Developing pest risk models of Buffel Grass using Unmanned Aerial Systems and Statistical methods

Why it matters

Buffel grass (Cenchrus ciliaris) is an introduced, perennial pasture grass that is found across much of the Australian continent, including arid and semi-arid regions. For many decades it has been widely planted for livestock production and land rehabilitation. In the cattle industry, it is well regarded as pasture because it grows rapidly under warm, moist conditions and persists under heavy grazing and drought.

However, buffel grass has spread well beyond planted areas and dominates the ground layer in many native plant communities. It reduces native plant diversity and can affect vegetation structure by changing fire regimes. In arid Australian areas such as Uluru-Kata Tjuta National Park, buffel grass invades some of the wetter, more fertile parts of the landscape, limiting the chances of survival of native plant and animal populations. Although it was planted for dust control in central Australia, it also imposes economic costs through the need to manage fire risks and to protect biodiversity assets and infrastructure.

Buffel grass has been identified as a significant threat to biodiversity in some regions of Australia. On Barrow Island (WA) for example, Chevron Australia has embarked on a significant campaign to map and eradicate buffel grass as a part of the environmental commitments related to the Gorgon Gas Project.

Project Overview

The aim of this project is to develop a system which combines new detection methods (UAVs and specialised sensors) with advanced modelling techniques to determine high-risk areas for pest risk surveillance, namely buffel grass. The project requires advanced engineering, ecological and statistical skills. In essence, the idea is to fly a UAV with sensors across a swathe of the landscape, identify pests of interest, develop a predictive risk model based on the received information, apply this model to a broader scene acknowledging uncertainties in prediction, and design corresponding surveillance protocols for pest detection.

This project has both immediate and longer-term strategic benefits. The protocols and methodology arising from this project will focus on detection of a specific invasive species, Buffel Grass, on Barrow Island (BI). However, the general skills, know-how and methods that are developed in the project will be much more widely applicable to other pest risk problems and more broadly to other geographic detection and modelling projects.

Real world impact

• Flight campaign using UAVs and imaging sensors onboard unmanned aircraft for invasive species detection and identification at exemplary site completed.
• Delivery of review of existing methods for detection and progress report assessing the feasibility of using imaging sensors onboard unmanned aircraft for invasive species detection and identification, based on limited datasets.
• Delivery of progress reports on the review of predictive modelling approaches combined with the feasibility of using imaging sensors onboard unmanned aircraft for invasive species detection and identification.
• Development of a pipeline process and software that offers automatic pixel-wise segmentation and mapping on aerial imagery of invasive grasses and other vegetation (e.g. Cenchrus ciliaris and Triodia), achieving ground sampling distances (GSD) up to 1.01 cm/pixel.

Publications

1. Sandino, J.; Gonzalez, F.; Mengersen, K.; Gaston, K.J. UAVs and Machine Learning Revolutionising Invasive Grass and Vegetation Surveys in Remote Arid Lands. Sensors2018, 18, 605. Link: https://doi.org/10.3390/s18020605
2. Pipeline process overview video: https://youtu.be/Hlgf-oO48BA

Partners