From Temperate to Boreal -- Untangling the Importance of Environment, Space, and Species Co-distributions in Forest Plant Distributions

Gabrielle Bibeau Lapierre, Marilou Cournoyer, Guillaume Blanchet

Aim
Quantify how environmental conditions, spatial autocorrelation, and species co-distributions shape forest plant community at regional scale.

Location
Southern Québec, Canada, between the 45th and 53rd parallels.

Period
From 1986 to 2000.

Major taxa studied
Vascular plants, mosses, and lichens.

Methods
We analyzed the presence-absence of 224 plant species across 28,210 ecological observation points using a multivariate hierarchical model. Environment was characterized by bioclimatic, edaphic, and topographic variables, while regional scale spatial structure was defined by distance-based Moran's eigenvector maps. Sampling unit level latent variables were used to characterize residual species co-distributions. Model parameters were estimated using Bayesian inference through Markov chain Monte Carlo (MCMC) and were studied using ordinations (a principal component analysis calculated on the parameter matrix) for environment and spatial autocorrelation or a correlation matrix to quantify species co-distributions. We then quantified the relative contributions of environmental conditions, spatial structure, and species co-distributions to the community using variation partitioning.

Results
Environment, spatial autocorrelation, and species co-distributions all contributed substantially to plant distribution patterns although broad scale spatial autocorrelation contributed slightly more to structuring the community. Environmental responses varied considerably among species, with drainage emerging as an important differentiation component. Residual species co-distributions were predominantly positive (56.7%), although these patterns may reflect unmeasured environmental conditions or ecological processes involving species co-distributions. Using, ordinations we could also quantify uncertainty in the parameter estimated for the environment and spatial autocorrelation. Using the ordination results, we highlighted the species for which we had the most confidence in the results and notably found a gradient in confidence (or uncertainty) in the species spatial autocorrelation.

Main conclusions
Environmental conditions, spatial autocorrelation, and species co-distributions all contributed to a similar extent in explaining regional patterns of plant distribution, with regional-scale spatial structure having the largest overall contribution. This study highlights the importance of taking a multifactorial perspective to better understand how vascular plants, mosses, and lichens are structured on the landscape.