Aerial analytics: Investigating the use of drones and spatial data to assess urban plantings

Summary

How do you know whether a planting has actually worked? Most urban landscape projects are assessed by eye, once, shortly after handover — long before the period that determines whether plants survive. This Fellowship report tests whether drones and spatial data can close that gap.

Michael White, the LFA's 2024 Fellow, applied drone imagery, multispectral analysis, point-cloud capture and GIS across five live pilot sites, monitoring plantings from installation through the critical first year of establishment. The sites were deliberately varied — light rail corridors to rooftop meadows, native and exotic palettes, different substrates, budgets and installation methods — so that the findings would transfer rather than describe one project.

The central proposition is that drone imagery can generate 'as-built' and '12-month established' GIS planting plans: a standardised, geolocated record of how vegetation actually developed. That record serves two purposes — judging an individual project, and building a shared evidence base across the industry.

For asset managers and councils, the practical interest is cost. Plant failure is expensive and usually discovered late. Monitoring that detects stress early changes when you can intervene, and gives a defensible basis for the maintenance budget.

Key findings

• Multispectral drone survey can quantify what visual inspection only estimates. Vegetation cover, plant counts and survival were tracked across all five pilots.
• Weed detection is promising but not yet reliable. Machine learning was trialled at Orchard Hills: mature weeds were identifiable, but seedlings of different species proved very difficult to separate, even with good training data.
• Substrate variation is a real confound. Differing soil reflectance complicated separating vegetation from background — a practical caution for mixed sites.
• The deeper problem is not technical. Consultation across industry, academia and the public sector identified the absence of clearly defined performance goals, and the difficulty of sustaining long-term monitoring, as the binding constraints.
• The value compounds only if data is shared. A single monitored site informs one project; a network of them supports evidence-based planting decisions across Australian conditions.

Key lessons

• Plan for success — define what good looks like, and how it will be measured, at the outset: a percentage of cover at six months, say, or plants per square metre surviving establishment.
• Create consistent data structures — checklists, GIS base files and reporting templates set up before monitoring begins.
• Validate against manual measurements — soil colour and moisture, weather and time of day all affect results, so check aerial analysis against the ground.
• Collaborate with experts in fields such as soil science and horticulture.
• Record site conditions — wind, rain and temperature — so they can be correlated with plant performance.
• Confirm site access early, and allow for weather delays in the program.
• Drones are powerful and accessible, but software, hardware and skills can be a barrier for small organisations.
• Plan how outcomes will be used — how learning will flow back into design and future projects.

Pilot sites

• Sydney Metro Trials, Castle Hill, NSW
• Green Track Prototype, Canberra, ACT
• AMRF First Building, Bradfield, NSW
• MAPT Trials, Burnley, Vic
• Sydney Metro Trials, Orchard Hills, NSW

Watch the interview

In the first of LFA's Fellows Interview Series, LFA Ambassador Crosbie Lorimer talks with Michael White about his Fellowship project.

Watch on YouTube

Downloads

Download: Report (46 pp)

Cite as

White, Michael G. “Aerial analytics: Investigating the use of drones and spatial data to assess urban plantings.” LFA Fellowship report. Landscape Foundation of Australia, 2025.