Burrell A1
1University Of Sydney, , Australia
Biography:
Dr Arden Burrell is a data scientist specialising in remote sensing and machine learning. He completed his PhD at the University of New South Wales in 2019, studying global-scale land degradation through remote sensing and statistical modelling. From 2020 to 2024 he worked at Woodwell Climate Research Center on NASA's Arctic-Boreal Vulnerability Experiment, using field data, satellite imagery, climate models and machine learning to assess forest vulnerability and recovery from wildfire and drought. At the Australian Plant Phenomics Network he develops data-quality and reproducibility systems for drone-based phenotyping, building version-controlled, citable standards and automated quality-control pipelines for the national network.
Abstract:
Across the Australian Plant Phenomics Network, drone-collected hyperspectral imagery underpins a growing body of earth and environmental research. Yet when two nodes fly the same crop, or one node flies it twice, the numbers often disagree. The Airborne Phenomics Excellence project set out to find, quantify, and remove these sources of error. Systematic validation flights and shared-data experiments revealed that the dominant, fixable cause was not the sensors but the process: reflectance over shared calibration panels varied by more than ten percent between nodes, repeat flights drifted within ninety minutes, and operator choices during calibration shifted the final result. The variance lived in field and processing procedure, and in whether the agreed method was written down, current, and followed. This presentation describes our response: managing scientific protocols the way software teams manage code. Field procedures are authored as version-controlled documents in a shared repository, revised through reviewed pull requests, locked as a published read-only wiki, and issued as tagged, citable releases with permanent identifiers, openly licensed and aligned with FAIR (Findable, Accessible, Interoperable, Reusable) principles. Alongside the protocols we are building automated quality-control checks, including a spatial accuracy test with a ten-centimetre threshold and emerging spectral diagnostics. The system is live today, governing the network's field protocols at revision one. We will demonstrate the repository, the publication pipeline, and the quality reports, then reflect on what governed, reproducible standards mean for the future of shared research infrastructure: a modest idea, borrowed from software, with outsized consequences for trustworthy data.
