Panel Discussion 1: The role of ML4EO in Biodiversity Net Gain
Date: Tuesday, 23rd June 2026
Panel Members: Donna Lyndsay, Dan Bloomfield, David Ferguson, Dan Carpetener
Co-Chairs: Andy Cunliffe and Brianna Pickstone
Brief
The panel “The Role of ML4EO in Biodiversity Net Gain” will examine how machine learning for Earth observation (ML4EO) can assist biodiversity assessments, monitoring and decision-making within the BNG framework. As BNG drives demand for scalable, repeatable and evidence-based ecological information, a growing number of commercial tools now combine remote sensing and machine learning to deliver outputs such as baseline habitat maps, biodiversity unit calculations, condition assessments and long-term site monitoring. These innovations offer exciting opportunities to completement traditional field surveys and address challenges of cost, access and consistency. At the same time, the rapid expansion of these commercial production raises important questions about transparency, validation and real-world adoption. Panellists will explore how these tools are being adopted within the ecological sector, how practitioners perceive their reliability, and what level of transparency and ecological grounding is needed to ensure confidence in ML enabled assessments that may inform legally binding BNG outcomes. They will also examine key technical challenges (data scarcity, bias and model transferability), and pathways to operational uptake by practitioners, conservationists and policymakers. The goal is to spark discussion about how ML4EO can be responsibly adopted across the sector, creating transparent tools that support credible BNG implementation
Panel Discussion 2: Valuing Reproducibility in ML4EO
Date: Wednesday 24th June 2026
Panel Members: Andy Cunliffe (Chair), Mark Kelson, Johan Wahlström, Katie Murray, Tomislav Hengl
Brief
The panel “Valuing Reproducibility in ML4EO” will examine why rigorous reproducibility practices are essential for credible Earth observation machine learning research and operational deployment. The discussion will touch on technical standards, computational workflows and institutional incentives that enable transparent, repeatable and extensible ML4EO science. Key themes include open data and code, version control, containerisation, benchmark datasets, documentation standards and provenance tracking across complex geospatial pipelines. Panellists will also address barriers such as proprietary data, computational cost, and misaligned academic reward structures. By highlighting practical tools and community norms, the session aims to clarify how valuing reproducibility strengthens trust, accelerates innovation and supports scalable, policy-relevant applications of ML4EO across environmental monitoring and decision-making contexts.
ML4EO 