Invasive weeds can threaten native biodiversity and negatively impact on agriculture. Being able to track the spread of invasive weeds, and to identify at-risk areas in need of careful monitoring, are essential for managing incursions. This is especially true in the early stages of an invasion when it is possible to prevent widespread damage.
A group of researchers at QUT and CSIRO, including an ACEMS member, designed and implemented a new framework for rapidly modelling the invasion risk of weeds using Bayesian belief networks.
The method is based on a weed’s species ecology and has simple data requirements. The spatial risk of invasion for any weed can be calculated using spatial risk proxies derived from open GIS data about environmental conditions in an area of interest (e.g. rainfall, soil moisture, land use or road networks), a set of records for where the weed has been detected in the past, and an understanding of the relative importance of the factors affecting the spread of the weed. They tested their risk modelling method on field data collected for Hudson pear and Mexican bean tree, two emerging problem species in Queensland, Australia. The aim was to see if they could retrace an ecological invasion as it happened in the past. Results showed that locations they predicted as being highly susceptible to invasion were later preferentially invaded by these weeds.
The details of the new rapid spatial risk modelling method are accompanied by an open source implementation in a series of R Shiny apps called riskmapr. These apps allow researchers and biosecurity managers to easily plug in their own spatial data to generate risk maps. The riskmapr apps and source code are available at https://github.com/apear9/riskmapr or through Zenodo at http://doi.org/10.5281/zenodo.3347110, along with instructions for deploying them.
This research was published in Methods in Ecology and Evolution: https://besjournals.onlinelibrary.wiley.com/doi/full/10.1111/2041-210X.13284