Details

Autor(en) / Beteiligte

• Klaus, Kristina V
• Matzke, Nicholas J

Titel

Statistical Comparison of Trait-Dependent Biogeographical Models Indicates That Podocarpaceae Dispersal Is Influenced by Both Seed Cone Traits and Geographical Distance

Ist Teil von

• Systematic biology, 2020-01, Vol.69 (1), p.61

Ort / Verlag

England

Erscheinungsjahr

2020

Link zum Volltext

Beschreibungen/Notizen

• The ability of lineages to disperse long distances over evolutionary timescales may be influenced by the gain or loss of traits adapted to enhance local, ecological dispersal. For example, some species in the southern conifer family Podocarpaceae have fleshy cones that encourage bird dispersal, but it is unknown how this trait has influenced the clade's historical biogeography, or its importance compared with other predictors of dispersal such as the geographic distance between regions. We answer these questions quantitatively by using a dated phylogeny of 197 species of southern conifers (Podocarpaceae and their sister family Araucariaceae) to statistically compare standard, trait-independent biogeography models with new BioGeoBEARS models where an evolving trait can influence dispersal probability, and trait history, biogeographical history, and model parameters are jointly inferred. We validate the method with simulation-inference experiments. Comparing all models, those that include trait-dependent dispersal accrue 87.5% of the corrected Akaike Information Criterion (AICc) model weight. Averaged across all models, lineages with nonfleshy cones had a dispersal probability multiplier of 0.49 compared with lineages with fleshy cones. Distance is included as a predictor of dispersal in all credible models (100% model weight). However, models with changing geography earned only 22.0% of the model weight, and models submerging New Caledonia/New Zealand earned only 0.01%. The importance of traits and distance suggests that long-distance dispersal over macroevolutionary timespans should not be thought of as a highly unpredictable chance event. Instead, long-distance dispersal can be modeled, allowing statistical model comparison to quantify support for different hypotheses.