Steelhead in space: integrating spatial ecological processes into stream fish conservation and management

Transcription

1 Steelhead in space: integrating spatial ecological processes into stream fish conservation and management Jeff Falke USGS, AK Cooperative Fish & Wildlife Research Unit Jason Dunham USGS, Forest and Rangeland Ecosystem Science Center Kris McNyset Department of Fisheries and Wildlife, Oregon State University Chris Jordan NOAA Fisheries Service, Northwest Fisheries Science Center Gordie Reeves USFS, Pacific Northwest Research Laboratory

2

3 Complex Life Cycles and Multiple Habitat Types Spawning habitat Movement to spawn Movement to grow Movement to mature Ocean Juvenile habitat Northcote (1978); Schlosser and Angermeier (1995)

4 Habitat complementation = Spatial proximity of nonsubstitutable habitat types Spatial Ecological Processes Expectations: 1. Energetic dispersal cost = Dunning et al. (1992) Growth and survival 2. Travel time = Predation mortality Schlosser (1995)

5 Habitat Complementation in Stream Networks? Spawning habitat Rearing habitat

6 Objectives 1)Develop continuous measures of spatial ecological processes that may be important to steelhead 2)Evaluate their relative importance in predicting distribution and abundance of spawning 3)Assess the utility of models for conservation and management applications

7 Study Area John Day River, OR 30 km PNW Natural Features Classification (T. Whittier, OSU)

8 John Day Steelhead (O. mykiss gairdnerii) ODFW Sampling Universe ~4300 stream-km Photo: J. Monroe/Freshwaters Illustrated North Fork Lower Mainstem Middle Fork Photo: J. McMillan Upper Mainstem South Fork Photo: J. McMillan

9 John Day Steelhead Redd Survey Data Photo: ODFW Status and Trend Monitoring Fifty - 2 km long surveys per year N = 209 sites (~10% of pop n) Response Presence/Absence Maximum Count *Data courtesy ODFW John Day River Basin Summer Steelhead Environmental Monitoring & Assessment Program Photo: ODFW Redds present Redds absent

10 Steelhead Freshwater Life-Cycle: Key Processes 1) Adult survival and spawning habitat accessibility Stream size Energetic cost 2) Spawning success Substrate suitability Scour likelihood 3) Juvenile growth and survival Thermal regime

11 Survival and Access Stream size John Day steelhead distribution ~ stream size (Mills et al. 2011) VIC hydrologic model (Wenger et al. 2010) Mean annual flow (m 3 sec-1 ) Excluded 0 900

12 Survival and Access Energetic cost Energetic work (Hinch et al. 1996; Crossin et al. 2004) Distance to outlet (km) x elevation (m) Travel time, range of velocities 0 100

13 Steelhead Freshwater Life-Cycle: Key Processes 1) Adult survival and spawning habitat accessibility Energetic cost Stream size 2) Spawning success Substrate suitability Scour likelihood 3) Juvenile growth and survival Thermal regime

14 Spawning Success Substrate Suitability D50 = Median grain size (mm) Reach-scale model (200-m) John Day-specific parameterization Bank-full depth (α, β) Channel classifications (n, k) Buffington et al. 2004: D50 = (ραaβ S) 1 n ρ s ρ kg n Classified reaches for steelhead Kondolf & Wolman 1993: Steelhead D50: P 25 :10-18 mm P 50 :18-34 mm P 75 :34-48 mm

15 What is a neighborhood for juvenile steelhead? Spawning habitat Rearing habitat

16 Dispersal kernel for Bridge Creek Watershed Neighborhood D50: D50NEXP = 1 j 1 d ij D50 j x 100 J. McMillan photo

17 Spawning Success Substrate Suitability Site D50 Neighborhood D50

18 Spawning Success High Flow Events (Scour) Scour likelihood (Montgomery et al. 1996; Fausch et al. 2001) # days in Spring among highest 5% flow Spring = Feb 1 June 30 Frequency of high spring flows (S95) 7 d 27 d

19 Steelhead Freshwater Life-Cycle: Key Processes 1) Adult survival and spawning habitat accessibility Energetic cost Stream size 2) Spawning success Substrate suitability Scour likelihood 3) Juvenile growth and survival Thermal regime

20 Thermal regime Water Temperature Data MODIS Satellite Data (NASA) 1km 2 spatial resolution, daily Spatially and temporally continuous John Day River basin Stream temp logger dataset ~1757 loggers Spatially and temporally patchy Land Surface Temperature (LST) (McNyset et al. in prep)

21 Predicted water temperature ( C) (McNyset et al. in prep) Modeling thermal regimes Temp ~ LST + Julian day 2003 R 2 = 0.94 RMSE = 1.39 C Observed water temperature ( C)

22 Mean daily water temperature ( C) <predicted> < 0 Predictions > 20

23 Mean daily water temperature ( C) <predicted> < 0 Predictions > 20

24 Mean daily water temperature ( C) <predicted> < 0 Predictions > 20

25 Juvenile habitat-thermal regime Riverscape temperature model (McNyset et al. in prep) Cumulative growing season degree-days ( C) May 1 Oct 31 Averaged across 10 years Growing season degree days (GSDD)

26 Spatial Structure / Hatchery Effects? visitidaho.org North Fork US ACOE Lower Mainstem Middle Fork Upper Mainstem South Fork columbiariverimages.com

27 Candidate Models Life stage Covariate P/A Count Maternal survival and access Stream size X Distance x Elevation X Spawning success Site D50 X Neighborhood D50 X Scour likelihood X Juvenile growth/survival Thermal regime X X Population TRT (Categorical) X Site-level predictors (2-km reaches) Neighborhood-level predictors (5+ km surrounding sites) Hurdle regression model (Colin & Trivedi 1998; 2005; Zuur et al. 2009): f hurdle y; β, γ = f binomial y = 0; γ 1 f binomial y = 0; γ x f negbin (y; β) 1 f negbin (y = 0; β) y = 0 y > 0

28 Model selection results Model AIC c K AIC c w i Neighborhood only Site only Neighborhood + Sub-basin Site + Sub-basin Mixture Site-level only model Neighborhood-level only model Mixture of local and neighborhood Photo: J. Monroe Falke et al PLoS ONE

29 Model-Averaged Parameter Estimates Redd occurrence = Stream size + 2x Site D50 + 2x Neighborhood D50 + 2x Thermal regime (GSDD) + 6x Distance x Elevation - 1x* Redd abundance = Scour likelihood - 1x Thermal regime (GSDD) + 7x Site-level Landscape-level *95% CI overlaps zero Falke et al PLoS ONE

30 Thermal Regime (GSDD) Falke et al PLoS ONE

31 Little suitable D50 nearby Much suitable D50 nearby Falke et al PLoS ONE

32 Scour Likelihood / TRT Falke et al PLoS ONE

33 Among-Model Predictive Diagnostics Binomial Count* Model PCC AUC r ρ AVE error RMSE Composite Neighborhood only Site only PCC = Proportion correctly classified AUC = Area under-the-curve statistic r = Pearson s correlation coefficient ρ = Spearman s rank correlation coefficient AVE error = 1 ( y n i y i ) i=1 RMSE = n n 1 ( y n i y i ) 2 i=1 *Based on bootstrap evaluation (Steyerberg et al. 2001; Potts & Elith 2006) Falke et al PLoS ONE

34 Lower Camas Creek (HUC-8) Observed absent Observed present P/A Predictions km

35 Discussion and Implications General guidelines for modeling spatial process 1. Think like a fish! 2. Predictors important across life stages 3. Don t ignore local habitat effects 4. Balanced, robust survey design

36 Future Directions Potamodromy Matanuska Susitna River Basin, Alaska Photos: Kevin Fraley Cook Inlet Image courtesy Alaska TNC

37 Acknowledgments Funding: Data/Analyses/Ideas: ODFW Field Crews, Carol Volk (NOAA), Allen Brookes (EPA), Phil Kaufmann (EPA), Marc Weber (EPA), Tom Kincaid (EPA), John Van Sickle (EPA), Ian Tattam (ODFW), Jim Ruzycki (ODFW), Seth Wenger (TNC), Agnes Przeszlowska (USFS), Scott Heppell (OSU)

38 Thanks! Photo: J. McMillan