Digital Data in Biodiversity Research Conference

Twenty-first century ecologists have an extensive suite of analyses and data types available to answer complex questions about endangered species decline, potential consequences of climate change across ecosystems, and the progression of biological invasions. Decisions in data selection and processing could create significant downstream consequences; however, many of these decisions are hidden in modeling algorithms or not reported in literature, potentially hindering reproducibility and biological interpretations. Species distribution models (SDMs), which bridge multiple data types to model relationships between species and their environments, provide an interesting example of the interaction of data complexity with model results. We selected several points throughout the SDM process to investigate their influence on model performance, outcomes, and interpretation using invasive European frog-bit (Hydrocharis morsus-ranae L.) in North America as a case study. Concerns about the impacts of European frog-bit on native ecosystems catalyzed targeted monitoring efforts across the invasive range, which have produced heterogenous occurrence data that are well-suited to the questions addressed. Our data-centric analysis pipeline integrated data discovery throughout the modeling process to identify and account for the effects of input data features on model outcomes, which ultimately impacted our interpretations relative to European frog-bit biology and ecology.