Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
UP: Fachverband Umweltphysik
UP 1: Oceanography and Climate Modelling
UP 1.7: Vortrag
Donnerstag, 2. September 2021, 12:30–12:45, H3
Bayesian parameter estimation for EBMs: What can we learn about climate variability? — •Maybritt Schillinger1,2, Beatrice Ellerhoff1, Kira Rehfeld1, and Robert Scheichl2 — 1Institute of Environmental Physics, INF 229 — 2Institute of Applied Mathematics, INF 205, 69120 Heidelberg, Germany
Reliable climate projections in the face of global warming require an improved understanding of the internally-generated and externally-forced variability of Earth's climate. To this end, energy balance models (EBMs) provide a conceptual tool for studying climate dynamics. However, EBMs are typically based on a set of parameters with considerable uncertainties across empirical data and model hierarchies. To incorporate these uncertainties, we describe the global mean temperature as an inverse problem: We model the observed data as a function of the unknown parameters, given through the EBM's solution, and stochastic noise, representing the internal variability. With a Bayesian approach and a MCMC algorithm, we estimate the parameters as well as the best model fit to the data. In particular, we investigate how this estimate depends on the strength of internal variability compared to the response to external forcing. We discuss results for the zero-dimensional linear EBM and possible extensions with time-dependent feedback parameters. Our approach represents an application of state-of-the-art analytical and numerical techniques to the complex dynamics of Earth's climate. It can help to elaborate the potential, but also limitations, of the inverse problems approach and be readily applied to other dynamical systems with uncertain parameters.