 0.04 ( ) ( ) 0.62 ( ) 0.50 ( ) 0.50 ( ) ( ) 0.26 ( ) 0.21 ( ) 0.18 ( ) ( ) 0.21 ( ) 0.22 ( ) 0.16 ( ) 8-yr Avg. Std_error Not applicable since some sites have no adult returns for estimating a site-specific SAR. 2 Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/

40 SAR Estimate Dworshak Hatchery Spring Chinook sart0 sarc Migration Year Figure 7. Estimated transport and inriver SARs for PIT-tagged Dworshak Hatchery spring Chinook for migration years 1997 to 2004 (latter with 2-salt adult returns). Migration year 2004 (2-salt returns) is producing a very low SAR LGR-to-LGR for the second consecutive year for Catherine Creek AP spring Chinook with all study categories having SARs at or below 0.35% (Table 7 and Figure 8). Relative to the 4-year average SAR(C 0 ) of Catherine Creek AP Chinook that passed the three collector dams undetected, a 84% higher transportation average SAR 2 (T 0 ) and 4% higher bypass average SAR(C 1 ) was estimated. Table 7. Estimated SAR LGR-to-LGR (%) for PIT-tagged spring Chinook from Catherine Creek AP for each study category from 2001 to 2004 (with 90% confidence intervals). Mig. Year SAR 1 (T 0 ) SAR 2 (T 0 ) SAR(C 0 ) SAR(C 1 ) 2001 NA ( ) {Assume =SAR(C 1 )} 0.04 ( ) 2002 NA 0.89 ( ) 0.49 ( ) 0.32 ( ) 2003 NA 0.36 ( ) 0.25 ( ) 0.36 ( ) ( ) 0.35 ( ) 0.20 ( ) 0.32 ( ) 4-yr Avg. Std_error Not applicable since some sites have no adult returns for estimating a site-specific SAR. 2 Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/

41 SAR Estimate Catherine Creek Hatchery Spring Chinook sart0 sarc Migration Year Figure 8. Estimated transport and inriver SARs for PIT-tagged Catherine Creek Acclimation Pond spring Chinook for migration years 2001 to 2004 (latter with 2-salt adult returns). Migration year 2004 (2-salt returns) is producing a very low SAR LGR-to-LGR for the second consecutive year for McCall Hatchery spring Chinook with all study categories having SARs at or below 0.31%, a level lower than last year (Table 8). These current estimated SARs are far below the magnitudes of 1998 to 2000 (Figure 9). Relative to the 8-year average SAR(C 0 ) of McCall Hatchery Chinook that passed the three collector dams undetected, a 77% higher transportation average SAR 2 (T 0 ) and 15% lower bypass average SAR(C 1 ) was estimated. Table 8. Estimated SAR LGR-to-LGR (%) for PIT-tagged summer Chinook from McCall Hatchery for each study category from 1997 to 2004 (with 90% confidence intervals). Mig. Year SAR 1 (T 0 ) SAR 2 (T 0 ) SAR(C 0 ) SAR(C 1 ) ( ) 1.52 ( ) 1.09 ( ) 1.10 ( ) ( ) 2.71 ( ) 1.38 ( ) 0.73 ( ) ( ) 3.61 ( ) 2.40 ( ) 2.05 ( ) ( ) 3.91 ( ) 2.06 ( ) 2.08 ( ) ( ) 1.24 ( ) {Assume =SAR(C 1 )} 0.04 ( ) ( ) 1.49 ( ) 1.03 ( ) 1.02 ( ) ( ) 0.79 ( ) 0.54 ( ) 0.35 ( ) NA ( ) 0.25 ( ) 0.12 ( ) 8-yr Avg. Std_error Not applicable since some sites have no adult returns for estimating a site-specific SAR. 2 Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/

42 SAR Estimate McCall Hatchery Summer Chinook sart0 sarc Migration Year Figure 9 Estimated transport and inriver SARs for PIT-tagged McCall Hatchery summer Chinook for migration years 1997 to 2004 (latter with 2-salt adult returns). Migration year 2004 (2-salt returns) is producing a very low SAR LGR-to-LGR for the second consecutive year for Imnaha AP spring Chinook with all study categories having SARs at or below 0.35%, a level lower than last year (Table 9). These current estimated SARs are far below the magnitudes of 1999 to 2000 (Figure 10). Relative to the 8-year average SAR(C 0 ) of McCall Hatchery Chinook that passed the three collector dams undetected, a 58% higher transportation average SAR 2 (T 0 ) and 23% lower bypass average SAR(C 1 ) was estimated. Table 9. Estimated SAR LGR-to-LGR (%) for PIT-tagged summer Chinook from Imnaha River AP for each study category from 1997 to 2004 (with 90% confidence intervals). Mig. Year SAR 1 (T 0 ) SAR 2 (T 0 ) SAR(C 0 ) SAR(C 1 ) 1997 NA ( ) 0.86 ( ) 0.69 ( ) 1998 NA 0.86 ( ) 0.55 ( ) 0.30 ( ) ( ) 2.72 ( ) 1.43 ( ) 1.22 ( ) ( ) 3.15 ( ) 2.41 ( ) 1.64 ( ) 2001 NA 0.62 ( ) {Assume =SAR(C 1 )} 0.06 ( ) ( ) 0.80 ( ) 0.45 ( ) 0.54 ( ) ( ) 0.58 ( ) 0.48 ( ) 0.38 ( ) ( ) 0.35 ( ) 0.23 ( ) 0.11 ( ) 8-yr Avg. Std_ error Not applicable since some sites have no adult returns for estimating a site-specific SAR. 2 Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/

43 SAR Estimate Imnaha Hatchery Summer Chinook sart0 sarc Migration Year Figure 10. Estimated transport and inriver SARs for PIT-tagged Imnaha River Acclimation Pond summer Chinook for migration years 1997 to 2004 (latter with 2-salt adult returns). The trend in annual SAR LGR-to-LGR for each hatchery and wild Chinook is presented in Figure 11. These annual SARs are computed using the estimated proportion transported and migrating in-river for the run-at-large as weights with the study specific SARs shown in Appendix F Table F-1. A general pattern of increasing SARs from 1997 to 1999 and decreasing SARs from 1999 to 2001 is shown for hatchery Chinook from McCall, Rapid River, and Dworshak hatcheries. Unlike the other three hatcheries, SAR (LGR-to-LGR) SAR (LGR-to-LGR) for Hatchery and Wild Chinook Wild DWOR RAPH MCCA IMNA CATH Figure 11. Trend in estimated annual SAR LGR-to-LGR for hatchery and wild sp/su Chinook; based on SAR estimates in transport and inriver categories weighted by estimated proportion of run-at-large in each category for migration years 1994 to 2004 (only 2-salt adult returns for 2004). 20

44 the SARs of Imnaha Hatchery Chinook dipped in 1998 and peaked in The annual trends observed for the PIT-tagged wild sp/su Chinook aggregate was similar to that of Imnaha Hatchery Chinook from 1997 to 1999 and similar to that of Rapid River Hatchery Chinook from 1999 to A slight increase in annual overall SAR was seen in 2002 after the drought year of 2001, followed by a large drop again in 2003 and low levels continuing in From the patterns of annual SARs the Rapid River Hatchery Chinook had the most similar trend as the PIT-tagged wild Chinook aggregate across the 8 years of adult returns for the four hatcheries continuously used in the CSS since Estimated in-river survival rates from Lower Granite Dam tailrace to Bonneville Dam tailrace (parameter Vc) were low in 2004, ranging between 0.33 and 0.44 for hatchery Chinook from Rapid River, Catherine Creek, Imnaha, and McCall facilities, whereas Dworshak Hatchery Chinook had an in-river survival rate estimate of 0.50 for 2004, which is close in magnitude to its 7-yr geometric mean (0.54) of covering and (Tables 10 to 14). Although not as low at the in-river survival estimates during the drought year 2001, the 2004 estimates for the other four hatcheries were well below their 7-yr geometric means ranging between 0.49 and The individual reach survival estimates for each migration year and hatchery used to compute Vc are presented in Appendix C Tables C-2 to C-6. Annual trends in Vc over the period 1994 to 2004 (hatchery Chinook beginning 1997) are presented in Figure 12 for both wild and hatchery Chinook. Table 10. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged Rapid River Hatchery spring Chinook for 1997 to 2004 (with 90% confidence intervals). Mig. Year V C SAR 2 (T 0 )/SAR(C 0 ) D (77% expansion) A 1.73 ( ) 0.61 ( ) B (25% expansion) 1.66 ( ) 1.01 ( ) B ( ) 1.28 ( ) 0.79 ( ) B ( ) 1.32 ( ) 0.82 ( ) B ( ) 1.52 ( ) 1.14 ( ) B ( ) 1.07 ( ) 0.75 ( ) 2004 C 0.35 ( ) 1.79 ( ) 0.65 ( ) Geometric mean D 0.33 ( ) 21.7 ( ) 7.3 ( ) B A Expansion shows percent of reach with a constant per/mile survival rate applied. B Migration year 1997 to 2002 confidence intervals for D are shifted 0.02 to 0.1 higher than reported in 2005 CSS to adjust for the corrected D presented in that report. C Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/2006. D For migration year 2001, the SAR(C 1 ) value is used in the denominator of the T/C ratio. 21

45 Table 11. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged Dworshak Hatchery spring Chinook for 1997 to 2004 (with 90% confidence intervals). Mig. Year V C SAR 2 (T 0 )/SAR(C 0 ) D (77% expansion) A 1.75 ( ) 0.88 ( ) B (25% expansion) 0.72 ( ) 0.37 ( ) B ( ) 0.99 ( ) 0.60 ( ) B ( ) 0.99 ( ) 0.53 ( ) B ( ) 1.24 ( ) 0.84 ( ) B ( ) 1.20 ( ) 0.87 ( ) 2004 C 0.50 ( ) 0.95 ( ) 0.49 ( ) Geometric mean D 0.24 ( ) 8.76 ( ) 2.2 ( ) B A Expansion shows percent of reach with a constant per/mile survival rate applied. B Migration year 1997 to 2002 confidence intervals for D are shifted 0.0 to 0.06 higher than reported in 2005 CSS to adjust for the corrected D presented in that report. C Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/2006. D For migration year 2001, the SAR(C 1 ) value is used in the denominator of the T/C ratio. Table 12. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged Catherine Creek AP spring Chinook for 2001 to 2004 (with 90% confidence intervals). Mig. Year V C SAR 2 (T 0 )/SAR(C 0 ) D ( ) 1.82 ( ) 1.23 ( ) B (25% expansion) A 1.44 ( ) 0.93 ( ) 2004 C 0.33 ( ) 1.75 ( ) 0.59 ( ) Geometric mean D 0.25 ( ) 5.32 ( ) 1.38 ( ) B A Expansion shows percent of reach with a constant per/mile survival rate applied. B Migration year 2001 to 2002 confidence intervals for D are shifted 0.03 higher than reported in 2005 CSS to adjust for the corrected D presented in that report. C Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/2006. D For migration year 2001, the SAR(C 1 ) value is used in the denominator of the T/C ratio. Table 13. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged McCall Hatchery summer Chinook for 1997 to 2004 (with 90% confidence intervals). Mig. Year V C SAR 2 (T 0 )/SAR(C 0 ) D (77% expansion) A 1.39 ( ) 0.64 ( ) B (25% expansion) 1.97 ( ) 1.17 ( ) B ( ) 1.50 ( ) 0.87 ( ) B ( ) 1.90 ( ) 1.25 ( ) B ( ) 1.45 ( ) 0.88 ( ) B ( ) 1.46 ( ) 1.08 ( ) 2004 C 0.44 ( ) 1.23 ( ) 0.55 ( ) Geometric mean D 0.27 ( ) 31.9 ( ) 9.0 ( ) B A Expansion shows percent of reach with a constant per/mile survival rate applied. B Migration year 1997 to 2002 confidence intervals for D are shifted 0.03 to 0.2 higher than reported in 2005 CSS to adjust for the corrected D presented in that report. C Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/2006. D For migration year 2001, the SAR(C 1 ) value is used in the denominator of the T/C ratio. 22

46 Table 14. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged Imnaha AP summer Chinook for 1997 to 2004 (with 90% confidence intervals). Mig. Year V C SAR 2 (T 0 )/SAR(C 0 ) D (77% expansion) A 1.37 ( ) 0.45 ( ) B (25% expansion) 1.56 ( ) 0.87 ( ) B ( ) 1.90 ( ) 1.12 ( B ( ) 1.30 ( ) 0.82 ( ) B ( ) 1.76 ( ) 0.96 ( ) B (25% expansion) 1.21 ( ) 0.91 ( ) 2004 B 0.37 ( ) 1.50 ( ) 0.58 ( ) Geometric mean C 0.37 ( ) 10.8 ( ) 4.2 ( ) B A Expansion shows percent of reach with a constant per/mile survival rate applied. B Migration year 1997 to 2002 confidence intervals for D are shifted 0.02 to 0.08 higher than reported in 2005 CSS to adjust for the corrected D presented in that report. C Migration year 2004 is incomplete with Age 2-salt adult returns through 8/9/2006. D For migration year 2001, the SAR(C 1 ) value is used in the denominator of the T/C ratio. Vc V c for Hatchery and Wild Chinook Vc-Wild Vc-DWOR Vc-RAPH Vc-MCCA Vc-IMNA Vc-CATH Figure 12. Trend in in-river survival (V c ) for PIT-tagged Snake River wild and hatchery spring/summer Chinook in migration years 1994 to Excluding migration year 2001, which had T/C ratios exceeding 5 in all hatchery groups, geometric mean T/C ratios covering the seven years from and have been around 1.5 for Rapid River, Imnaha, and McCall Hatchery Chinook (Tables 10, 13, and 14). For Dworshak Hatchery Chinook, the 7-yr geometric mean T/C ratio was less than 1.1 (Table 11). Although Catherine Creek AP hatchery Chinook have a shorter time series of data (Table 12), its T/C ratios tend to follow the former three hatcheries closer than Dworshak Hatchery. Trends in T/C ratio (log transformed) are presented in Figure 13. A significant increase in the transport SAR over the inriver SAR is found when the lower limit of the 90% confidence interval of the T/C ratio estimates is greater than 1. This did not occurring with any of the five hatcheries in In prior years, estimated T/C ratios significantly greater than 1 were observed in most years for Rapid River and McCall hatchery Chinook and about half the time for Imnaha AP 23

47 hatchery Chinook. Significant T/C ratios have never been observed for Dworshak Hatchery Chinook. In the absence of this differential delayed mortality, D should average close to 1. However, except for 2001 when estimated D was greater than 1 at each hatchery, the remaining years have seen a 7-yr geometric mean D of 0.62 at Dworshak, 0.78 at Imnaha, 0.81 at Rapid River, and 0.89 at McCall hatcheries. Trends in D (log transformed) are presented in Figure 14. ln(t/c) ln(t/c) for Hatchery and Wild Chinook ln(t/c)-wild ln(t/c)-dwor ln(t/c)-raph ln(t/c)-mcca ln(t/c)-imnh ln(t/c)-cath Figure 13. Trend in ratio of SAR 2 (T 0 )/SAR(C 0 ) (log-transformed) for PIT-tagged Snake river hatchery and wild Chinook for migration years 1994 to ln(d) ln(d ) for Hatchery and Wild Chinook ln(d)-wild ln(d)-dwor ln(d)-raph ln(d)-mcca ln(d)-imna ln(d)-cath Figure 14. Trend in D (log-transformed) for PIT-tagged Snake River hatchery and wild Chinook in migration years

48 While wild and hatchery populations demonstrated differences in magnitude for some parameters (T/C, D, and SARs), the annual patterns of these parameters were similar among wild and hatchery populations. In-river survival (Vc) of the wild population tracked closely with survival of hatchery populations across years (Figure 12). While T/C ratios were higher for Snake River hatcheries than for wild fish, the T/C pattern for the wild population tracked well with those of the hatchery populations across years (Figure 13). Similarly, Snake River hatchery fish had higher D values than wild fish, but wild and hatchery Ds also tracked well across years (Figure 14). SARs for wild Snake River spring/summer Chinook were intermediate to the different hatcheries, but like other metrics, hatchery SARs tracked wild SAR patterns (Figure 11). Given the high variability in survival for Snake River Chinook populations, we will most likely encounter future years when the abundance of wild juveniles is too low for generating a reliable SAR estimate. In that situation, we will need to rely on surrogate estimates from hatchery-produced fish. This provides a rationale for establishing the relationship between survival of wild and hatchery populations under similar migration, climate, and ocean conditions. CONCLUSIONS 1. The annual SARs (indexed LGR smolts-to-lgr adults) for wild Snake River sp/su Chinook has been highly variable, rising from below 0.5% before 1997 to highs of 2.4% in 1999 before dropping each year to below 0.35 % in 2003 and 2004 (2-salt returns). Current overall annual SAR LGR-to-LGR estimates are far below the minimum 2% recommended in the NPCC Fish and Wildlife Program mainstem amendments (NPCC 2003), and estimated as needed for keeping the stocks stable (Marmorek et al. 1998). 2. Transportation provided little or no benefit to wild sp/su Chinook during the conditions experienced in most years during , except during the severe drought year The 10-year geometric mean (excluding 2001) SAR ratio transported to inriver migrants (T/C) was 0.98, while in 2001, the T/C was approximately 9-fold higher. The T/C ratio was significantly > 1 in only Delayed mortality of transported wild sp/su Chinook smolts was substantial most years relative to that of inriver migrants, based on a 10-yr geometric mean D estimate (excluding 2001) of 0.49, indicating transported smolts died at twice the rate as inriver migrants once they passed BON tailrace. In 2001, D was greater than 2, indicating inriver migrants died at twice the rate of transported smolts in the estuary and ocean. 4. The estimated inriver survival of wild sp/su Chinook from LGR tailrace to Bonneville Dam (BON) tailrace averaged 0.46 (geometric mean) for (excluding 2001, when estimated survival was 0.23). 5. During the 11-yr period 1994 to 2004, SAR(C 1 ) averaged approximately 32% lower than SAR(C 0 ) for wild sp/su Chinook. 25

49 6. SARs (LGR-to-LGR) for hatchery Snake River spring/summer Chinook have shown similar patterns as wild Chinook during , although the actual survival rates have differed among hatcheries and between spring and summer runs. For spring Chinook hatcheries, SARs for Rapid River Hatchery have exceeded those of Dworshak Hatchery, and SARs of hatchery summer Chinook (particularly from McCall) have exceeded those of hatchery spring Chinook. SARs of most hatchery Chinook (except Dworshak) have equaled or exceeded the SARs of wild Chinook in migration years In general, transportation provided benefits most years to Snake River hatchery sp/su Chinook , however benefits varied among hatcheries. Omitting 2001 (when all T/C ratios exceeded 5), the 7-year geometric mean T/C ranged from 1.08 at Dworshak, 1.46 at Rapid River, 1.50 at Imnaha and 1.54 at McCall hatcheries, indicating a higher return rate for the transported Chinook from these latter three hatcheries. Although having a shorter time series, annual T/C ratios at Catherine Creek AP hatchery Chinook have remained greater than Delayed mortality of transported hatchery spring and summer Chinook smolts was evident most years relative to that of inriver migrants, based on estimated values of D. Except for 2001 when all D values exceeded 1, the other seven years produced geometric mean D values of 0.62 at Dworshak, 0.78 at Imnaha, 0.81 at Rapid River, and 0.89 at McCall hatcheries. 9. The 7-yr ( , ) geometric mean of the estimated inriver reach survival rate of hatchery sp/su Chinook from LGR tailrace to BON tailrace ranged from 0.49 to 0.54 across hatcheries. In 2001, the estimated reach survival rate ranged from 0.27 to 0.37 across hatcheries. 10. During the 8-yr period 1997 to 2004, SAR(C 1 ) has remained lower than lower than SAR(C 0 ) for Chinook from Rapid River, Dworshak, Imnaha, and McCall hatcheries. 11. While wild and hatchery populations demonstrated differences in magnitude for some parameters (T/C, D and SARs), the annual patterns of these parameters were highly correlated among wild and hatchery populations. 26

50 CHAPTER 4 SARs, T/C ratios, and D for Wild and Hatchery Steelhead Wild Steelhead The wild PIT-tagged juvenile steelhead used in the CSS analyses are obtained from all available marking efforts in the Snake River basin above Lower Granite Dam. Wild steelhead smolts from each tributary (plus fish tagged at the Snake River trap near Lewiston) were represented in the PIT-tag aggregates for migration years 1997 to 2003 (Table 15). A list of the locations within these tributaries where the PIT-tagged wild steelhead were released is provided in Appendix H. Table 15. Number of PIT-tagged wild steelhead smolts from the four tributaries above Lower Granite Dam (plus Snake River trap) used in the CSS for migration years 1997 to Migr. Year Number of PIT-tagged wild steelhead (>130 mm) utilized in CSS by location of origin Total PIT Clearwater Snake River Grande Salmon Tags River trap 1 Ronde River River Imnaha River (Rkm 308) (Rkm 224) (Rkm 225) (Rkm 271) (Rkm 303) ,703 5, , ,512 4,131 1, ,683 2, ,763 5, ,628 5,569 2, ,254 8,688 1,211 3,618 6,245 4, ,487 8, ,370 7,844 3, ,183 10,206 2,368 3,353 6,136 3, ,284 5,885 1,197 4,261 6,969 5,972 Average % of total 36.6% 5.8% 13.1% 26.9% 17.6% 1 Snake River trap located at Lewiston, ID, collects wild steelhead originating in Grande Ronde, Salmon, and Imnaha rivers. The estimated number of PIT-tagged wild steelhead smolts (with bootstrapped 90% confidence intervals) arriving at LGR for each CSS study category, T 0, C 0, and C 1, is presented in Appendix B Table B-7 along with the associated number of returning adults in each study category. Through migration year 2002, few PIT-tagged wild steelhead are in the T 0 study category due to the default operation of routing most PIT-tagged fish back to the river at the Snake River collector dams. Until 2003, the number of PIT-tagged wild steelhead actually transported has been relatively small relative to the number of untagged wild steelhead transported (Figure 15 and Appendix E Table E-5). Beginning in 2003, more PIT-tagged wild steelhead have become available in the transport group as state and tribal research programs allowed a portion of their PIT-tagged wild steelhead smolts to be routed to the raceways at Snake River transportation facilities. 27

51 Proportion of PIT-tagged smolts being transported relative to the untagged fish collected and transported prop(lgr) prop(lgs) prop(lmn) 0.20 Proportion Migration year Figure 15. Trend in proportion of PIT-tagged wild steelhead transported at each Snake River collector dam, 1997 to The age breakdown of the returning wild steelhead adults detected at Lower Granite Dam adult fish ladder from smolts outmigrating in 1997 to 2003 is shown in Appendix D Table D-4. For migration years 1997 to 2003, the average age of return to Lower Granite Dam is 42.7% salt, 55.0% 2 + -salt, and 2.3% 3 + -salt (hereafter the + notation on age will be dropped in both text and tables). Steelhead returning in the fall of the same year that they migrate to the ocean are not included in the adult return data. Because of the small percentage of age-3 salt returning wild steelhead, the return data for migration year 2003 should be close to complete with regard to computing SARs, T/C ratios, or D estimates for that migration year. Obtaining a valid estimate of the number of PIT-tagged wild steelhead in Category C 0 in 2001 is problematic due to apparent large amount of residualism that year. This is based on the finding that most inriver migrants with an adult return were hold-overs. Six of the eight adult returns of Category C 1 wild steelhead from migration year 2001 were actually detected in the lower river in For the three PIT-tagged wild steelhead adult returns with no detection in 2001, it was more likely these fish either completed their smolt migration in 2002 or passed undetected into the raceways during a computer outage in mid-may at LGR than traversed the entire hydrosystem undetected in 2001, when <1% of the wild steelhead run-at-large was estimated to be destined to ever pass all three Snake River collector dams through turbines (no spill route available). Because of the uncertainty in passage route and timing of the undetected PIT-tagged wild steelhead smolts in 2001, the in-river SAR utilizing fish from Category C 1 rather than Category C 0 will be used in comparisons with the transport SARs that year. 28

52 The site-specific transportation SAR estimates [e.g., SAR(T LGR ), SAR(T LGS ), and SAR(T LMN )] used in estimating SAR 1 (T 0 ) for wild steelhead are presented in Appendix E Table E-6. Because of the low number of PIT-tagged smolts transported and small number of returning adults, this study s ability to detect potential differences in site-specific SARs was limited. The 90% confidence intervals of the site-specific SARs are extremely wide and overlapping across all three dams in all years of study. However, this does not impact the conduct of this study since our goal is to create an overall multi-dam estimate of transportation SAR for comparison with the SARs of in-river migrants. The SARs for wild steelhead from migration year 2003 remained in the 2% vicinity for transported fish, but was around 0.5% for the in-river migrants (Table 16). These differences were significant based on non-overlapping 90% confidence intervals. Significant differences in estimated SARs between transported and in-river migrants were also observed for migration years 2001 and 2002 (Figure 16). Relative to the 7-year average SAR(C 0 ) of wild steelhead that passed the three collector dams undetected, a 138% higher transportation average SAR 2 (T 0 ) and 27% lower bypass average SAR(C 1 ) was estimated. Table 16. Estimated SAR LGR-to-LGR (%) for PIT-tagged wild steelhead in annual aggregate for each study category from 1997 to 2003 (with 90% confidence intervals). Mig. Year SAR 1 (T 0 ) SAR 2 (T 0 ) SAR(C 0 ) SAR(C 1 ) 1997 NA ( ) 0.66 ( ) 0.23 ( ) 1998 NA 0.21 ( ) 1.07 ( ) 0.23 ( ) ( ) 3.07 ( ) 1.35 ( ) 0.77 ( ) ( ) 2.79 ( ) 1.92 ( ) 1.82 ( ) 2001 NA 2.49 ( ) {Assume =SAR(C 1 )} 0.07 ( ) ( ) 2.84 ( ) 0.67 ( ) 0.94 ( ) ( ) 1.99 ( ) 0.48 ( ) 0.52 ( ) 7-yr Avg. Std_error Not applicable since some sites have no adult returns for estimating a site-specific SAR. 2 Migration year 2003 is incomplete until 3-salt adult returns occur at GRA. 29

53 0.05 Wild Steelhead SAR Estimate sart0 sarc Migration Year Figure 16. Estimated transport and inriver SARs (with 90% confidence intervals) for PIT-tagged wild steelhead aggregate for migration years 1997 to 2003 (incomplete 2003 returns). Annual estimates of SAR LGR-to-LGR for Snake River wild steelhead have dropped each year from the high of 2.86% estimated in 1999 to 1.57% estimated in Although the wild steelhead estimated annual SARs for migration years 1999 to 2002 were at or above the NPCC interim objective for a minimum SAR of 2%, they remain below the recommended average of 4% SAR (Appendix F Table F-4 and Figure 17). The pattern of decreasing estimated annual SARs for wild steelhead is following that of the wild Chinook, just not dropping as rapidly over the migration years 1999 to Annual SAR Estimate Estimated Annual SAR LGR-toLGR for Wild Steedhead Compared to Wild Chinook Steelhead Chinook Migration Year Figure 17. Estimated annual SAR for wild steelhead compared to wild Chinook with 90% confidence intervals; based on SAR estimates in transport and inriver categories weighted by estimated proportion of run-at-large in each category for migration years 1997 to

54 The estimated inriver survival from LGR tailrace to BON tailrace (V C ), transport SAR to inriver SAR (T/C) ratio, and delayed mortality D for the PIT-tagged wild steelhead aggregate group are presented in Table 17 for migration years 1997 to The individual reach survival estimates for each migration year used to obtain V C are presented in Appendix Table C Table C-7. The geometric mean of V C for 1997 to 2002, excluding 2001, was Over these same six years, the wild Chinook V C estimates had a geometric mean of 0.56, which was 27% higher. Figure 18 shows the tend in annual Vc estimates for wild steelhead compared to wild Chinook for A T/C estimate > 2 with a corresponding D estimate > 1 occurred in 5 of 7 years for PIT-tagged wild steelhead (3 of the 5 T/C ratios and 2 of the 5 D estimates were significant based on the lower limit of the 90% confidence interval). Whereas the PIT-tagged wild Chinook had only a single T/C estimate > 2 (and significant) for drought year 2001, but the corresponding D estimate, though > 1, was not significant based on the lower limit of the 90% confidence intervals (Figures 19 and 20). Excluding 2001, the geometric mean T/C of 1.72 for wild steelhead was double that computed for wild Chinook over these same six years (geometric mean of 0.82). The resulting D estimates for and had a geometric mean of 0.80 for wild steelhead and 0.50 for wild Chinook (trend across years shown in Figure 20). These data suggest a very different response to transportation as a recovery tool for listed wild Chinook and wild steelhead. Table 17. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged wild steelhead for migration years 1997 to 2003 (with 90% confidence intervals). Mig. Year V C SAR 2 (T 0 )/SAR(C 0 ) D (25% expansion) A 2.20 ( ) 1.18 ( ) B (25% expansion) 0.20 ( ) 0.11 ( ) B ( ) 2.28 ( ) 1.07 ( ) B (25% expansion) 1.45 ( ) 0.50 ( ) B ( ) 4.25 ( ) 2.24 ( ) 2003 C 0.37 ( ) 4.13 ( ) 1.64 ( ) Geometric Mean ( ) 37.0 ( ) 1.46 ( ) B A Expansion shows percent of reach with a constant per/mile survival rate applied. B Migration year 1997 to 2001 confidence intervals for D are shifted 0.0 to 0.05 higher than reported in 2005 CSS to adjust for the corrected D presented in that report. C Migration year 2003 is incomplete until 3-salt adult returns occur at GRA. 31

55 Vc V c for Wild Steelhead Compared to Wild Chinook Steelhead Chinook Figure 18. Trend in in-river survival (V c ) for PIT-tagged Snake River wild steelhead and wild Chinook for migration years 1997 to ln(t/c) for Wild Steelhead Compared to Wild Chinook ln(t/c) Steelhead Chinook Figure 19. ln(sar 2 (T 0 )/SAR(C 0 )) for PIT-tagged wild steelhead and wild Chinook from migration years 1997 to

56 ln(d ) for Wild Steelhead Compared to Wild Chinook ln(d ) Steelhead Chinook Figure 20. Trend in D (log-transformed) for PIT-tagged Snake River wild steelhead and wild Chinook in migration years

57 Hatchery Steelhead The PIT-tagged hatchery steelhead used in the CSS are obtained from all available marking efforts in the Snake River basin above Lower Granite Dam. Hatchery steelhead from each tributary, plus PIT-tag releases in the mainstem Snake River at the Lewiston trap and below Hells Canon Dam, were represented in the PIT-tag aggregates for migration years 1997 to 2003 (Table 18). A list of the locations within the tributaries where the PIT-tagged hatchery steelhead were released is provided in Appendix H. The hatchery steelhead comprising the PIT-tag aggregate appear to be well spread across the drainages above LGR. Table 18. Number of PIT-tagged hatchery steelhead smolts from the four tributaries above Lower Granite Dam (plus mainstem Snake River) used in the CSS for migration years 1997 to Migr. Year Number of PIT-tagged hatchery steelhead utilized in CSS by location of origin Total PIT Tags Clearwater River (Rkm 224) Snake River trap 1 (Rkm 225) Grande Ronde River (Rkm 271) Salmon River (Rkm 303) Imnaha River (Rkm 308) Snake River at Hells Canyon Dam (Rkm 397) ,705 12, ,039 9,394 6, ,913 8,451 4,209 4,904 8,457 4, ,968 11,486 3,925 5,316 9,132 6, ,000 8,488 3,290 5,348 8,173 6, ,099 9,155 3,126 4,677 7,859 3, ,573 7,819 4,722 3,888 7,011 2, ,379 4,912 4,171 3,113 7,764 6, Average % of total 29.0% 11.1% 15.3% 26.6% 17.1% 0.9% 1 Snake River trap located at Lewiston, ID, collects hatchery steelhead released in Grande Ronde, Salmon, and Imnaha rivers, and below Hells Canyon Dam. The estimated number of PIT-tagged wild steelhead smolts (with bootstrapped 90% confidence intervals) arriving at LGR for each CSS study category, T 0, C 0, and C 1, is presented in Appendix B Table B-7 along with the associated number of returning adults in each study category. Through migration year 2002, few PIT-tagged wild steelhead are in the T 0 study category due to the default operation of routing most PIT-tagged fish back to the river at the Snake River collector dams. Until 2003, the number of PIT-tagged hatchery steelhead actually transported has been relatively small relative to the number of untagged hatchery steelhead transported (Figure 21 and Appendix E Table E-7). Beginning in 2003, more PIT-tagged wild steelhead have become available in the transport group as hatchery research programs started routing a portion of their PIT-tagged hatchery steelhead smolts to the raceways at Snake River transportation facilities. 34

58 Proportion of PIT-tagged smolts being transported relative to the untagged fish collected and transported prop(lgr) prop(lgs) prop(lmn) 0.25 Proportion Migration year Figure 21. Trend in proportion of PIT-tagged hatchery steelhead transported at each Snake River collector Dam, 1997 to The age breakdown of the returning hatchery steelhead adults detected at Lower Granite Dam adult fish ladder from smolts outmigrating in 1997 to 2003 is shown in Appendix D Table D-5. For migration years 1997 to 2003, the average age of return to Lower Granite Dam is 55.8% 1 + -salt, 43.9% 2 + -salt, and 0.3% 3 + -salt (hereafter the + notation on age will be dropped in both text and tables). Steelhead returning in the fall of the same year that they migrate to the ocean are not included in the adult return data. Because of the small percentage of age-3 salt returning hatchery steelhead, the return data for migration year 2003 should be close to complete with regard to computing SARs, T/C ratios, or D estimates for that migration year. Obtaining a valid estimate of the number of PIT-tagged hatchery steelhead in Category C 0 in 2001 is problematic due to residualism just as it was for PIT-tagged wild steelhead. One of the 3 adult returns of Category C 1 hatchery steelhead from migration year 2001 was actually detected in the lower river in There were two PIT-tagged hatchery steelhead adult returns with no smolt detection in As noted with wild steelhead, these two never detected hatchery steelhead also were more likely to have completed their smolt migration in 2002 or have been inadvertently transported from Lower Granite Dam without detection there. Because of the uncertainty in passage route and timing of the undetected PIT-tagged hatchery steelhead smolts in 2001, fish from Category C 1 will be used in the transport versus inriver migration comparisons for that year. The site-specific transportation SAR estimates [e.g., SAR(T LGR ), SAR(T LGS ), and SAR(T LMN )] used in estimating SAR 1 (T 0 ) for hatchery steelhead are presented in Appendix E Table E-8. Because of the low number of PIT-tagged smolts transported and small number of 35

59 returning adults, this study s ability to detect potential differences in site-specific SARs was limited. The 90% confidence intervals of the site-specific SARs are extremely wide and overlapping across all three dams in all years of study. However, this does not impact the conduct of this study since our goal is to create an overall multi-dam estimate of transportation SAR for comparison with the SARs of in-river migrants. The SARs for hatchery steelhead from migration year 2003 was 1.81% for transported fish, but below 0.7% for the in-river migrants (Table 19). These differences were significant based on non-overlapping 90% confidence intervals. Significant differences in estimated SARs between transported and in-river migrants were also observed for migration years 2001 and 2002 (Figure 22). Relative to the 7-year average SAR(C 0 ) of hatchery steelhead that passed the three collector dams undetected, a 72% higher transportation average SAR 2 (T 0 ) and 31% lower bypass average SAR(C 1 ) was estimated. The pattern and relative magnitude of these differences between the study categories was similar for both the wild and hatchery steelhead (Figure 23). Table 19. Estimated SAR LGR-to-LGR (%) for PIT-tagged hatchery steelhead in annual aggregate for each study category from 1997 to 2003 (with 90% confidence intervals). Mig. Year SAR 1 (T 0 ) SAR 2 (T 0 ) SAR(C 0 ) SAR(C 1 ) 1997 NA ( ) 0.24 ( ) 0.17 ( ) ( ) 0.51 ( ) 0.89 ( ) 0.22 ( ) 1999 NA 0.90 ( ) 1.04 ( ) 0.59 ( ) ( ) 2.10 ( ) 0.95 ( ) 1.05 ( ) 2001 NA 0.94 ( ) {Assume =SAR(C 1 )} ( ) 2002 NA 1.06 ( ) 0.70 ( ) 0.73 ( ) ( ) 1.81 ( ) 0.68 ( ) 0.37 ( ) 7-yr Avg. Std_error Not applicable since some sites have no adult returns for estimating a site-specific SAR. 2 Migration year 2003 is incomplete until 3-salt adult returns occur at GRA. 36

60 0.04 Hatchery Steelhead SAR Estimate sart0 sarc Migration Year Figure 22. Estimated transport and inriver SARs for PIT-tagged hatchery steelhead aggregate for migration years 1997 to 2003 (incomplete returns for 2003). The 2003 overall estimate of SAR LGR-to-LGR for Snake River hatchery steelhead was 1.46%, which is not as high as the estimate for 2000, but still above the other five years since 1997 (Appendix F Table F-4 and Figure 23). The annual time series of aggregate hatchery steelhead SARs were lower than the corresponding annual time series of aggregate wild steelhead SARs in all but one year between 1997 and 2002 (Figure 23), but only significantly lower in 1999 based on non-overlapping 90% confidence intervals (Appendix F Tables F-3 and F-4). SAR (LGR-to-LGR) SAR (LGR-to-LGR) for Hatchery and Wild Steelhead 0.05 Wild Hatchery Figure 23. Trend in estimated annual SAR for hatchery and wild steelhead with 90% confidence intervals; based on SAR estimates in transport and inriver categories weighted by estimated proportion of run-at-large in each category for migration years 1997 to 2003 (incomplete 2003 returns). 37

61 The estimated inriver survival from LGR tailrace to BON tailrace (V C ), transport SAR to inriver SAR (T/C) ratio, and delayed mortality D for the PIT-tagged hatchery steelhead aggregate group are presented in Table 20 for migration years 1997 to The individual reach survival estimates for each migration year used to obtain V C are presented in Appendix Table C Table C-8. The geometric mean of V C for 1997 to 2002, excluding 2001, was 0.41, similar to what was estimated for wild steelhead. Figure 24 shows the tend in annual Vc estimates for wild steelhead compared to hatchery steelhead for A T/C estimate > 2 with corresponding estimated D >1 occurred in 2 of 7 years for PIT-tagged hatchery steelhead (only 2003 was significant based on the lower limit of the 90% confidence interval) (Figures 25 and 26). Migration year 2001 had very large estimated T/C and D, but the precision in these estimates was extremely low. Excluding 2001, the geometric mean T/C of 1.46 for hatchery steelhead was approximately 15% lower than that estimated for wild steelhead. The D estimates for and had a geometric mean of 0.64 for hatchery steelhead, approximately 20% lower than the geometric mean D of 0.80 estimated for wild steelhead. Although differences arise between the estimates for wild and hatchery steelhead, these data suggest that steelhead as a whole have a very different response to transportation as a recovery tool than do the listed wild Chinook. Table 20. Estimated inriver survival LGR to BON (V C ), T/C ratio, and D of PIT-tagged hatchery steelhead for migration years 1997 to 2003 (with 90% confidence intervals). Mig. Year V C SAR 2 (T 0 )/SAR(C 0 ) D (25% expansion) A 2.21 ( ) 0.92 ( ) B ( ) 0.58 ( ) 0.39 ( ) B ( ) 0.87 ( ) 0.41 ( ) B (25% expansion) 2.20 ( ) 0.55 ( ) B ( ) 1.51 ( ) 0.60 ( ) 2003 C 0.51 ( ) 2.65 ( ) 1.43 ( ) Geometric Mean ( ) 59.7 ( ) 2.40 ( ) B A Expansion shows percent of reach with a constant per/mile survival rate applied. B Migration year 1997 to 2001 confidence intervals for D are shifted 0.01 to 0.08 higher than reported in 2005 CSS to adjust for the corrected D presented in that report. C Migration year 2003 is incomplete until 3-salt adult returns occur at GRA. 38

62 0.7 V c for Hatchery and Wild Steelhead Vc Wild Hatchery Figure 24. Trend in in-river survival (V c ) for PIT-tagged Snake River hatchery and wild steelhead for migration years 1997 to 2003 ln(t/c) ln(t/c) for Hatchery and Wild Steelhead Wild Hatchery Figure 25. Trend in ratio of SAR 2 (T 0 )/SAR(C 0 ) (log-transformed) for PIT-tagged Snake River hatchery and wild steelhead in migration years 1997 to

63 ln(d ) ln(d ) for Hatchery and Wild Steelhead Wild Hatchery Figure 26. Trend in D (log-transformed) for PIT-tagged Snake River hatchery and wild steelhead in migration years CONCLUSIONS 1. Wild steelhead from the Snake River basin had higher estimated annual SARs (indexed LGR to LGR) than hatchery steelhead in 6 of the 7 migration years (1997 to 2003). Wild steelhead had four years with annual SARs > 2%. 2. The pattern of decreasing estimated annual SARs for wild steelhead is following that of the wild Chinook, just not dropping as rapidly over the migration years 1999 to Transportation seems to provide benefit to wild and hatchery Snake River steelhead; the geometric mean T/C ratio ( , ) was 1.72 wild stocks and 1.46 for hatchery stocks. Migration year 2001 had very high, but imprecise T/C ratios, for both wild and hatchery steelhead. 4. Delayed mortality was evident with transported wild and hatchery steelhead relative to inriver migrants as the geometric mean D for (excluding 2001) was 0.80 for wild stocks and 0.64 for hatchery stocks. Migration year 2001 estimated Ds were >1 for wild and hatchery steelhead. Confidence intervals were wide due to small sample size. 5. Given small sample sizes and wide confidence intervals for both wild and hatchery steelhead, it is premature to conclude whether hatchery steelhead can serve as surrogates for wild steelhead. However, trends in Vc and T/C ratios were similar between wild and hatchery steelhead. 40

64 CHAPTER 5 Relationships between wild and hatchery Chinook salmon smolt-to-adult survival and inriver, estuary/early ocean, and off-shore marine environmental conditions Introduction Patterns observed in recruits-per-spawner (R/S) and smolt-to-adult survival (SAR) data collected as part of the CSS, as well as studies done by other researchers (e.g., Pyper et al. 2005), indicate that strong covariation in performance exists among anadromous salmon populations in the Pacific Northwest. Such synchronized population behavior is believed to be driven primarily by large-scale climate variables or year effects. Thus, towards a more complete understanding of factors influencing inter-annual patterns in PIT-tag-based SARs and other performance measures used by the CSS (i.e., T/C ratios and D), we evaluated relationships between SARs and selected environmental parameters for our 2006 report. In doing this analysis, our goal was to lay a foundation for future study-group comparisons that will explicitly consider the role of environmental drivers in determining population performance. We provide an analysis of wild and hatchery Chinook salmon SAR (Lower Granite-to-Lower Granite) variation due to inriver, estuary/early ocean, and off-shore marine environmental conditions. Further, in order to determine whether or not CSS SAR environmental variable relationships are consistent with and representative of those existing for wild Snake River spring/summer Chinook salmon historically, we simultaneously analyzed relationships between run-reconstruction-based SARs and environmental factors. Methods SAR estimates -- For PIT-tagged Chinook, we quantified relationships between environmental variables and smolt-to-adult survival using annual weighted SAR estimates for both wild and hatchery fish (Tables F-1 and F-2, Appendix F). These values incorporate SARs of both transported (T 0 ) and inriver (C 0, C 1 ) study groups, with the contribution of each category to the overall estimate being weighted by its relative abundance in the run at large (during outmigration). SARs for wild salmon are derived from PIT-tag releases occurring in natal streams and at smolt traps. The wild Chinook SAR time series extends from migration year (MY) 1994 to 2004 (11 years). SARs were estimated for hatchery Chinook salmon populations based on PIT-tag releases occurring at Dworshak National Fish Hatchery, Imnaha Hatchery, McCall Hatchery, and the Rapid River Hatchery. Our hatchery Chinook salmon SAR time series extends from MY 1997 to 2004 (8 years). Though hatchery and wild Chinook salmon SAR estimates are not statistically independent across populations, they were treated as such in our analysis (i.e., n = 43). In addition, we included historical run-reconstruction-based SARs in our analysis. These values were taken from the appendix of Scheuerell and Williams (2005), which incorporate information on smolt and spawner abundance from Raymond (1988), Petrosky et al. (2001), and Williams et al. (2005), and represent an aggregate of all wild Snake River spring/summer Chinook populations. In contrast to Scheuerell and Williams (2005), however, we did not use 41

65 those years when estimated smolt abundance was based on spawner-recruit model predictions (i.e., MYs ); thus, our historical time-series extends from and (i.e., n = 31 years). A time series plot of contemporary (PIT-tag-based) and historic (runreconstruction-based) SARs appears in Figure SAR (%) MY Groups Hatch-PIT Wild-PIT Wild-RR Figure 27. Time series plot of annual weighted PIT-tag SARs for wild ( Wild-PIT ) and hatchery ( Hatch-PIT, hatchery averaged) Chinook salmon across migration years Runreconstruction-based SARs (from Scheuerell and Williams 2005; Wild-RR ) extending from and are presented for comparison. Note, SARs were interpolated between 1984 and 1991 for illustrative purposes only. Environmental variables -- Though we could have selected many environmental variables for this analysis, we focused on two classes of variables inriver variables (flow and water transit time) and ocean environment variables (indices describing coastal upwelling intensity and sea surface temperature/atmospheric pressure anomalies). We did this for two reasons: 1) preexisting knowledge on factors influencing ocean salmon production and downstream migration survival provide a clear basis for a priori screening of possible variables; and 2) our dataset was relatively small, effectively supporting an evaluation of models containing only a few variables. Additionally, among the multitude of possible inriver and ocean variables available for analysis, preliminary analyses narrowed our focus to these particular descriptors. Given that outmigrant survival and the potential delayed consequences of outmigration experience may be flow- and migration-duration-related, we described inriver migration conditions based on river discharge and water-particle transit time (WTT). Thus, using flow measurements (in kcfs) made by the USGS on the mainstem Columbia River (at the Dalles Dam, USGS site ), we described outmigration flow conditions as the two month (April-May) average for each MY. Similarly, using dam-specific April-May mean discharge values for dams occurring between the Snake Basin and the Columbia River estuary (i.e., from LGR to BON) and existing discharge velocity conversions, we estimated the time (in days; WTT) for water to travel from the Clearwater-Snake confluence to below Bonneville Dam. 42

66 We included in our analysis two variables describing environmental conditions existing during the early-ocean phase of Chinook salmon. First, after Scheuerell and Williams (2005), we described conditions existing immediately off shore using an index of coastal upwelling intensity (i.e., the Bakun Index, CUI hereafter) estimated at 45 N and 125 W during the months of April (i.e., approximately when Snake River spring/summer Chinook arrive at sea) and October. Upwelling in April and downwelling in October have been associated with increased primary productivity and recruitment of juvenile anadromous salmonids in their early ocean phase. CUI data were obtained from NOAA (Pacific Fisheries Environmental Lab website, Second, we described conditions existing in the off-shore marine environment using the Pacific decadal oscillation index (PDO hereafter), given existing knowledge on associations between salmon production and PDO regimes (e.g., Hare et al. 1999). PDO is a large-scale ocean-climate index based on sea surface temperature and pressure anomalies measured at stations in the North Pacific Ocean (poleward of 20 N). Negative values indicate cold-pdo and positive values warm-pdo phases; production of Columbia River salmon is believed to be greatest during cold-pdo phases due to the increased primary production encountered by these fish while at sea. We obtained a PDO time series from University of Washington and NOAA s Joint Institute for the Study of the Atmosphere and Ocean website ( As used in our analysis, PDO values represent mean summer (June-August) values for the MY of ocean entry. See Figure 28 for time series plots of all environmental variables for the period extending from Data analysis -- We explored relationships between contemporary and historic SARs (lntransformed for normalization) and inriver and estuary/early ocean environmental conditions, separately, through a multi-stage linear regression modeling exercise. That is, we fit several models (Tables 21 and 22) to both PIT-tag- and run-reconstruction-based SAR environmental variable datasets in order to identify the most parsimonious model accounting for the greatest degree of variance in SARs. Thus, we started with a set of bivariate single-predictor inriver models (i.e., selecting between discharge and WTT) and single-predictor ocean environment models (i.e., distinguishing between CUI-April, CUI-October, and PDO) and progressively built towards our most fully parameterized model one including a single inriver and 2 marine variables (i.e., including the best upwelling variable and PDO). Additionally, because data were available for both hatchery and wild Chinook salmon in our CSS PIT-tag dataset, we also added a group variable to our analysis to determine whether or not both groups respond similarly to inter-annual variation in environmental conditions. For each dataset, we completed all analyses using linear regression and distinguished between candidate models at each stage using the least-squares version of 43

67 Water Transit Time (days) Discharge (kcfs) Pacific Decadal Oscillation MY MY MY April Upwelling October Upwelling MY MY Figure 28. Time series plots of inriver (Water transit time; Columbia River discharge) and estuary/ocean environmental variables (PDO, CUI-April, CUI-October) across migration years 1964 to See text and Table 21 caption for variable definitions. Akaike s Information Criterion (AIC c ; also corrected for small sample size), following the information-theoretic approach advocated by Burnham and Anderson (2002). Although we completed a separate model selection and fitting exercise for both historic (i.e., runreconstruction-based) and contemporary (i.e., PIT-tag-based) SAR datasets, we ultimately contrasted results between groups in order to understand the generality of patterns existing in each. To do this, we qualitatively compared model selection results, contrasted bivariate slope parameters (i.e., estimates +/- 95% CIs), and examined associated scatter plots. All analyses were completed using SYSTAT version 9. Results PIT-tag-based contemporary SARs -- Both SARs and inriver and marine environmental conditions varied considerably across migration years (Figures 27 and 28). SARs spanned a range of over an order of magnitude across observations (min to max: 0.21 to 3.28 %). Further, SARs were highly correlated across PIT-tag release groups (mean pair-wise Pearson correlation, R = 0.88, range: ), and when evaluated on a group-by-group basis, with the same environmental variables in the same manner (Figure 29). There was no statistical evidence suggesting that separating hatchery and wild Chinook salmon PIT-tag groups was warranted; both AIC c -based model selection and regression parameter estimates identify the group effect as being negligible (i.e., models lacking group effects ranked better using AICc; across models, the 44

68 group parameter did not differ from zero [P > 0.40 in all cases]). Thus, only aggregate hatcherywild results are presented and discussed from here on. Table 21. Model selection results for SAR environmental variable regression models fitted using MY PIT-tag-based annual weighted SARs. The bold-faced entry corresponds to the model with the lowest AIC c value (corrected for sample size) score. Q is discharge, WTT is water transit time, CUI-April is April upwelling, CUI-Oct is October upwelling, and PDO is Pacific decadal oscillation. Contemporary Models K AIC c ΔAIC c R 2 SAR = Q SAR = WTT SAR = CUI-Apr SAR = CUI-Oct SAR = PDO SAR = Q + CUI-Apr SAR = Q + CUI-Apr + PDO Our PIT-tag SAR model-selection exercise indicated that among bivariate relationships, discharge and WTT described survival variation comparably (though discharge slightly better); also, both April upwelling and PDO bivariate models received greater support than October upwelling based on ΔAIC c (Table 21). Overall, the best regression (inclusive of bivariate and multivariate possibilities) was a three-variable model including discharge, April upwelling, and PDO effects (Table 21). Based on slope parameter estimates for this model, SARs tended to be highest in those MYs characterized by high flows during outmigration, coastal upwelling during ocean entry, and cool-phase PDO conditions during a Chinook s first summer at sea. Table 22. Model selection results for SAR environmental variable regression models fitted using and MY run-reconstruction-based SARs. The bold-faced entry corresponds to the model with the lowest AIC c value (corrected for sample size) score. See Table 21 description for variable definitions. Model description K AIC c ΔAIC c R 2 SAR = Q SAR = WTT SAR = CUI-Apr SAR = CUI-Oct SAR = PDO SAR = WTT + CUI-Apr SAR = WTT + CUI-Apr + PDO Run-reconstruction-based historic SARs -- Historic SAR estimates based on run reconstruction varied considerably from 1964 to 2001, ranging between 3-5% in the late 1960s to less than 45

69 0.5% in the early 1990s. Also, estimates were highly correlated with (R = 0.93) and were comparable to those measured using PIT-tags (i.e., ; Figure 27), though there was some deviation past Environmental variables, especially WTT and both month-specific upwelling indices, exhibited much greater variation across the complete time series than over the contemporary window; thus, the domain for these variables was greater in the historic than the PIT-tag SAR analysis. Run-reconstruction SAR model-selection results indicate that among bivariate relationships: 1) WTT described survival variation better (ΔAIC c > 2) than discharge; and 2) April and October upwelling and summer PDO similarly described SAR variability, with April CUI SAR regression having the lowest AIC c value of the three models (Table 22). Of all models fitted, the best of all was a two-variable function including WTT and April upwelling. This result suggests that from , SARs were highest in MYs with short travel times and positive upwelling conditions. When PDO was added as a predictor (i.e., the 3-variable model in Table 22), it indicates that SARs are lowest in those MYs with warm-phase PDO conditions LN(SAR) 0 LN(SAR) Discharge (kcfs) April Upwelling 1 LN(SAR) Pacific Decadal Oscillation Group Hatchery Wild Figure 29. Scatter plots of hatchery and wild PIT-tag annual weighted SARs versus discharge during outmigration, April upwelling, and summer PDO. 46

70 Contemporary vs. historic SAR environment relationships -- Generally speaking, Snake River spring/summer Chinook salmon smolt-to-adult survival varied as a function inriver and ocean conditions in a similar manner during both historic and contemporary time periods (Figure 30; Table 23). There were, however, some subtle differences. First, given its minimal variation from (compared to the longer time series), October upwelling explained little variation in PIT-tag-based SARs; for the historic time series, however, October upwelling was a significant predictor of SARs (Table 23). Additionally, whereas discharge was a significant predictor of contemporary survival, it was not for the historic time series either by itself (Table 23) or relative to WTT (Table 22). However, WTT accounted for a significant portion of SAR variation in the historic time series. Despite these differences, however, both historic and contemporary slope parameter estimates were comparable (i.e., they had overlapping 95% CIs and similar point estimates; Table 23). Both analyses suggest that SARs vary as a positive function of discharge/wtt, April upwelling, and as a negative function of PDO and October upwelling. Table 23. Least-squares slope parameter estimates (+/- 95% CIs) for bivariate regressions between PIT-tag- ( Contemp. in table) and run-reconstruction-based ( Historic in table) SARs and environmental factors. Bold-faced cell entries correspond to those estimates differing significantly from zero. β 1 estimate Variable Contemp. Historic Discharge (0.001 to 0.007) ( to 0.006) WTT ( to ) ( to ) CUI-April (0.012 to 0.040) (0.008 to 0.031) CUI-Oct ( to 0.023) ( to ) PDO ( to ) ( to ) Discussion and Conclusions We found moderate-to-strong and convergent relationships between Chinook salmon smolt-to-adult survival and both inriver and ocean environmental conditions across contemporary ( ) and historic ( ) time periods. Further, we documented similar environmental variable SAR relationships for stocks of different origin (i.e., hatchery and wild populations) and based on datasets generated using different survival estimation techniques (PIT-tag mark-recapture vs. run reconstruction). Based on both historic and contemporary datasets, SARs were highest for those MYs when fish emigrated rapidly and/or during high-flow conditions, arrived at the coast during periods of increased upwelling, and completed their off-shore migration under cool-phase PDO conditions. 47

71 LN(SAR) Discharge (kcfs) Water Travel Time (days) Data source Contemp. (PIT Historic (RR) 1 1 LN(SAR) April Upwelling Pacific Decadal Oscillation Data source Contemp. (PIT) Historic (RR) Figure 30. Scatter plots of PIT-tag- ( Contemp. ) and run-reconstruction-based ( Historic ) bivariate SAR environmental variable (discharge and WTT during outmigration, April upwelling, and summer PDO). The relationships we document herein are consistent with those reported for previous evaluations of environmental controls on stock productivity based on other metrics. For example, R/S residual (Schaller and Petrosky In Review) and fishery catch based (Hare et al. 1999) evaluations suggest that these measures of stock productivity respond similarly to upwelling and PDO trends. The associations between both discharge and WTT and SARs, similarly, are in agreement with the results of Schaller and Petrosky (In Review) who found that increased WTT a function of discharge and reservoir volume existing between Lower Granite Pool and Bonneville Dam tailrace was associated with a decrease in an estimated survival-rate index. In sum, it appears that observed inter-annual patterns in PIT-tag-based SARs, as estimated for the CSS, are robust. In addition to the analyses described in this section of our report, we plan to further analyze relationships between SARs and environmental variables for our upcoming 10-year report. Planned analyses include: 1) as the data permit, we will consider other and more complex model structures, including those using study-category-specific SARs with associated model effects (i.e., T 0 vs. C 0 dummy variables) and interactions (e.g., study category water travel time) in a model-selection framework; 2) we will evaluate relationships between reach survival estimates (i.e., Vc s) and inriver migration conditions across our 11+ migration year (MY) 48

72 dataset; and 3) as more MYs become available for wild and hatchery steelhead, we will pursue similar analyses for these fish. 49

73 CHAPTER 6 Associations between smolt outmigration experience and adult Chinook salmon Bonnevilleto-Lower-Granite-Dam apparent survival rates Introduction Given that estimates of T/C ratios and D both rely on smolt-to-adult survival rates (SARs) based on adult detections at Lower Granite Dam (LGR), these values include both an ocean mortality component and one occurring during upstream migration (i.e., between Bonneville Dam, BON, and LGR) in the year of adult return. Using data collected at PIT-tag interrogation systems on adult fishways, the latter quantity can be directly estimated and compared between CSS s transport (T 0 and T 1 ) and inriver (C 0 and C 1 ) study categories. By quantifying upstream survival rates, it may be possible to more precisely identify mechanisms responsible for a portion of the observed study-category SAR differential. Accordingly, we initiated an analysis/comparison of the inter-dam drop out (i.e., mortality) rates of hatchery and wild Chinook salmon for our 2005 annual report. For the present report, we extended analyses reported in 2005 in several ways: 1) we updated our dataset with return year (RY) 2006 adult PIT-tag detections; 2) we quantified adult migration (BON-LGR) survival for both study groups; and 3) we formally tested for differences in adult migration survival, timing, and duration between groups. Additionally, we evaluated the role of environmental factors (i.e., flow, spill, and temperature) on the upstream survival of adult salmon. Methods Approach -- We tested for an effect of juvenile transportation on upstream adult migration timing, duration, and success for Chinook salmon through three separate analyses: 1) we tested whether or not BON-LGR migration success was independent of juvenile outmigration history using χ 2 -tests (Note: given the ~100% detection probability at LGR, we take detection at LGR [i.e., BON-LGR migration success] to be synonymous with upstream-migration apparent survival [i.e., inclusive of both mortality and straying]); 2) we modeled individual survival, a binary response, using logistic regression; within this analysis, we tested for transportation and environmental variables effects using an Akaike s Information Criterion (AIC)-based modelselection exercise and based on significance tests for fitted model parameters and associated odds ratios; 3) we contrasted adult return timing (i.e., arrival at BON) and BON-LGR upstream travel time (i.e., passage duration, in days) across outmigration histories using analysis of variance. Dataset description -- We evaluated relationships between outmigration experience and upstream survival and migration characteristics for hatchery and wild Chinook salmon, separately. For hatchery Chinook salmon, we used available adult PIT-tag detections for fish released from Catherine Creek (CATH), Dworshak (DWOR), Imnaha (IMNA), McCall (MCCA), and Rapid River (RAPH) hatcheries; for wild salmon, we relied on PIT-tag releases 50

74 from CSS-affiliated smolt traps and from tagging efforts occurring in natal streams throughout the Snake River Basin. We included in our analysis >1-salt adults (i.e., we excluded jacks) from MYs that were detected as adults at BON, McNary (MCN), Ice Harbor (IHR), and LGR PIT-tag interrogation sites in RYs Also, we excluded those adult that were not initially detected at BON during their respective upstream migration. We determined each adult s juvenile outmigration experience based its smolt capture history and grouped individuals in a manner similar to Marsh et al. (2005). Thus, we included categories for the following juvenile outmigration histories: 1) inriver outmigrants (i.e., undetected or detected but bypassed = inriver group hereafter); 2) transported individuals that were collected at and transported from LGR ( LGR group hereafter); and 3) transported individuals that were collected at and transported from LGS or another downstream project ( LGSdown group hereafter). Sample sizes, by migration year, transport history, and BON-LGR passage success are provided in Tables 24 (hatchery; aggregate n = 3,649) and 25 (wild; aggregate n = 539). Table 24. Counts of hatchery Chinook salmon adults that failed ( F ) or were successful ( S ) in surviving their BON-LGR migration in return years , grouped by migration year and outmigration experience (see Methods for group definitions). There was evidence for a significant association between transport history and migration success where sufficient observations-per-cell were available (see Table 26 for details). MY2001 MY2002 MY2003 MY2004 Combined Outmigration history F S F S F S F S F S Inriver LGR LGSdown Table 25. Counts of wild Chinook salmon adults that failed ( F ) or were successful ( S ) in surviving their BON-LGR migration in return years There was evidence for a significant association between transport history and migration success where sufficient observations-per-cell were available (i.e., > 5; MY2002: χ 2 = 8.74, df = 2, P = 0.013; Combined: χ 2 = 7.94, df = 2, P = 0.019; MY2001, MY2003-4, not applicable). MY2001 MY2002 MY2003 MY2004 Combined Outmigration History F S F S F S F S F S Inriver LGR LGSdown Environmental variables -- Within the context of our logistic regression-based assessment of transportation effects, we also wished to account for variation in BON-LGR survival that could be attributed to inriver migration conditions. Specifically, given the results from the University of Idaho s radio telemetry work (Keefer et al. 2004; Naughton et al. 2006), we quantified the influence of discharge, spill (%), and water temperature on adult passage success. We summarized these variables using records from the Fish Passage Center and USACE s websites. 51

75 Discharge and temperature data were summarized for Lower Granite (i.e., used as a proxy for Snake River hydrological and thermal conditions) and Bonneville dam (i.e., as a proxy for Columbia River conditions) sites and averaged across 2-week time blocks in each RY. Similarly, spill was summarized as an average Lower Columbia (BON, TDA, JDA, and MCN, averaged) and Lower Snake (IHR, LMN, LGS, and LGR, averaged) value for the same time blocks. Environmental variables were matched with individual fish records based on their Bonneville arrival date. However, given that the majority (hatchery: 570/714 or 80%; wild: 64/78 or 82%) of adults that failed to arrive at LGR dropped out before McNary Dam and that variables are correlated across sites, we used only Lower Columbia environmental variables in our final analysis. Statistical analysis -- For both wild and hatchery Chinook salmon, we analyzed relationships between outmigration experience and adult migration success according to the following steps. First, we ran a separate χ 2 -test (2 3 table; success/failure inriver/lgr/lgsdown categories) for each migration year (MY) and RY, when sufficient observations per cell were available (i.e., > 5); we also performed a single χ 2 -test, pooling individuals across years. We additionally performed hatchery-specific tests for hatchery Chinook. Second, we evaluated the effects of both transportation history and environmental conditions (i.e., Lower Columbia flow, spill, and temperature) on the upstream migration survival of individual fish using logistic regression. Thus, we fit 11 a priori models (Tables 27 and 29) describing an individual s survival response (0 = unsuccessful; 1 = successful) as a function of a combination of transportation (i.e., dummy variables for LGR and LGSdown histories; intercept = inriver) and/or environmental predictor variables. Thus, we evaluated the possibilities that individual upstream passage success was determined by transportation history or environmental conditions alone, or in combination. We used an AIC-based model selection approach to determine the level of support for different models (i.e., hypotheses) and subsequently assessed slope parameter sign (+/-) and significance (using a t-test), as well as success odds ratio estimates (i.e., O LGR /O inriver and O LGSdown /O inriver, where O i = p success /p fail for group i) and associated 95% CIs from our top model. For the final component of our analysis, we contrasted BON arrival timing (i.e., date of adult return, measured as the Julian calendar date) and BON-LGR upstream travel times (in days, log 10 -transformed for normality purposes) between inriver, LGR, and LGSdown groups. To do this, we performed ANOVAs on both hatchery and wild Chinook salmon data sets, separately. Factors included in both arrival timing and travel time analyses were transport history (i.e., inriver, LGR, LGSdown groups), RY (i.e., as a blocking factor), and their interaction. We evaluated model-effect significance based on F-tests (Type-III sums-of-squares) and subsequently contrasted responses between categories using Tukey s HSD test. All statistical analyses were performed using SYSTAT v. 9 (SPSS 1998). We evaluated statistical significance at α =

76 Table 26. Summary of MY-, RY-, and hatchery-specific χ 2 -tests for hatchery Chinook salmon. The P-values listed are not corrected for multiple tests. The success rate ranking corresponds to the ordering of % successful upstream migrants by juvenile outmigration history. The entry NA corresponds to table values that are not applicable because either a test was not performed due to low cell counts (i.e., RY2002) or the resulting test statistic was not significant (α = 0.05). df = 2 for all tests. Table P-value Success Rate Ranking Aggregate <0.001 Inriver > LGSdown > LGR MY NA MY Inriver > LGSdown > LGR MY Inriver > LGSdown > LGR MY NA RY2002 NA NA RY Inriver > LGSdown > LGR RY Inriver > LGSdown > LGR RY Inriver > LGSdown > LGR RY NA CATH Inriver > LGR > LGSdown DWOR <0.001 LGSdown > Inriver > LGR IMNA NA MCCA NA RAPH Inriver > LGSdown > LGR Results Hatchery Chinook χ 2 tests -- The results from the aggregate, MY-, RY-, and hatchery-specific χ 2 -tests are summarized in Table 26. Though there was some variability in which of these tests indicated a significant departure from the null expectation (i.e., that migration success was independent of outmigration experience), on average 77% of LGR adults passed from BON to LGR; in contrast, 81% and 84% of all LGSdown and inriver outmigrants, respectively, made a successful BON-LGR migration (Figure 31). This pattern was generally consistent across χ 2 - tests conducted on a MY, RY, or aggregate basis. Hatchery-specific χ 2 -tests also suggest a transportation effect. However, there appeared to be a distance-to-lgr effect on the results for the different hatcheries. That is, the disparity in migration success between inriver and LGR adults was generally less for those individuals originating from hatcheries that were further upstream (Pearson R = -0.61, correlation between the LGR vs. inriver success-rate difference and distance from release to LGR). Also worth noting is the possible role of race type in survival patterns. χ 2 -tests for IMNA and MCCA hatcheries the only two releasing summer-run Chinook smolts were not significant. The association between outmigration experience and adult migration success for spring-run Chinook hatcheries, in contrast, was statistically significant across all sites. Wild Chinook χ 2 tests -- Given the small sample size for wild CSS Chinook salmon adults, we focused primarily on the pooled χ 2 -test for inferential purposes (i.e., MY2002 was the only year 53

77 with >5 observations per cell for all MY- and RY-specific analyses). Consistent with our findings for hatchery salmon, this analysis (see Table 2 caption for test-statistic details) suggests that wild adult Chinook salmon BON-LGR migration success is influenced by outmigration experience. Specifically, adults that were transported from LGR as smolts were consistently less successful at returning to their upstream tributaries than those that emigrated as inriver or LGSdown smolts (P = 0.019). Whereas only about 10% of inriver and LGSdown smolts did not survive (inclusive of mortality and straying) from BON and LGR, approximately 25% of those collected and transported from LGR as smolts did not reach LGR (Figure 31) Successful (%) Successful (%) LGR LGSdown Inriver LGR LGSdown Inriver Figure 31. Bar chart of the percent of hatchery (left) and wild (right) Chinook salmon that were successful in migrating from BON to LGR for inriver, LGR, and LGS-down outmigration histories across return years (i.e., combined counts). Error bars correspond 95% confidence intervals. Hatchery Chinook logistic regression analysis -- Consistent with hatchery χ 2 findings, our AICbased model-selection exercise also demonstrates an effect of transportation history on upstream adult migration success. The best model describing individual migration success included transport, temperature, and spill effects (Table 28). Model evidence ratios (i.e., w i -best overall model / w i -best environmental variables-only model; Table 27) indicate that the top model, which contained a combination of transportation and environmental effects, was > 6,000 times more likely than the best environmental variables-only model. Thus, based on these data and candidate models evaluated, there is clear evidence suggesting that patterns in individual survival are due to a combination of transportation history and environmental conditions. Considering the top logistic regression model in greater detail (i.e., the transport + temperature + spill model), all parameters differed significantly from zero, except for the dummy variable identifying an LGSdown-group effect (P = 0.085; Table 28). Parameter estimates indicate that the probability of an individual fish migrating successfully from BON to LGR was less for LGR individuals than for either inriver outmigrants and LGSdown individuals. Additionally, parameter estimates suggest that upstream migration success was lessened during periods characterized by high spill and cold temperatures in the Lower Columbia River. Further, the odds ratio estimate for the LGR group (estimate: 0.64; 95% CI: ) indicates that these adults had significantly lower odds of surviving their BON-LGR migration than inriver 54

78 outmigrants (i.e., the 95% CI did not include 1). The odds ratio for the LGSdown parameter did not differ from 1 (estimate: 0.81; 95% CI: ), suggesting that these individuals had a similar likelihood of making it to LGR as inriver-outmigrant adults. Table 27. Logistic regression model-selection results for CSS hatchery Chinook salmon. Note, Y = P(Success X), where X is the variable in question. The bold-faced model was the one most supported by the data, however those with a ΔAIC < 2 can be considered nearly equivalent. K is the number of estimated parameters (inclusive of variance). Model K AIC ΔAIC w i Y = Spill Y = Flow Y = Temperature Y = Spill + Temperature Y = Flow + Temperature Y = Transport Y = Transport + Spill Y = Transport + Flow Y = Transport + Temperature Y = Transport + Spill + Temperature Y = Transport + Flow + Temperature Table 28. Parameter estimates for the top logistic regression model describing BON-LGR migration success for CSS hatchery Chinook salmon returning in Parameter Estimate SE t P-value Intercept <0.001 LGR <0.001 LGSdown Spill Temperature Wild Chinook logistic regression analysis -- Our wild Chinook logistic regression analysis also demonstrates an effect of transportation history on upstream adult migration success. The best model describing individual migration success included transport effects alone (Table 29); every one of the closest competing models (i.e., those models with ΔAIC < 2) also included transportation effects. Model evidence ratios (i.e., w i -best model / w i -best environmental variable-only model; Table 29) indicate that a transport-effects-only model is 4 times more likely than the best environmental variables-only model. Thus, based on these data and candidate models, there is stronger support for a transportation-legacy hypothesis than any environmental conditions-only hypotheses. Of parameters estimated for our top model, only the LGR parameter differed significantly from zero (P = 0.003; Table 30). As expected, the probability of an individual fish migrating successfully from BON to LGR was lower for LGR individuals than for either inriver outmigrants or LGSdown individuals. At 0.46 (95% CI: ), the odds ratio estimate for this group indicates that LGR salmon were about half as likely to survive their migration from BON to LGR than inriver outmigrants. Similar to hatchery 55

79 models logistic regression results, the odds ratio for LGSdown adults did not differ from 1 (estimate: 1.24; 95% CI: ). Table 29. Logistic regression model-selection results for CSS wild Chinook salmon. Note, Y = P(Success X), where X is the variable in question. The bold-faced model was the one most supported by the data, however those with a ΔAIC < 2 were viewed as equivalent. K is the number of estimated parameters (inclusive of variance). Model K AIC ΔAIC w i Y = Spill Y = Flow Y = Temperature Y = Spill + Temperature Y = Flow + Temperature Y = Transport Y = Transport + Spill Y = Transport + Flow Y = Transport + Temperature Y = Transport + Spill + Temperature Y = Transport + Flow + Temperature Table 30. Parameter estimates for the top logistic regression model describing BON-LGR migration success for CSS wild Chinook salmon returning from Parameter Estimate SE t P-value Intercept <0.001 LGR LGSdown Hatchery Chinook arrival and travel time ANOVAs -- Analysis of variance results for hatchery Chinook salmon suggest that no consistent trend exists in either BON arrival date or BON-LGR travel time across the three outmigration histories, though there was considerable variation in both responses across RYs. Significant effects in the arrival date ANOVA include RY (F = 35.1, P < 0.001) and its interaction with outmigration history (F = 6.2, P < 0.001). The model effect outmigration by itself did not account for a significant portion of arrival date variation (F = 2.2, P = 0.12). Given the significant RY outmigration history interaction effect, we evaluated differences between groups within years using Tukeys HSD test. Of all within-year, acrossgroup comparisons, the only significant difference observed was between LGR and inriver fish during 2003 (P < 0.001); in this case, LGR fish arrived at BON 10 days earlier than inriver adults. Across years, however, all groups returned to BON within a 3-day window of each other, with inriver, LGR, and LGSdown mean arrival dates being 21-May, 23-May, and 19-May, respectively. Similar to BON arrival timing, travel times varied significantly across years (RY F-test, F = 71.7, P < 0.001) and there were some differences between study categories that varied by year (RY outmigration history F-test, F = 3.3, P = 0.001). However, the outmigration effect by itself was not significant (F = 0.4, P = 0.662). As with arrival timing, the only significant within-year difference was between LGR and inriver fish in 2003; inriver migrants passed from BON to LGR 2 days faster than LGR study fish. All other year-group comparisons indicate 56

80 negligible differences occur in upstream travel times due to outmigration history, though LGR fish tended towards a more skewed distribution (i.e., at the slow end of travel times; Figure 32). On average, all groups passed from BON to LGR in 14 days. Wild Chinook arrival and travel time ANOVAs -- Similar to the hatchery Chinook BON arrival timing and BON-LGR travel time analysis, there was considerable variability in both responses across RYs but not groups. For the BON arrival timing ANOVA, the only significant model effect was RY (F = 7.1, P <0.001), with arrival dates tending to be earlier in than Arrival dates averaged later than those for hatchery Chinook, with inriver, LGS, and LGSdown adults groups averaging 30-May, 27-May, and 28-May across the 5-year record, respectively. Thus, return timing did not differ as a function of outmigration experience. Similarly, BON- LGR travel times varied considerably (and slightly increasing in time) across years (RY F-test, F = 8.0, P < 0.001), but not as a function of outmigration experience, either across or within years (outmigration history F-test, F = 0.5, P = 0.623; RY*outmigration history F-test, F = 1.3, P = 0.247). All study groups migrated upstream at a similar rate (i.e., in 14.8, 14.0, and 13.3 days, aggregate means for LGR, LGSdown, and inriver groups, respectively); however, as with hatchery Chinook, there was a tendency towards a more skewed and slower travel time distribution for LGR adults (Figure 32) Travel time (days) LGR LGSdown Study category inriver 0 LGR LGSdown Study category Figure 32. Box-and-whisker plot of BON-LGR travel times for hatchery (left) and wild (right) Chinook salmon, by outmigration experience (pooled across RYs ). Lower and upper box bounds correspond to 25 th and 75 th percentiles, respectively; the mid line represents the median; the upper and lower whiskers encompass 1.5 times the inter-quartile range (IQR); values beyond 3 times the IQR appear as circles, those within as asterisks. Discussion and Conclusions inriver For both wild and hatchery Chinook salmon, our analysis demonstrates a significant effect of outmigration experience on the upstream migration success or apparent survival of returning adults. However, our analysis also illustrates that this effect was most pronounced for fish that were transported from LGR as smolts, with these individuals surviving at an approximately 10% lower rate than those with either an inriver or LGSdown smolt history. 57

81 Further, our results suggest that outmigration experience does not affect the timing of adult return (based on all BON detections) or the upstream travel times of those salmon surviving to LGR. Previous research suggests that transportation can affect adult apparent survival rates in the direction we observed in several ways. First, it has been suggested that smolt transportation can disrupt the imprinting process, which typically occurs during smoltification (e.g., Quinn 2005), and thus lead to increased straying of spawners upon return (e.g., Pascual et al. 1995; Bugert et al. 1997; Chapman et al. 1997). In the case where successful migration is defined by an individual s arrival at LGR, inter-dam straying is equivalent to mortality. Additionally, elevated fallback rates and extensive downstream forays by adult salmon have been attributed to juvenile transportation (Keefer et al. 2006). Given that mortality can increase with the number of fallback events and reascension attempts that are made by individuals (Keefer et al. 2005), transport-related fallback may also explain a portion of the observed disparity between study categories. Though less clear, other possible mechanisms may account for the mortality differential we observed. For instance, if increased fallback and impaired homing increase an individual s residence time between BON and MCN dams, transported fish may be more vulnerable to the zone-6 tribal fishery. This possibility, however, has not been evaluated to any great extent. Regardless of the precise mechanisms involved, our results have important implications worth noting: 1) A portion of deviation in both T/C and D from their null expectations may be attributed to survival differences occurring in the mainstem Columbia and Snake rivers after adults return to the freshwater environment to spawn. As a simple example, if we alter wild Chinook SAR(T 0 ) and SAR(C 0 ) values for MY2004 to represent a marine-only post-bon mortality component (i.e., using the average BON-LGR survival for adults with LGR (0.76) and inriver (0.87) outmigration histories), T/C ratios change from 0.97 to 1.11 and D from 0.39 to ) The effect of outmigration experience on upstream adult survival appears to be tempered by a distance-from-release effect. Although we provide only a preliminary analysis of this issue in the present report, we observed two results supporting this conclusion: a) in contrast to LGRtransported fish, the differential between transported and inriver outmigrants was considerably less for those fish collected and transported from LGS or sites even further downstream (i.e., LMN, MCN); and b) the survival discrepancy between LGR and inriver outmigrants tended to be less for hatcheries existing higher in the watershed. This finding is consistent with the results Solazzi et al. (1991), who documented an increase in the straying rates of adult coho salmon that were transported and released as smolts at differing distances from their hatchery rearing site. Further, the lack of a transportation effect on homing for adults transported from IHR as smolts (Ebel et al. 1973) prior to the completion of LGR suggests that sufficient distance for imprinting may exist between LGR and IHR. 3) Finally, using project-specific PIT-tag detections has become the standard for estimating interdam conversion rates for use in in-season fisheries management. While a PIT-tag approach has permitted managers to avoid some of the pitfalls associated with traditional count-based approaches towards conversion rate estimation (Dauble and Mueller 2000), our data suggest that such estimates may be biased (relative to the run at large) if transportation history is not 58

82 considered in the estimation process. For example, if upstream survival is greater for inriver than transported groups as we demonstrate here and the majority of the untagged run-at-large was transported as smolts, a raw PIT-tag estimate of adult conversion rates will be biased relative to the run-at-large. While we intend to further explore these conclusions, their implications, as well as perform additional supporting analyses for future reports, we document a clear inriver, upstreammigrant mortality effect resulting from different juvenile outmigration experiences. 59

83 CHAPTER 7 Upstream-downstream comparisons: Differential mortality for upriver and downriver PIT-tagged wild and hatchery sp/su Chinook Background The upstream/downstream stock comparison was initiated primarily to provide information relevant to the patterns observed in recruit/spawner patterns between upriver and downriver stream-type Chinook (e.g., Schaller et al. 1999, Deriso et al. 2001). The PATH comparison of R/S patterns indicated Snake River stocks productivity and survival rates declined coincident with development and operation of the FCRPS. The R/S comparisons also provided evidence of delayed mortality of inriver migrants from the Snake River, after accounting for direct mortality, differential delayed mortality of transported smolts (D), and the common year effect (CSS Delayed Mortality Workshop proceedings, Marmorek et al. 2004). Analyses by coauthors Schaller and Petrosky in the 2005 CSS Annual Report showed that the differential mortality rates (between upstream and downstream populations) from migration years 2000 to 2002 estimated from PIT-tag SAR data corresponded with differential mortality rates estimated from historic recruit/spawner ratios covering smolt migration years 1993 to They found that differential mortality between Snake River and downriver stocks averaged 1.47; thus Snake River populations survived only 1/4 (i.e., e ) as well as the downriver populations since hydrosystem completion in the mid-1970s. In the current annual report, we will add two additional migration years to the time series of PIT-tagged fish generated SARs for upriver and downriver stocks. SARs computed with upriver wild and hatchery Chinook stocks from LGR as smolts to BOA as adults, downriver wild Chinook from JDA as smolts to BOA as adults, and downriver hatchery Chinook from BON as smolts to BOA as adult. The downriver wild Chinook originate in John Day River and the downriver hatchery Chinook originate at Carson NFH on the Wind River. Rather than compare the downstream SAR to upriver fish in the transport and in-river study categories separately as was done last year, we have used an overall weighted SAR for the upriver stocks, where the study-specific SARs get weighted by their estimated proportion in the overall run-at-large. Recovery year 2002 is the first year when BOA had sufficient coverage of all fish ladders with PIT-tag detectors. Therefore, all comparisons of upriver and downriver stocks will be limited to migration years 2000 and later. We do not attempt to expand the BOA counts for any harvest occurring below Bonneville Dam and assume that any harvest occurring will be affecting the upriver and downriver stocks equally. Carson NFH Spring Chinook Although the CSS has PIT-tagged a given number of Carson Hatchery production in each year since 1997 (see Appendix G Table G-2 for the number of Carson NFH Chinook PITtagged, median length, and percentage of production tagged in each year from 1997 to 2004), an adult PIT-tag system was not fully installed at BON until the 2002 return season. Therefore, we will limit discussion in the annual report of Carson Hatchery PIT-tag releases to migration years 60

84 2000 to 2004 for purpose of the upriver and downriver SAR comparison. SAR data from 1997 to 1999 may be seen in the 2005 CSS Annual Report (Berggren et al. 2005). For Carson Hatchery spring Chinook, BON is the primary evaluation site. BON is the only project these fish pass on their way to the ocean, and juvenile survival estimates must rely on a recapture site(s) below the project to estimate survival to Bonneville Dam and thereby the number of PIT-tagged Carson Hatchery Chinook smolts index at that dam. NOAA Fisheries operates a trawl located at River KM 74 near Clatskanie, OR, that is equipped with PIT-tag detection equipment in the cod-end of the net. Only a specific amount of sets can be made during the season, and catch rate will vary based on river flow, velocity of the flow, and debris and other factors that might reduce sampling time during a given year. Since these recapture numbers can be low, we explored in the 2003/04 CSS Annual Report (Berggren et al. 2005) the additional use of PIT tags decoded from the tern and cormorant nesting sites at Rice Island (Rkm 34) and East Sand Island (Rkm 8) in the lower Columbia River estuary. We found that the CJS reach survival estimate from Carson Hatchery to BON for migration years 1998 to 2002 were more stable (fluctuating only 10 percentage points over these years) when both the tag detections at the trawl and tag recoveries on the bird colonies as two final recovery sites below BON. However, along with utilizing the PIT-tags recovered from bird colony comes the unproven assumption that the birds did not capture PIT-tagged fish above Bonneville Dam. Table 31 presents the resulting survival estimates to BON. Table 31. Number of PIT-tagged Carson Hatchery Chinook released in the Wind River, estimated survival and resulting smolt population arriving Bonneville Dam in migration years 2000 to 2004 (with 90% confidence intervals) with detected adults at BOA. Migration Release Survival rate A Smolt est. Smolts at BON Adults at year number Estimate (95% CI) at BON 90% CI BOA , ( ) 12,945 11,015 15, , ( ) 12,506 11,244 14, , ( ) 12,349 10,096 15, , ( ) 12,709 10,855 15, B 14,973 Estimate > 1, so use (avg of ) 12,622 NA 79 A Survival estimates and 95% confidence intervals from hatchery to Bonneville Dam (BON) tailrace based on trawl site and bird colony sites as the downstream PIT-tag detection sites. B Migration year 2004 is incomplete with jacks and Age 2-salt adult returns through 8/9/2006; including 226 PIT tags found on East Sand Island bird colony, estimated release-to-bon survival >1 was obtained, so average survival rate of prior 4 years is used for In determining SARs indexed on adult returns at (BOA), we need an estimate of the number of smolts passing BON and number of PIT-tagged adults passing BOA in the fish ladders. Only 2-salt and older adult returns are used in the computations of the SARs (the full age composition of the returning jacks and adults for each migration year is shown in Appendix Table D-4). Beginning with return year 2002 there was the capability to detect nearly all PITtagged adult fish passing the three ladders at BOA. However, since a portion of the fish swim over the weir crests and don t pass through the orifices where the detection equipment is installed, the detection rate for PIT-tagged adult fish at BON remains less than 100%. To expand the number of adult PIT-tag detections at BON to account for missed fish, we computed BOA 61

85 adult PIT-tag detection efficiency estimates for migration years 2000 (see Table 46 of Berggren et al. 2005) and 2001 to 2004 (Table 32). The combined hatchery/wild detection efficiency estimates were used for all wild and hatchery Chinook groups in the estimation of SARs. Table 32. PIT-tag detections of returning adult Chinook (ages 2- and 3-salt) at Bonneville and Lower Granite dams with percentage of fish undetected at Bonneville Dam returns from smolts that outmigrated in 2001 to Smolt Dam for adult Age 2-and 3-Salt Returning Adult Chinook Migr. Year detections 1 Hatchery Chinook 2 Wild Chinook 3 Combined Chinook 2001 BOA GRA, MCA, IHA BOA detection efficiency % 97.8% 98.3 % 2002 BOA 1, ,258 GRA, MCA, IHA 1, ,305 BOA detection efficiency % 96.7% 96.4 % BOA GRA, MCA, IHA/ICH BOA detection efficiency % 93.3% 94.5 % BOA GRA, MCA, ICH BOA detection efficiency % 97.7% 97.6% 1 BOA covers Bonneville Dam ladders (detectors BO1, BO2, and BO3), MCA covers McNary Dam ladders (detectors MC1 and MC2), IHA/ICH covers Ice Harbor Dam ladders, and GRA covers the Lower Granite Dam ladder. 2 Hatchery Chinook contains the combination of PIT-tagged fish from Rapid River, Dworshak, Catherine Creek AP, Imnaha AP, and McCall hatcheries. 3 Wild Chinook contain the aggregate of PIT-tagged fish originating above LGR used in the CSS. 4 Migration year 2004 is incomplete with 2-salt adult returns as of 8/9/ Calculated as p = (N detected at BOA) / (N detected at BOA + N passing BOA undetected that were later detected upstream) The SARs from first-dam encountered as smolts to Bonneville Dam as adults was higher across migration years 2000 to 2004 for Carson NFH Chinook (downriver group) than for the upriver spring Chinook hatchery releases, but not always higher for the upriver summer Chinook (Table 33). The SAR computations used BOA adult numbers expanded by the reciprocal of the PIT-tag detection efficiency estimated for that site. The PIT-tag hatchery Chinook from the upriver Snake River hatcheries and the downriver hatchery both had a decreasing trend in SARs from migration year 2000 to The ratio of the upriver SAR to downriver SAR ranged was highest among all five upriver hatcheries in migration year 2003, and lowest in 2001 for Dworshak, Catherine Creek, and Imnaha hatcheries and lowest in 2004 for Rapid River and McCall hatcheries (Table 33). The higher upriver/downriver ratios in 2003 were significant higher than prior years based on non-overlapping 90% confidence intervals for the two summer stocks (McCall and Imnaha hatcheries). Confidence intervals were not available for migration year 2004 data, because the estimation of the population of PIT-tagged smolts at BON for that year could only be indirectly estimated using the average survival rate from release to BON tailrace of the prior four years (see Table 31). 62

86 Table 33. Estimates of SAR from first dam encountered 1 as smolts to Bonneville Dam (BOA) as adults 2 for the upriver PIT-tagged wild Chinook aggregate and the downriver PIT-tagged John Day River wild Chinook that outmigrated in 2000 to Hatchery Run Type RAPH Sp Ch DWOR Sp Ch CATH Sp Ch MCCA Su Ch IMNA Su Ch Migr. Upriver Hat. Chinook 3 SAR LGR-to-BOA Carson NFH Chinook SAR BON-to-BOA Upriver/Downriver Ratio Year Est. 90% CI Est. 90% CI % Est. 90% CI % % % N/A 0.50 N/A N/A 0.63 N/A N/A 0.66 N/A N/A 0.69 N/A N/A 0.78 N/A 1 First dam encounter is LGR for upriver wild Chinook and JDA for downriver wild Chinook 2 Estimated SARs use adults detected at BOA that have been expanded by reciprocal of the PIT-tag detection efficiency estimates of for migration year 2000 from Table 46 in Berggren et al. 2005, and 0.983, 0.964, 0.945, and for migration years 2001 to 2004 from Table 32 in this chapter. 3 Upriver SAR is weighted average of study-specific SARs when weight is estimated proportion of study group in run-at-large for migration year. 4 Migration year 2004 is incomplete with 2-salt adult returns as of 8/9/2006. Estimates of differential mortality, calculated as ΔM = -ln(sar snake /SAR downriver ), for each of the five PIT-tagged hatchery Chinook upriver population relative to the Carson NFH Chinook downriver population are graphically presented in Figure 33. Although the estimated ΔM differ among hatcheries, there is a common annual pattern among the five upriver hatcheries. 63

87 -ln (U/D) Differential mortality upriver/downriver stocks of hatchery sp/su Chinook DWOR RAPH MCCA IMNA CATH Migration year Figure 33. Trend in differential mortality ΔM=-ln(U/D) for hatchery Chinook (Snake River basin stocks [U] versus Carson NFH stock [D]) for smolt migration years 2000 to John Day River Wild Chinook In the lower Columbia River basin, the CSS utilizes the PIT-tagged wild spring Chinook from John Day River (tagged under a separate contract between ODFW and BPA) for the upstream/downstream comparison. ODFW crews have PIT-tagged the following number of juvenile Chinook within the John Day River basin (methods and locations of this PIT-tagging are found in Carmichael et al. [2002]). Estimating SAR from first dam encountered as smolts to BOA as adults requires an estimate of the number of PIT-tagged John Day River wild Chinook smolts passing JDA. This smolt estimate (Table 34) was obtained by multiplying the tag release number by estimated survival from release to JDA tailrace. In estimating this survival, we did not include the PIT-tag recoveries from the bird colonies, since the detections at BON and the trawl alone provided sufficient precision in the survival estimate to JDA tailrace. The number of adult returns (2-salt and older) detected at BOA are also shown in Table 34 (the full age composition of returning jacks and adults for each migration year is shown in Appendix Table D-2). Table 34. Number of PIT-tagged wild Chinook released in John Day River basin, estimated survival and resulting smolt population arriving John Day Dam in migration years 2000 to 2004 (with 90% confidence intervals) with detected adults at BOA. Migration year Release number Survival estimate A Survival 90% CI Smolt est. at JDA JDA # 90% CI Adults at BOA , ,312 1,199 1, , ,721 2,617 2, , ,555 2,279 2, , ,203 3,919 4, B 4, ,755 2,359 3, A Survival of aggregate from release sites to John Day Dam (JDA) tailrace based on Bonneville Dam and trawl sites as downstream PIT-tag detection sites. B Migration year 2004 is incomplete with jacks and Age 2-salt adult returns through 8/9/

88 The SARs from first-dam encountered as smolts to Bonneville Dam as adults was substantially higher across migration years 2000 to 2004 for the John Day River wild Chinook (downriver group) than aggregate Snake River stocks (upriver group)(table 35). The SAR computations used BOA adult numbers expanded by the reciprocal of the PIT-tag detection efficiency estimated for that site. The PIT-tag aggregate of wild Chinook from the John Day River and the PIT-tag aggregate of wild Chinook from the Snake River basin above LGR both had a decreasing trend in SARs from migration year 2000 to The ratio of the upriver SAR to downriver SAR was significantly higher for migration years 2001 and 2002 compared to 2003 and 2004 based on non-overlapping 90% confidence intervals. The U/D ratio for migration year 2000 was intermediate to the other years. Table 35. Estimates of SAR from first dam encountered 1 as smolts to Bonneville Dam (BOA) as adults 2 for the upriver PIT-tagged wild Chinook aggregate and the downriver PIT-tagged John Day River wild Chinook that outmigrated in 2000 to Migr. Upriver Wild Chinook Downriver Wild Chinook Ratio Upriver/Downriver Year Weighted 3 SAR % SAR LGR-to-BOA 90% CI % Estimated SAR % SAR JDA-to-BOA 90% CI % Estimated U/D Ratio U/D Ratio 90% CI First dam encounter is LGR for upriver wild Chinook and JDA for downriver wild Chinook 2 Estimated SARs use adults detected at BOA that have been expanded by reciprocal of the PIT-tag detection efficiency estimates of for migration year 2000 from Table 46 in Berggren et al. 2005, and 0.983, 0.964, 0.945, and for migration years 2001 to 2004 from Table 32 in this chapter. 3 Upriver SAR is weighted average of study-specific SARs when weight is estimated proportion of study group in run-at-large for migration year. 4 Migration year 2004 is incomplete with 2-salt adult returns as of 8/9/2006. Estimates of differential mortality, calculated as ΔM = -ln(sar snake /SAR downriver ), for the six years of SAR data (smolt migration years 2000 to 2004) from PIT-tagged wild populations (Snake and John Day rivers) are presented in Table 36 with associated 95% confidence intervals for comparison with the historic differential mortality estimates from Deriso et al. (2001). Wider confidence intervals (95% instead of 90%) are used to match those of the historic data set. In the one year of overlap between the two data series, the PIT-tag wild Chinook SAR-based differential mortality estimate (ΔM) for 2000 agreed well with the differential mortality estimated from the spawner-recruit analysis (Figure 34). A benefit of the SAR-based ΔM estimate appears to be a much narrower 95% confidence interval than obtained from the spawner-recruit analysis see the trend in confidence interval spread from 2000 to

89 Table 36. Conversion of estimated upriver/downriver ratios to differential mortality rates for comparison to differential mortality rates computed by spawner-recruit analyses, 95% confidence intervals shown with each method. Migr. Ratio Upriver/Downriver Differential Mortality (ΔM) Year Estimated U/D Ratio U/D Ratio 95% CI Estimated -ln(u/d) -ln(u/d) 95% CI Differential mortality LCL_mu mu UCL_mu LCL_M M=-ln(SAR ratio) UCLM Differential mortality Migration year Figure 34. Differential mortality estimates from the Deriso et al. (2001) model updated through smolt year 2000 (Marmorek et al. 2004) compared to estimates based on SARs of wild Snake River and John Day River sp/su Chinook, smolt migration years Differential mortality estimates (ΔM) between upriver and downriver Chinook stocks based on the SAR ratios of PIT-tagged fish over the five migration years were greater for the wild Chinook stocks than for the hatchery populations (Figure 35), and showed different annual patterns. The ΔM trend was similar across years for the five upriver hatcheries with the lowest value occurring for migration year Migration year 2003 produced the lowest overall annual SARs for each upriver hatchery stock and even a lower annual SAR for Carson NFH Chinook. The upriver wild Chinook stocks also had their lowest overall annual SAR estimated for migration year 2003, but the SAR for wild Chinook from John Day River was only slightly lower than the prior three years. CSS has proposed adding more wild and hatchery downriver stocks in future years to aid in the assessment of differential mortality and a common year effect. 66

90 As the time series of SAR data increases, we should improve our ability to calibrate PIT-tag generated SAR data to the recruit/spawner derived differential mortality, common year effect and other metrics. -ln (U/D) Differential mortality upriver/downriver stocks of wild and hatchery Chinook WILD DWOR RAPH MCCA IMNA CATH Migration year Figure 35. Trend in differential mortality ΔM=-ln(U/D) for wild Chinook (Snake River basin stocks [U] versus John Day stocks [D]) and hatchery Chinook (Snake River basin stocks [U] versus Carson NFH stock [D]) for smolt migration years 2000 to CONCLUSIONS 1. Differential mortality rates (between upstream and downstream populations) estimated from SAR data appear to correspond well with differential mortality rates estimated from recruit/spawner ratios for wild Chinook populations. 2. Differential mortality estimates based on SAR ratios of hatchery populations were generally less than those based on SAR ratios of wild populations. 67

91 Chapter 8 Upstream-downstream comparisons: contrasting smolt life histories between Snake River and John Day River stream-type Chinook salmon populations Introduction The use of an upstream-downstream stock-comparison approach towards evaluating the effects of the FCRPS on endangered anadromous salmonids (e.g., Chapter 7; Schaller et al. 1999; Deriso et al. 2001; Schaller and Petrosky In Review) has been criticized for a number of reasons (Zabel and Williams 2000; Williams et al. 2005). Critics suggest that downstream stocks, which pass through fewer dams than upstream stocks (i.e., 3 vs. 8 projects), are not appropriate controls for evaluating the effects of hydropower development because a number of confounding issues are at play. For instance, downstream smolts may migrate to sea at a different time than upstream stocks and therefore experience different (more favorable) conditions during estuary/early ocean residence (Zabel and Williams 2000; Williams et al. 2005); also, they may be less exposed to ocean fisheries than their upstream counterparts (Zabel and Williams 2000). More recently, it has been suggested that smolts produced by upstream populations may be smaller than those originating from downstream stocks (Williams et al. 2005), thereby suffering greater (size-selective) mortality at sea (Zabel and Williams 2002). Overall, critics argue that the existence of systematic differences in upstream and downstream population life history attributes precludes the ability to ascribe stock viability differences to the FCRPS. Previous responses to this criticism (Schaller et al. 2000; Deriso et al. 2001; Budy et al. 2002) have stressed that life-history differences would need to explain the systematic change in relative performance existing for upstream and downstream populations coincident with, but unrelated to, the development and operation of the FCRPS. Thus, the relevant issue is not whether or not genetic or life history differences exist between upstream and downstream groups, but rather whether or not differences (if present) were manifested contemporaneously with the completion of the FCRPS. For this reason, upstream-downstream criticisms may be best evaluated using a historic time series comparison approach (i.e., where parameters describing various life history attributes are contrasted between groups as a function of time). Though we are attempting to assemble such a historical dataset, contemporary data (i.e., from the last decade) are all that is available for a quantitative evaluation. For our present purpose, we explore whether or not there are any observable (presenttime) differences between upriver and downriver populations that could explain the observed differential mortality. We focused on life history characteristics associated with the active outmigrant, or smolt, life stage. For both upstream and downstream populations, we quantified and compared outmigration attributes in order to understand the possible confounding effects of smolt life history differences on the results reported in Chapter 7 and elsewhere (Schaller et al. 1999; Schaller and Petrosky In Review). To do this, we exploited a six-year time series of outmigrant smolt data collected at juvenile traps affiliated with the wild Chinook salmon tagging component of the CSS. We contrasted size-at-tagging (fork length, in mm), emigration timing (using the trap site as a reference point for emigration), downstream migration rates (in km / day, to Bonneville Dam, BON), and estuary arrival timing (taken as arrival at BON) between 68

92 wild/natural Chinook salmon smolts captured, tagged, and released at upstream (above Lower Granite Dam, LGR) trap sites and the John Day River mainstem trap site for migration years (MY) 2000 through Methods We used five upstream smolt trap sites in our comparison of upstream-downstream life histories: (1) the Snake River trap (SNKTRP); (2) the Salmon River trap (SALTRP); (3) the Clearwater River trap (CLWTRP); (4) the Grande Ronde River trap (GRNTRP);and (5) the Imnaha River trap (IMNTRP). Our primary downstream reference for wild Chinook salmon smolt collection and tagging is the John Day River mainstem trap site (JDAR1). Our analysis of smolt life history characteristics was based on daily smolt collections for the primary period of juvenile outmigration (March 15 th to May 20 th ; i.e., our evaluation is inclusive of spring outmigrants only) during migration years 2000 to 2005 (Note: CLWTRP operations were not initiated until 2002). Smolt size analysis -- We tested for differences in smolt size across the six release sites under two approaches. First, we tested for differences in size while explicitly accounting for acrosssite differences in relative abundance (i.e., using per-kilometer redd density as a surrogate measure of abundance to account for density dependent effects; See Chapter 8 Appendix, Table 42 for details) using analysis of covariance (ANCOVA). Second, we used an ANOVA approach where we implicitly accounted for inter-annual variation in in-stream conditions relating to juvenile growth and size (i.e., by incorporating MY as a factor). We evaluated ANOVA and ANCOVA model-effect significance based on F-tests (Type-III sums-of-squares); we contrasted density- and year-adjusted mean fork length between John Day smolts and those collected at other release sites using Tukey s post-hoc HSD test. To further explore the effects of density on smolt size, we inspected slope parameters and their associated significance tests and examined plots of mean fork length against redd density, for each site. As a final note, because the sample sizes involved were quite large (Table 37) and statistical significance was therefore virtually guaranteed for all tests, we judged biological significance when betweengroup size differences were greater than 5 mm in magnitude. Outmigration timing -- Assuming that daily tag releases were proportional across the outmigration period and that collected individuals were actively migrating smolts, we estimated passage distribution statistics for each wild/natural Chinook salmon trap site described above. That is, we plotted cumulative passage distributions for each site and MY, as well as for the 6- year average. Additionally, we computed the median passage date for each trap site and MY. Downstream migration rate -- We estimated downstream migration rates, in kilometers per day (km / d) for fish tagged and released at upstream and downstream sites. For distance estimation, the upstream reference was the location of release (i.e., the trap site) and the downstream reference was BON (inclusive of all juvenile interrogation sites); migration duration was estimated for each individual as the difference between release date/time and final date/time of detection at BON (if detected). Migration distances used in computations were 512, 564, 603, 170, 694, and 513 for CLWTRP, GRNTRP, IMNTRP, JDAR1, SALTRP, and SNKTRP release 69

93 sites, respectively. Ultimately, we tested for a difference in migration rates between upstream and downstream populations using ANOVA (as described above for our smolt size evaluation). Table 37. Summary statistics for wild Chinook salmon smolts captured, tagged, and released at CSS trap sites between March 15 th and May 20 th during migration years Release site MY Trap releases (n) Mean fork length, mm (SD) BON detections (n) JDAR , (9) , (8) , (9) , (10) , (10) , (9) 307 SNKTRP , (10) (16) , (10) (11) (11) (9) 8 SALTRP , (11) , (13) , (10) , (11) , (10) , (9) 203 CLWTRP NA NA NA NA (9) (9) , (10) , (10) 22 IMNTRP , (9) , (10) , (11) , (10) , (10) , (9) 72 GRNTRP , (10) (11) , (9) , (12) , (11) , (12) 43 Given the different distances traveled by upstream and downstream fish prior to reaching downstream detection sites and the distance acceleration relationships that have been documented for Snake-origin spring/summer Chinook salmon (i.e., migration speeds increase as fish progress through the hydrosystem; Williams et al. 2005), we also compared migration rates between populations for a comparable (developmentally speaking) segment of their mainstem FCRPS hydrosystem migration corridor, on an exploratory basis. As dictated for downstream detection opportunities for JDAR1 fish, we compared mean first-to-third dam (John Day Dam- Bonneville Dam for downstream, LGR-Lower Monumental Dam for upstream fish) migration 70

94 durations (in days) between populations. Because different river reaches (of comparable length JDA-BON = 116 km; LGR-LMN = 158 km) had to be used for this analysis by design, we evaluated whether or not populations differed as a function of reach- and/or year-specific water velocities, as measured water travel time values (WTT; the average duration in days it takes water particles to travel from the upstream end of a reservoir to the tailrace of another dam; a function of observed river flow and estimated reservoir volume). Estuary arrival timing -- Using the same methods as for outmigration timing, we quantified arrival timing distribution statistics for those fish detected at BON, assuming that passage at this site is equivalent to estuary arrival. That is, for those fish that survived and were detected at BON, we plotted cumulative passage distributions and estimated dates of 50% passage (i.e., median passage dates) for both upstream and downstream release groups Fork Length (mm) Fork Length (mm) Fork Length (mm) CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Release Site CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Release Site CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Release Site Fork Length (mm) Fork Length (mm) Fork Length (mm) CLWTRP GRNTRP IMNTRP Release Site JDAR1 SALTRP SNKTRP 75 CLWTRP GRNTRP IMNTRP Release Site JDAR1 SALTRP SNKTRP 75 CLWTRP GRNTRP IMNTRP Release Site JDAR1 SALTRP SNKTRP Figure 36. Wild Chinook salmon smolt size (mean fork length +/- 1 SD) for fish tagged and released during migration years (between 15 March and 20 May). From left to right, trap sites are: CLWTRP = Clearwater R., GRNTRP = Grande Ronde R., IMNTRP = Imnaha R., JDAR1 = John Day R., SALTRP = Salmon R., SNKTRP = Snake R. Note: there were no wild Chinook smolt size data available for CLWTRP prior to As a final note, due to the small number of fish released and subsequently detected at BON in 2001 (n = 4), 2004 (n = 17), and 2005 (n = 8) for the SNKTRP site, we did not estimate migration rate or estuary arrival timing for this site in these years (Table 37). Additionally, to understand the potential influence of disparate mortality levels imposed upon upstream- relative 71

95 to downstream-originating smolts prior to BON arrival, we computed the BON detection rate as a proxy for survival (i.e., n BON detects / n released at trap site). Results Summary -- In total, we evaluated differences between upstream and downstream smolt life histories based on a sample of over 100,000 individual fish collected across the 6-year time series. Based on these data, we observed that smolt size and outmigration timing were generally similar across upstream and downstream sites. However, we also observed that upstreamoriginating smolts that survived to and were detected at BON migrated downstream at a faster rate but arrived in the estuary at a later time later than downstream-origin smolts. Of JDAR1 fish tagged and released, 13% were detected at BON; 7% of upstream-origin smolts were detected at BON. Smolt size analysis -- Our analysis demonstrates that smolt size varies considerably across migration years, both within and across sites (Table 37; Figure 36). Within these data, however, there was no clear indication of a systematic size difference between the John Day fish relative to those captured at upstream trap sites. During some years, JDAR1 smolts were larger than those captured at upstream sites whereas in other years they were considerably smaller. The only clear and consistent trend indicated that those fish captured at the GRNTRP site were generally the largest whereas those captured at the CLWTRP site were the smallest of all sites in question. More importantly, with the exception of GRNTRP and CLWTRP sites, JDAR1 fish were generally within 5 mm of upstream sites. Table 38. Results from an ANCOVA-based comparison of smolt size across upstream and downstream release sites, using redd density as a covariate. Effect Sum-of-squares df MSS F P Rel_site 311, , < Redds 48, , < Rel_site*Redds 137, , < Error 11,417, , Analysis of Covariance (ANCOVA) results indicate that fork length varies across sites, but as a site-specific function of redd density (Table 38). With the exception of GRNTRP, smolt size redd density regressions all had negative, non-zero (P < for all parameter significance tests) slopes (Figure 37). Given that the density effect was site specific, we contrasted least-squares adjusted mean fork length between release sites at both the average density and at 4 redds per km a level of abundance common to all sites (i.e., to avoid extrapolating for low-escapement sites). At an average level of density (8.9 redds per km), density-adjusted mean fork lengths differed significantly between all release sites (P < for all pairwise contrasts); values were 74, 121, 106, 106, 100, and 100 mm for CLWTRP, GRNTRP, IMNTRP, JDAR1, SALTRP, and SNKTRP fish. At 4 redds per km, density-adjusted sizes for the same release groups (respectively) were 90, 117, 108, 107, 100, and 104 mm. Thus, though there is evidence for statistically significant differences between fish across release sites, the magnitude of departure may not be biologically profound. However, it should be noted that 72

96 this model accounted for only a minor proportion of fork length variation and that the majority was due to the release site effect (not redd density). In addition to explicitly incorporating density effects, we also contrasted fork lengths between release sites using ANOVA with MY as a factor. This approach accounted for a greater proportion of overall fork length variation than the density-specific model (i.e., Table 39 vs. Table 38). Similar to the ANCOVA results, ANOVA results indicate that significant differences exist among release sites, but that the general pattern varies depending on the migration year in question (Tables 38 and 39; Figure 36). Post-hoc pair-wise comparisons indicate the rank of JDAR1 fish size relative to Fork length (mm) Redd density (no. / km) Release Site CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Figure 37. Scatter plot of mean fork length (mm) against redd density (redds / km) for wild Chinook salmon smolts collected, tagged, and released at CSS trap sites during migration years (between 15 March and 20 May). See Figure 36 caption for release site abbreviation definitions. upstream sites varied across years (P < for all contrasts): 1) in 2000, JDAR1 fish were between 2 and 8 mm larger than those collected at upstream sites; 2) in 2001, they were between 5 and 17 mm smaller than those captured at all other sites; 3) JDAR1 smolts were smaller than all but SALTRP and CLWTRP fish in 2002; 4) excluding CLWTRP and GRNTRP in 2004 and GRNTRP and IMNTRP in 2005, JDAR1 fish were within 5 mm of those collected at upstream sites in both of these years. Table 39. Results from an ANOVA evaluating smolt size variation across release sites and migration years. Effect Sum-of-squares df MSS F P Rel_site 1,145, , , <0.001 my 93, , <0.001 Rel_site*my 704, , <0.001 Error 10,411, , Outmigration timing -- Outmigration timing varied considerably across sites and migration years, particularly so for upstream-origin smolts. In most years, the 50% passage date occurred in mid 73

97 April, but was as early as March 27 th (SALTRP, MY 2004) and as late as May 17 th (SNKTRP, MY 2005). Variability in JDAR1 outmigration timing was considerably less than that observed for upstream release groups. Table 40 details median passage dates for each site and migration year. Despite the wide range of variability in outmigration timing, there was no evidence for any systematic difference between upstream and downstream populations that is, in some years downstream populations emigrated earlier than upstream populations whereas in other years they emigrated later. Despite the variability within sites across years, it appears that upstream and downstream populations initiate emigration from tributary streams within a similar time window, on average (Figure 38); both the upstream aggregate (i.e., all traps together) and the JDAR1 6- year average date of 50% passage was April 13 th (across ). Thus, in terms of trap catch data, we found no evidence for a disparity in outmigration timing for upstream and downstream groups. Table 40. Dates of 50% passage (i.e., median emigration date) for Chinook salmon captured, tagged, and released at CSS-affiliated trap sites during MYs Median emigration date 6-y Site mean JDAR1 18-Apr 11-Apr 14-Apr 11-Apr 13-Apr 15-Apr 13-Apr SNKTRP 20-Apr 27-Apr 16-Apr 17-Apr 28-Apr 17-May 25-Apr SALTRP 12-Apr 25-Apr 9-Apr 4-Apr 27-Mar 12-Apr 9-Apr CLWTRP NA NA 2-May 31-Mar 29-Mar 3-Apr 8-Apr IMNTRP 1-Apr 28-Mar 19-Apr 4-Apr 12-Apr 10-Apr 7-Apr GRNTRP 20-Apr 19-Apr 17-Apr 3-Apr 12-Apr 29-Apr 16-Apr Cumulative passage Julian date Release Site CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Figure year mean trap passage (i.e., emigration) distributions for JDAR1, SNKTRP, SALTRP, CLWTRP, IMNTRP, and GRNTRP release sites. Note: Julian date 75 is March 16 th, 100 is April 10 th, 125 is May 5 th, and 150 is May 30 th. See Figure 36 caption for release site abbreviation definitions. Downstream migration rates -- Based on those fish tagged, released, and later detected at BON, we also estimated total downstream migration rates (km / d) and compared them between upstream and downstream populations. This comparison demonstrates that smolts from 74

98 upstream populations actually migrated faster than downstream-origin smolts, once their differing migration distances were accounted for. As illustrated in Figure 39, JDAR1 fish migrated to the estuary at a rate of approximately 5-10 km / d whereas upstream fish did so at twice the rate (10-20 km / d). Further, the across-site pattern was consistent and statistically significant (ANOVA with my, rel_site, and my*rel_site effects, P < for all F-tests, and for all pair-wise contrasts between JDAR1 and upstream sites) for the MYs in question. As an aside, CLWTRP fish were Migration rate (km / d) Migration rate (km / d) Migration rate (km / d) CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Release site CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Release site CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Release site Migration rate (km / d) Migration rate (km / d) Migration rate (km / d) CLWTRP GRNTRP IMNTRP Release site JDAR1 SALTRP SNKTRP 0 CLWTRP GRNTRP IMNTRP Release site JDAR1 SALTRP SNKTRP 0 CLWTRP GRNTRP IMNTRP Release site JDAR1 SALTRP SNKTRP Figure 39. Wild Chinook salmon smolt downstream migration rates (km / d, +/- 1 SD) for those fish captured, tagged, and released at CSS trap sites during migration years (between 15 March and 20 May). See Figure 36 caption for release site abbreviation definitions. Note, CLWTRP operations did not begin until 2002; also, too few tags were available for SNKTRP estimation in 2001, the slowest of all upstream-origin smolts and had the migration rate closest to that of JDAR1 fish. It should re-emphasized, however, that relative survival to BON differed between upstream and downstream release groups by ~6% (Table 37) and that relative detection rates (i.e., non- CJS estimates) at BON were low for both groups (upstream: 7%, downstream: 13%). Despite their different overall trap-bon migration rates, we found evidence of similar and WTT-influenced first-to-third dam migration lengths (in days) for both upstream and downstream populations (Figure 40). In particular, analysis of covariance (with site and WTT effects) suggests a strong positive influence of WTT (F 1,27 = 71.3, P < 0.001) but no effect of release site on migration duration, once upstream-downstream WTT differences are considered 75

99 (F 5,27 = 0.9, P = 0.485). The mean (WTT-adjusted) first-to-third dam migration duration (+ 2SE) for JDAR1 was 12+2 days; for upstream populations, durations averaged 10+2 days. Given that this statistical comparison relied partially on extrapolation for both upstream and downstream populations (Figure 40), however, this result can only be taken as suggestive. Estuary arrival timing Despite the contemporaneous natal stream departure schedule and the faster downstream migration rate of upstream relative to downstream fish, upstream-origin smolts generally reached the estuary (taken as BON) later than downstream fish (Table 41; Figure 41). That is, while upstream release groups reached BON within roughly a day of each other on average (based on 6-year average of 50% passage date), they arrived 9-10 days after the downstream release group. On average, downstream fish arrived at the estuary on May 9 th whereas upstream fish arrived on May 18 th. Further, this pattern of delayed arrival was consistent across years. 20 First-to-third dam migration (days) Water travel time (days) Release site CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Figure 40. Scatter plot of first-to-third dam migration duration as a function of water travel time. Each dot reflects the mean value for a year-site combination. See Figure 36 caption for release site abbreviation definitions. 76

100 Table 41. Median estuary arrival (i.e., BON detection) dates for Chinook salmon smolts captured, tagged, and released at CSS-affiliated trap sites during MYs Median estuary arrival date Site y mean JDAR1 8-May 10-May 11-May 14-May 7-May 5-May 9-May SNKTRP 12-May NA 18-May 16-May NA NA 15-May SALTRP 12-May 5-Jun 19-May 15-May 15-May 18-May 19-May CLWTRP NA NA 28-May 22-May 18-May 17-May 21-May IMNTRP 8-May 2-Jun 22-May 18-May 17-May 18-May 19-May GRNTRP 14-May 4-Jun 19-May 9-May 16-May 23-May 19-May 1.0 Cumulative passage Julian date Release Site CLWTRP GRNTRP IMNTRP JDAR1 SALTRP SNKTRP Figure year mean estuary arrival (measured at BON) timing distributions for JDAR1, SNKTRP, SALTRP, CLWTRP, IMNTRP, and GRNTRP release sites. Note: Julian date 100 is April 10 th, 125 is May 5 th, 150 is May 30 th, and 175 is June 24 th. See Figure 36 caption for release site abbreviation definitions. Discussion and Conclusions Our comparison of upstream and downstream Chinook salmon population-specific life history attributes yielded several important results: 1) We found no evidence for a consistent and/or systematic difference in size-at-migration existing between upstream and downstream populations. That is, both upstream and downstream production areas yielded smolts of similar, but variable (on an inter-annual basis) size. We also demonstrated that a portion of fork length variation could be attributed to density-dependent effects. 2) Our analysis of trap-passage timing distributions illustrates that both upstream and downstream populations depart from natal streams within a similar timeframe. We also found evidence for greater variation in outmigration timing for upstream relative to downstream 77

101 populations. This finding is consistent with that of Williams et al. (2005), who reported greater variation in passage timing (at BON) for unmarked, upstream-origin yearling Chinook salmon. 3) Across all years in question, we found that upstream-origin smolts migrated to the estuary at a faster rate (~ twice as fast) than those emigrating from the John Day system. This result was not surprising given that upstream-origin fish spend a greater amount of time en route to sea (i.e., they travel from 3-4 times as far away as downstream stocks) and that smoltification status increases and travel times decrease as an increasing function of time spent in migration (e.g., Berggren and Filardo 1993; Williams et al. 2005). 4) Based on a comparison of migration rates between upstream and downstream populations for similar sections of their respective mainstem migration corridors (i.e., between the first and third dams encountered by each group), we found that hydrosystem migration rates did not differ between groups but were strongly influenced by water travel time. 5) Despite their similar size, similar emigration timing, and faster downstream migration rate, upstream-origin smolts arrived at the estuary later (~7-10 days) than John Day River Chinook salmon smolts. Given conclusions 2, 3, and 4 above and the historical increase in water transit times due to hydropower dam development, however, the observed discrepancy in arrival timing at BON is more likely a result of the FCRPS than some innate life history difference existing between upstream and downstream Chinook populations. In summary, our analysis illustrates that although subtle differences occur within and across Chinook salmon populations, there is no indication that a systematic smolt life history difference exists between upstream and downstream production areas. Thus, while our use of an upstream-downstream comparison relies on a natural experiment approach and is therefore imperfect by design, the analysis we present here illustrates that the potential confounding effects due to life history differences are negligible, if not non-existent. However, to address these matters further, we will conduct additional analyses for our 10-year report. 78

102 Chapter 8 Appendix Redd Density Estimation In order to account for the effects of density on smolt size, we used wild and natural Chinook salmon redd counts as an index of the relative abundance of juveniles emigrating in a given migration year. To do this, we summed redd counts from the majority of trend-monitoring reaches occurring upstream of each of the CSS-affiliated traps. For Oregon sites (JDAR1, IMNTRP, GRNTRP), we acquired redd count data from ODFW reports (Wilson et al. 2005) and unpublished data sources (accessed via Streamnet); for Idaho sites (SALTRP, CLWTRP), we compiled redd abundance data from Brown (2002) and unpublished IDFG sources. For the SNKTRP site, we summed counts compiled for sites upstream of IMNTRP, GRNTRP, and SALTRP. In each case, we also estimated the total stream length surveyed (in km) so that density estimates could be standardized across sites. Survey-length information was obtained directly from reports when available; otherwise, it was accessed via queries of Streamnet s database. This approach is meant to provide a coarse, but relative picture of juvenile density across the different production areas and years. However, it relies on several assumptions. Among the more important ones are: 1) there is a strong relationship between spawner abundance and juvenile production and rearing density; and 2) trend monitoring areas are where the majority of spawning/production occurs, or if not, they approximate redd density in non-indexed reaches. The redd data, their sources, and some minor comments/clarifications appear in Table 42. Table 42. Redd abundance and surveyed kilometers for production areas upstream of CSS trap sites used to contrast smolt size between upstream and downstream populations. Counts by brood year (migration year - 2) Trap Site km Comments John Day Inclusive of all indexed reaches that are reported in Wilson et al. (2005) and upstream of trap site. Salmon Inclusive of all indexed reaches that are reported in Brown (2002) and upstream of trap. Clearwater Inclusive of all indexed reaches that are reported in Brown (2002) and upstream of trap. Snake Sum of survey lengths and redd numbers from reaches associated Salmon, Imnaha, and Grande Ronde traps. Imnaha Accessed via Streamnet. Grande Ronde Accessed via Streamnet; does not include the Vey Meadows area, nor other production areas in the Grande Ronde Basin upstream of the trap (due to data limitations). 1. Lengths were estimated based on the maps presented in Brown (2002) and a query of Streamnet s database for transect details; thus, a minor amount of measurement error exists in these values. 79

103 CHAPTER 9 Understanding the implications of smolt size detection probability relationships for CSS study-group comparisons Introduction Recent analyses demonstrate the existence of negative relationships between wild and hatchery Chinook salmon and steelhead smolt size (fork length, FL, in mm) and detection probabilities at Little Goose (LGS) and Lower Monumental (LMN) dams (Williams et al. 2005; Zabel et al. 2005). Given that the primary opportunity for smolts to be detected at these projects is via bypass systems, the implications of consistent size detection probability relationships for studies relying on bypassed (i.e., inclusive of C 1 and T 0 study groups) and inriver (i.e., undetected, C 0 ) outmigrant categories for drawing inference on the effects of the FCRPS on salmon may be considerable. Specifically, in their discussion Zabel et al. (2005) suggest that a study design like that used by CSS is inherently confounded; because smolt-to-adult survival (SAR) is size-dependent (e.g., Zabel and Williams 2002) and individuals may be sorted into transported and inriver treatment groups on the basis of size (i.e., larger fish stay inriver, smaller individuals end up in the bypass system), it may not be possible to separate the effects of the treatment (i.e., bypass and/or transport) from the pre-existing effect of size on performance. Before size-related confounding can be concluded, several issues need to be considered. First, while Zabel et al. (2005) demonstrate significant size detection probability relationships for LGS and LMN collection and transport sites, the majority (>50%, Appendix E, Table 1) of CSS transport-group fishes are collected at Lower Granite Dam (LGR). Thus, the strength and sign (i.e., +/-) of size detection probability association at LGR, which have not been previously quantified, are perhaps most influential on any realized size-related confounding. Secondarily, the practical (i.e., biological) significance of LGS-, LMN-, and LGR-specific size detection probability relationships needs to be evaluated. For instance, while Zabel et al. s relationships for wild steelhead were consistent, those estimated for both hatchery and wild Chinook salmon appeared to be weaker and more variable. Thus, it may be necessary to estimate the realized size discrepancy between study categories to fully appreciate the influence of size detection probability relationships on studies like the CSS. Based on the results of Zabel et al. (2005) and on NOAA s comments on our 2005 annual report (Appendix D in Berggren et al. 2005), we consider these issues in detail in this section of our report. In doing so, we focused exclusively on wild Chinook salmon for our evaluation, given that this group exhibits the largest transport vs. inriver post-bonneville differential delayed mortality difference (i.e., D = SAR BON- LGR(T 0 )/SAR BON-LGR (C 0 ); Chapter 3) of all species and rearing type combinations evaluated as part of the CSS. Our approach relied on the following steps: 1) Using an AIC-based model-selection procedure, we evaluated the level of empirical support for size detection probability relationships at LGR, LGS, and LMN among smolts tagged and released immediately upstream of Lower Granite pool during migration years as part of the CSS; 80

104 2) We estimated size detection probability function parameters (i.e., the slope and intercept of fitted logistic functions) and their associated uncertainty for LGR, LGS, and LMN bypass/collection sites; 3) We contrasted FL (at release) between detected and undetected smolts (minus known removals made at upstream projects) through year and project-specific t-tests. In the following pages, we detail the methods and results associated with each of these steps. We conclude with a brief discussion about the implications of our findings and those of Zabel et al. for current and future efforts of CSS. Also, we identify other analyses that we hope to complete in order to more fully evaluate size detection probability relationship issues in the future. Methods Dataset details -- We evaluated the level of empirical support for fork length detection probability relationships using a dataset consisting of wild Chinook salmon smolts that were measured, PIT-tagged, and released at the Snake River trap (SNKTRP) and Clearwater River trap (CLWTRP) during migration years (MY) Because fish size was a variable of primary concern, we selected these sites to ensure that size-at-release could be reasonably assumed (i.e., given their close proximity to LGR) to reflect that existing at dam arrival and/or bypass opportunity. Initially, we queried PIT-tag releases occurring in a period encompassing the peak of smolt outmigration and tagging at the SNKTRP and CLWTRP release sites (15 March-20 May) so as to obtain as large of a sample size as possible (minimum used for survivaldetection modeling, n = 1,000). However, due to the existence of temporal trends in discharge, spill, and fish size-at-release within migration years, we limited our analysis to the 30-day period extending from 11 April-10 May. Based on preliminary analyses, both hydrological variables and size were reasonably stable across this period during the years in question. Table 43. Sample sizes for PIT-tagged release groups (sum of SNKTRP and CLWTRP releases between 11 April and 10 May) used in our estimation of P(det FL) relationships (1999, 2000, 2002, ) and comparison of size between detected and undetected study categories, by migration year (MY). Bold-faced values correspond to MYs included in our survival/detection probability modeling exercise. Species/rear type MY n wild Chinook wild Chinook ,592 wild Chinook ,320 wild Chinook wild Chinook ,150 wild Chinook wild Chinook wild Chinook ,026 wild Chinook ,022 81

105 Based on the above restrictions, we obtained data for use in our evaluation of wild Chinook salmon survival and recapture probabilities for five of the eight years queried (1999, 2000, 2002, ; Table 43); for detected vs. undetected t-tests, we included all years available (i.e., ). Using PIT-tag detections made at mainstem Snake and Columbia river dam sites, we constructed 5-digit binary (0 = not detected; 1 = detected) capture histories for each individual, with the initial digit corresponding to the SNKTRP release site, the second to LGR, the third to LGS, the fourth to LMN, and the fifth to a combined McNary-John Day- Bonneville Dam (MCN-BON) detection site (i.e., 3 detection opportunities were collapsed into a single one to bolster sample sizes; after Zabel et al. 2005). Also, fork length measurements made at the trap site were paired with individual capture histories. Analytical approach -- Using the dataset described above, we evaluated relationships between individual sizes and bypass probability using a three-tiered approach. First, we modeled survival and recapture probabilities for marked fish as a function of individual size and site effects using a modified Cormack-Jolly-Seber (CJS) framework. While maintaining a constant survival probability structure (see Table 44 caption for details), we fit eight candidate models, reflecting various hypotheses about whether or not size influenced individual recapture probability, and if so at which projects (Table 44). Note that our approach did not consider size effects for the final joint survival-detection probability (i.e., for MCN-BON) parameter, given that it was a pooled 3- site estimate and that we were interested in site-specific effects only. We evaluated the level of empirical support for competing hypotheses using an information-theoretic approach (Burhnam and Anderson 2002); we ranked models according to their ΔAIC values, and considered the model with the lowest AIC score to be our top model. If the best detection probability model included size effects and was measurably better than closely competing models (i.e., separated by > 2.0 ΔAIC units), we concluded that evidence existed for a relationship between FL and detection probability. Table 44. Candidate detection probability (p) models fitted for fish groups released in migration years , 2002, and For detection-probability model selection, the survival (φ) model structure was held constant based on the recommendations of Lebreton et al. (1992), in the most global form [i.e., φ(site FL, all), survival varies across sites as a site-specific function of length]. Model structure p(.) p(site only) p(site FL, LGR,LGS,LMN) p(site FL, LGR, LGS) p(site FL, LGR,LMN) p(site FL, LGS,LMN) p(site FL, LGR) p(site FL, LGS) p(site FL, LMN) Description of associated biological hypothesis p is constant across sites and does not vary as a function of size. p varies by site, but irrespective of size. p varies across sites as a site-specific function of size. p varies across sites, but as a function of size at LGR and LGS only. p varies across sites, but as a function of size at LGR and LMN only. p varies across sites, but as a function of size at LGS and LMN only. p varies across sites, but as a function of size at LGR only. p varies across sites, but as a function of size at LGS only. p varies across sites, but as a function of size at LMN only. Independent of the results from our model-selection phase, we also evaluated fitted slope parameters (and SEs) for site-specific (i.e., LGR, LGS, LMN) size detection probability 82

106 functions. To do this, we assessed the sign and precision of all slope parameters and inspected plots of size detection probability relationships. We also computed across-year slope estimates using a meta-analysis framework. That is, for each collection site we estimated a pooled slope as β pool = [Σβ i *w i ]/Σw i, where w i = 1/σ i 2 (i.e., inverse of variance around β i ) and i is one of the study years; standard errors were computed as SE(β pool ) = (1/Σw i ). For all slope estimates, we deemed statistical significance (i.e., H 0 : β = 0) when the absolute value of estimates exceeded two standard errors, after Zabel et al. (2005). Further, we contrasted our estimates with those reported previously by NOAA-Fisheries in order to understand the generality of patterns seen across years and sites common to both our datasets. Finally, it should be noted that while we did estimate survival and detection probabilities at LGR, LGS, LMN, and MCN-JDA projects, we relied on a modest trap-release dataset for doing so and thus emphasize primarily those results for our upper-most sites only, particularly LGR. In addition to estimating size detection probability relationships at LGR and downstream, we also compared the size distributions of detected and undetected wild Chinook salmon that were tagged and released at the CLWTRP and SNKTRP sites. For LGR, LGS, and LMN, we compared log 10 -transformed FLs between groups using t-tests, on a MY-by-MY basis. However, we excluded individuals that were known to be removed (i.e., placed in a barge) at an upstream site for subsequent downstream comparisons. Note that while we could account for known removals in this process, this approach did not allow us to account for inter-dam mortality. All survival/detection probability analyses were completed using Program MARK. Other statistical analyses were completed using SYSTAT, version 9. We evaluated significance at α = Table 45. Model-selection results for wild Chinook salmon release groups with sufficient tags for survival and recapture probability estimation (i.e., >1,000), by migration year. ΔAIC values appear in cells. Top models (i.e., those with the lowest AIC value) are identified with bold-faced font and underlining; near-top models (i.e., those with a ΔAIC value < 2) also appear as underlined, but in italics. See Table 44 for description of survival and detection probability model structures. Results Model structure p(.) p(site) p(site FL, all) p(site FL, LGR, LGS) p(site FL, LGR,LMN) p(site FL, LGS,LMN) p(site FL, LGR) p(site FL, LGS) p(site FL, LMN) Detection probability modeling exercise -- Based on fish tagged and released above LGR, our survival-detection probability model-selection exercise provided no clear indication of a strong relationship between individual size and detection probability at LGR, or any site downstream 83

107 (Table 45). In two of the five years analyzed (2000 and 2002), the model lacking FL effects (i.e., the null case) at all sites was the top model, though not unambiguously so; in two other years (2005 and 2006), the non-size model was virtually equivalent to any containing FL effects. The only year with measurable support for a size detection probability relationship was 1999, during which the top model indicated a positive relationship exists between FL and detection probability at LGR (Figure 42). In contrast to site versus site size model ambiguity, separation between a constant detection probability model (i.e., p(.)) and all others was unequivocal. 1.0 Detection probability Fork length (mm) Figure 42. Estimated fork length (mm) detection probability relationships for wild Chinook salmon at LGR for MYs 1999, 2000, 2002, 2005, and Our evaluation of site-specific size detection probability slope parameter estimates resulting from our most fully specified models matched model-selection results. The only slope parameter differing significantly from zero out of all MY-site combinations was that for the 1999 LGR size detection probability relationship. Estimates from all other years, though of similar magnitude to those reported for LGS and LMN by NOAA (i.e., where comparison was possible; Table 46), did not differ significantly from zero on a year-specific basis. However, given the limited amount of data available for estimation below LGR, we most emphasize our LGR findings. At this site, relationships varied qualitatively from positive to negative to neutral across the MYs in question (Figure 42). Further, pooled slope values (β LGR = 0.001, SE(β LGR ) = 0.003; β LGS = , SE(β LGS ) = 0.004; β LMN = , SE(β LMN ) = 0.004) based on an inverse-variance-weighted, meta-analysis estimation corroborate year-specific patterns; only LGS had a non-zero significant (and negative) pooled size detection probability slope value. Thus, overall there was no strong evidence for a consistent size-related bias in detection probability for our primary transport site (LGR), but some evidence for an effect at LGS. Detected vs. undetected FL contrasts -- t-tests comparing size distributions between detected and undetected smolts after release at SNKTRP and CLWTRP provide an additional indication that size-sorting is not a major concern for CSS wild Chinook salmon (Table 47). For LGR, detected fish had significantly greater log 10 -transformed FLs than did undetected fish (by 2 mm, based on back-transformed means) in two years (1998, 1999). At LGS and LMN, detected fish were significantly smaller than undetected fish in 1 and 2 years, respectively. In all cases, size 84

108 differences between detected and undetected fish, where statistically significant, were less than or equal to 2 mm; in all other years, median sizes of detected and undetected fish were virtually identical and not statistically distinguishable (Table 47). It should be noted that even though we could not directly account for mortality in this analysis, if survival was positively and detection probability (and subsequent removal) was negatively size dependent, we would expect size differences to progressively increase in a downstream direction. However, this is not the case. Thus, these results conservatively indicate differences existing in size distributions between detected and undetected fish are virtually nonexistent. Table 46. Maximum likelihood slope parameter estimates from detection probability fork length relationships for wild Chinook salmon captured, PIT-tagged, and released at the Snake River and Clearwater River smolt traps (rel_site = SNKTRP, CLWTRP). Bold-faced values correspond to those parameters with point estimates that were greater than twice the value of their standard errors (after Zabel et al. 2005). Estimates delineated by NOAA correspond to the values reported in Zabel et al., CSS corresponds to our upstream-of-lgr release analysis. p 2 (LGS) p 3 (LMN) MY slope SE slope SE 1999-NOAA NOAA NOAA mean CSS CSS CSS mean Discussion and Conclusions Despite its limitations, this analysis suggests that on average size detection probability relationships are likely of negligible importance for wild Chinook salmon study group comparisons currently made as part of the CSS. The following observations support this conclusion: 1) First, model-selection results provided only marginal support for any size detection probability relationship, across sites and MYs. For LGR in particular, the bypass site where the majority of our study fish are collected and assigned to their respective treatment groups estimated relationships were weak to nonexistent. At LGS and LMN, relationships were quite variable across the 5-year record and of comparable magnitude to those estimated by Zabel et al. (2005). Given the high survival values estimated for release to LGR (φ 1 = 90-95% in all MYs) and high detection probabilities estimated for lower-river sites (>60-70% at LGS, LMN, MCN- BON), we are confident that this result is not simply a statistical power issue. This contention is supported by the fact that the data clearly discriminated between site-specific and constant detection probability (i.e., p(.)) models in all MYs. 85

109 Table 47. Summary statistics for detected and undetected wild Chinook salmon captured, tagged, and released from the Snake River and Clearwater River smolt traps during MYs Rows with bold-faced font are those MYs where a significant difference (α = 0.05) was detected between categories using a t-test. Detected Undetected Site Year n median SD n median SD LGR LGS LMN ) Considering realized size distributions, there were no clear differences between detected and undetected fish, across projects and years. This is especially true at LGR, the site for which our data are most reliable, where sizes were virtually identical for both groups. For downstream sites (LGS, LMN), our t-test results similarly suggest a lack of size separation. In sum, both survival/detection probability modeling results and basic comparisons of size distributions between detected and undetected wild Chinook salmon suggest that realized size differences between transported and inriver study groups are negligible. While this section constitutes our first attempt at addressing size-sorting issues in the CSS, we plan to conduct additional analyses to further understand the confounding effects of size detection probability relationships on our results to date. Specifically, we intend to perform simulations evaluating realized distributions of fish entering bypass and remaining inriver given the range of possible size-related detection and survival probability functions (and associated uncertainty) 86

110 demonstrated herein and in Zabel et al. (2005). Also, we intend to compare SARs between study groups containing fish that were detected and not detected in the Lower Snake (conditioned upon being seen at a site in the Lower Columbia River), as an explicit function of length (e.g., using ANCOVA), in order to evaluate this issue on a total life-cycle basis. Also, we intend to explore similar analyses for other species and rear types used in the CSS. With this information, we hope that we will be able to separate the effects of transport from those due to a size-related treatment-group assignment bias, if such a phenomenon indeed exists for these other study groups. 87

111 CHAPTER 10 Computer program to create simulated PIT tag input files for testing robustness of CJS survival estimates The CJS methodology assumes that all members of a tagged group of interest have a common underlying probability of survivability and collectability. When these conditions (along with other assumptions mentioned in Appendix A) are met, the CJS estimates for reach survival between dams and collection efficiency at dams will be unbiased with minimum variance. A key purpose of the simulation studies will be to determine how unequal the underlying survivability and collectability may become among members of a population before the CJS estimates are compromised. An initial demonstration of how the simulation program may be used to investigate this question is conducted in this chapter using a set of default values for parameter inputs (described below). These default values were established to reflect the variable conditions of survivability and collectability occurring with real populations of migrating smolts. In the 2002/03 CSS Annual Report (Berggren et al. 2005), we discussed the need to conduct simulation studies to evaluate the robustness of the CJS inriver survival estimates. Also, we wanted to compare the results of using the CJS on a population from the full season to weighted survival rates created by running the CJS on temporal subsamples of tagged fish from the season (i.e., subcohorts). Since then, development and refinement of a sophisticated computer program to create simulated data sets as an input to the bootstrap program has been underway. The resulting simulator program will provide simulated data sets for investigating the extent to which changing survival and detection probabilities may impact CJS estimates computed for full season samples. It does not allow the post-stratifying, based on date of detection at LGR, of this full season sample into temporal blocks for assessing the subcohort method, which was attempted, but abandoned in earlier CSS reports. Early on, it became apparent that a large sample size of PIT-tagged fish was needed to estimate in-river survival rates from Lower Granite Dam to Bonneville Dam and so applying the CJS method to the full sample rather than subcohorts was essential. The simulator creates a single population of tagged fish that moves through the hydrosystem experiencing user defined changing patterns of survivability and collectability over the migration season. The simulator program accounts for travel time and temporal spread of the passage distributions of migrating fish as they move thorough the hydrosystem in order to reflect how real fish pass the monitored dams. Capture history codes are created as these fish are split between undetected, detected and bypassed, or detected and transported routes of passage at these dams. The resulting simulated population of fish with associated capture history codes may then be run through the bootstrap program to obtain the CJS reach survival estimates. Estimates of reach survival rates between Lower Granite and Lower Monumental dams are used in expanding study category smolt numbers to Lower Granite Dam equivalents. Estimates of inriver survival rates between Lower Granite Dam and Bonneville dams are used in calculating the Vc term in the computation of D. In contract year 2006, there have been accomplishments in this direction, and some preliminary simulation results will be presented herein. However, the major activity in 2006 has been in the area of continued computer program enhancements to make it more efficient for the 88

112 end-user when multiple scenarios of changing survival and capture probabilities are being run. Additionally, a new shell program is being developed to allow the end-user to create many (1000 or multiples of 1000) independent dataset as random draws from a specific underlying population defined by a unique set of input parameters to produce an environment in which temporally changing survival and capture probabilities and smolt inter-dam travel times occurs. This later enhancement is needed to address whether the confidence intervals created by the bootstrapping program for the various survival rate parameters and combinations of these parameters presented in the CSS closely approximate the nominal coverage around the point estimate of interest. Simulator Input Running the simulator program creates a dataset that may then become the input file to the bootstrap program. To demonstrate how this program works, I will first start with the default set of input parameter values and show examples of samples drawn from the population of fish defined by the default parameter values. The default parameter values were calibrated to reflect conditions seen with real data of past years (particularly smolt migration year 2000). Figures 43 to 49 show the default parameter values entered into the seven input screens of the simulator program for an individual simulator run. Figure 43. First input screen of simulator program initial settings including release number and survival to LGR, travel time related parameters, and assumed SAR levels. 89

113 Figure 44. Second simulator input screen arrival population characteristics, collection efficiency and removal rates at LGR, and smolt travel time and survival to LGS. Figure 45. Third simulator input screen collection efficiency and removal rates at LGS, and smolt travel time and survival to LMN. 90

114 Figure 46. Fourth simulator input screen collection efficiency and removal rates at LMN, and smolt travel time and survival to MCN. Figure 47. Fifth simulator input screen collection efficiency and removal rates at MCN, and smolt travel time and survival to JDA. 91

115 Figure 48. Sixth simulator input screen collection efficiency and removal rates at JDA, and smolt travel time and survival to BON. Figure 49. Seventh simulator input screen collection efficiency and removal rates at BON, smolt travel time to trawl site, and trawl collection rate (joint survival-collection efficiency). 92