Strengthening a national standard for UAV-based field phenomics

One of the unique strengths of the Australian Plant Phenomics Network (APPN) is its distributed national field network, operating across diverse agricultural environments yet unified by a shared purpose: delivering high quality, comparable phenomics data to researchers across Australia. Central to this capability are the GRDC funded Mobile Phenotyping Units (MPUs) – advanced, field deployable platforms designed to bring cutting edge sensing technologies directly into agricultural trials, wherever they are located.

The long-term vision is ambitious but clear: a common set of sensors deployed nationwide, operated in exactly the same way by highly skilled teams, producing data that is directly comparable across locations, seasons and research programs. In such a system, variability driven by environment or management can be interrogated with confidence, rather than being obscured by differences in data collection or processing. Realising this vision is essential if UAV based phenotyping is to support robust, multi-site agricultural research at scale.

Recognising Both Potential and Challenge

Over recent seasons, APPN’s UAV capabilities have expanded rapidly. Hyperspectral imaging, in particular, provides rich spectral information capable of revealing subtle differences in crop status, stress response and physiological function that are invisible to conventional sensing approaches. Hyperspectral imagery is both powerful and complex, however, it is challenging to do it well. 

While working through data recently collected across the network, the University of Sydney APPN node identified that small differences in flight planning, sensor configuration, calibration approaches, and downstream processing pipelines had introduced variation that made good cross site comparisons difficult.

To tackle this challenge, the APPN Aerial Phenomics Excellence (APEx) program was established. Led by the University of Sydney node, the program is internally funded through APPN and involves cooperation across all of APPN’s field nodes.

Establishing a Baseline: Understanding the Scale of Variation

The first phase of APEx was about understanding the level of variation that existed between hyperspectral datasets across the nodes. To do this, Arden Burrell and Richard Harwood, data scientists within the University of Sydney node, undertook a coordinated analysis exercise using data from the 2025 season.

Two complementary approaches were used. First, datasets collected by different APPN nodes were reprocessed using a common workflow at the University of Sydney to quantify how much variation arose from differences in data collection alone. Second, a single common dataset was distributed to each field node and processed using each node’s existing pipelines, allowing the team to quantify variation introduced during the processing stage, from raw data through to final data products. This was important to do, as it highlighted avoidable sources of error existed, both in collection approaches and in processing pipelines. 

A Flagship Opportunity: APPN Calibration Week 2026

The 2026 APPN Calibration Week in Adelaide and Roseworthy, South Australia represented a unique opportunity for the network. Representatives across all six field nodes, including operators and analysts were able to come together with their hyperspectral sensors and work through what excellence really looks like for UAV hyperspectral data collection, processing and analysis.

Calibration Week became a focal point for the APEx program, not just as a technical exercise, but as a deliberate effort to enhance shared understanding across the field network. Formal daytime sessions provided the structure for this, with detailed technical discussions focused on flight protocols, calibration strategies, sensor configuration and processing pipelines for the network’s flagship UAV hyperspectral sensors, GOBI and CALViS which are at each field node. These sessions helped align thinking around best practice and ensured difficult methodological questions were explored openly and constructively.

Just as importantly, many of the most insightful conversations happened outside the formal agenda. Informal discussions over dinner, on bus rides to and from the flight sites, and during downtime between flights allowed ideas to be tested, challenged and refined in ways that are difficult to achieve in structured meetings alone.

The ideas generated through these discussions were then put into practice in the field. Across more than 40 calibration and experimental flights conducted over four days of clear blue skies, the team systematically compared performance system‑to‑system, and across flight heights, speeds and hyperspectral exposure settings. 

What emerged was not just a clearer picture of sensor performance, but a shared sense of ownership over what true excellence looks like in hyperspectral data collection. Calibration Week underscored that achieving consistency at a national scale depends as much on trust, openness and shared learning as it does on technical documentation.

Looking Ahead: Refining Excellence Through Evidence

Calibration Week marked a major milestone for the APEx program, because it strengthened and operationalised principles that have long underpinned the network. 

Over the coming months, the protocols and data processing pipelines refined through APEx will be made available via GitHub, supported by clear documentation, version control and traceability. This transparency is intentional. Providing shared, accessible workflows ensures that standard APPN approaches can be readily adopted across all nodes, supporting consistency for operators and clarity for users.

Through the APEx initiative, APPN has strengthened its national capability and reinforced its leadership in best practice UAV‑based hyperspectral phenomics, both within Australia and internationally.

Pictured above are VNIR hyperspectral orthomosaics from four APPN nodes (UOA, USYD, UQ, CSU), animated to compare spatial alignment. Red dots mark ground control point locations surveyed with gold-standard GPS. The next stage will assess spectral accuracy — comparing reflectance measurements extracted over known spectral targets to validate spectral consistency

12 May 2026