Inferring community assembly processes from macroscopic patterns using dynamic eco‐evolutionary models and Approximate Bayesian Computation (ABC)

Abstract

Statistical techniques exist for inferring community assembly processes from community patterns. Habitat filtering, competition, and biogeographical effects have, for example, been inferred from signals in phenotypic and phylogenetic data. The usefulness of current inference techniques is, however, debated as a mechanistic and causal link between process and pattern is often lacking, and evolutionary processes and trophic interactions are ignored. Here, we revisit the current knowledge on community assembly across scales and, in line with several reviews that have outlined challenges associated with current inference techniques, we identify a discrepancy between the current paradigm of eco‐evolutionary community assembly and current inference techniques that focus mainly on competition and habitat filtering. We argue that trait‐based dynamic eco‐evolutionary models in combination with recently developed model fitting and model evaluation techniques can provide avenues for more accurate, reliable, and inclusive inference. To exemplify, we implement a trait‐based, spatially explicit eco‐evolutionary model and discuss steps of model modification, fitting, and evaluation as an iterative approach enabling inference from diverse data sources. Through a case study on inference of prey and predator niche width in an eco‐evolutionary context, we demonstrate how inclusive and mechanistic approaches—eco‐evolutionary modelling and Approximate Bayesian Computation (ABC)—can enable inference of assembly processes that have been largely neglected by traditional techniques despite the ubiquity of such processes. Much literature points to the limitations of current inference techniques, but concrete solutions to such limitations are few. Many of the challenges associated with novel inference techniques are, however, already to some extent resolved in other fields and thus ready to be put into action in a more formal way for inferring processes of community assembly from signals in various data sources.

Dos and don'ts when inferring assembly rules from diversity patterns

“…Pigot and Etienne (2015) developed a dynamic null model of assembly that allows the estimation of the effect of allopatric speciation, colonization and local extinction. Ultimately, the idea is to build more mechanistic, dynamic models of community assembly (Connolly, Keith, Colwell, & Rahbek, 2017; Pontarp, Brännström, et al, 2019) that are general enough to include and contrast different ecological theories and processes and can be parameterized inversely with a selection of complementary diversity patterns (Cabral, Valente, & Hartig, 2017). The logic of this inverse parameterization, in simple terms, is to run the model across the relevant parameter space, to compare simulated patterns with observed patterns using appropriate summary statistics and to choose the parameter combinations that lead to the best match between simulated and observed patterns (Grimm et al, 2005; Hartig, Calabrese, Reineking, Wiegand, & Huth, 2011).…”

“…Here, we address this question by pinpointing the major pitfalls linked to the different steps of the standard filtering approach (Figure 1). Although many of the limitations of this framework have already been pointed out in previous reviews with various foci and levels of detail, and sometimes also in combination with possible solutions (e.g., Gerhold, Cahill, Winter, Bartish, & Prinzing, 2015; Lopez et al, 2016; Pontarp, Brännström, & Petchey, 2019), an overarching synthesis and a set of general guidelines for correctly applying the filtering framework is still lacking. Building on existing work, we provide a new comprehensive and structured overview of the different pitfalls and the solutions that have been developed (Table 1).…”

Identifying models of trait‐mediated community assembly using random forests and approximate Bayesian computation

“…These aspects may be more or less relevant depending on the taxonomic scale of the community being investigated (Weiher et al, 2011). Furthermore, the inference power could expand by making CAMI an individual-based model of community assembly (Pontarp et al, 2019;Rosindell, Harmon, & Etienne, 2015), where individuals can diverge to speciate and harbor intraspecific diversity among phenotypes (Jung et al, 2014;Jung, Violle, Mondy, Hoffmann, & Muller, 2010), all while abundance distributions and population demographics are being tracked (HilleRisLambers et al, 2012;Overcast, Emerson, & Hickerson, 2019). A spatially explicit model (see Pontarp et al, 2019) could allow for the exploration of how geography, or even local topography, impacts the assembly process.…”

“…Several approaches have implemented model‐based inference procedures for community assembly already (Munoz et al, ; Pontarp, Brännström, & Petchey, ; van der Plas et al, ), paving the way to measuring the relative impact of different processes on community assembly. However, we still lack a method that integrates both phylogenetic and phenotypic information in a species‐based model where the strength of the non‐neutral processes can be estimated.…”

A unified model of species abundance, genetic diversity, and functional diversity reveals the mechanisms structuring ecological communities

“…For example Cabral et al (2019) unify population-level and evolutionary timescales to investigate the dynamic relationship between community age, competition, and local richness. Likewise, Pontarp et al (2019a) develop a trait-based, spatially explicit eco-evolutionary model to make inferences about prey and predator niche width with potentially diverse data types. Incorporating temporal dynamics can help to distinguish among processes (Azaele et al 2006;Chisholm & O'Dwyer 2014;Jabot et al 2018;Kalyuzhny et al 2015;Nee 2005;Ricklefs 2006) , yet current theory fails to generalize across levels of biological organization.…”