Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Dissertation defense

Dissertation defense

My dissertation defense talk in its entirety. This talk was presented on April 14, 2017.

Peter D Smits

April 14, 2017
Tweet

More Decks by Peter D Smits

Other Decks in Science

Transcript

  1. Remodeling the fossil record analysis of emergent evolutionary and ecological

    patterns Peter D Smits Committee on Evolutionary Biology, University of Chicago 1 / 74
  2. Macroevolution and macroecology Structured data and modelling emergent patterns Patterns

    in survival Background extinction and expected differences in species survival Interplay between extinction intensity and extinction selectivity Patterns in functional diversity Mammal species pool functional composition Conclusions and commentary 2 / 74
  3. Macroevolution and macroecology Structured data and modelling emergent patterns Patterns

    in survival Background extinction and expected differences in species survival Interplay between extinction intensity and extinction selectivity Patterns in functional diversity Mammal species pool functional composition Conclusions and commentary 3 / 74
  4. Definition macroevolution: study of patterns which emerge when considering the

    evolutionary history of multiple species. macroecology: study of patterns which emerge when considering the ecology of multiple species. in both time and space 5 / 74
  5. Traits as conceptual and operational link Definition trait: identifiable property

    of an organism e.g. pelage color, body mass, beak depth, tooth shape functional trait: trait that strongly influences performance/means of interacting with environment species trait: identifiable property assignable to a species 6 / 74
  6. Species selection Rabosky and McCune 2010 TREE Species selection is

    the outcome of heritable variation in speciation and extinction rates among taxa. avoids selection versus sorting, “strict” species selection versus effect macroevolution 7 / 74
  7. Species fitness Cooper 1984 J. Theoretical Biology Expected time till

    extinction. logic: if more fit, more likely to be present distribution based definition (population) other definitions can be derived based on definition of extinction 8 / 74
  8. Law of Constant Extinction Van Valen 1973 Evol. Theory Extinction

    risk (species fitness), in a given adaptive zone, is taxon–age independent. 9 / 74
  9. Survival of the unspecialized Simpson 1944 Tempo and Mode in

    Evolution p. 143 When related phyla die out . . . more specialized phyla tend to become extinct before less specialized. This phenomenon is also far from universal, but it is so common that it does deserve recognition as a rule or principle in evolutionary studies: the rule of the survival of the relatively unspecialized. 10 / 74
  10. Macroevolution and macroecology Structured data and modelling emergent patterns Patterns

    in survival Background extinction and expected differences in species survival Interplay between extinction intensity and extinction selectivity Patterns in functional diversity Mammal species pool functional composition Conclusions and commentary 14 / 74
  11. Inference The theory of probability is the only mathematical tool

    available to help map the unknown and the uncontrollable. (Mandelbrot) 16 / 74
  12. Bayesian inference and statistics flexible, expressive, intuitive regularize, partial pooling,

    external information Stan probabilistic programming language Hamiltonian Monte Carlo Automatic Differentiation Variational Inference 19 / 74
  13. Macroevolution and macroecology Structured data and modelling emergent patterns Patterns

    in survival Background extinction and expected differences in species survival Interplay between extinction intensity and extinction selectivity Patterns in functional diversity Mammal species pool functional composition Conclusions and commentary 21 / 74
  14. Motivating questions How do mammal species traits affect extinction risk?

    How do shared time of origination or evolutionary history relate to extinction risk? How do my findings compare to current risk factors? Is species extinction risk age-independent? 23 / 74
  15. Survival model diagram y i y i ∼Weibull(σ ,α) η

    j[i] ∼Normal(0,σ c ) σ c ∼half-Cauchy(2.5) h i ∼MultiNormal(0,Σ) Σ=σ p 2 V phy σ p ∼half-Cauchy(2.5) β∼Normal(0,10) α∼half-Cauchy(2.5) σ ,α ~ exp(−(η j [i ] +h i +∑ βT X i ) α ) ⏟ μ σc ~ γ ~ ~ γ μ σ ~ MultiNormal(0,σp 2 V phy ) γ ~ ~ ⋯ ⋯ ⋯ 29 / 74
  16. Effect of locomotor category on extinction risk −0.2 0.0 0.2

    0.4 0.6 βarb − βgrd βarb − βscn βgrd − βscn Estimated difference A (Smits 2015 PNAS) 31 / 74
  17. Effect of dietary category on extinction risk −0.50 −0.25 0.00

    0.25 0.50 βcrn − βhrb βcrn − βist βcrn − βomn βhrb − βist βhrb − βomn βist − βomn Estimated difference B (Smits 2015 PNAS) 32 / 74
  18. Difference in risk between origination cohorts q q q q

    q q q q q q q q q q q q q q q q q q q q q q q q q q q q −0.5 0.0 0.5 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Time (Mya) Estimated cohort effect (Smits 2015 PNAS) 33 / 74
  19. Three sources of variance 0 3 6 9 0 3

    6 9 0 3 6 9 individual cohort phylogeny 0.00 0.25 0.50 0.75 1.00 Variance partition coefficient Prob. Density (Smits 2015 PNAS) 34 / 74
  20. Summary of results Survival of the unspecialized as time-invariant generalization.

    Decrease in extinction risk with time. Both cohort/temporal and phylogenetic effect. Some incongruence with risk factors in the Recent. e.g. effect of body size, trophic category, phylogenetic clustering. 35 / 74
  21. Macroevolution and macroecology Structured data and modelling emergent patterns Patterns

    in survival Background extinction and expected differences in species survival Interplay between extinction intensity and extinction selectivity Patterns in functional diversity Mammal species pool functional composition Conclusions and commentary 36 / 74
  22. Jablonski 1986 Science At K/Pg mass extinction, biological traits (except

    geographic range) have no effect on bivalve taxonomic survival. 37 / 74
  23. Questions and analysis How do the effect of traits on

    duration (extinction selectivity) vary with expected duration (extinction intensity)? 40 / 74
  24. Post-Cambrian Paleozoic brachiopod genera and covariates time range approx. 488-252

    Mya. stage as time unit; duration measured in stages (2-5 My each) effect of traits varies by origination cohort geographic range body size environmental preference (v, v2) gap statistic as measure of sampling (Foote and Raup 1996 Paleobio), imputed for taxa with short durations 42 / 74
  25. Hierarchical survival model sampling E[taxon duration for cohort j] body

    size for j environmental preference for j environmental breadth for j geographic range for j “age” correlation of effects over time 43 / 74
  26. Sampling statement for the joint posterior probability yi,t ∼ Weibull(σi,t,

    α) log(σi,t) = Xi Bj[i],t + δsi α Bj ∼ MVN(µ, Σ) Σ = diag(τ)Ωdiag(τ) si ∼ Beta(φi , λ) φi = logit−1(Wi γ) µintensity ∼ N(0, 5) µrange ∼ N(−1, 1) µenvpref ∼ N(0, 1) µenvcurve ∼ N(1, 1) µsize ∼ N(0, 1) δ ∼ N(0, 1) τ ∼ C+(1) Ω ∼ LKJ(1) λ ∼ Pareto(0.1, 1.5) γ ∼ N(0, 1) Note: Calculation of log probability of right and left censored observations is modified from the above 44 / 74
  27. Variation in trait effects between cohorts q q q q

    q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q −4.0 −3.5 −3.0 −2.5 −2.0 −1.5 −2.0 −1.5 −1.0 −0.5 0.0 −3 −2 −1 0 1 0.0 2.5 5.0 −1.0 −0.5 0.0 0.5 1.0 intensity range env_pref env_curv size 250 300 350 400 450 Time (My) Effect estimate for... 45 / 74
  28. Overall effect of environmental preference 1 2 −0.5 0.0 0.5

    Environmental preference (open−ocean <−−> epicontinental) log(approx. expected duration in t) 46 / 74
  29. Change in effect of environment between cohorts 14. Emsian 15.

    Eifelian 16. Givetian 17. Frasnian −1 0 1 2 3 −0.5 0.0 0.5 −0.5 0.0 0.5 −0.5 0.0 0.5 −0.5 0.0 0.5 Environmental preference (v) log(approx. expected duration in t) 47 / 74
  30. Change in effect of environment between cohorts 1. Tremadoc 2.

    Floian 3. Dapingian 4. Darriwilian 5. Sandbian 6. Katian 7. Hirnantian 8. Llandovery 9. Wenlock 10. Ludlow 11. Pridoli 12. Lochkovian 13. Pragian 14. Emsian 15. Eifelian 16. Givetian 17. Frasnian 18. Famennian 19. Tournaisian 20. Visean 21. Serpukhovian 22. Bashkirian 23. Moscovian 24. Stephanian 25. Asselian 26. Sakmarian 27. Artinskian 28. Kungurian 29. Roadian 30. Wordian 31. Capitanian 32. Wuchiapingian 33. Changhsingian −2 0 2 4 −2 0 2 4 −2 0 2 4 −2 0 2 4 −2 0 2 4 −0.5 0.0 0.5 −0.5 0.0 0.5 −0.5 0.0 0.5 −0.5 0.0 0.5 −0.5 0.0 0.5 Environmental preference (v) log(approx. expected duration in t) 48 / 74
  31. Summary of results Effect of geographic range consistent with prior

    expectations; low variance. No effect of body size; low variance. Epicontinental environmental preference slightly favored on average; high variance. Strong support for survival of unspecialized as generalization wrt environmental preference; medium variance. 50 / 74
  32. Macroevolutionary process Magnitude of effect of geographic range and environmental

    preference increase with extinction intensity. As extinction risk decreases, the differences between taxa matter less. Evidence for qualitative difference between mass and background extinction. 51 / 74
  33. Macroevolution and macroecology Structured data and modelling emergent patterns Patterns

    in survival Background extinction and expected differences in species survival Interplay between extinction intensity and extinction selectivity Patterns in functional diversity Mammal species pool functional composition Conclusions and commentary 52 / 74
  34. relative expected species duration short long arboreal ground dwelling scansorial

    insectivore herbivore carnivore omnivore locomotor diet (Smits 2015 PNAS) 55 / 74
  35. Paleontological fourth-corner model true presence observed presence observation probability traits

    environment trait species species species time time time factor 57 / 74
  36. Covariates of interest individual-level (species i at time unit t)

    ecotype: combination diet and locomotor categories effect is function of group-level covariates body size (rescaled log body mass) group-level (2 My time unit t) temperature record based on Mg/Ca estimates mean and range (rescaled log degrees) plant community phase following Graham 2011 58 / 74
  37. Paleontological fourth-corner model true presence observed presence observation probability traits

    environment trait species species species time time time factor 59 / 74
  38. Model and sampling statement definition yi,t ∼ Bernoulli(pi,t zi,t )

    pi,t = logit−1(α0 + α1mi + rt ) rt ∼ N(0, σ) α0 ∼ N(0, 1) α1 ∼ N(1, 1) σ ∼ N+(1) zi,1 ∼ Bernoulli(φi,1) zi,t ∼ Bernoulli zi,t−1πi,t + t x=1 (1 − zi,x )φi,t φi,t = logit−1(aφ t,j[i] + bφ 1 mi + bφ 2 m2 i ) πi,t = logit−1(aπ t,j[i] + bπ 1 mi + bπ 2 m2 i ) aφ ∼ MVN(Uγφ, Σφ) aπ ∼ MVN(Uγπ, Σπ) Σφ = diag(τφ)Ωφdiag(τφ) Σπ = diag(τπ)Ωπdiag(τπ) ρ ∼ U(0, 1) bφ 1 ∼ N(0, 1) bπ 1 ∼ N(0, 1) bφ 2 ∼ N(−1, 1) bπ 2 ∼ N(−1, 1) γφ ∼ N(0, 1) γπ ∼ N(0, 1) τφ ∼ N+(1) τπ ∼ N+(1) Ωφ ∼ LKJ(2) Ωπ ∼ LKJ(2). 60 / 74
  39. Summary of results changes to ecotype composition driven by origination,

    not extinction specific ecotypes source of most variation in overall origination arboreal taxa decrease through Paleogene, all but absent by Neogene digitigrade and unguligrade herbivores only groups with sustained increase environmental covariates virtually always affect origination, not survival 68 / 74
  40. Macroevolution and macroecology Structured data and modelling emergent patterns Patterns

    in survival Background extinction and expected differences in species survival Interplay between extinction intensity and extinction selectivity Patterns in functional diversity Mammal species pool functional composition Conclusions and commentary 69 / 74
  41. High level review macroevolution and macroecology devoted to explaining emergent

    patterns in evolution and ecology emphasis on functional traits yields strong and intuitive results because of obvious selective importance 70 / 74
  42. High level review macroevolution and macroecology devoted to explaining emergent

    patterns in evolution and ecology emphasis on functional traits yields strong and intuitive results because of obvious selective importance Gelman: “big data are messy. messy data need large models. large models need Bayesian inference.” 70 / 74
  43. Synthesis law of constant extinction neither study of survival supports

    this evidence for increasing risk with duration 71 / 74
  44. Synthesis law of constant extinction neither study of survival supports

    this evidence for increasing risk with duration survival of the unspecialized strong support as generality; should be our “null” brachiopods: when extinction intensity high this pattern breaks qualitative difference between mass and background extinction 71 / 74
  45. Synthesis law of constant extinction neither study of survival supports

    this evidence for increasing risk with duration survival of the unspecialized strong support as generality; should be our “null” brachiopods: when extinction intensity high this pattern breaks qualitative difference between mass and background extinction functional diversity hypotheses from macroevolution can inspire macroecological study model unifies macroevolutionary and macroecological frameworks 71 / 74
  46. Acknowledgements Angielczyk lab David Grossnickle, Dallas Krentzel, Jackie Lungmus, Jonathan

    Mitchell Foote lab Marites Villarosa Garcia, Samuel Miller, Nadia Pierrehumbert, Kathleen Ritterbush Gregory Wilson Alistair Evans Jeff Bradley, Donald Grayson, Jim Kenagy, Nancy Simmons Sandy Carlson, Christine Janis Carolyn Johnson, Elizabeth Eakin 2012 Darwinian/Paleo cohorts Stewart Edie, Amy Henry, Katherine Silliman, Sarah Tulga, Max Winston Jessica Escott, Ben Frable, Brian Goodrich, Colin Kyle, Darcy Ross, Elizabeth Sander, Laura Southcott, Julie Symaszek, Brian Waligorski Jean Leahy 73 / 74