| Summary: | Abstract Wetlands are among the most vulnerable ecosystems, stressed by habitat loss and degradation from expanding and intensifying agricultural and urban areas. Climate change will exacerbate the impacts of habitat loss by altering temperature and rainfall patterns. Wetlands within Australia's Great Barrier Reef (GBR) catchment are not different, stressed by extensive cropping, urban expansion, and alteration for grazing. Understanding how stressors affect wildlife is essential for the effective management of biodiversity values and minimizing unintended consequences when trading off the multiple values wetlands support. Impact assessment is difficult, often relying on an aggregation of ad hoc observations that are spatially biased toward easily accessible areas, rather than systematic and randomized surveys. Using a large aggregate database of ad hoc observations, this study aimed to examine the influence of urban proximity on machine‐learning models predicting taxonomic richness and assemblage turnover, relative to other habitat, landscape, and climate variables, for vertebrates dwelling in the wetlands of the GBR catchment. The distance from the nearest city was, by substantial margins, the most influential factor in predicting the richness and assemblage turnover of all vertebrate groups, except fish. Richness and assemblage turnover was predicted to be greatest nearest the main urban centers. The extent of various wetland habitats was highly influential in predicting the richness of all groups, while climate (predominately the rainfall in the wettest quarter) was highly influential in predicting assemblage turnover for all groups. Bias of survey records toward urban centers strongly influenced our ability to model wetland‐affiliated vertebrates and may obscure our understanding of how vertebrates respond to habitat loss and climate change. This reinforces the need for randomized and systematic surveys to supplement existing ad hoc surveys. We urge modelers in other jurisdictions to better portray the potential influence of survey biases when modeling species distributions.
|
|---|
