Data from 'Wetland plant hydrological zonation predicted from observed water depth'

This repo included four files: two datasets (one used to build, the other to validate, a model of wetland plant zonation), the fitted model object ('topmod') of class(brmsfit), and, an R script to reproduce the analysis.

Each dataset consists of presence-absence of plants within quadrats, within wetlands, where plant species identity has been used to assign one of three hydrological response groups: T = terrestrial, A = amphibious, S = submerged. These are based on the water plant functional group classification of Casanova & Brock (2000, Plant Ecology V 147, pp 237-250). Accompanying the species / hydrological group composition for each quadrat is an observed depth measurement. All quadrats were observed in the local (Austral) Spring. The build data (n = 813 observations of 1 x 1 m quadrats across 12 wetland complexes in temperate Southeast Australia) were collected in October 2013; the validation data (n = 198, 23 wetland complexes) are completely independent of the model build data and were compiled from multiple monitoring programs over the period 2008-2012. Wetlands were from the same region, but were largely not sampled in the surveys used in the model build dataset.

The central premise of the study is that wetland hydrological niche space can be conceptualized as comprising three distinct zones, respectively dominated by fringing terrestrial, amphibious, and submerged species. This offers a framework to predict high-level impacts on wetland vegetation, characterizing the hydrological niche using the observed water depth in the Spring growing season and defining vegetation zones as the relative number of species adapted to terrestrial, amphibious, and submerged conditions (hydrological group richness, HGR).

A regression model was first developed using the build data to predict the hydrological conditions (ie depth) marking transition points between these zones. The model explained 0.72 of variation in HGR, predicting submerged-terrestrial (median ± [95% credible intervals] = 13 [7, 22] cm) and submerged-amphibious transitions (62 [50, 72] cm) directly and estimating amphibious-terrestrial transitions (predicted to occur below ground; -8 [-18, -2] cm) from extrapolation.

These thresholds were tested using the validation data and predictive performance decreased in the order submerged-terrestrial > amphibious-terrestrial > submerged-amphibious transitions. Only the latter proved unreliable (Cohen’s kappa = 0.73, 0.59 and 0.10 respectively). Results suggest it could be possible to identify critical wetland tipping points at the dry end of the hydrological niche using only simple summary statistics and show improved precision in predicting terrestrial-amphibious transitions (a critical tipping point in climatic drying risk) will require a focus shallow sub-surface (e.g., 0 to -20 cm) saturation dynamics.