A neutral sampling formula for multiple samples and an 'exact' test of neutrality

As the utility of the neutral theory of biodiversity is increasingly being recognized, there is also an increasing need for proper tools to evaluate the relative importance of neutral processes (dispersal limitation and stochasticity). One of the key features of neutral theory is its close link to data: sampling formulas, giving the probability of a data set conditional on a set of model parameters, have been developed for parameter estimation and model comparison. However, only single local samples can be handled with the currently available sampling formulas, whereas data are often available for many small spatially separated plots. Here, I present a sampling formula for multiple, spatially separated samples from the same metacommunity, which is a generalization of earlier sampling formulas. I also provide an algorithm to generate data sets with the model and I introduce a general test of neutrality that does not require an alternative model; this test compares the probability of the observed data (calculated using the new sampling formula) with the probability of model-generated data sets. I illustrate this with tree abundance data from three large Panamanian neotropical forest plots. When the test is performed with model parameters estimated from the three plots, the model cannot be rejected; however, when parameter estimates previously reported for BCI are used, the model is strongly rejected. This suggests that neutrality cannot explain the structure of the three Panamanian tree communities on the local (BCI) and regional (Panama Canal Zone) scale simultaneously. One should be aware, however, that aspects of the model other than neutrality may be responsible for its failure. I argue that the spatially implicit character of the model is a potential candidate.     

A new sampling formula for neutral biodiversity

The neutral model of biodiversity, proposed by Hubbell (The Unified Neutral Theory of Biodiversity and Biogeography, Princeton University Press, Princeton, NJ, 2001) to explain the diversity of functionally equivalent species, has been subject of hot debate in community ecology. Whereas Hubbell studied the model mostly by simulations, recently analytical treatments have yielded expressions of the expected number of species of a particular abundance in a local community with dispersal limitation. Moreover, a formula has been offered for the joint likelihood of observing a given species‐abundance dataset in a local community with dispersal limitation, but this formula is too complicated to allow practical applications. Here, I present a much simplified expression that can be regarded as an enhanced version of the famous Ewens sampling formula. It can be used in maximum likelihood methods for quick estimation of the model parameters, using all information in the data, and for model comparison. I also show how to rapidly generate examples of species‐abundance distributions for a given set of model parameters and how to calculate Simpson's diversity index.     

The zero-sum assumption in neutral biodiversity theory

The neutral theory of biodiversity as put forward by Hubbell in his 2001 monograph has received much criticism for its unrealistic simplifying assumptions. These are the assumptions of functional equivalence among different species (neutrality), the assumption of point mutation speciation, and the assumption that resources are continuously saturated, such that constant resource availability implies constant community size (zero-sum assumption). Here we focus on the zero-sum assumption. We present a general theory for calculating the probability of observing a particular species-abundance distribution (sampling formula) and show that zero-sum and non-zero-sum formulations of neutral theory have exactly the same sampling formula when the community is in equilibrium. Moreover, for the non-zero-sum community the sampling formula has this same form, even out of equilibrium. Therefore, the term "zero-sum multinomial (ZSM)" to describe species abundance patterns, as coined by Hubbell [2001. The Unified Neutral Theory of Biodiversity and Biogeography, Princeton University Press, Princeton, NJ], is not really appropriate, as it also applies to non-zero-sum communities. Instead we propose the term "dispersal-limited multinomial (DLM)", thus making explicit one of the most important contributions of neutral community theory, the emphasis on dispersal limitation as a dominant factor in determining species abundances.     

A dispersal-limited sampling theory for species and alleles

The importance of dispersal for biodiversity has long been recognized. However, it was never advertised as vigorously as Stephen Hubbell did in the context of his neutral community theory. After his book appeared in 2001, several scientists have sought and found analytical expressions for the effect of dispersal limitation on community composition, still in the neutral context. This has been done along two relatively independent lines of research that have a different mathematical approach and focus on different, yet related, types of results. Here, we study both types in a new framework that makes use of the sampling nature of the theory. We present sampling distributions that contain binomial or hypergeometric sampling on the one hand, and dispersal limitation on the other, and thus views dispersal limitation as ubiquitous as sampling effects. Further, we express the results of one line of research in terms of the other and vice versa, using the concept of subsamples. A consequence of our findings is that metacommunity size does not independently affect the outcome of neutral models in contrast to a previous assertion (Ecol. Lett., 7, 2004, p. 904) based on an incorrect formula (Phys. Rev. E, 68, 2003, p. 061902, eqns 11-14). Our framework provides the basis for development of a dispersal-limited non-neutral community theory and applies in population genetics as well, where alleles and mutation play the roles of species and speciation respectively.     

The structure of allelic diversity in the presence of purifying selection

In the absence of selection, the structure of equilibrium allelic diversity is described by the elegant sampling formula of Ewens. This formula has helped to shape our expectations of empirical patterns of molecular variation. Along with coalescent theory, it provides statistical techniques for rejecting the null model of neutrality. However, we still do not fully understand the statistics of the allelic diversity expected in the presence of natural selection. Earlier work has described the effects of strongly deleterious mutations linked to many neutral sites, and allelic variation in models where offspring fitness is unrelated to parental fitness, but it has proven difficult to understand allelic diversity in the presence of purifying selection at many linked sites. Here, we study the population genetics of infinitely many perfectly linked sites, some neutral and some deleterious. Our approach is based on studying the lineage structure within each class of individuals of similar fitness in the deleterious mutation-selection balance. Consistent with previous observations, we find that for moderate and weak selection pressures, the patterns of allelic diversity cannot be described by a neutral model for any choice of the effective population site. We compute precisely how purifying selection at many linked sites distorts the patterns of allelic diversity, by developing expressions for the likelihood of any configuration of allelic types in a sample analogous to the Ewens sampling formula.     

Effect of sampling strategy on estimation of fine-scale spatial genetic structure in Androsace tapete (Primulaceae), an alpine plant endemic to Qinghai-Tibetan Plateau

The fine-scale spatial genetic structure (SGS) of alpine plants is receiving increasing attention, from which seed and pollen dispersal can be inferred. However, estimation of SGS may depend strongly on the sampling strategy,including the sample size and spatial sampling scheme. Here, we examined the effects of sample size and three spatial schemes, simple-random, line-transect, and random-cluster sampling, on the estimation of SGS in Androsace tapete, an alpine cushion plant endemic to Qinghai-Tibetan Plateau. Using both real data and simulated data of dominant molecular markers, we show that: (i) SGS is highly sensitive to sample strategy especially when the sample size is small (e.g., below 100); (ii) the commonly used SGS parameter (the intercept of the autocorrelogram) is more susceptible to sample error than a newly developed Sp statistic; and (iii) the random-cluster scheme is susceptible to obvious bias in parameter estimation even when the sample size is relatively large (e.g., above 200). Overall,the line-transect scheme is recommendable, in that it performs slightly better than the simple-random scheme in parameter estimation and is more efficient to encompass broad spatial scales. The consistency between simulated data and real data implies that these findings might hold true in other alpine plants and more species should be examined in future work.     

Effect of sampling strategy on estimation of fine‐scale spatial genetic structure in Androsace tapete (Primulaceae), an alpine plant endemic to Qinghai‐Tibetan Plateau

Abstract The fine‐scale spatial genetic structure (SGS) of alpine plants is receiving increasing attention, from which seed and pollen dispersal can be inferred. However, estimation of SGS may depend strongly on the sampling strategy, including the sample size and spatial sampling scheme. Here, we examined the effects of sample size and three spatial schemes, simple‐random, line‐transect, and random‐cluster sampling, on the estimation of SGS in Androsace tapete, an alpine cushion plant endemic to Qinghai‐Tibetan Plateau. Using both real data and simulated data of dominant molecular markers, we show that: (i) SGS is highly sensitive to sample strategy especially when the sample size is small (e.g., below 100); (ii) the commonly used SGS parameter (the intercept of the autocorrelogram) is more susceptible to sample error than a newly developed Sp statistic; and (iii) the random‐cluster scheme is susceptible to obvious bias in parameter estimation even when the sample size is relatively large (e.g., above 200). Overall, the line‐transect scheme is recommendable, in that it performs slightly better than the simple‐random scheme in parameter estimation and is more efficient to encompass broad spatial scales. The consistency between simulated data and real data implies that these findings might hold true in other alpine plants and more species should be examined in future work.     

Age structure in neutral theory resolves inconsistencies related to reproductive-size threshold

Aims Neutral theory consists of a suite of models that assume ecological equivalence among individual organisms. They have been most commonly applied to tropical forest tree communities either as null models or as approximations. Neutral models typically only include reproductive adults; therefore, fitting to empirical tree community data requires defining a reproductive-size threshold, which for trees is usually set arbitrarily to a diameter at breast height (DBH) of 100 mm. The inevitable exclusion of some reproductive adults and inclusion of some saplings cause a non-random sampling bias in neutral model fits. Here, we investigate this problem and resolve it by introducing simple age structure into a neutral model.Methods We compared the performance and sensitivity of DBH threshold of three variants of a spatially explicit neutral model: the traditional model, a model incorporating random sampling and a model with two distinct age classes—reproductive adults and saplings. In the age-structured model, saplings are offspring from adults that disperse according to a Gaussian dispersal kernel around the adults. The only extra parameter is the ratio of adults to saplings, which is not a free parameter but directly measurable. We used species–area relationships (SARs) to explore the predicted effect of saplings on the species richness at different scales in our model. We then evaluated the three model variations to find the parameters required to maintain the observed level of species richness in the 50-ha plot on Barro Colorado Island (BCI). We repeated our analysis filtering the data at different minimum tree-size thresholds in order to find the effect this threshold has on our results. Lastly, we used empirical species–individual relationships (SIRs) to test the pre-existing hypothesis that environmental filtering is the primary cause of differences between the assemblage of saplings and that of adults on BCI.Important findings Our age-structured neutral model was characterized by SARs that were insensitive to the presence of saplings at large scales and highly sensitive to them at small scales. Both models without age structure were highly sensitive to the DBH threshold chosen in a way that could not be explained based on random samplings alone. The age-structured neutral model, which allowed for non-random sampling based on life stage, was consistent with species richness observations. Our analysis of empirical SIRs did not support environmental filtering as a dominant force, but it did show evidence for other differences between age classes. Age can now be easily incorporated into future studies of neutral models whenever there is a concern that a sample is not entirely composed of reproductive adult individuals. More generally, we suggest that modeling studies using tree data subject to a minimum size threshold should consider the sensitivity of their results to that threshold.     

An under-appreciated difficulty: sampling of plant populations for analysis using molecular markers

Results of studies using molecular markers for determining demographic and genetical population parameters especially in plants or sessile animals under field conditions are strongly dependent on the sampling strategy adopted. There are two critical decisions to make when determining this strategy: (i) what is the unit to be sampled?, (ii) how should units to be sampled in the field be selected? For the first decision, there are two conceptually different approaches: sampling ramets of clonal plants as units (to get information about within-genet parameters, such as genet sizes or numbers) and sampling genets of clonal or non-clonal plants as units (to get information of the genetic structure of the population). For the second decision, it is critically important to make the goal of the study explicit. We argue that in this case fully random sampling is needed only when an estimate of the true value of the population parameter is needed; if a comparison between populations is the goal, however, other sampling schemes may be adopted. The efficiency of different types of sampling strategies to recover relative values in a spatially extended population is studied by means of a spatially explicit simulation model. The results show that a regular pattern of sampling is best for obtaining information on genet sizes or inbreeding coefficients; in contrast, random or hierarchical sampling strategies are better for obtaining information on parameters that are based on comparison of pairs of individuals, such as distribution of genet sizes or autocorrelation in genetic structure. A set of recommendations is provided for designing a good sampling strategy.     

The distribution of rare alleles

Population geneticists have long been interested in the behavior of rare variants. The definition of a rare variant has been the subject of some debate, centered mainly on whether alleles with small relative frequency should be considered rare, or whether alleles with small numbers should be. We study the behavior of the counts of rare alleles in samples taken from a population genetics model that allows for selection and infinitely-many-alleles mutation structure. We show that in large samples the counts of rare alleles — those represented once, twice, ... — are approximately distributed as a Poisson process, with a parameter that depends on the total mutation rate, but not on the selection parameters. This result is applied to the problem of estimating the fraction of neutral mutations.     

A case-control follow-up study for disease-specific mortality

Case-control studies often rely on subjects to report their own screening or exposure information: this information is often obtained from cases after the event of interest has occurred. This is problematic for mortality outcomes, because dead subjects cannot report the desired information. To avoid this problem, Weiss and Lazovich (1996, American Journal of Epidemiology 143, 319-322) proposed obtaining exposure or screening information from potential cases, i.e., subjects diagnosed with disease, at the time of disease diagnosis, and also from a referent series. The design is best viewed as a new scheme for sampling from a cohort. I review estimation of the effects of time-varying screening or exposure in cohort studies, using a new factorization. I then show how this factorization, together with ignorability assumptions, allows valid estimation from these new designs. Even when the sampling fraction of nondiseased subjects is unknown, causal risk ratios are estimable if diagnosis is rare in the cohort. I illustrate and compare conventional and new methods with data from the Health Insurance Plan study.     

Likelihood and approximate likelihood analyses of genetic structure in a linear habitat: performance and robustness to model mis-specification

We evaluate the performance of maximum likelihood (ML) analysis of allele frequency data in a linear array of populations. The parameters are a mutation rate and either the dispersal rate in a stepping stone model or a dispersal rate and a scale parameter in a geometric dispersal model. An approximate procedure known as maximum product of approximate conditional (PAC) likelihood is found to perform as well as ML. Mis-specification biases may occur because the importance sampling algorithm is formally defined in term of mutation and migration rates scaled by the total size of the population, and this size may differ widely in the statistical model and in reality. As could be expected, ML generally performs well when the statistical model is correctly specified. Otherwise, mutation rate estimates are much closer to mutation probability scaled by number of demes in the statistical model than scaled by number of demes in reality when mutation probability is high and dispersal is most limited. This mis-specification bias actually has practical benefits. However, opposite results are found in opposite conditions. Migration rate estimates show roughly similar trends, but they may not always be easily interpreted as low-bias estimates of dispersal rate under any scaling. Estimation of the dispersal scale parameter is also affected by mis-specification of the number of demes, and the different biases compensate each other in such a way that good estimation of the so-called neighborhood size (or more precisely the product of population density and mean-squared parent-offspring dispersal distance) is achieved. Results congruent with these findings are found in an application to a damselfly data set.     

Landscape genetic inferences vary with sampling scenario for a pond‐breeding amphibian

A critical decision in landscape genetic studies is whether to use individuals or populations as the sampling unit. This decision affects the time and cost of sampling and may affect ecological inference. We analyzed 334 Columbia spotted frogs at 8 microsatellite loci across 40 sites in northern Idaho to determine how inferences from landscape genetic analyses would vary with sampling design. At all sites, we compared a proportion available sampling scheme (PASS), in which all samples were used, to resampled datasets of 2–11 individuals. Additionally, we compared a population sampling scheme (PSS) to an individual sampling scheme (ISS) at 18 sites with sufficient sample size. We applied an information theoretic approach with both restricted maximum likelihood and maximum likelihood estimation to evaluate competing landscape resistance hypotheses. We found that PSS supported low‐density forest when restricted maximum likelihood was used, but a combination model of most variables when maximum likelihood was used. We also saw variations when AIC was used compared to BIC. ISS supported this model as well as additional models when testing hypotheses of land cover types that create the greatest resistance to gene flow for Columbia spotted frogs. Increased sampling density and study extent, seen by comparing PSS to PASS, showed a change in model support. As number of individuals increased, model support converged at 7–9 individuals for ISS to PSS. ISS may be useful to increase study extent and sampling density, but may lack power to provide strong support for the correct model with microsatellite datasets. Our results highlight the importance of additional research on sampling design effects on landscape genetics inference.     

Spatial auto-correlation and auto-regressive models estimation from sample survey data

Maximum likelihood estimation of the model parameters for a spatial population based on data collected from a survey sample is usually straightforward when sampling and non-response are both non-informative, since the model can then usually be fitted using the available sample data, and no allowance is necessary for the fact that only a part of the population has been observed. Although for many regression models this naive strategy yields consistent estimates, this is not the case for some models, such as spatial auto-regressive models. In this paper, we show that for a broad class of such models, a maximum marginal likelihood approach that uses both sample and population data leads to more efficient estimates since it uses spatial information from sampled as well as non-sampled units. Extensive simulation experiments based on two well-known data sets are used to assess the impact of the spatial sampling design, the auto-correlation parameter and the sample size on the performance of this approach. When compared to some widely used methods that use only sample data, the results from these experiments show that the maximum marginal likelihood approach is much more precise.     

Reconciling neutral community models and environmental filtering: theory and an empirical test

It is widely believed that the neutral theory of biodiversity cannot be used for parameter inference if the assumption of neutrality is not met. The goal of this work is to extend this neutral framework to quantify the intensity of recruitment limitation (limited dispersal plus environmental filtering) in natural species assemblages. We model several local communities as part of a larger metacommunity, and we assume that neutrality holds in each local community, but not in the metacommunity. The immigration rate m does not only reflect dispersal limitation into a given local community, but also the intensity of environmental filtering. We develop a novel statistical method to infer the immigration parameter m in each local community. Using simulated datasets, we show that m indeed depends on both dispersal limitation and on the intensity of environmental filtering. We then apply this method to a network of tropical tree plots in central Panama. Inferred recruitment rates m were positively correlated with the fraction of trees dispersed by mammals, and with annual rainfall, possibly due to a weaker environmental filtering as rainfall increases. Finally, m, as estimated from trees greater than 1 cm trunk diameter, were significantly larger than an estimation based on trees greater than 10 cm trunk diameter. This suggests a cumulative effect of environmental filtering upon trees throughout their ontogeny.     

Accuracy of coalescent likelihood estimates: do we need more sites, more sequences, or more loci?

A computer simulation study has been made of the accuracy of estimates of Theta = 4Nemu from a sample from a single isolated population of finite size. The accuracies turn out to be well predicted by a formula developed by Fu and Li, who used optimistic assumptions. Their formulas are restated in terms of accuracy, defined here as the reciprocal of the squared coefficient of variation. This should be proportional to sample size when the entities sampled provide independent information. Using these formulas for accuracy, the sampling strategy for estimation of Theta can be investigated. Two models for cost have been used, a cost-per-base model and a cost-per-read model. The former would lead us to prefer to have a very large number of loci, each one base long. The latter, which is more realistic, causes us to prefer to have one read per locus and an optimum sample size which declines as costs of sampling organisms increase. For realistic values, the optimum sample size is 8 or fewer individuals. This is quite close to the results obtained by Pluzhnikov and Donnelly for a cost-per-base model, evaluating other estimators of Theta. It can be understood by considering that the resources spent collecting larger samples prevent us from considering more loci. An examination of the efficiency of Watterson's estimator of Theta was also made, and it was found to be reasonably efficient when the number of mutants per generation in the sequence in the whole population is less than 2.5.     

Linking dispersal, immigration and scale in the neutral theory of biodiversity

In the classic spatially implicit formulation of Hubbell's neutral theory of biodiversity a local community receives immigrants from a metacommunity operating on a relatively slow timescale, and dispersal into the local community is governed by an immigration parameter m . A current problem with neutral theory is that m lacks a clear biological interpretation. Here, we derive analytical expressions that relate the immigration parameter m to the geometry of the plot defining the local community and the parameters of a dispersal kernel. Our results facilitate more rigorous and extensive tests of the neutral theory: we conduct a test of neutral theory by comparing estimates of m derived from fits to empirical species abundance distributions to those derived from dispersal kernels and find acceptable correspondence; and we generate a new prediction of neutral theory by investigating how the shapes of species abundance distributions change theoretically as the spatial scale of observation changes. We also discuss how our main analytical results can be used to assess the error in the mean-field approximations associated with spatially implicit formulations of neutral theory.     

Building anisotropic sampling schemes for the estimation of anisotropic dispersal

Anisotropy, a structural property of dispersal, is observed in dispersal patterns occurring for a wide range of biological systems. While dispersal models more and more often incorporate anisotropy, the sampling schemes required to collect data for validation usually do not account for the anisotropy of dispersal data. Using a parametric model already published to describe the spatial spread of a plant disease, the wheat yellow rust, we carry out a study aimed at recommending an appropriate sampling scheme for anisotropic data. In a first step, we show with a simulation study that prior knowledge of dispersal anisotropy can be used to improve the sampling scheme. One of the main guidelines to be proposed is the orientation of the sampling grid around the main dispersal directions. In a second step, we propose a sequential sampling procedure (SSP) used to automatically build anisotropic sampling schemes adapted to the actual anisotropy of dispersal. The SSP is applied to simulated and real data. The proposed methodology is expected to be adapted easily to any kind of organisms with wind-borne propagule dispersal because it does not require the inclusion of biological features specific of the considered organism.     

LIKELIHOOD-BASED INFERENCE IN ISOLATION-BY-DISTANCE MODELS USING THE SPATIAL DISTRIBUTION OF LOW-FREQUENCY ALLELES

Estimating dispersal distances from population genetic data provides an important alternative to logistically taxing methods for directly observing dispersal. Although methods for estimating dispersal rates between a modest number of discrete demes are well developed, methods of inference applicable to "isolation-by-distance" models are much less established. Here, we present a method for estimating ρσ², the product of population density (ρ) and the variance of the dispersal displacement distribution (σ²). The method is based on the assumption that low-frequency alleles are identical by descent. Hence, the extent of geographic clustering of such alleles, relative to their frequency in the population, provides information about ρσ². We show that a novel likelihood-based method can infer this composite parameter with a modest bias in a lattice model of isolation-by-distance. For calculating the likelihood, we use an importance sampling approach to average over the unobserved intraallelic genealogies, where the intraallelic genealogies are modeled as a pure birth process. The approach also leads to a likelihood-ratio test of isotropy of dispersal, that is, whether dispersal distances on two axes are different. We test the performance of our methods using simulations of new mutations in a lattice model and illustrate its use with a dataset from Arabidopsis thaliana .     

Considering sampling bias in close‐kin mark–recapture abundance estimates of Atlantic salmon

Genetic methods for the estimation of population size can be powerful alternatives to conventional methods. Close‐kin mark–recapture (CKMR) is based on the principles of conventional mark–recapture, but instead of being physically marked, individuals are marked through their close kin. The aim of this study was to evaluate the potential of CKMR for the estimation of spawner abundance in Atlantic salmon and how age, sex, spatial, and temporal sampling bias may affect CKMR estimates. Spawner abundance in a wild population was estimated from genetic samples of adults returning in 2018 and of their potential offspring collected in 2019. Adult samples were obtained in two ways. First, adults were sampled and released alive in the breeding habitat during spawning surveys. Second, genetic samples were collected from out‐migrating smolts PIT‐tagged in 2017 and registered when returning as adults in 2018. CKMR estimates based on adult samples collected during spawning surveys were somewhat higher than conventional counts. Uncertainty was small (CV < 0.15), due to the detection of a high number of parent–offspring pairs. Sampling of adults was age‐ and size‐biased and correction for those biases resulted in moderate changes in the CKMR estimate. Juvenile dispersal was limited, but spatially balanced sampling of adults rendered CKMR estimates robust to spatially biased sampling of juveniles. CKMR estimates based on returning PIT‐tagged adults were approximately twice as high as estimates based on samples collected during spawning surveys. We suggest that estimates based on PIT‐tagged fish reflect the total abundance of adults entering the river, while estimates based on samples collected during spawning surveys reflect the abundance of adults present in the breeding habitat at the time of spawning. Our study showed that CKMR can be used to estimate spawner abundance in Atlantic salmon, with a moderate sampling effort, but a carefully designed sampling regime is required.