The importance of biological plausibility for data poor models in the face of an immediate threat by an emerging infectious disease: a reply to Katz and Zellmer (2018)

Abstract

Species distribution models (SDM) are an important tool to predict the invasion risk of alien species and emerging infectious diseases. However, building reliable models in early stages of invasions is a challenging task. Katz and Zellmer (Biol Invasions 20:2107–2119, 2018) addressed this problem and presented a framework for model selection for data-poor newly invasive species. Based on data of the recently discovered and invasive amphibian chytrid fungus Batrachochytrium salamandrivorans (Bsal), they built SDM and gave concluding implications for model selection in general and for Bsal in its invasive range in particular. In our opinion the authors’ SDM and their final processing show severe flaws and lead to a biologically implausible (although statistically best) model, which they used to promote their framework. We here intend to remind readers of the importance of biological relevance and plausibility of SDM, especially its input data, by highlighting some deficiencies and their implications for the biological relevance of the predictions. We further emphasize considerations and recommendations to improve the biological relevance and to make model evaluation more comprehensible. Though biological relevance is a basic SDM rationale, it is of particular importance when limited data impede a ‘proper’ model evaluation. If a prediction is evaluated as good or best by some statistical measure, but is obviously implausible in a biological sense, we should not ignore the biology of our model species. Especially in the face of an imminent threat such as the spreading Bsal, when immediate conservation or management actions are needed, wrong implications based on biologically implausible models may have severe counterproductive consequences.