Paleoclimate Modeling and Data Assimilation

Wed, 22 Oct 2025, 17:00 CEST

Conveners: Julien Emile-Geay & Francesco Muschitiello

Improving future climate change projections requires a better understanding of past climate change, but this is challenging due to incomplete proxy records and imperfect models. This session brings together two complementary perspectives on addressing these challenges. The first talk explores an approach to paleoclimate modeling that embraces uncertainty as a productive resource. This approach proposes flexible model design to explore a range of plausible climate scenarios rather than seeking a single definitive outcome. The second presentation introduces paleoclimate data assimilation, a method that integrates proxy data with Earth system model outputs to create physically consistent reconstructions of past climates. By combining the strengths of these approaches, this session offers a compelling look at how scientists advance the study of Earth's past climate systems despite complexity and ambiguity.

Guest speakers:

• Ruza Ivanovic (University of Leeds, UK)
• Quentin Dalaiden (Nansen Environmental and Remote Sensing Center Bergen, Norway)

Modelling the Unknowable: Embracing Uncertainty in Climate Change

Ruza Ivanovic

Climate models are useful tools for understanding the dynamics of a physical system, which may relate to the real world and known past changes. However, even for the last glacial period and deglaciation, which present some of the most recent case studies for 'observed' climate change, it remains challenging to accurately simulate known physical behaviors in the interactions between different earth system components for specific chains of events. Moreover, amidst all the uncertainty in the tools (models), constraints (climate records), and even the definition of the problem (what is it we are trying to understand and how can this be useful?); how do we design relevant experiments that offer genuine insight to climate change? In this webinar, I will explore an approach to paleoclimate modelling that aims to give space to multiple avenues of scientific interest and method; embracing uncertainty to more robustly understand the multiplicity of plausible realities and de-tune models to increase their flexibility to project change.

Dynamical climate reconstructions from proxy records and Earth System Models through Data Assimilation

Quentin Dalaiden

Paleoclimate proxies and Earth System Models (ESMs) offer complementary insights into historical climate variability, though each comes with inherent limitations. To leverage the strengths of both, data assimilation techniques—originally developed for weather prediction—have been adapted, over the past two decades, to produce dynamically consistent reconstructions of past climates, ranging from the last few centuries to deep-time epochs. This approach, known as Paleoclimate Data Assimilation (PDA), integrates sparse proxy networks with outputs from ESMs. PDA provides a coherent multivariate reconstruction constrained by physical laws, offering a powerful framework to investigate the dynamical drivers behind proxy-observed climate changes. This talk will outline the fundamental principles of PDA, highlight recent methodological advances, and discuss promising directions for future research.