Ecology of spatially structured populations

Transcription

1 Ecology of spatially structured populations Jean-François Le Galliard CNRS iees Paris CNRS/Ens CEREEP/Ecotron IleDeFrance

2 Introduction

3 Habitat fragmentation and spatial structure Habitat fragmentation describes a state (or a process) of discontinuities (fragments) within the preferred living area (habitat) of a species. The classical paradigm of population ecology is that of a single, large and homogeneous population, but it is widely recognised that most populations are fragmented and heterogeneous implications for ecological processes? effects on population viability and extinction dynamics?

4 Habitat destruction and habitat fragmentation Habitat destruction is associated with massive habitat loss, fragmentation and habitat degradation ~ 83 % land surface affected by human activities Forest fragmentation (green area) in Finland from 1752 to 1990 Habitat destruction includes several processes: Reduction in the total area of the habitat Increase in number of habitat patches Decrease in habitat patches area Increase in isolation of habitat fragments Possibly, a decrease in habitat quality Fahrig. Ann. Rev. Ecol. Syst

5 Spatially-structured vs. unstructured populations Unstructured population one habitat patch often treated like a closed population (I=E=0) single habitats may have spatial heterogeneity and internal structure (e.g. age class) R t = B t + I t D t E t Spatially structured population several habitat patches dynamics of local habitat patches depend not only on B and D but also on E and I habitat patches may be heterogeneous (size, distance, and quality)

6 Examples of spatially structured populations Telfer et al Parentage assignment detects frequent and large-scale dispersal in water voles. Molecular Ecology 12:

7 Examples of spatially structured populations Saccheri I, Kuussaari M, Kankare M, Vikman P, Fortelius W, Hanski I, Inbreeding and extinction in a butterfly metapopulation. Nature 392:

8 Examples of spatially structured populations

9 Dispersal in animals and plants Dispersal is the process of going or distributing in different directions or over a wide area (Oxford English Dictionary) and is a critical animal behaviour and life history trait in spatially structured populations Dispersal is a behaviour involving key steps emigration transfer and (in animals) prospection immigration (also called settlement) Dispersal can occur at may different times during the life cycle natal dispersal in animals and seed-fruit dispersal in plants breeding dispersal in animals and vegetative dispersal in plants Dispersal can also occur at many different spatial scales but usually decreases with distance = distance-limited dispersal

10 Emi-immigration: two sides of the same coin Jordano P, What is long-distance dispersal? And a taxonomy of dispersal events. Journal of Ecology 105:75-84.

11 Examples of dispersal strategies

12 Dispersal distances in animals

13 Dispersal distances in plants Jordano P, What is long-distance dispersal? And a taxonomy of dispersal events. Journal of Ecology 105:75-84.

14 Habitat loss and degradation are leading causes of species extinction Habitat destruction is considered as one of the main cause of species loss on earth with overexploitation and species invasion according to the latest 2008 IUCN statistics of the Red List of Threatened species: 44,837 species assessed 16,928 species are threatened with extinction including 3,246 critically endangered >99% of threatened species are at risk from human activities: humans are the main cause of extinction and the principle threat to extinctions of species with low or declining population sizes Habitat loss and degradation are the leading threats: they affect 93% of all threatened birds, 45% of the threatened mammals assessed and 87% of the threatened amphibians, and most of the documented threatened plant species.

15 Habitat loss does not necessarily imply habitat fragmentation Expected effects of habitat fragmentation are 1. An increase in the number of patches 2. A decrease in mean patch size 3. An increase in mean patch isolation Fahrig L, Effects of habitat fragmentation on biodiversity. Annu Rev Ecol Syst 34:

16 Habitat fragmentation has some immediate effects on biodiversity loss and ecosystems Haddad et al Habitat fragmentation and its lasting impact on Earth's ecosystems. Science Advances 1.

17 Habitat fragmentation generates an extinction debt in grassland biodiversity Haddad et al Habitat fragmentation and its lasting impact on Earth's ecosystems. Science Advances 1.

18 Example of specific effects of habitat fragmentation from a large scale experiment Large-scale experimental habitat destruction experiment in Brazil (13 years, 23 patches) using 12 pristine forest patches versus 11 isolated patches ranging from 10 to 600 ha. Monitoring of the bird community and analysis with a statistical model of patch turnover in species presence/absence Extinction rate estimated according to the «best» statistical model

19 Habitat destruction and species decline: large scale experiment Positive effect of fragmentation on extinction rates, but results are highly variable and many species are insensitive to habitat fragmentation Negative effect of patch size on extinction rate Ferraz G, Nichols JD, Hines JE, Stouffer PC, Bierregaard Jr. RO, Lovejoy T, A large scale deforestation experiment: effects of patch area and isolation on Amazon birds. Science 315:

20 Short conclusions Most natural populations have a spatial structure meaning that local processes (local population growth) depends not only on birth and death rates but also on regional processes of migration Habitat loss implies habitat fragmentation but does not equate with habitat fragmentation defined by an increase in the number of patches, decrease in path area and increase in patch isolation Understanding of spatially structured populations requires good knowledge of dispersal behavior Habitat fragmentation can cause population decline and species losses but not all species are affected in the same way

21 Dynamical properties of spatially structured populations

22 Demographic properties of fragmented habitats Populations are characterized by a spatial structure in addition to any other form of life history structure (age, stages, etc) : existence of discrete, localised patches of preferred habitat separated by a matrix of non-preferred habitat patchy distribution homogeneous/heterogeneous patch area homogeneous/heterogeneous patch quality Whole population dynamics is partly determined by local demography, and therefore driven by dynamics of smaller populations : e.g., small patches are more likely to go extinct and more variable than large populations Whole population dynamics is also partly determined by landscape connectivity and permeability: patches are separated by a matrix of nonpreferred habitat putting limits on dispersal abilities

23 Example of spatial dynamics in wild species Landscape ecology of the Granville fritillary butterfly in Finland Hanski I, Metapopulation ecology. Oxford: Oxford University Press.

24 Example of spatial dynamics in wild species Local population dynamics often go wild showing strong stochastic components and potential regional differences Hanski I, Metapopulation ecology. Oxford: Oxford University Press.

25 Example of spatial dynamics in wild species Extinction are common and more likely in small habitats occupied by smaller populations Hanski I, Metapopulation ecology. Oxford: Oxford University Press.

26 Example of spatial dynamics in wild species Colonization depends on patch connectivity (a complex measure of patch isolation taking into account distance and dispersal behaviour) and on patch area such that isolated patches of small areas are less likely to be occupied Hanski I, Metapopulation dynamics. Nature 396:41-49.

27 The Levins model of habitat fragmentation c P Occupied (P) Empty (1-P) ε 1. Local populations are identical and have the same behaviour 2. Extinctions occur independently in different patches and local dynamics are asynchronous 3. Colonisation is independent of patch properties because there is a global dispersal pool spreading across the landscape 4. All patches are equally connected to all other patches Levins. Bull. Ent. Soc. Entom. USA

28 Forecasting occupancy probabilities 1. At equilibrium, the population is in balance between extinction and colonization 2. Colonization rate must be greater than extinction rate for the persistence of the population 3. Habitat fragmentation can cause a decrease of the colonization rate which imperils population persistence Levins. Bull. Ent. Soc. Entom. USA

29 Metapopulation dynamics: a class of model derived from Levins model of fragmentation In metapopulation ecology, landscapes are viewed as networks of idealized habitat patches in which occur as discrete local populations connected by migration Hanski I, Metapopulation dynamics. Nature 396:41-49.

30 Two connected patches with large population size and deterministic dynamics Local Ricker-logistic model Local Ricker-logistic model Homogeneous migration between the two habitat patches Hanski I, Metapopulation ecology. Oxford: Oxford University Press.

31 Migration tends to stabilize local population dynamics and metapopulation dynamics Rockwood. Introduction to population ecology. 2006

32 Rescue effect in metapopulations Immigration can reduce the probability of local extinction by boosting numbers of inhabitants under adverse conditions, thus buffering local populations against demographic and environmental stochasticity. This reduction in local extinction probability was termed the rescue effect by Brown & Kodric- Brown (1977). The rescue effect implies a positive feedback from metapopulation occupancy to local population persistence and may also involve local Allee effects (e.g., intrinsic growth rate increases locally with the number of colonizers due to mate limitation). Brown JH, Kodric-Brown A, Turnover rates in insular biogeography: effect of immigration on extinction. Ecology 58:

33 Rescue effect in a butterfly metapopulation Survival of local populations in a number of large metapopulations of this butterfly species increases with habitat quality (climate, ground and plant cover), patch area but also direct connectivity with occupied patches (rescue effect) Brown JH, Kodric-Brown A, Turnover rates in insular biogeography: effect of immigration on extinction. Ecology 58:

34 Infinite number of connected patches with density-dependent local dynamics Disp. pool K Infinite number of discrete patches of size [0,K] individuals (demographic stochasticity) Local stochastic birth and death processes (density-dependence included) Local catastrophes (environmental stochasticity) Global dispersal towards a dispersal pool and partial settlement assuming a given cost of dispersal (lower survival) Casagrandi R, Gatto M, A mesoscale approach to extinction risk in fragmented habitats. Nature 400:

35 Infinite number of connected patches with density-dependent local dynamics Casagrandi R, Gatto M, A mesoscale approach to extinction risk in fragmented habitats. Nature 400:

36 Metapopulations of heterogeneous patches: the source-sink metapopulation model Metapopulations are often collection of heterogeneous patches with differences in habitat quality that influence birth and death rates. Local populations can be ranked by their habitat qualities and therefore their intrinsic growth rates. Ignoring density-dependence, populations can be classified as naturally growing patches (r>0) or naturally declining patches (r<0). For metapopulations with many different patches, Pulliam (1988) coined the terms sources and sinks to distinguish these two kinds of habitat patches. Since local growth depends also on emigration and immigration, it is possible that sink patches persist because they are supplemented with immigrants from sources patches. This is the basis of the source-sink metapopulation model.

37 Source-sink metapopulation model with two habitat patches and density-dependence Source habitat (λ 1 >1) Sink habitat (λ 2 <1) Saturated at N= n Pulliam HR, Sources, sinks, and population regulation. Am Nat 132:

38 Example of source-sink metapopulation Manier MK, Arnold SJ, Population genetic analysis identifies source sink dynamics for two sympatric garter snake species (Thamnophis elegans and Thamnophis sirtalis). Molecular Ecology 14:

39 Towards spatially explicit approaches: dispersal distances and incidence function models Levins model Incidence function models 1. Patches are identical and are not geolocated (no spatial dimension) 2. Dispersal is homogeneous and unconditional 3. Populations are occupied or empty Dynamics are dominated by extinction and colonization processes involving all patches. This model is spatially implicit. 1. Patches differ in area/quality which influence local extinction and colonization probability and are geolocated 2. Dispersal is distance-limited and costly (mortality costs) 3. Populations are occupied or empty but the model can be extended to track local densities Dynamics are dominated by extinction and colonization processes involved all patches. This model is more spatially realistic.

40 Example of incidence function model: butterflies in fragmented landscapes Negative exponential distribution of dispersal distances Number Parameter β: number of migrants Parameter α: dispersal capacity of migrants Rockwood. Introduction to population ecology Hanski I, Kuussaari M, Niemenen M, Metapopulation structure and migration in the butterfly Melitaea cinxia. Ecology 75:

41 Butterflies in fragmented landscapes: local dynamics of individual patches C i Probability of colonization per unit time: Ci E i Probability of extinction per unit time without rescue effect (1-C i )E i Probability of extinction per unit time with rescue effect Long-term probability of patch occupancy or incidence

42 Determinants of local patch dynamics Extinction probability depends on a constant parameter, patch size/area and a scaling factor Colonization probability depends on distance to all occupied patches in the metapopulation and includes a competition among immigrants such that colonization plateaus at high densities Incidence function

43 Empirical study: shrub patches in a small lizard population Hokit DG, Stith BM, Branch LC, Comparison of Two Types of Metapopulation Models in Real and Artificial Landscapes. Conservation Biology 15:

44 Empirical study: habitat patches in American Pika Moilanen A, Hanski I, Smith AT, Long-term dynamics in a metapopulation of the American Pika. Am Nat 152:

45 Parameter estimations from field data Parameter Interpretation Estimation Estimate b A 0 α x Allometric scaling of population size with patch perimeter Minimum patch area where E=1 Effect of distance on migration success Allometric scaling of extinction with patch area Regression with population size 0.74 Smallest patch occupied by pikas Mark-recapture study of 11 pikas fitted with distribution 15 m 5.28 Observed colonization events 2.34 Incidence function fitting 1.28 e Basal extinction rate Incidence function fitting 0.01 y Competition among colonizers Incidence function fitting 1.5 Moilanen A, Hanski I, Smith AT, Long-term dynamics in a metapopulation of the American Pika. Am Nat 152:

46 Simulation of the different parts of the network Moilanen A, Hanski I, Smith AT, Long-term dynamics in a metapopulation of the American Pika. Am Nat 152:

47 Metapopulations with different migration potential (due to connectivity) Low colonization potential = very low metapopulation occupancy (extinction) due to negative metapopulation growth Positive effect of migration through rescue effect and recolonozation of extinct populations = high to very high occupancy Bistability = existence of two alternative metapopulation equilibria at intermediate colonization potentials Predicted (theory) Observed (66 networks) Predicted (empirical model) Hanski I, Metapopulation dynamics. Nature 396:41-49.

48 Alternative to incidence function models: spatially explicit structured population models Metapopulation of 20 habitat patches monitored by CRM techniques during 10 years. Design of a spatially explicit structured population model based on field data assuming (1) local density-dependence, (2) dispersal between sites and (3) spatial correlation of local dynamics Schtickzelle N, Baguette M, Metapopulation viability analysis of the bog fritillary butterfly using RAMAS/GIS. Oikos 104:

49 Model assumptions (field parameterization) Local density-dependence based on a logistic growth model, carrying capacity increases with patch size, with large (significant) effects of climate conditions on population growth modeled by a stochastic component Significant spatial autocorrelation of population growth rates fitted by a negative exponential function (most correlations occur at scales below 1000 m) Virtual model for dispersal assuming a decrease of dispersal with distance and some fat-tail dispersal kernel Schtickzelle N, Baguette M, Metapopulation viability analysis of the bog fritillary butterfly using RAMAS/GIS. Oikos 104:

50 Local population growth Substantial range of variation caused by fluctuating climate conditions Characteristic logistic Ricker decline Schtickzelle N, Baguette M, Metapopulation viability analysis of the bog fritillary butterfly using RAMAS/GIS. Oikos 104:

51 The structured population model predicts well occupancy and extinctions

52 Calculation of quasi-extinction risks: PVA Schtickzelle N, Baguette M, Metapopulation viability analysis of the bog fritillary butterfly using RAMAS/GIS. Oikos 104:

53 Short conclusions Spatially structured populations behave differently than the average local population because regional processes of migration can synchronize and/or stabilize local population dynamics and rescue local populations from high extinction risks Classical models of spatially structured populations envision a system dominated by small, discrete patch entities connected by more or less random dispersal. Such classical metapopulations can be described by models where there is a balance between migration and extinction. These models predict in general that low connectivity is a threat to whole metapopulation viability

54 Alternative models and some complications

55 Not all spatially structured populations are classical metapopulations with extinction-colonization Thomas CD, Kunin WE, The spatial structure of populations. J Anim Ecol 68:

56 Example of mainland-island metapopulation Harrison S, Murphy DD, Ehrlich PR, Distribution of the bay checkerspot butterfly, Euphydrias editha bayensis: evidence for a metapopulation model. Am Nat 132:

57 Local populations can be ranked in different categories using the BIDE equation components Mobility axis Local growth axis Mobility axis compensates local growth axis (B+I-D-E=0)

58 Compensation and mobility axis allow to rank metapopulations in different categories

59 Case study Large number of small patches with most populations slightly positively compensated but several small populations acting like dispersal sinks of larger, nearby patches. More patches are net exporters than importers due to strong dispersal costs. Large number of populations with most populations well compensated but dominance of larger populations acting like net exporters of migrants. More patches are net exporters than importers due to strong dispersal costs.

60 Ignorance of landscape features may lead to severe biases in model predictions Classical metapopulation theory assume idealized landscapes with discrete patches of suitable habitats connected by aerial/terrestrial dispersal through unsuitable habitats and/or discrete corridors. Most landscapes are made out of complex matrices of more or less suitable breeding habitats connected by more or less suitable dispersal habitats. Spatially explicit landscape models emphasize the importance of accounting for spatially realistic combinations of breeding and dispersal habitats and usually are based on individual based-simulations of empirical data.

61 Example: landscape permeability to dispersal Land cover maps are translated into maps of least-cost dispersal using empirical estimates of land type s resistance to dispersal. Least-costs dispersal maps can then be translated intro connexion maps allowing to separate habitat networks and identify spatial patterns of connectivity across all spatial scales This usually requires good movement data from the field Stevenson-Holt CD, Watts K, Bellamy CC, Nevin OT, Ramsey AD, PLoS ONE 9:e

62 Behavioral rules of dispersal are often complex leading to non-random habitat selection Classical metapopulation theory assume random dispersal or distancelimited dispersal but most behavioural studies demonstrate that dispersal is not random because it depends on local environmental and social conditions it depends on individual attributes How individuals disperse in complex landscapes depends on habitat selection rules and individual attributes, such that for example high quality habitats may not be preferred, isolated patches may not be colonized, individual quality of dispersers may facilitate long-distance dispersal and colonization more than expected at random... A good example of how complex dispersal rules may change metapopulation dynamics involves density-dependent dispersal

63 Density-dependent dispersal and rescue effect Density-independent dispersal = causes some rescue at low population density but tends to synchronize local population dynamics (spatial autocorrelation, also called Moran s effect) Negative density-dependent dispersal = precipitates population extinction (can be due to conspecifics attraction) but tends to limit spatial synchronization Positive density-dependent dispersal = increases the rescue effect at low population density (dispersal through colonization) but tends to increase spatial synchronization

64 Behavioral plasticity in dispersal Field experiments with root voles (Microtus oeconomus) in Norway = fence effects with less local dispersal at high population densities Ims RA, Andreassen HP, Density-dependent dispersal and spatial population dynamics. Proc R Soc B-Biol Sci 272:

65 Short conclusions The classical metapopulation ecology is only one (simple and very efficient) approach to model complex spatially structured populations. It requires a balance between local extinction and colonization which may not hold when patches are large, some patches act like a mainland or connectivity is too low. Landscapes are more complex than assumed by classical metapopulation ecology and landscape features may be important to take into account to predict dispersal and therefore colonization. Habitat selection can lead to non-random dispersal with various consequences for metapopulation dynamics

66 Evolutionary consequences of habitat fragmentation and evolutionary rescue

67 Levels of selection in fragmented populations Examples of antagonistic selective pressures Selection within demes (intrademic selection) social interactions kinship structures Selection between demes (interdemic selection) dispersal and colonisation migration and founder effects «Metapopulation effect» Olivieri and Gouyon Cooperation = selected for between demes but counterselected within each deme Dispersal in plants = counterselected within the deme but selected between demes Virulence = selected for within the deme but can be selected against between demes Habitat fragmentation causes selection due Genetic heterogeneity : inbreeding and kinship structure. Demographic heterogeneity : e.g. density-dependence. Environmental heterogeneity : e.g. habitat quality.

68 Dispersal evolution: kin selection Basic assumptions homogeneity in deme sizes homogeneity in deme structures kin selection due to genetic heterogeneity Interactions with Philopatry Dispersal Relatives Many Few Conspecifics Some Some Kin competition Dispersal Hamilton & May Nature Kin cooperation Philopatry Perrin & Goudet. Oxf Univ Press 2001

69 Dispersal evolution: demographic heterogeneity Basic assumptions no kinship structure variance in patch occupancy due to local extinction selection due to demographic heterogeneity (avoidance of competition) Model of successional dynamics and plant dispersal More colonization opportunities Fast succession Less local competition Slow succession Ronce et al. Am Nat

70 Dispersal evolution: environmental heterogeneity Basic assumptions : habitat heterogeneity selection due to environmental heterogeneity two traits : dispersal and local adaptation traits Habitat variation alone two habitats - no kin selection local maladaptation = cost of dispersal = loss of migration local adaptation = benefits of specialization = evolution of specialist strategies with two non-dispersive specialist strategies inside each habitat Habitat + temporal variation - no kin selection temporal variation = risk spreading benefits = evolution of partial migration co-evolution of local adaptation can lead to various patterns of existence and coexistence between the two non-dispersive specialists and a generalist dispersive strategy Kisdi. Am Nat

71 Evolution of plant dispersal on islands «Mainland» «Island» Comparative analysis of dispersal abilities for two plant species based on morphological measurements The loss of migration abilities is a common evolutionary syndrome of island species / populations Cody and Overton. J. Ecol. 1996

72 Evolution of flight behaviour in butterflies «Woodland» butterflies «Agricultural» butterflies Raised in a common garden and investigated for their flight behaviour in the laboratory Pararge aegeria Observed differences between the fragmented and non-fragmented landscapes: females from woodland habitats travel longer distances per unit time females from woodland habitats cross more often boundary females from woodland habitats more often seen at flight females from woodland habitats traverse more often between their preferred habitats males from woodland do not differ from male from agricultural landscapes Conclusion Observed differences restricted to females Counter-selection of dispersal behaviour in females from agricultural landscapes Merckx et al. Proc. Roy. Soc. London 2003

73 Dispersal behaviour and landscape in spiders Isolated Connected Continuous Raised in a common garden and investigated for the «tip-toe» behaviour in the laboratory Bonte et al. Anim. Behav Passive dispersal seems to be selected against in more fragmented habitats! This could be explained by dominant effects of the cost of dispersal or some form of habitat specialization

74 Dispersal and habitat specialization across species Dispersive species are habitat generalists dispersal may be counterselected in isolated landscapes due to habitat specialization Index of habitat specialization based on local recordings and literature review in Europe Bonte et al. Proc. Roy. Soc. Lond. 2003

75 Short conclusions Evolution of life history and dispersal strategies involves a metapopulation effect where selection operates at two distinct levels: local within-population selection and regional betweenpopulation selection. Habitat fragmentation thus not only changes ecological dynamics but can modify evolutionary dynamics. Rapid evolution of dispersal and life history strategies in more fragmented habitats has been observed in a few study systems suggesting that evolutionary dynamics could interact with ecological dynamics