1. In review Do NOT CITE The role of climate change in forecasts of Pacific salmon population dynamics Mark Scheuerell mark.scheuerell@noaa.gov

2. In review Do NOT CITE DISCLAIMER The results presented herein are currently "in review" and therefore should not be distributed or cited until further notice. If you have any questions, please contact Mark Scheuerell Northwest Fisheries Science Center NOAA Fisheries (206) 302-2437 mark.scheuerell@noaa.gov September 30, 2004

3. In review Do NOT CITE Acknowledgments John Williams (NOAAF) Countless others from NOAAF, USFWS, IDFG, ODFW, WDFW

4. In review Do NOT CITE Looking toward the future Society is faced with an uncertain future from increasing global change Scientists and policy makers agree that future “success” rests with the capacity to anticipate Increasingly important as the human population grows Ecological forecasting represents a step toward predicting ecosystem services using specified uncertainties under future scenarios Clark et al. (2001) Science

5. In review Do NOT CITE A caution on forecasting Pielke & Conant (2003) Ecology In Feb 1997, forecasters predicted that the Red River of the North would see flooding greater than anything previously recorded At Grand Forks, ND forecasters predicted a flood crest of 49 ft. In April, the river crested at 54 ft. & inundated several cities, causing $2 billion in damages Local, state, & federal officials cited the inaccuracy of the forecast as the problem Reality: The forecast was within the long-term 10% error Bottom line: everyone needs to understand the uncertainty involved

6. In review Do NOT CITE Forecasting in fisheries It’s done all the time in fisheries management (but not very well) We often use simple models like stock-recruit relationships More recent incorporation of more complex mathematics & environmental effects (e.g., Logerwell et al. 2003; Lawson et al. 2004) Salmon represent a good case study because of their high economic, social, and ecological value (Ruckelshaus et al. 2002)

7. In review Do NOT CITE Columbia R. Snake R. Oregon Idaho Washington NE Pacific Canada

8. In review Do NOT CITE Spring/Summer Chinook Salmon “Stream type” life history Adults return to spawn & die in April-July (year t) Parr emerge and rear in natal creeks & rivers (year t+1) Smolts emigrate in April-June (year t+2) Adults at sea (years t+3 to t+5) freshwater ocean

9. In review Do NOT CITE The slide toward extinction Year Returning adults (1000s) ESA listing

10. In review Do NOT CITE Possible reasons for decline Generally grouped under the “4 H’s” Harvest Hatchery operations Habitat degradation Hydroelectric (& other) dams …but there are others too Exotic species Climate Marine-derived nutrients

11. In review Do NOT CITE

12. Smolt Adult Assessing stock productivity

13. In review Do NOT CITE Smolt-to-adult survival rate (SAR) Count smolts emigrating past dam Count adults returning 1-3 years later

14. In review Do NOT CITE Early trends in survival Number of dams SAR (%)

15. In review Do NOT CITE It’s the dams, dummy!

16. In review Do NOT CITE Recent trends in survival Number of dams SAR (%)

17. In review Do NOT CITE Is it really just the dams?

18. In review Do NOT CITE The Pacific Decadal Oscillation* *Mantua et al. (1997); cited 568 times as of Sept 2004 A “shot in the arm” for fisheries, oceanography, climatology PDO

19. In review Do NOT CITE A dynamic ocean environment Salmon survival related to climate PDO (Mantua et al. 1997) ALPI (Beamish et al. 1997) AFI (McFarlane et al. 2000) Upwelling (Botsford & Lawrence 2002) Various (Logerwell et al. 2003) Other trophic levels as well Zooplankton (Brodeur et al. 1999) Crabs (Zheng & Kruse 2000) Intertidal inverts (Sagarin et al. 1999) Seabirds (Jones et al. 2002)

20. In review Do NOT CITE SAR (%) An ocean-climate effect? Regime shift Regime shift

21. In review Do NOT CITE Columbia R. Snake R. Oregon Idaho Washington NE Pacific Canada

22. In review Do NOT CITE The environmental driver Also known as the Bakun Index (Bakun 1990) Generated monthly by NOAA PFEL based on naval oceanographic data Spatially referenced at every 3° of lat from 21-60 N Spring upwelling promotes 1° & 2° production (Pearcy 1992, Brodeur & Ware 1992) Fall downwelling may decrease advection of important zooplankton prey (Mackas 2001) Related to salmon survival (Nickelson 1986, Botsford & Lawrence 2002, Logerwell et al. 2003) Pacific Coastal Upwelling Index (CUI)

23. In review Do NOT CITE Narrowing the search Choosing candidate predictor variables Exhaustive search over all possible combinations of 12 months is daunting Chose index from 45N, 48N & 46.25N (interpolated) Used stepwise multiple regression to choose potential predictor months for time series model The results The index from 45N was far superior  reflects early ocean distribution? April, September & October were significant  transition periods important?

24. In review Do NOT CITE Time series of the CUI m 3 seawater / 100 m shoreline / sec April September October

25. In review Do NOT CITE Time Series Analysis Observation equation Y t = X t ´  t + v t v t ~N[0,V t ] Evolution equation  t = G t  t-1 + w t w t ~N[0,W t ] Dynamic Linear Models Recipe for DLMs 1)Make forecast using information up through previous year 2)Wait for current-year observation and then update all priors 3)Repeat steps 1-2 to the end of the time series 4)Assess overall model performance through Bayes Factors

26. In review Do NOT CITE Dynamic Linear Models A note on information discounting At each time step, there is a decay of information This leads to greater uncertainty Address this through discounting of the Bayesian priors V[  t |D t-1 ] =  -1 ·V[  t-1 |D t-1 ]where  (0,1] Choose appropriate  by minimizing NLL of model When  small, parameters “evolve” quickly, but with decreased precision of the prediction In practice, 0.99 <  < 0.8

27. In review Do NOT CITE Columbia R. Snake R. Oregon Idaho Washington Canada NE Pacific CUI SAR

28. In review Do NOT CITE Forecasting climate-induced survival 19601970198019902000 Year of ocean entry 3 1 0 4 5 6 2 model observed R 2 = 0.71 SAR (%)

29. In review Do NOT CITE The best statistical description 19601970198019902000 Year of ocean entry 3 1 0 4 5 2 model observed R 2 = 0.91 SAR (%)

30. In review Do NOT CITE Linking all environments

31. In review Do NOT CITE Columbia R. flow at The Dalles Flow (kcfs)

32. In review Do NOT CITE Smolts Adults Eggs Parr An early view of the life cycle OceanFreshwater

33. In review Do NOT CITE Adding up the drivers The 4 H’s Marine-derived nutrients Exotic species Climate change in the oceans & on land

34. In review Do NOT CITE An improved view? Atmosphere Smolts Adults Human influence Smolts Adults Eggs Parr OceanFreshwater

35. In review Do NOT CITE Conclusions Effective conservation and management requires ecological forecastsEffective conservation and management requires ecological forecasts Environmental science has largely failed to produce theseEnvironmental science has largely failed to produce these Pacific salmon provide a good case studyPacific salmon provide a good case study We can use simple ocean-climate metrics to predict salmon survivalWe can use simple ocean-climate metrics to predict salmon survival We need to examine the “big picture” with respect to both life history & environmental processesWe need to examine the “big picture” with respect to both life history & environmental processes It’s time to move forward with an eye on the pastIt’s time to move forward with an eye on the past