Slides from ESA 2015
- 1. Integrated population models for predicting spatial and temporal dynamics at range margins Richard Chandler and Samuel Merker Warnell School of Forestry and Natural Resources University of Georgia
- 2. Trailing-edge populations Motivation Methods Results 2 / 20
- 3. Trailing-edge populations Motivation Methods Results 2 / 20
- 4. Trailing-edge populations Motivation Methods Results 2 / 20
- 5. Trailing-edge populations Motivation Methods Results 2 / 20
- 6. Trailing-edge populations Motivation Methods Results 2 / 20
- 7. Trailing-edge populations Motivation Methods Results 2 / 20
- 8. Trailing-edge populations Motivation Methods Results 2 / 20
- 9. Trailing-edge populations Motivation Methods Results 2 / 20
- 10. Motivating questions What are the mechanisms governing range shifts? Motivation Methods Results 3 / 20
- 11. Motivating questions What are the mechanisms governing range shifts? • Phenological mismatch Motivation Methods Results 3 / 20
- 12. Motivating questions What are the mechanisms governing range shifts? • Phenological mismatch • Competition/Predation Motivation Methods Results 3 / 20
- 13. Motivating questions What are the mechanisms governing range shifts? • Phenological mismatch • Competition/Predation • Direct physiological effects Motivation Methods Results 3 / 20
- 14. Motivating questions What are the mechanisms governing range shifts? • Phenological mismatch • Competition/Predation • Direct physiological effects • Allee effects Motivation Methods Results 3 / 20
- 15. Motivating questions What are the mechanisms governing range shifts? • Phenological mismatch • Competition/Predation • Direct physiological effects • Allee effects What will be the consequences for trailing-edge populations? Motivation Methods Results 3 / 20
- 16. Motivating questions What are the mechanisms governing range shifts? • Phenological mismatch • Competition/Predation • Direct physiological effects • Allee effects What will be the consequences for trailing-edge populations? How can we draw inferences, make reliable forecasts, and inform conservation efforts? Motivation Methods Results 3 / 20
- 17. What would be the ideal model? Motivation Methods Results 4 / 20
- 18. What would be the ideal model? Components • Survival, reproduction, dispersal Motivation Methods Results 4 / 20
- 19. What would be the ideal model? Components • Survival, reproduction, dispersal • Spatial and temporal variation Motivation Methods Results 4 / 20
- 20. What would be the ideal model? Components • Survival, reproduction, dispersal • Spatial and temporal variation • Density dependence Motivation Methods Results 4 / 20
- 21. What would be the ideal model? Components • Survival, reproduction, dispersal • Spatial and temporal variation • Density dependence • Species interactions Motivation Methods Results 4 / 20
- 22. What would be the ideal model? Components • Survival, reproduction, dispersal • Spatial and temporal variation • Density dependence • Species interactions • Demographic and environmental stochasticity Motivation Methods Results 4 / 20
- 23. What would be the ideal model? Components • Survival, reproduction, dispersal • Spatial and temporal variation • Density dependence • Species interactions • Demographic and environmental stochasticity • Individual heterogeneity Age Sex Size Other traits Motivation Methods Results 4 / 20
- 24. Ecological process model Spatio-temporal point process model Motivation Methods Results 5 / 20
- 25. Ecological process model Spatio-temporal point process model Initial distribution and abundance si,1 ∝ λ(s) E(N1) = Λ = S λ(s) ds Motivation Methods Results 5 / 20
- 26. Ecological process model Spatio-temporal point process model Initial distribution and abundance si,1 ∝ λ(s) E(N1) = Λ = S λ(s) ds zi,1 ∼ Bernoulli(Λ/M) N1 = M i=1 zi,1 Motivation Methods Results 5 / 20
- 27. Ecological process model Spatio-temporal point process model Initial distribution and abundance si,1 ∝ λ(s) E(N1) = Λ = S λ(s) ds zi,1 ∼ Bernoulli(Λ/M) N1 = M i=1 zi,1 Survival and recruitment zi,t ∼ Bernoulli(zi,t−1φ(s) + ai,t−1γ(s) ) Motivation Methods Results 5 / 20
- 28. Ecological process model Spatio-temporal point process model Initial distribution and abundance si,1 ∝ λ(s) E(N1) = Λ = S λ(s) ds zi,1 ∼ Bernoulli(Λ/M) N1 = M i=1 zi,1 Survival and recruitment zi,t ∼ Bernoulli(zi,t−1φ(s) + ai,t−1γ(s) ) Dispersal si,t ∼ k(si,t−1) Motivation Methods Results 5 / 20
- 29. Observation process models Capture-recapture data Distance sampling data Motivation Methods Results 6 / 20
- 30. Observation process models Capture-recapture data pi,j,k,t = p0 exp( si,t − xj /(2σ2 )) yi,j,k,t ∼ Bern(pi,j,k,t × zi,t) Distance sampling data Motivation Methods Results 6 / 20
- 31. Observation process models Capture-recapture data pi,j,k,t = p0 exp( si,t − xj /(2σ2 )) yi,j,k,t ∼ Bern(pi,j,k,t × zi,t) Distance sampling data ui,t,k ∼ Normal(si,t, σ) qi,j,k,t = exp( ui,t,k − xj /(2τ2 )) yi,j,k,t ∼ Bern(qi,j,k,t × zi,t) Motivation Methods Results 6 / 20
- 32. Study area Motivation Methods Results 7 / 20
- 33. Study area Motivation Methods Results 8 / 20
- 34. Study design Motivation Methods Results 9 / 20
- 35. Study design Motivation Methods Results 10 / 20
- 36. Study design Motivation Methods Results 10 / 20
- 37. Study design Motivation Methods Results 10 / 20
- 38. Study design Motivation Methods Results 10 / 20
- 39. Study design Motivation Methods Results 11 / 20
- 40. Data Motivation Methods Results 12 / 20
- 41. Data Motivation Methods Results 12 / 20
- 42. Data Motivation Methods Results 12 / 20
- 43. Data Motivation Methods Results 12 / 20
- 44. Model Initial abundance • Elevation • Rhododendron index (NDVI) Motivation Methods Results 13 / 20
- 45. Model Initial abundance • Elevation • Rhododendron index (NDVI) Survival – constant Motivation Methods Results 13 / 20
- 46. Model Initial abundance • Elevation • Rhododendron index (NDVI) Survival – constant Recruitment • Elevation • Rhododendron index • Density dependence Motivation Methods Results 13 / 20
- 47. Model Initial abundance • Elevation • Rhododendron index (NDVI) Survival – constant Recruitment • Elevation • Rhododendron index • Density dependence Detection • Distance • Observer • Playback Motivation Methods Results 13 / 20
- 48. Results Process Parameter Median Lower CI Upper CI Initial abundance Intercept 0.171 -0.190 0.501 Elevation 0.983 0.711 1.239 NDVI 0.056 -0.239 0.339 Motivation Methods Results 14 / 20
- 49. Results Process Parameter Median Lower CI Upper CI Initial abundance Intercept 0.171 -0.190 0.501 Elevation 0.983 0.711 1.239 NDVI 0.056 -0.239 0.339 Survival — 0.432 0.229 0.803 Motivation Methods Results 14 / 20
- 50. Results Process Parameter Median Lower CI Upper CI Initial abundance Intercept 0.171 -0.190 0.501 Elevation 0.983 0.711 1.239 NDVI 0.056 -0.239 0.339 Survival — 0.432 0.229 0.803 Recruitment Intercept -1.614 -4.340 -0.910 Elevation 1.270 0.508 2.523 NDVI 0.336 -0.943 0.908 Density -0.178 -0.008 -0.639 Motivation Methods Results 14 / 20
- 51. Results – Initial distribution Motivation Methods Results 15 / 20
- 52. Results – Initial distribution Motivation Methods Results 15 / 20
- 53. Results – Population growth Motivation Methods Results 16 / 20
- 54. Results – Forecasts Motivation Methods Results 17 / 20
- 55. Conclusions (1) Spatial IPMs allow for inference on mechanisms governing range shifts. Motivation Methods Results 18 / 20
- 56. Conclusions (1) Spatial IPMs allow for inference on mechanisms governing range shifts. (2) Underlying point process model can be informed by numerous data types. Motivation Methods Results 18 / 20
- 57. Conclusions (1) Spatial IPMs allow for inference on mechanisms governing range shifts. (2) Underlying point process model can be informed by numerous data types. (3) Canada Warbler results are preliminary but suggest a strong effect of climate on recruitment. Motivation Methods Results 18 / 20
- 58. Conclusions (1) Spatial IPMs allow for inference on mechanisms governing range shifts. (2) Underlying point process model can be informed by numerous data types. (3) Canada Warbler results are preliminary but suggest a strong effect of climate on recruitment. (4) Future work will allow for causal inference by combining observational and experimental data. Motivation Methods Results 18 / 20
- 59. Thanks • Robert Cooper • Jeff Hepinstall-Cymerman • Len Reitsma • Samuel Hach´e • Heather Abernathy • Brittney Ferrari • Anna Joy Lehmicke • Carly Chandler • Jared Feura (photos) • Alex Wilson • Vincent Weber • Angela Hsiung
- 60. Questions?