Status

Contracting

Project Leader

Dr Angus Carnegie, NSWDPI

Background

Several reviews have highlighted a need for improved biosecurity surveillance at first points-of-entry, commonly called high risk site surveillance (HRSS), such as at major ports and approved arrangement facilities. Government- and industry-funded surveillance programs utilize insect traps and host-tree surveillance. Currently in most jurisdictions, host trees are identified and mapped via ground surveillance (i.e. driving the streets), supplemented by examining GoogleMaps. This is very inefficient, and as such has only been conducted for a relatively small area of high-risk sites. During an emergency response to a forest pest detection, tree-host mapping is required for surveillance to delimit the spread and determine feasibility of eradication. Without accurate host-tree location data, this process is time-consuming and resource-intensive, possibly delaying a timely response to an exotic incursion. Remote sensing technologies combined with machine learning applications show promise in being able to locate and identify individual trees in urban areas (e.g. citrus canker response in Darwin). Some local councils already capture such data (e.g. to map greenspace), thus allowing biosecurity agencies to piggy-back on data acquisition to assist in high risk site and emergency response surveillance of forest, amenity and environmental pests.

Objective

  1. Assess the feasibility of remote sensing technologies and machine learning applications for detection and mapping of urban trees to assist in forest biosecurity surveillance.
  2. Liaise with local councils to develop a collaborative agreement to improve urban tree biosecurity surveillance, linking in the DPI/Local Land Services Peri-Urban Biosecurity Program.

Methods/Activities

  • Utilise existing relationships with local councils to establish collaborative agreement(s).
  • Analyse existing remote sensing data captured by ArborCarbon (e.g. City of Sydney) and tree-location data captured by DPI to investigate likely sensors and resolution for further acquisition.
  • Acquisition of aerial imagery over Bayside Council (Port Botany); image processing; machine learning application; generation of derived products.
  • Utilise tree location data already captured by DPI and supplement with higher resolution ground capture (i.e. differential GPS) to feed into machine learning process; includes tree location and species identification.

Outputs

  • Spatial data of individual tree locations for key hosts, e.g.: Pinus (forestry), Platanus (amenity), Magnolia and Ailanthus (brown marmorated stink bug) and Eucalyptus.
  • Investigate options for detecting red imported fire ant nests (e.g. via thermal imagery).