Developing and validating species distribution models for wetland plants
across Europe
Abstract
Wetlands and their unique biodiversity are threatened by drainage,
agricultural conversion, and climate change. Spatial species-level
biodiversity modelling help identifying effective conservation measures.
However, existing species-level models for wetland vegetation are often
geographically limited, miss variables for hydrological conditions and
neglect moss species, essential to many wetlands. Thus, we aimed to
develop and validate a species-level biodiversity model for wetland
vegetation across Europe. We fitted and cross-validated species
distribution models (SDMs) for 265 vascular plant and moss species
characteristic of European wetlands, using environmental variables
representing climate, soil, hydrology and anthropogenic pressures. We
validated the spatial predictions of the SDMs with independent dataset
from the Global Biodiversity Information Facility (GBIF) and the niche
optima of the species, as obtained from the modelled species response
curves, with empirical niche optima. The cross-validation and validation
with GBIF revealed good predictive power of the SDMs, especially for
diagnostic mosses. Median cross-validated values of AUC and TSS equalled
0.93 and 0.73, respectively, and median true positive rate (TPR)
equalled 0.77. SDMs of diagnostic vascular plants performed well too,
with median AUC, TSS and TPR of 0.91, 0.69 and 0.67, respectively. SDMs
of non-diagnostic plants had the lowest performance, with median AUC,
TSS and TPR values of 0.84, 0.54 and 0.62, respectively. Correlations
between modelled and empirical niche optima were typically in the
expected direction. Climate variables, in particular the mean
temperature of the coldest month, were the most important predictors of
species occurrence. In addition, groundwater table depth was an
important predictor for diagnostic vascular plants, but not for mosses.
We conclude that our SDMs are suitable to predict broad-scale patterns
of wetland plant species distributions as governed by climatic
conditions. Alternative or additional variables or a different modelling
approach might be needed to better represent the local heterogeneity in
hydrological conditions of wetlands