Biodiversity Information Science and Standards : Conference Abstract
Print
Conference Abstract
Using Global Biodiversity Information Facility Occurrence Data for Automated Invasive Alien Species Risk Mapping
expand article infoAmy J.S. Davis‡, Tim Adriaens§, Rozemien De Troch|, Peter Desmet§, Quentin Groom, Damiano Oldoni§, Lien Reyserhove§, Sonia Vanderhoeven#, Diederik Strubbe‡
‡ Ghent University, Ghent, Belgium
§ Research Institute for Nature and Forest (INBO), Brussels, Belgium
| Royal Meteorological Institute, Brussels, Belgium
¶ Meise Botanic Garden, Meise, Belgium
# Belgian Biodiversity Platform, Brussels, Belgium
Corresponding author: Amy J.S. Davis (amyjsdavis2@gmail.com)
Received: 30 Sep 2020 | Published: 01 Oct 2020
© 2020 Amy Davis, Tim Adriaens, Rozemien De Troch, Peter Desmet, Quentin Groom, Damiano Oldoni, Lien Reyserhove, Sonia Vanderhoeven, Diederik Strubbe
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation: Davis AJ.S, Adriaens T, De Troch R, Desmet P, Groom Q, Oldoni D, Reyserhove L, Vanderhoeven S, Strubbe D (2020) Using Global Biodiversity Information Facility Occurrence Data for Automated Invasive Alien Species Risk Mapping . Biodiversity Information Science and Standards 4: e59172. https://doi.org/10.3897/biss.4.59172
Open Access

Abstract

To support invasive alien species risk assessments, the Tracking Invasive Alien Species (TrIAS) project has developed an automated, open, workflow incorporating state-of-the-art species distribution modelling practices to create risk maps using the open source language R. It is based on Global Biodiversity Information Facility (GBIF) data and openly published environmental data layers characterizing climate and land cover. Our workflow requires only a species name and generates an ensemble of machine-learning algorithms (Random Forest, Boosted Regression Trees, K-Nearest Neighbors and AdaBoost) stacked together as a meta-model to produce the final risk map at 1 km2 resolution (Fig. 1). Risk maps are generated automatically for standard Intergovernmental Panel on Climate Change (IPCC) greenhouse gas emission scenarios and are accompanied by maps illustrating the confidence of each individual prediction across space, thus enabling the intuitive visualization and understanding of how the confidence of the model varies across space and scenario (Fig. 2). The effects of sampling bias are accounted for by providing options to:

Figure 1.

Risk map for presence of Cyperus eragrostis in Belgium generated by the TrIAS modeling workflow. The map is scaled from 0-1, with 0 indicating the lowest risk and 1 indicating the highest risk.

Figure 2.

Confidence of the predicted risk for Cyperus eragrostis generated by the TrIAS modeling workflow. Confidence is scaled from 0 (lowest confidence) to 1 (highest confidence).

  1. use the sampling effort of the higher taxon the modelled species belongs to (e.g., vascular plants), and
  2. to thin species occurrences.

The risk maps generated by our workflow are defensible and repeatable and provide forecasts of alien species distributions under further climate change scenarios. They can be used to support risk assessments and guide surveillance efforts on alien species in Europe. The detailied modeling framework and code are available on GitHub: https://github.com/trias-project.

Keywords

species distribution models, automated workflow, risk assessment, machine learning, sampling bias

Presenting author

Amy J.S. Davis

Presented at

TDWG 2020

Grant title

Tracking Invasive Alien Species (TrIAS)

Hosting institution

Ghent University, Belgium

login to comment