1 ' $ Neutral Model of Vegetation – An Introduction to Hubbell – Jari Oksanen Department of Biology University of Oulu 6622 for 2007 & % 2 ' $ The Book Hubbell, S. P. (2001) The Unified Neutral Theory of Biodiversity and Biogeography. Princeton University Press. & % 3 ' $ The World without Competition • Radically neutral model • No competitiion • No differences among species • No environmental differences or “gradients” • Neutral model is certainly wrong but this does not matter because the model works! & % 4 ' $ The Lectures • Introduction into Hubbell’s Unified Neutral Theory and related aspects of plant community ecology • Species abundance relations, (bio)diversity, species–area relationship, spatial pattern, and species pool: All in the light of Neutral Theory • How to estimate Hubbell’s Fundamental Parameter (θ) with the Data • Simulation of Hubbell models and parameter estimation in statistical environment • Warning: Some new material will be added to these lectures! & % CONTENTS 5 ' $ Contents 1 A Gentle Introduction to Hubbell 6 2 Species Abundance Models 18 3 Hardcore Hubbell 45 4 Hubbell’s Game 63 5 Species Richness and Sample Size 74 6 Hubbell’s Second Game: Metacommunity Landscape 88 7 Species Pool: What is a Metacommunity? & 107 % 1 A GENTLE INTRODUCTION TO HUBBELL ' 1 6 $ A Gentle Introduction to Hubbell In “The Unified Neutral Theory of Biodiversity and Biogeography’ S. P. Hubbell explains his neutral model that can explain most things that are worth explaining in Biogeography. A crucial parameter is Hubbell’s “fundamental biodiversity number” θ. In this lecture I give a gentle introduction to the model — no matrix algebra at all! & % 1 A GENTLE INTRODUCTION TO HUBBELL ' 7 $ Neutral model • All organisms are equal, irrespective of species or position in the community • Individuals can interact, as long as the rules are identical for all individuals and species • Recognized mechanisms 1. Random ecological drift 2. Stochastic, albeit limited distribution 3. Random speciation • Neutral models are “wrong”, but if they work — who cares? & % 1 A GENTLE INTRODUCTION TO HUBBELL ' 8 $ Island Biogeography: Another Neutral Model • Only the sizes and distances of islands important: No differences among species, but the the toal number of species is interesting • Logical consequence: Random collection of species • Applications negated neutrality, and looked for differences among species. • Hubbell more radical: Neutral at the level of individual. & % 1 A GENTLE INTRODUCTION TO HUBBELL ' 9 $ The Earth is Neutral and Crowded • Species multiply and fill the Earth, but communities won’t grow without limit • The number of individuals J increases (linearly) with area • Community assembled in a zero-sum game: Total number of individuals unchanged • Species succeeds only if some other species perish & % 1 A GENTLE INTRODUCTION TO HUBBELL ' 10 $ Local Community Dynamics 1. Assume we have a local community of J individuals 2. D randomly picked individuals die 3. Surrounding Metacommunity sends in M immigrants 4. Remaining free lots are replaced from within the local community with probabilities proportional to abundances after disturbance & % A GENTLE INTRODUCTION TO HUBBELL 11 ' $ 60 80 Ecological Drift in Local Community 0 20 40 N 1 0 50 100 150 200 t Absorbing states: Monodominance, extinction to others & % A GENTLE INTRODUCTION TO HUBBELL 12 ' $ Isolated Communities Island: extinction 500 J=1600, D=400, m=0 100 50 t=0 t=10 20 t=500 10 • The ultimate fate of all communities is monodominance. t=1000 200 • Isolation leads to extinction and lower richness. Abundance t=25 t=100 2 • Local extinction is the fate of most species. 5 t=250 1 1 t=50 0 10 20 30 40 Rank & % de 1 1 A GENTLE INTRODUCTION TO HUBBELL • Single Local Community in the Metacommunity Sea: θ defines ' Metacommunity composition Pi and it is filtrated into Local Communities asRescue: m allows:Metacommunity 13 $ m Ji Pi θ = 2JM ν • Metacommunity as a Network of Local Communities: Migration •mLocal communityrates J is surrounded by aCommunities, Metacommunity and speciation ν create Local andP which can send in new isspecies Metacommunity their sum: • Immigration from a Metacommunity restores diversity and cures θ Jki Jki ν= ! ! from monodominance m 2 Jki k i ! entity: We only • At the moment, Metacommunity is a mythical Jki Jki Pi = Jki know that it is there because we see its effectsk in local "community, but we do not know what is a Metacommunity • At least Metacommunity is big so that extinctions are slow, but still they should happen & $ % 1 A GENTLE INTRODUCTION TO HUBBELL 14 ' $ Speciation in Metacommunity • Monodominance is the ultimate fate of a community • In Metacommunity, random speciation balances extinction and leads into stable species abundance relations • The size JM of a Metacommunity is large, and (nearly) impossible to measure, the rate of speciation ν is low, and impossible to measure, but together they produce Hubbell’s fundamental diversity number θ: θ = 2JM ν which is moderate in range and can be estimated & % 1 A GENTLE INTRODUCTION TO HUBBELL ' 15 $ Derivation of Ultimate Diversity. . . • What is the probability F2t+1 that two random individuals belong to the sampe species in generation t + 1 when this probability was F2t in the previous generation? • Probability that neither speciated at birth is (1 − ν)2 and probability that they have a common parent is 1/JM . • In the beginning of Times, all had one parent and recursively 1 1 t+1 2 F2 = (1 − ν) + 1− F2t JM JM and in stable community F2t+1 = F2t . . . & % 1 A GENTLE INTRODUCTION TO HUBBELL ' 16 $ . . . And The Solution is • In stable community F2t+1 = F2t = F2 , with an approximate solution: F2 = ∼ = ∼ = & (1 − ν)2 JM − (1 − ν)2 (JM − 1) 1 1 + 2JM ν 1 1+θ % 1 A GENTLE INTRODUCTION TO HUBBELL ' 17 $ Things to explain • Species abundance distribution • Slope of species–area curve • Similarity decay with distance (“beta diversity”) • Things to describe = things to estimate: – Fundamental diversity number θ, and perhaps immigration between communities (m) or its flip side, isolation ω – Fitted models depend on community size J, but that we take as known & % 2 SPECIES ABUNDANCE MODELS ' 2 18 $ Species Abundance Models The Neutral Theory predicts some kind of abundance model and we can —in principle— estimate Neutral Theory parameters from abundances models. But what kind of model do we expect? & % 2 SPECIES ABUNDANCE MODELS ' 19 $ Abundance Distribution models • Logarithmic series: Rare species never end. • Log-Normal model: There is modal commonness, rare species get rare. • Pre-emption model: Species have a natural, hierarchic ordering. • Brokenstick: Species divide community randomly. • Zipf–Mandelbrot: Rare species have larger costs to invade the community because they must cope with conditions and abundant species. • Hubbell’s model: Commonness from a zero-sum game & % SPECIES ABUNDANCE MODELS 20 ' $ Logarithmic series R.A. Fisher in 1940’s: Barro Colorado Island, Site 1, alpha = 35.7 25 20 15 0 • In larger samples, you may find more individuals of rare species, but you find new rare species, too 5 10 • Most species are rare, and species found only once are the largest group 30 αxn fn = n Species 2 0 5 10 15 20 25 Frequency & % SPECIES ABUNDANCE MODELS 21 ' $ Sample size and Log-series 10 5 0 • When you collect a larger sample, you will find new individuals of rare species, Species 15 Barro Colorado Island, N = 12, alpha = 34.3 0 500 1000 1500 Frequency 20 15 Species 10 5 0 • But you will find new, previously unknown species — and these are rare. 25 Barro Colorado Island, N = 25, alpha = 37.1 0 500 1000 1500 Frequency 10 Species 15 Barro Colorado Island, N = 50, alpha = 35.0 5 • Log-series: Shape and parameter α should remain constant with increasing sample size. 0 2 0 500 1000 1500 Frequency & % SPECIES ABUNDANCE MODELS 22 ' $ Fisher’s alpha 200 Barro Colorado Island, Accumulation of Plots 50 20 • Another parameter (x) is an uninteristing nuisance parameter: x = N/(N + α) 10 • Should be independent of sample size. Fisher alpha S exp(H) 1/Simpson Berger−Parker 100 • Fisher’s α is used as a diversity index. Diversity 2 0 5000 10000 15000 20000 Number of Trees Rug: Plots & % SPECIES ABUNDANCE MODELS 23 ' $ Log-Normal model Preston did not accept Fisher’s log-series, but assumed that rare species end with sampling • Canonical standard model of our times & 35 30 25 20 15 10 5 • Modal class in higher octaves, and not so many rare species 0 • Plotted number of species against ‘octaves’: doubling classes of abundance Barro Colorado Island, All Sites Species 2 1 2 4 8 16 32 64 128 512 2048 Frequency % SPECIES ABUNDANCE MODELS 24 ' $ Sample Size and Log-Normal Model 20 10 0 • Mode moves to the right. Species 30 Barro Colorado Island, N = 12 1 • The veil moves and reveals new, rare species. 8 16 32 64 128 256 512 20 10 0 Species 30 Barro Colorado Island, N = 25 1 2 4 8 16 32 64 128 256 512 1024 Frequency 25 35 Barro Colorado Island, N = 50 15 • The shape of the fitted model should remain unchanged. 4 Frequency Species • Fewer rare species found as sampling proceeds. 2 0 5 2 1 2 4 8 16 32 64 128 256 512 1024 2048 Frequency & % SPECIES ABUNDANCE MODELS 25 ' $ Preston vs. Fisher 10 15 20 25 30 Fisher Preston Octaves • Higher octaves lump number of frequencies. large 0 5 • Actually Fisher’s logseries becomes humped if you plot it by octaves. 35 • “Small data sets follow logseries, but when you collect more data, they are log-Normal”. Species 2 1 2 4 8 16 32 64 128 256 512 2048 Frequency & % SPECIES ABUNDANCE MODELS 26 ' $ Extrapolated Species Richness 250 200 150 100 • If Log-Normal height is S0 and width σ, the total richness, including species behind the veil, √ is ST = S0 σ 2π. Observed Extrapolated 50 • Normal density has a finite area: It is possible to estimate the number of species behind the veil integrating the LogNormal model. 300 Barro Colorado Island, Accumulation of Plots Number of Species 2 0 5000 10000 15000 20000 Number of Trees Rug: Plots & % SPECIES ABUNDANCE MODELS 27 ' $ Ranked abundance diagrams • Linear: Pre-emption model • Sigmoid: Log-normal or brokenstick & 100 50 10 5 The shape of abundance distribution clearly visible: 1 • Vertical axis: Logarithmic abundance 500 • Horizontal axis: ranked species Runsaus 2 0 50 100 150 200 250 Rank Brokenstick Preemption Lognormal Zipf % SPECIES ABUNDANCE MODELS 28 ' $ Log-Normal RAD model 20 Preemption Lognormal Veiled.LN Zipf Mandelbrot • Veiled Log-Normal: Cut out rare species. 5 1 2 • Sigmoid: excess of both abundant and rare species to preemption model. 10 • Ranked abundances Normal Abundance 2 5 10 15 Rank Brokenstick Preemption Lognormal Zipf & % SPECIES ABUNDANCE MODELS 29 ' $ Preemption and Mandelbrot RAD • Preemption pˆr = k(1 − k)r−1 – Zipf (β = 0): Strong dominance and a long tail of rare species. 20 5 2 • Mandelbrot pˆr = pˆ1 (r + β)γ 10 decay – Line in RAD. Preemption Lognormal Veiled.LN Zipf Mandelbrot 1 – Ranked abundances proportionally. Abundance 2 5 – Mandelbrot: Weaker dominance but a long tail. 10 15 Rank Brokenstick Preemption Lognormal Zipf & % 2 SPECIES ABUNDANCE MODELS 30 ' $ Selecting RAD • Adding parameter means better fit: Pre-emption 1, Log-Normal and Zipf 2, Veiled Log-Normal and Zipf–Mandelbrot 3 parameters. • Fit to minimize −` or negative log-Likelihood, but penalize for each estimated parameter p, and select model with smallest penalized criterion. • Akaike’s Information Criterion AIC = −2` + 2p, and Bayesian Information Criterion BIC = −2` + log(S) × p. • The penalty in AIC is mild, and BIC often more appropriate (but Information Criteria have different goals than statistical testing). & % SPECIES ABUNDANCE MODELS 31 ' $ Best fitting RADs 0 5 Preemption Lognormal 10 15 20 Veiled.LN Zipf 0 Mandelbrot 5 10 15 20 0 5 10 R4P9V2 R1P12V3 R4P11V3 R3P8V2 R1P6V2 R4P6V2 R3P10V3 R2P16V4 R1P10V3 R1P1V1 R2P3V1 R1P3V1 15 20 100 30 10 3 1 100 Abundance 2 30 10 3 1 R4P4V1 R2P8V2 R1P4V1 R2P13V4 R2P5V2 R1P2V1 100 30 10 3 1 0 & 5 10 15 20 0 5 10 15 20 0 5 Rank Carabids, Sites Ordered by Species Richness, Criterion BIC 10 15 20 % SPECIES ABUNDANCE MODELS 32 ' $ Broken Stick • Result looks sigmoid, and can be fitted with log-Normal model. 10 1 2 • No real hierarchy, but chips arranged in rank order: 5 • Species ‘break’ a community (‘stick’) simultaneously in S pieces. 20 Carabids, Trap 5 Abundance 2 5 10 15 Rank Brokenstick Preemption Lognormal Zipf & % SPECIES ABUNDANCE MODELS 33 ' $ Hubbell’s abundance model Ultimate diversity parameter θ Carabids, Trap 5 Zipf–Mandelbrot usually the best model (but Hubbell says Lognormal) & 2 • Simulations can be used for estimating θ. 1 • θ and J define the abundance distribution 5 10 20 • θ = 2JM ν, where JM is metacommunity size and ν evolution speed Abundance 2 5 10 15 Rank % SPECIES ABUNDANCE MODELS 34 ' $ High Diversity 10 5 2 1 1 2 5 Abundance 10 20 Barro Colorado Island, Site 1 & Hubbell 20 Barro Colorado Island, Site 1 & Brokensticks Abundance 2 0 20 40 60 Rank & 80 0 20 40 60 80 Rank % 2 SPECIES ABUNDANCE MODELS ' 35 $ Hubbell’s Species Generator Species generator θ θ+j−1 gives the probability that jth collected individual belongs to a new species 1. Collect J individuals, or j = 1 . . . J 2. Each individual is a new species at probability θ/(θ + j − 1): If this happens, add the species to the community 3. In other case, add abundance 1 to some of the old species, with probabilities proportional to abundances & % 2 SPECIES ABUNDANCE MODELS ' 36 $ R code is crystal clear h u b b e l l . b u i l d <− function ( t h e t a , J ) { comm <− NULL for ( j in 1 : J ) { i f ( runif ( 1 ) < t h e t a / ( t h e t a+j −1)) comm <− c (comm , 1 ) else { sp <− sample ( length (comm ) , 1 , prob=comm/ ( j −1)) comm [ sp ] <− comm [ sp ] + 1 } } return (comm) } & % SPECIES ABUNDANCE MODELS 37 ' $ Expected RAD from simulations Simulated, J=64 Mean of 100 simulations θ=∞ 100 Abundance 1000 5.00 10000 50.00 J=100000 0.50 Abundance 0.05 10 θ = 100 θ=3 θ=1 θ = 10 1 0.01 2 θ = 0.1 0 5 10 15 Rank & 20 25 30 θ = 2 θ = 5 θ = 10 0 100 θ = 50 θ = 20 200 300 400 500 Rank % SPECIES ABUNDANCE MODELS 38 ' $ Supports Preston. . . or Fisher? θ = 100 Simulation J=100000 θ = 20 80 60 5 10 Species 60 40 20 Species 15 80 20 θ = 100 6 8 10 12 2 4 6 Octave 8 10 12 14 Octave θ=5 θ = 10 40 4 Species 2 20 5 4 Species 3 6 2 4 Species 8 6 10 7 All from simulation J=100000 2 4 6 8 Octave 10 12 14 0 1 2 2 2 4 6 8 Octave 10 12 14 0 500 1000 1500 2000 2500 3000 Abundance & % SPECIES ABUNDANCE MODELS 39 ' $ Difference is in the eye of the beholder The same data look like logarithmic series or log-Normal, depending how you plot them 15 0 0 5 5 10 10 Species 20 25 15 30 35 Barro Colorado Island, All Sites Species 2 0 500 1000 Frequency & 1500 1 2 4 8 16 32 64 128 512 2048 Frequency % SPECIES ABUNDANCE MODELS 40 ' $ The Hubbell RAD Carabids Average rad (line) vs. All traps (points) 100 50 1e+00 Abundance 500 1e+01 1e+02 100 Simulations, theta=4, J=100 Abundance 5 10 1e−01 Individual Communities Mean 1 1e−02 2 5 10 15 Rank 20 0 10 20 30 40 Rank Hubbell suggests averaging RADs, but biased because averages ranks instead of species. & % SPECIES ABUNDANCE MODELS 41 ' $ Typical RAD • From preemption to Zipf–Mandelbrot, perhaps through Zipf with increasing J, and faster with high θ. • Lognormal rare: Often an averaging artefact. Preemption Lognormal Veiled.LN Zipf 32 128 theta=4 Mandelbrot 512 2048 theta=15 theta=40 40 Counts 2 30 20 10 0 32 128 512 2048 32 128 512 2048 J & % SPECIES ABUNDANCE MODELS 42 ' $ Diversity Indices 1nverse 10 8 6 4 2 m C Carabids M Fisher F IInverse M 0 isher arabids alpha Simpson • It is best to see diversity indices as variance estimator of abundances • Fisher’s α is also a diversity index, and it is pretty close to Simpson’s index & 4 2 • All sensible indices are pretty similar: it does not matter which you use, unless you use species richness 6 8 10 Carabids Inverse Simpson 2 2 4 6 8 10 Fisher alpha % 2 SPECIES ABUNDANCE MODELS 43 ' $ Hill numbers Common measures of diversity are special cases of R´enyi entropy; S X 1 Ha = log pai 1−a i=1 Mark Hill proposed using Na = exp(Ha ) or the “Hill number”: H0 = H1 = H2 = log(S) PS − i=1 pi log p1 PS − log i=1 p2i H∞ = − log(max pi ) N0 = S Number of species N1 = Shannon N2 = exp(H1 ) PS 1/ i=1 p2i Simpson N∞ = 1/ max pi Berger–Parker All Hill Numbers in the same units: Number of “virtual species”, assuming all species are equal. & % SPECIES ABUNDANCE MODELS 44 ' $ Choice between indices 1.5 2.0 2.5 0.5 1.0 1.5 3.0 1.0 2.0 2.5 1.5 2.0 0 1.5 1 1.5 2.0 1.0 • It is not so important which index is used, since all sensible indices are very similar. 2.5 • Diversity indices are only variances of species abundances. 0.5 1.5 1.0 • Sensitivy to rare species decreases with increasing scale of R´enyi diversity. 1.0 2 Inf 0.5 2 1.5 2.0 2.5 3.0 0.5 1.0 1.5 2.0 Carabids & % 3 HARDCORE HUBBELL ' 3 45 $ Hardcore Hubbell In this section we have a more formal introduction into Hubbell’s radical theory. This chapter is heavy. . . you were warned. . . & % 3 HARDCORE HUBBELL ' 46 $ Isolated local community: No immigrants • Species have two absorbing states with no return: 1. Monodominance: all replacements from one species. 2. Extinction: no replacement – fate of all but one species. & % 3 HARDCORE HUBBELL 47 ' $ Transition matrix for µ = 1 and D = 1 M= & 1 0 0 1 J J−2 J 1 J .. . .. . .. . ··· 0 ··· .. . (Jk )−2(J−2 k−1 ) (Jk ) .. . (J−2 k−1 ) (Jk ) .. . 0 .. . 0 0 .. . 0 .. . ··· 0 .. . .. . 0 0 ··· .. . .. . (J−2 k−1 ) (Jk ) .. . 0 0 0 ··· 1 J J−2 J 1 J 0 0 0 ··· 0 0 1 % 3 HARDCORE HUBBELL ' 48 $ The analysis of the transition matrix • Matrix M can be arranged into a canonical form: I 0 M= R Q • I is the identity matrix for absorbing states, R gives the probability of transitions to absorbing states, Q gives transitions between non-absorbing states, and 0 is zero matrix. & % 3 HARDCORE HUBBELL 49 ' $ Ecological Random Drift • The matrix of ecological random drift is A = (I − Q)−1 • Fixation time is T (N ) = A1T , and in our very special case T (Ni ) = " # N J−1 i X X −1 (J − 1) (J − Ni ) (J − k) + Ni k −1 k=1 & k=Ni +1 % 3 HARDCORE HUBBELL 50 ' $ Numerical example A very small community (J = 8) 1 0 0 0 & 0 0 0 0 0 0.125 0.75 0.125 0 0 0 0 0 0 0 0.214 0.571 0.214 0 0 0 0 0 0 0 0.286 0.429 0.286 0 0 0 0 0 0 0 0.268 0.464 0.268 0 0 0 0 0 0 0 0.268 0.429 0.268 0 0 0 0 0 0 0 0.214 0.571 0.214 0 0 0 0 0 0 0 0.125 0.750 0.125 0 0 0 0 0 0 0 0 1 % 3 HARDCORE HUBBELL 51 ' $ Small Community • Row is the population size at time t, and column at time t + 1 • First row, first column: size 0. • Time step chosen so that D = 1: one change per time step. • Principal diagonal gives the probability that nothing changes, the other elements give the probability of increase or decrease by one individual. J • The binomial coefficient is k : how many ways do we have to 8 8 8 8 pick k individuals from J: 1 = 8, 2 = 28, 3 = 56, 4 = 70, 8 5 = 56 . . . & % 3 HARDCORE HUBBELL 52 ' $ Further numerical analysis • Let’s take the submatrix Q (7 × 7) of previous (9 × 9) matrix. • Matrix A = (I − Q)−1 ={ajk } : If species has abundance Nj how many times it passes abundance Nk before absorptions. A= 7 3.5 2.33 1.75 1.4 1.67 1 6 7 4.67 3.5 2.8 2.33 2 5 5.83 7 5.25 4.2 3.5 3 4 4.67 5.6 7 5.6 4.67 4 3 3.5 4.2 5.25 7 5.83 5 2 2.33 2.8 3.5 4.67 7 6 1 1.17 1.4 1.75 2.33 3.5 7 • The expected time to fixation T (N ) = A1T is T = [ 18.2 & 28.3 33.8 35.5 33.8 28.3 18.2 ] % 3 HARDCORE HUBBELL 53 ' $ Back to business (forget math) • Time to extinction or to monodominance is dependent on community size J and population size Ni of species i. • For rare species (Ni J) time to fixation (extinction or monodominance) is: T (Ni ) ∼ = Ni (J − 1)[1 + ln(J)] • Three multiplications for J: Potentially a large number. • Based on special case D = 1: If mortality is higher, things can develop faster. & % 3 HARDCORE HUBBELL 54 ' $ Open community • New species can immigrate from the metacommunity. • Monodominance no longer an absorbing state: “ergodic” community. • The transition matrix for ergodic community B gives the probability that species remains at abundance Ni (principal diagonal) or that its abundance changes with one individual (subdiagonal elements). & % 3 HARDCORE HUBBELL 55 ' $ Descriptive Statistics The first eigenvector ψ of the matrix B describes species dynamics: E(Ni ) = J X n · ψ(n) n=1 Var(Ni ) = j X (n − E{Ni })2 · ψ(n) n=1 & % 3 HARDCORE HUBBELL 56 ' $ Ergodic community • The simplest possible ergodic community is of size J = 1: 1 − mPi mPi B= m(1 − Pi ) 1 − m(1 − Pi ) where ψ = [1 − Pi , Pi ] • “After six weeks of algebra, we find that the general solution eigenvector is. . . ” (Hubbell, p. 88) & % 3 HARDCORE HUBBELL 57 ' $ . . . And the solution is • Let Pi bet the expected probability of occurrence in the metacommunity, so its expected abundance in the local community will be: E(Ni ) = J X ψ(k) · k = JPi k=0 or equal to the metacommunity, and independent of immigration rate. • Variance, though, is dependent on immigration rate and community size. & % 3 HARDCORE HUBBELL 58 ' $ Many deaths • Previously D = 1: one death per time step. • If more deaths per time step, everything is much more complicated, so we take a shortcut and go directly to conclusions • Everything is faster, but similar. • In closed communities, time to fixation inversely related to mortality in large community, and in small communities even stronger relation. • In ergodic communities D does not influence exptected population size E(Ni ) = JPi , but variance increases. & % 3 HARDCORE HUBBELL ' 59 $ Speciation and Metacommunity • Metacommunity feeds new species to local community to replace local extinctions. • With time, species go extinct even in metacommunity. • Speciation maintains diversity. • Hubbell uses a simple, mechanistic model – in lack of theory – of random point mutation • Rate of speciation ν: Probablilty that offspring belongs to different species (!) than her parent – we may guess this is very small (e.g. ν ≈ 10−10 ) • The universal occurrence of log-Normal abundance model follows from speciation in metacommunity. & % 3 HARDCORE HUBBELL 60 ' $ Stable abundances in metacommunity • Let us inspect stable abundaces (when t → ∞) in a metacommunity of size JM . • What is the probability F2t+1 that two random individuals belong to the same species in generation t + 1, when this probability was F2t in the previous generation? • Probability that neither speciated in birth is (1 − ν)2 • Probability that they have a common parent is & 1 JM % 3 HARDCORE HUBBELL 61 ' $ Balance of Speciation and Extinctions • In the beginning of Times, all had one parent, and so we get a recursive equation: 1 1 F2t+1 = (1 − ν)2 + 1− F2t JM JM • In stable community nothing changes F2t+1 = F2t = F2 or speciation balances extinctions • An approximate solution is: F2 = ∼ = & (1 − ν)2 JM − (1 − ν)2 (JM − 1) 1 1 + 2JM ν % 3 HARDCORE HUBBELL 62 ' $ Hubbell’s Fundamental Diversity Number • Hubbell’s fundamental number θ is θ = 2JM ν • This number is central in the rest of the Theory. • JM is very large and ν is very small, but θ is moderate. • JM is difficult to measure and ν may be impossible to measure, but θ can be estimated. & % 4 HUBBELL’S GAME ' 4 63 $ Hubbell’s Game Most useful results of Hubbell’s theory are best found with simulations, or playing Hubbell Games. & % 4 HUBBELL’S GAME ' 64 $ Metacommunity and local community • If species is very common in Metacommunity, it is dominant in the local community as well. . . most of the time. • Isolated local community: local abundance profile U-shaped: (Temporary) dominance or (temporary) extinction. • With increased migration, local abundance profile approaches metacommunity probability & % 4 HUBBELL’S GAME 65 ' $ Make your own local communities To estimate both θ and migration m, you must play many rounds of the Hubbell game to estimate resulting abundance distribution and its variance • Explicit simulation of a local community with known size J • Run several simulation timesteps with immigration from a Metacommunity • Look for abundance patterns similar to the observed community 1. Vary migration rate m 2. Vary Metacommunity θ & % Hubbell’s Two Models 4 HUBBELL’S GAME ' • Single Local Community in the Metacommunity Sea: θ defines Metacommunity composition Pi and it is filtrated into Local Parameters in Simulation Communities as m allows: 66 $ m Ji de 1 Pi θ = 2JM ν • Metacommunity as a Network of Local Communities: Migration •mJ:and Sizespeciation of local community (fixed) rates ν create Local Communities, and is theirper sum: •Metacommunity D: Number of deaths time step (fixed) θ • m: Probability of immigration from the Metacommunity Jki Jki ν= ! ! m 2 k i Jki (random) ! Jkiproportions Jki in the Metacommunity Pi = k Jki • Pi : Species (fixed) — these may be simulated for a given Metacommunity size JM and " Ultimate Diversity θ & $ % HUBBELL’S GAME 67 ' $ 1 5 10 50 100 500 Start with any community Abundance 4 0 10 20 30 40 Rank & % HUBBELL’S GAME 68 ' $ Isolated Community 1 5 10 50 100 500 θ = 20, m = 0.005, J = 1600 Abundance 4 0 10 20 30 40 50 60 Rank & % HUBBELL’S GAME 69 ' $ Open Community 1 5 10 50 100 500 θ = 20, m = 0.5, J = 1600 Abundance 4 0 10 20 30 40 50 60 Rank & % HUBBELL’S GAME 70 ' $ Expected Abundance in Local Community • Expected proportion in a local community = proportion in Metacommunity • Incidences vary, and variance higher in isolated communities (low m). 0.0 0.1 0.2 0.3 Proportion in Metacommunity & 0.4 0.8 0.6 0.4 0.0 0.2 Proportions in Local Communities 0.8 0.6 0.4 0.0 0.2 Proportions in Local Communities 0.0 0.2 0.4 0.6 0.8 Observed Mean 1:1 1.0 m = 0.005 1.0 m = 0.05 1.0 m = 0.5 Proportions in Local Communities 4 0.0 0.1 0.2 0.3 Proportion in Metacommunity 0.4 0.0 0.1 0.2 0.3 0.4 Proportion in Metacommunity % HUBBELL’S GAME 71 ' $ • The level of variance, or the height of the parabola, increases with decreasing immigration rate m. & 10 m=0.005 m=0.05 m=0.5 5 • Metacommunity variance is a parabolic function of P . 0 • Variance of local abundances related to metacommunity variances JPi (1 − Pi ). 15 Variance of Local Abundance SD in Local Communities 4 0.0 0.2 0.4 0.6 0.8 1.0 Proportion in Metacommunity % HUBBELL’S GAME 72 ' $ Incidence Function & 0.8 0.0 0.2 0.4 Incidence 0.6 0.8 0.6 0.4 0.2 0.0 10 15 20 25 30 0 5 10 15 20 Frequency Frequency m=0.005 m=0.005 25 30 25 30 0.6 0.4 0.0 0.2 0.4 Incidence 0.6 0.8 5 0.8 0 0.2 • In isolated communities, ecological random drift dominates, and causes U–shaped incidence functions: Species either in all plots or in none. m=0.5 0.0 • Ecological Random Drift towards monodominance and extinctions. m=0.5 Incidence • The average incidence = proportion in Metacommunity (Pi ). Incidence 4 0 5 10 15 20 Frequency 25 30 0 5 10 15 20 Frequency Simulated, J=32 % HUBBELL’S GAME 73 ' $ Isolation and RAD m=1 m=0.005 0 20 Observed theta=4 40 m=0.05 60 80 m=0.5 100 Observed theta=15 Observed theta=40 2^ 6 2^ 4 2^ 2 Abundance 4 2^ 0 Fitted theta=4 Fitted theta=15 Fitted theta=40 2^ 6 2^ 4 2^ 2 2^ 0 0 20 40 60 80 100 0 20 40 60 80 100 Rank J=400, D=100, timelag=500 & % 5 SPECIES RICHNESS AND SAMPLE SIZE ' 5 74 $ Species Richness and Sample Size Species richness increases with sample size, and Neutral Theory predicts a specific relationship with number of individuals and number of species. Moreover, we can try to estimate the joint effect of isolation and Ultimate Diversity from Species—Area curves. & % 5 SPECIES RICHNESS AND SAMPLE SIZE ' 75 $ Species richness: The trouble begins • Species richness increases with sample size: can be compared only with the same size. • Rare species have a huge impact in species richness. • Rarefaction: Removing the effects of varying sample size. • Sample size must be known in individuals: Equal area does not imply equal number of individuals. • Plants often difficult to count. & % SPECIES RICHNESS AND SAMPLE SIZE 76 ' $ Rarefied Species Richness • Results may not be clear: Rarefaction curves can cross. 10 8 2 4 • Diversity indices do about the same. 6 • Rarefaction compares species richness in samples rarefied to equal number of individuals. 12 Carabids, Traps 12 and 63 Species Richness 5 0 50 100 150 200 250 300 Sample Size & % SPECIES RICHNESS AND SAMPLE SIZE 77 ' $ Assume that plants are cubes. . . • Sparse stand: Biomass increases when number of plants increases • Crowded stand: Biomass increases when size of plants increases and number decreases 5 10 15 20 25 Kokemäenjoki meadows Species Richness 5 + ++ + + ++ + + + + ++ + + + ++ + + + + + + ++++ + ++ ++ +++++ +++ ++++ + ++ ++ +++++ + + + +++++++ + + + ++++++ +++ + +++ ++ +++++++++++++++++ + +++ + + +++++++ ++ +++ + ++++ + ++++++ ++++++ +++ + + +++ + +++++++ +++ ++ + ++++ + + ++ + ++ ++++++ + + +++ ++++++++ ++++++ ++ +++ +++ + + + + ++++++ +++ +++ + + ++ ++ ++ +++++++ + + + +++++ + + + ++ + ++ + ++ + + + 0 500 + + + + + + ++ 1000 ++ + 1500 Biomass g/m2 & % SPECIES RICHNESS AND SAMPLE SIZE 78 ' $ Species Richness and Area: Classical models 200 150 65.3Lakes0.3 100 • Doubling area brings along a constant number of new species: E(S) = c + log(A) 36.2 + 31.8log2(Lakes) 50 – Usually the only model used, therefore regarded as universally best 250 • Arrhenius model most popular: E(S) = cAz , probably because easy to fit S 5 0 20 40 60 80 100 Lakes & % SPECIES RICHNESS AND SAMPLE SIZE 79 ' $ Species Richness and Fisher Carabids Species never end, but the rate of increase slows down. 5 10 Fisher log-series predicts: N S = α ln 1 + α 15 20 α = 3.82 S 5 0 100 200 300 400 500 600 700 Number of individuals & % 5 SPECIES RICHNESS AND SAMPLE SIZE 80 ' $ Unified, Neutral prediction JM X θ E(S|θ, J) = θ+j−1 j=1 JM − 1 ≈ 1 + θ · log 1 + θ • Exact estimator and its variance estimator available • Often applied on area basis: The shape similar, but θ can be estimated only with counts • Hubbell’s estimator very close to Fisher’s JM E(S|α, J) = α · log 1 + α indicating α ∼ = θ = 2JM ν & % SPECIES RICHNESS AND SAMPLE SIZE 81 ' $ 80 60 40 20 • Both indicate “infinite” species richness: Fisher simply said that rare species never end, Hubbell said that new species evolve at rate ν, since θ = 2JM ν. 0 • Fisher’s α is almost identical to Hubbell’s θ: Approximate θ by estimating α. 100 Hubbell’s theta and Fisher’s alpha Estimated Fisher alpha 5 0 20 40 60 80 100 theta in simulation & % SPECIES RICHNESS AND SAMPLE SIZE 82 ' $ Species Accumulation Curves Carabids • Simulation and exact method possible, like in rarefaction. 30 20 10 • Related to rarefaction but, number of individuals unknown. . . 0 • Related to Species–Area models, but area unknown. . . 40 • Number of species against number of sampling units Number of Species 5 0 10 20 30 40 50 60 Number of Traps & % SPECIES RICHNESS AND SAMPLE SIZE 83 ' $ Hubbell and Classical Models 200 • In large data sets similar to Fisher, in small similar to Arrhenius, says Hubbell 150 100 S 10 S S 20 100 50 150 100 θ = 30 50 50 5 θ = 10 2 θ=3 0 2000 4000 6000 J & 8000 10000 0 1 θ=1 0 5 1 10 100 J 1000 10000 1 10 100 1000 10000 J % 5 SPECIES RICHNESS AND SAMPLE SIZE ' 84 $ Local Communities & Dispersal Limitation • Basic model applies to the Metacommunity: Due to local extinctions and isolation (m < 1), Species–Area curve flatter • Hubbell claims (wrongly) that S–A curve asymptotically E(S) = J z , and immigration depends on the size of local community, so that a model for isolation would be m(J) = J −ω , where ω ≥ 0 is isolation • Modified Species Generator for Dispersal Limitation: J X θ · j −ω E(S|θ, ω, J) = θ+j−1 j=1 which cannot be solved in closed form & % SPECIES RICHNESS AND SAMPLE SIZE 85 ' $ Fitting Hubbell’s curves Myllylampi 50 40 30 θ = 105.7, ω = 0.26 20 • In principle, we can estimate isolation ω, but in practice too strongly correlated with θ θ = 18.6 10 • With Area instead of J, shape OK, but θ wrong 60 • Non-linear regression to find θ S 5 0 100 200 300 400 J & % SPECIES RICHNESS AND SAMPLE SIZE 86 ' $ All Models for Diatoms 250 • Hubbell bad, with θ = 33.8 200 • Hubbell with Dispersal Limitation good (θ = 7.8), but non-sensical ω = −0.20 • Log–Area excellent: 31.5 new species for doubling sample size 150 100 • Arrhenius poor, with z = 0.30 Hubbell ω Arrhenius log−A 50 S 5 0 10000 20000 30000 40000 N & % SPECIES RICHNESS AND SAMPLE SIZE 87 ' $ Diversity and Isolation • Diversity θ and Isolation ω correlated: Almost similar model with high θ and high ω, or with low θ and low ω. Point Estimate 95% Joint Confidence Region 0.5 1.0 1.5 + 0.0 + • The same problem with migration rate m and diversity θ earlier. & −1.0 −0.5 • Almost impossible to estimate θ and ω simultaneously: One should be known independently. Carabids omega 5 −100 −50 0 50 100 theta % 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 6 88 $ Hubbell’s Second Game: Metacommunity Landscape Hubbell has two different games. In the first game, we studied a single local community surrounded by a mythical Metacommunity. The definition of Metacommunity was left open, though. In the second game, Hubbell defines the Metacommunity as a network of local communities. & % 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 89 $ Building Communities from scratch Start with a Metacommunity made of a network of local communities, where species can migrate to their neighbour community 1. Kill D of J plants in each local community 2. Fill vacated slots with immigrant offspring from neighbour communities at probability m 3. Fill the remaining slots from within the local community using proportions of remaining species as probabilities 4. Mutate some of the new plants to a new species & % HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE 90 ' $ 1 10 100 1000 10000 In the beginning, Earth is gray Abundance 6 1.0 1.5 2.0 2.5 3.0 Rank & % HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE 91 ' $ Cells of Evolution in Isolated Landscape 1 5 10 50 100 500 5000 θ = 14, JM = 7056, m = 0.005, t = 100 Abundance 6 0 5 10 15 20 25 Rank & % HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE 92 ' $ Isolated Landscape 1 5 10 50 100 500 5000 θ = 14, JM = 7056, m = 0.005, t = 1000 Abundance 6 0 20 40 60 80 Rank & % HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE 93 ' $ Open Landscape 1 5 10 50 100 500 5000 θ = 14, JM = 7056, m = 0.5, t = 1000 Abundance 6 0 5 10 15 20 25 30 35 Rank & % List of Slides 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ! ' ! Slide 1 Hubbell’s Two Models Hubbell’s Two Models Hubbell’s Two Models 943 # $ # • Single Local Community in the Metacommunity Sea: θ defines • Single Local Community in the PMetacommunity Sea: θ defines Metacommunity composition i and it is filtrated into Local • Single Local Community in the Metacommunity Sea: θ defines Metacommunity Communities ascomposition m allows: Pi and it is filtrated into Local Metacommunity Pi and it is filtrated into Local Communities as mcomposition allows: Communities as m allows: m Ji Pi θ = 2JM ν m Slide 1 Pi of Local θ= 2JM ν • MetacommunityJas Communities: Migration i a Network • Metacommunity as arates Network of Local Communities: Migration and speciation ν create Local Communities: Communities,Migration and • m Metacommunity as a Network of Local m Metacommunity and speciation rates ν create Communities, and their sum: Local m and speciation is rates ν create Local Communities, and Metacommunity is their sum: Metacommunity is their sum: θ Jki Jki ν= ! ! m 2 θk i Jki Jki Jki ν= ! ! m 2 !k i Jki Jki Jki Pi = k Jki ! Jki Jki Pi = k Jki " $ " $ & % de 1 6 m HUBBELL’S SECOND GAME:JMETACOMMUNITY LANDSCAPE Pi θ = 2JM ν i ' • Metacommunity as a Network of Local Communities: Migration m and speciation rates ν Parameters create Local Communities, and Metacommunity is their sum: Jki " Jki m Jki Jki ν= 2 Pi = θ ! ! ! k k i 95 $ Jki Jki • k: Number of local communities (fixed) $ • Jk : Size of each local community (fixed) • D: Number of deaths in each local community (fixed) • m: Migration probability among neighbouring local communities (random) P • ν: Speciation probability ν = θ/(2JM ), JM = k Jk & % 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 96 $ Is World big enough for Hubbell? • Speciation rate ν can be found from θ and JM : θ ν= 2JM • In Hubbell’s metacommunity simulation θ = 10, JM = 7056 or ν = 10/2/7056 = 0.0007: one child of 1411 belongs to different species than her parent. • If JM is unknown, but we know the metacommunity area AM , we can get the density ρ needed for a realistic speciation rate, say ν = 10−10 • In Panama: θ = 50, AM = 4.3 · 105 ha: ρ = 2AθM ν = 2×4.3·10505 ×10−10 = 5.8 · 105 individuals per hectare (obvserved ca. 400 trees/ha). & % 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 97 $ Diversity and Metacommunity Landscape • Isolated communities (low m) – Local extinctions: Monodominance usual – Local communities unique: Rare species locally dominant – Local diversity low, regional high • Open communities (high m) – Local communities like miniature of Metacommunity – Local communities similar ot each other – Local diversity high, regional low & % HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE 98 ' $ • In multi-indivdiual community persistence makes sense: highest when m = 0 or m = 1. 80 60 40 20 • Counterintuitively immigration decreases persistence. P=0.9 P=0.5 P=0.01 0 • When J = 1, persistence to expected time to extinction is 1 TNi =1→Ni =0 = m(1−P i) 100 Persistence and Immigration Time to extinction T 6 0.0 0.2 0.4 0.6 0.8 1.0 Immigration rate m & % 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 99 $ Alpha and Beta Diversity • α–diversity is the local species richess (or Arrhenius c) • β–diversity is the change is species composition when exiting the local community (or Arrhenius z) • High migration m increases α but decreases β • Fundamental θ increases both α and β & % HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE 100 ' $ 0 y −3 −1 CCA2 1 1 0 −1 −3 −2 0 DCA1 & x 2 2 3 3 Ordination finds order in a Neutral Universe DCA2 6 2 4 −4 −2 0 2 CCA1 % 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 101 $ Putting things in even better order . . . Whether they want or not > anova(hubb.cca) Permutation test for CCA under reduced model Model: cca(X = hubb, Y = hubb.coord) Df Chisq F N.Perm Pr(>F) Model 2 0.6220 0.6845 100 < 0.01 *** Residual 46 20.8995 --Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 & % 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 102 $ Three-Phase Species–Area Curve 1. Local Scale. Local community abundance relations 2. Regional Scale: Speciation and dispersal rates 3. Continental Scale: Exceeds metacommunity limits & % 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 103 $ Metacommunity Diameter • Rate of increase in species number slows down with increasing area. • After crossing the metacommunity border, number of new species increases. • The limit is the correlation length: L intersection of tangents of species – area curve. • Varies among plant groups and areas, in trees of Panama d = 30 . . . 100 km & % 6 HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 104 $ Correlation Length & % HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE 105 ' $ Diameter of Lichen-rich Pine Forests 60 80 100 Lichen−rich pine forests + 40 Species richness 6 100 200 300 400 600 Circle diameter (km) & % HUBBELL’S SECOND GAME: METACOMMUNITY LANDSCAPE ' 106 $ Community Similarity and Distance • The lower is m, the faster species composition changes with distance. • The smaller the local communities, the faster the change in species composition. 0.2 0.3 0.4 0.5 0.7 Lichen−rich pine forests Jacccard index 6 + +++ ++ + + + + ++ + + +++ ++ +++ + + + ++ + +++ + + ++ ++ +++ ++ + +++ +++++ + ++++ + + ++ + + +++ ++ ++ +++++ +++++ + ++ + +++++++++ ++++ +++ +++ + +++ ++ ++ +++++ +++++++++++++ +++ + + + + + ++ ++ + + + + + + + + + + + + + + ++ +++++++++++ +++++ + ++ +++++++++++++ ++ ++ + +++++++++++ + ++ + + +++++++ + + + + ++++ + ++ + + + + ++++ ++++++ ++ ++++++ + ++++++++++ + + + + + + + + + + + + ++ + + + + + +++ + + + + + + + +++++++++ + + +++ +++++ + ++ ++ ++++++++++++ +++ ++++ + + + ++ ++ ++ + + + ++ +++++ + + + + ++++++ ++ + ++ + +++++++ + +++++ + + + + ++++ + + ++ ++++ + + + + + + + ++ + + + + + + + + + + + + + + + + + + + ++ + ++ + +++ ++++ + +++++++++ +++++ +++ + ++++++ +++++ +++ ++ + +++ + +++ +++ ++ + +++ + + + +++ + ++ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + ++ ++++ +++ + +++ ++++++ + +++ + + + + ++++ + + +++ ++ ++++ ++ ++ + + + ++ +++ ++++ +++++ ++ + + +++ + +++ ++ +++++ ++ ++ ++ ++ +++ ++++++++ +++++ + ++++ ++++ + + + + ++ + + + +++ + ++ + ++++++ + + + + + + + + + + + + + + ++ + + + + + + + + + + ++++ + + + + +++ +++ ++ + + ++ + + +++ ++ + ++++ +++ + + ++ ++++++ + + + + + + + ++ + + ++++ + + + +++ + + + +++ ++ + + ++++++ + + +++ + + ++ + ++ + + + + + + + + + ++ + +++ + ++ ++ + + + ++++ ++ + ++ + + + + + + + + + + + + ++ ++ + + + + + + ++ + + + + + + ++ + ++ + ++ + ++ + + + + + + ++ + + + + + +++ ++ + + 0 200 400 600 800 1000 Distance (km) & % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? ' 7 107 $ Species Pool: What is a Metacommunity? Metacommunity can be seen as the species pool feeding species to local communities. Neutral Model predicts a certain relationship between pool size and local richness — and that relation is similar to “niche limitation” model, although niches are not needed in Neutral Theory. . . & % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? ' 108 $ Species pool and local species richness • The local richness is determined by the available species pool 1. Actual species pool: The species really occurring in the community. 2. Local species pool: The species that can live in the community and occur in the landscape. 3. Regional species pool: The species that can live in the community and occur in the geographic region. • Habitat is a filter that selects species from the species pool. • There is no local control of richness, but the actual species richness follows from the pool size. & % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? 109 ' $ Filtering species from the pool Speciation migration Regional species Ecological filter Isolation Regional pool Local species Ecological filter Isolation Local pool Community & % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? ' 110 $ Null Model • Pa¨rtel – Zobel – Zobel – van der Maarel (PZZM) hypothesis: Community richness what ever between 0 and (local) species pool: Testing concerns the variance of species richness. • Pool limitation model: Constant proportion of species in the pool occur in the community: Tested by linear regression. • Niche-limitation model: Pool limitation at low richness, but niche limitation at high richness (“competition”), and this leads into saturating curve: Tested (approximately) by quadratic regression. & % SPECIES POOL: WHAT IS A METACOMMUNITY? 111 ' $ Alternative Null Models 0 50 100 150 200 Species Pool & 250 300 250 200 150 Species Richness 0 50 100 250 200 150 100 50 0 0 50 100 150 Species Richness 200 250 300 Niche Limitation 300 Pool Limitation 300 PZZM Species Richness 7 0 50 100 150 200 Species Pool 250 300 0 50 100 150 200 250 300 Species Pool % SPECIES POOL: WHAT IS A METACOMMUNITY? 112 ' $ Species richness in Hubbell communities • Theoretical form: Expected species richness in the Metacommunity of size JM against local community of size J & 40 30 20 • Indistinguishable from the “pool limitation”. JM = 64000, J = 64 Theoretical Fitted GAM Linear fit 10 • Species richness in a local community is dependent on the species richness in the Metacommunity: Species pool = Metacommunity. Local Richness 7 0 50 100 150 200 250 300 350 Pool Size % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? ' 113 $ Does species richness exist? • PZZM and pool limitation: Species richness is an incident and actually does not exist: There are as many species as there can live, but incidences vary. • Niche-limitation: Species richness is controlled and so strictly exists: • Removing a species gives space for immigration. • Adding a species causes local extinctions. & % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? ' 114 $ Which Species make up the Pool? • All species that could live in a site and could get there belong to the (regional) species pool. • The Estonian strategy: 1. Assess the “ecological character” of a site using Ellenberg indicator values for all species in the site 2. Find the species that are “close” to the assessed character of the site 3. Their number is the size of the species pool & % SPECIES POOL: WHAT IS A METACOMMUNITY? 115 ' $ Generating Communities from Gradients • Assume that gradients alone determine the species composition & 30 40 50 60 70 + + + + + + + + +++ + + +++ + + + + ++ + +++++ +++ + + + + + + ++ ++ + + + + + −− −− −− −− −− −− − − −− −− 20 • Interaction: Gradient information alone insufficient, but we must know which other species already are in the community. 10 • Adequate if gradient model is appropriate and there are no interactions between species. pH filter in Diatom species pool Species Richness 7 −− 4.5 −− 5.0 5.5 6.0 6.5 pH % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? ' 116 $ Biological Interactions. . . and Statistical • The variance of sum of independent random variates is the sum of their variances: Var(X + Y ) = Var(X) + Var(Y ) • If there are interactions, then Var(X + Y ) = Var(X) + Var(Y ) + 2Cov(X, Y ) • Covariance is a compund of correlation ρ and standard deviations σ: Cov(X, Y ) = ρX,Y σX σY • Positive interactions: Sum has larger variance. • Negative interactions (“competition”): Sum has lower variance. & % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? ' 117 $ Zobel’s quasineutral model • Filtering from regional pool to the actual (local) pool and filtering from actual pool to local community is a neutral and stochastic process. – Inter-specific competition typically asymmetric: Has a species ever replaced another of similar size? – Local community richness defined for number of shoots: Constant proportion from the pool. • Species are replaced from the community by light competition during succession: Gives a regular change in species richness ven per shoot. & % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? ' 118 $ Implications from the quasineutral model • Local communities are open and immigration is important • Succession is not a regular change in species compositon but change in growth form • Spatial scale is important. • Similar species co-exist and do not out-compete each other. • Species richness must be evaluated per shoot number, not per area. & % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? ' 119 $ Species richness and biomass: Neutral model • Low biomass because plants are small and sparse. • When the number of shoots increases, species richness grows lilnearly: N ∝ B • In closed vegetation, biomass can increase only if the number of shoots decreases because plant size incerases: Self thinning model predicts N ∝ B −2 • Species richness follows from shoot number, either from Fisher or Hubbell model. & % SPECIES POOL: WHAT IS A METACOMMUNITY? 120 ' $ Richness productivity hump 5 10 15 20 25 Kokemäenjoki meadows Species Richness 7 + ++ + + ++ + + + + ++ + + + ++ + + + + + + ++++ + ++ ++ +++++ +++ ++++ + ++ ++ +++++ + + + +++++++ + + + ++++++ +++ + +++ ++ +++++++++++++++++ + +++ + + +++++++ ++ +++ + ++++ + ++++++ ++++++ +++ + + +++ + +++++++ +++ ++ + ++++ + + ++ + ++ ++++++ + + +++ ++++++++ ++++++ ++ +++ +++ + + + + ++++++ +++ +++ + + ++ ++ ++ +++++++ + + + +++++ + + + ++ + ++ + ++ + + + 0 500 + + + + + + ++ 1000 ++ + 1500 Biomass g/m2 & % SPECIES POOL: WHAT IS A METACOMMUNITY? 121 ' $ Plant Size, Pool Size and Plot Richness • Grazing, e.g., can reduce plant size and/or change the spatial pattern. 2 3 4 5 15 20 25 30 1 10 • Plot level richness may increase although grazing introduces no new species to the site Given : grazing Species Richness 7 1 • Local species pool may remain unchanged. & 2 3 4 5 Year % 7 SPECIES POOL: WHAT IS A METACOMMUNITY? 122 ' $ Extrapolated Species Richness: Unseen Species • Jackknife: There are almost as many unseen species as there were species seen only once N −1 SP = SO + a1 N where SP is the pool size, SO the number of species seen, a1 number of species seen only once, and N the number of sites. • Bootstrap: Number of unseen species equals number of species not seen if sample resampled with replacement SP = SO + N X (1 − pi )N i=1 • Other ideas: Chao SP = S0 + & a1 2a2 , 2◦ Jackknife, . . . % SPECIES POOL: WHAT IS A METACOMMUNITY? 123 ' $ Unseen Species in Open Communities • Assumption:: There is a fixed number of all species — or a species pool & 260 240 220 200 180 • In open communities, all methods tend to add a constant number of “unseen” species to the observed species. Observed Chao Jackknife 1 Jackknife 2 Bootstrap 160 • Open communities: Accumulation of species never ends, but new rare species are found as sampling proceeds Barro Colorado Island Number of species 7 10 20 30 40 50 Number of plots % SPECIES POOL: WHAT IS A METACOMMUNITY? 124 ' $ Rarefaction: Biased Variance Barro Colorado Island Site 3 60 40 Rarefaction Its 95% CI Hubbell accumulator Its 95% CI • Number of unseen species: Unseen to variance, too. 0 20 • Current sample assumed errorfree: All error should come from the resampling 80 • Rarefaction: Sampling without replacement from the current sample Species 7 0 100 200 300 400 Trees & % SPECIES POOL: WHAT IS A METACOMMUNITY? 125 ' $ Species Accumulation Curves • Similar to rarefaction, but accumulates sites instead of individuals. & 200 + ++ ++++ + ++++ 100 150 + 0 • Variance biased. ++ + 50 • Again assumed that the complete collection of sites is errorfree. + + 0 • Randomized resampling without replacement or “exact” accumulator Barro Colorado Island exact 7 10 20 30 40 50 Sites %
© Copyright 2024