Abstracts

 

Kerstin Lux-Gottschalk (Eindhoven University of Technology)

Title: Uncertainty Quantification for Climate Tipping Points

Abstract: Several subsystems of the Earth are prone to sudden, large and irreversible changes, known as tipping points (TPs), under sustained anthropogenic forcing. We approach these tipping phenomena by taking a nonlinear dynamical systems’ perspective thereon. In particular, we use tools from bifurcation theory, which provides the mathematical framework for the creation and loss of equilibria as well as changes in their stability properties under variation of a deterministic control parameter.
To tackle real-world phenomena, it is important to account for uncertainties in critical thresholds and other system characteristics. Therefore, in this talk, I will address the question how uncertainty in model parameters affects the tipping behavior of the  dynamical system in terms of the location of TPs and how they are approached. I will present a workflow for climate tipping points that links several mathematical areas: starting from a Bayesian inference on the uncertain model input parameter, we  propagate the obtained input probability distribution forward through the nonlinear dynamics and perform a bifurcation analysis. As a result, we deal with probabilistic bifurcation curves [1]. I will show numerical results on a conceptual model of the  Atlantic Meridional Overturning Circulation, which is one of the identified tipping elements.

This talk covers joint work with Christian Kühn (Technical University of Munich, Germany), Peter Ashwin (University of Exeter, UK), Richard Wood (Met Office, UK), and Jonathan Baker (MetOffice, UK).

REFERENCES
[1] K. Lux, P. Ashwin, R. Wood, and C. Kuehn. Assessing the impact of parametric uncertainty
on tipping points of the atlantic meridional overturning circulation. Environmental Research
Letters, 17(7):075002, 2022.

 

Laura Scarabosio (Radboud University)

Title:  A seamless integration of models and data to predict and understand tumor cell dynamics

Abstract: Cancer cell dynamics is strongly affected by the microenvironment surrounding the tumor, with mechanisms that are not yet fully understood. After introducing a systematic approach to model the influence of different environmental factors on  tumor progression, I will discuss the methodology to calibrate the resulting model within a Bayesian framework. The latter will allow us to quantify the uncertainty in the parameter estimates and possible parameter correlations. We will also see how  the Bayesian framework can be used to compare different models. The results of the calibration procedure using different in-vitro laboratory data, where the environmental factors are either nutrient deprivation or oxygen level, tissue stiffness and the  presence of chemotherapeutic drugs, will allow us to shed some light on the underlying biological mechanisms affecting tumor progression. This is joint work with Sabrina Schönfeld (TU Munich), Christina Kuttler (TU Munich), Alican Ozkan (Harvard  University) and Marissa Nichole Rylander (UT Austin).

 

Lisa Kusch (Eindhoven University of Technology)

Title: A loss-based formulation for multi-objective optimization under uncertainties

Abstract: Finding optimal solutions to given problems in, e.g., engineering design often involves competing objectives like cost and quality of a solution. When accounting for uncertainties in this context it is advantageous to analyze the combined effect of uncertainties in the multi-objective setting. In this talk, we will present an overview of research aspects connected to multi-objective optimization under uncertainties. The focus will be on a proposed problem formulation to find solutions that minimize losses due to uncertain outcomes in multi-objective space. Different strategies how to quantify these losses and how to numerically solve the underlying problem formulation will be presented and applied in the context of aerodynamic shape optimization.

 

Ivan Kryven (Utrecht University)

Title: Branching process representation of evolutionary PDEs

Abstract: In their limiting regime, many discrete probabilistic models can be reduced to an effective macroscopic differential equation. A famous example of such dichotomy is the Brownian motion, which can be represented by a Laplace operator in the limit. In this talk, we will follow a reverse intuition: we will start with a certain evolutionary partial differential equation in one dimension and show that the random model that corresponds to it is a simple branching process (which in turn corresponds to Erdos-Renyi random graph.) We will then show that the analogy with (more general) branching processes is maintained, even when one considers an arbitrary nonlinear term in the PDE and some natural generalisations to multidimensional PDEs are possible as well. Interestingly, this method can be used to derive existence domains for PDEs without explicitly knowing their analytical solution.  This is a joint work with Jochem Hoogendijk,
https://arxiv.org/abs/2310.11338.