Modelling weed distribution under climate change

Choosing weed species to model

More than 25,000 non-native plant species have been introduced to Australia since European settlement. At least 3,000 of these are now known to be significant weed species. About 1,600 are known to have established in New South Wales and a recent assessment indicates that about 350 species pose a significant threat to native species.

Given the number of weed species present, key weed species were selected based on whether they were of National (eg Weeds of National Significance (WONS)) and/or State importance or if they were National Alert Listed species. Inclusion of other key weed species is also occurring where there is sufficient information available to model.

Why use models

Modelling of native and introduced species distribution is a widely-used method for predicting the occurrence of species in locations where surveys have not been made or where information is scant. More recently such modelling processes have also been used to make predictions on the distribution of species under future climatic conditions.

Modelling tools

A large array of modelling tools is available and this project examined several of them before selecting MaxEnt. This technique has been shown in peer-reviewed studies to be very reliable in its spatial predictions using current climate data, even when the number of spatial records is small, or where there is uncertainty in the spatial coordinates of historical data.

Data required to run the models

There are two types of data required:

• Weed species distribution data: This is obtained primarily from records contained in the major publicly-accessible data repositories such as Australia's Virtual Herbarium (AVH) and the Global Biodiversity Information Facility (GBIF). Additional data is also extracted from scientific publications, reports and other documents, as well as from unpublished information, localised databases and expert knowledge. All data is subjected to rigorous screening procedures to ensure that only records with reliable species identification and accurate geographical information are included.
• Climate data: Predicting the likely weed distribution in un-surveyed areas is achieved by using climate data for both the present-day climate and modelled future climate. Data is extracted from the WorldClim database which can be used to facilitate climate modelling in any part of the world at a high resolution.

Interim results

Several key weed species have now been modelled. The weed species were chosen to illustrate the predictions across a range of growth forms (eg grass, shrubs and trees).

Note: these are preliminary results only. They will be subject to refinement as additional data is obtained and improved climate models become available.

More detailed results will be published in peer-reviewed scientific papers. See the list of publications derived from this project to date.

 

	Potential distribution map based on climate for:
Weed species	Present day	2020	2050
Orange Hawkweed (Hieracium aurantiacum)
Lippia (Phyla nodiflora)
Chilean Needle Grass (Nassella neesiana)
Serrated Tussock (Nassella trichotoma)
Lantana (Lantana camara)
Madeira Vine (Anredera cordiflora)
African Olive (Olea europaea subspecies cuspidata)
Predicted probability of presence