Proposal Description

Artificial intelligence and machine learning are transforming environmental science by enabling the analysis of complex, high-dimensional data and the development of predictive tools for monitoring, forecasting, and decision-making. However, these advances also raise pressing questions about interpretability, uncertainty quantification, reproducibility, and the robustness of conclusions drawn from data-driven models.

This panel will explore how statistical principles can strengthen the scientific foundations of environmental AI. Bringing together perspectives from academia and industry, the discussion will focus on integrating statistical inference and probabilistic reasoning within AI pipelines to ensure transparency, reliability, and meaningful uncertainty communication. Topics will include probabilistic machine learning, Bayesian calibration, model validation, and the design of interpretable hybrid models that bridge mechanistic and data-driven approaches. Panellists will examine both methodological opportunities and practical challenges in building trustworthy AI systems for environmental applications. The session aims to stimulate dialogue between the statistics and AI communities, identifying shared priorities and pathways for future collaboration.