Hierarchical models for functional traits

Highlights

• There are increasing interests in measuring and modeling finescale variations in functional traits.
• With hierarchical Bayesian models, we showed in two cases that aggregation of traits might lead to similar inference with lower model complexity.
• We argue that hierarchical models should be used to explore the structure of data and the scale of process, in order to inform model complexity.

---

Motivation

Size traits vary with latitude (Joswig et al., 2021).

Climatic and soil variables explain variance in traits (Joswig et al., 2021).

Dispersion of species, populations and individuals in the trait space (Albert et al., 2010).

Model complexity can be decomposed into multiple components (Malmborg et al., 2024).

Case study 1: Trait-environment relation

Variations of specific leaf area between and within communities.

(Top) Individual-level model. (Middle) Community-level model. (Bottom) Hierarchical model.

Model complexity assessed following Albert et al., 2010.

Case study 2: Climate niche change over time

Changes in abundance of species with different temperature and temperature niches at grassland plots in Elkhorn Slough, California.

(Top) Community-level model. (Middle) Species-level model. (Bottom) Hierarchical model.

Model complexity assessed following Albert et al., 2010.