Pre-Conference Workshop: Probabilistic Graphical Models for Earth System Science (PGM-ESS 2026)

• Date: 8 September 2026
• Location: Valencia, Spain (in conjunction with PGM 2026)
• Contact: workshop organizers via the PGM 2026 chairs

Overview

Earth System Science (ESS) investigates the complex, interconnected, and dynamic components of our planet, including the atmosphere, oceans, biosphere, and cryosphere. Developing a deep understanding of these systems, characterising feedback loops, attributing extreme events, and projecting future climate scenarios requires methods that go beyond simple correlation-based machine learning.

Probabilistic Graphical Models (PGMs)—such as Bayesian networks, structural causal models, dynamic PGMs, and staged trees—offer a rigorous mathematical framework to represent complex multivariate dependencies, incorporate physical constraints, perform probabilistic inference under uncertainty, and discover causal pathways from observational and simulated data.

This workshop aims to bring together researchers from machine learning, statistics, causal inference, and earth system sciences to discuss recent methodological advances and novel applications of general and causal graphical models to environmental, ecological, atmospheric, and climate challenges.

Topics of Interest

We invite submissions on both theoretical developments and applied research at the intersection of Probabilistic Graphical Models (both causal and general frameworks) and Earth System Science. Topics of interest include, but are not limited to:

• Probabilistic Inference & Reasoning: Efficient exact and approximate inference algorithms, variational methods, and Monte Carlo approaches for complex environmental models.
• Structure & Parameter Learning: Learning graphical representations, parameter estimation under data sparsity, and handling missing data in spatiotemporal datasets.
• Causal Discovery: Algorithms for learning causal graphs from high-dimensional, spatiotemporal, or non-stationary environmental and climate data.
• Causal Inference & Attribution: Estimating causal effects and attributing extreme events (e.g., heatwaves, droughts, floods) to specific drivers.
• Physics-Informed PGMs: Incorporating physical laws, conservation principles, or numerical model outputs into graphical models.
• Spatiotemporal Graphical Models: Dynamic Bayesian networks, state-space models, and graphical representations of atmospheric/oceanic transport.
• Explainable & Trustworthy AI in Climate Science: Using PGMs to provide interpretable explanations of complex climate phenomena.
• Decision Support & Risk Assessment: Bayesian networks for ecological management, environmental policy-making, and natural hazard risk assessment.
• Asymmetric Graphical Models: Staged trees, context-specific independence, and asymmetric causal models applied to environmental and biological systems.
• Uncertainty Quantification: Quantifying epistemic and aleatoric uncertainty in climate projections using graphical frameworks.

Invited Speaker

We are pleased to announce our invited speaker for the workshop:

• Urmi Ninad (University of Potsdam)
  • Abstract: TBA

Submission Guidelines

We invite submissions in the form of a simple abstract (up to 500 words or 1 page). Abstracts should outline the core research topic, methodology, and key results or discussion points.

All accepted abstracts will be selected for oral or poster presentations during the workshop. Please note that workshop submissions are non-archival, meaning there will be no formal proceedings. This is intended to encourage the presentation of ongoing work, preliminary results, or research that is under review elsewhere.

Submission Instructions

Abstracts should be submitted via email to pgm2026@uv.es, either as plain text in the body of the email or as a PDF attachment. Please use the subject line “PGM-ESS 2026 Abstract Submission: [Your Name]”.

Important Dates

We accept submissions on a rolling basis to allow for early planning and travel arrangements.

• Submissions open: 15 June 2026
• Rolling notifications start: 15 July 2026
• Final submission deadline: 15 August 2026
• Workshop date: 8 September 2026

At least one author of each accepted abstract must register for the workshop/conference.

Workshop Organizers

The workshop is organized in coordination with the PGM 2026 Program Chairs:

• Gherardo Varando (Universitat de València)
• Manuele Leonelli (IE University, Madrid)
• Gustau Camps-Valls (Universitat de València)