Reproducible WiSDM : a workflow for reproducible invasive alien species risk maps under climate change scenarios using standardized open data

Abstract

Introduction: Species distribution models (SDMs) are often used to produce risk maps to guide conservation management and decision-making with regard to invasive alien species (IAS). However, gathering and harmonizing the required species occurrence and other spatial data, as well as identifying and coding a robust modeling framework for reproducible SDMs, requires expertise in both ecological data science and statistics. Methods: We developed WiSDM, a semi-automated workflow to democratize the creation of open, reproducible, transparent, invasive alien species risk maps. To facilitate the production of IAS risk maps using WiSDM, we harmonized and openly published climate and land cover data to a 1 km2 resolution with coverage for Europe. Our workflow mitigates spatial sampling bias, identifies highly correlated predictors, creates ensemble models to predict risk, and quantifies spatial autocorrelation. In addition, we present a novel application for assessing the transferability of the model by quantifying and visualizing the confidence of its predictions. All modeling steps, parameters, evaluation statistics, and other outputs are also automatically generated and are saved in a R markdown notebook file. Results: Our workflow requires minimal input from the user to generate reproducible maps at 1 km2 resolution for standard Intergovernmental Panel on Climate Change (IPCC) greenhouse gas emission representative concentration pathway (RCP) scenarios. The confidence associated with the predicted risk for each 1km2 pixel is also mapped, enabling the intuitive visualization and understanding of how the confidence of the model varies across space and RCP scenarios. Discussion: Our workflow can readily be applied by end users with a basic knowledge of R, does not require expertise in species distribution modeling, and only requires an understanding of the ecological theory underlying species distributions. The risk maps generated by our repeatable workflow can be used to support IAS risk assessment and surveillance.