TRR 181 Seminar “Stochastic parametrizations in climate models: Incorporation of uncertainty estimates and sub-grid scale variability”

The TRR 181 seminar is held by Dr. Stephan Juricke, University of Oxford on

“Stochastic parametrizations in climate models: Incorporation of uncertainty estimates and sub-grid scale variability”

at Jacobs University, on October 20 at 9.45 am.

Abstract

Global climate models are used to predict the future behaviour of Earth system components such as the atmosphere, the oceans, and sea ice. However, these models are approximations of the truth and as such their capability to represent certain aspects of reality is limited. Crucial physical processes are not resolved by climate models because their spatial and temporal resolution is too coarse to capture them explicitly. As a results model errors emerge. In combination with uncertainties in the model initialisation, these errors have the tendency to grow with time, reducing forecast accuracy on longer timescales. To provide a measure of confidence in forecasts, model uncertainty estimates are incorporated in climate models. One specifically successful approach is the use of stochastic parametrizations. While classical deterministic parametrizations aim to reproduce the averaged effect of unresolved physical processes on the resolved circulation patterns, a stochastic parametrization generates single possible realisations of a sub-grid scale process. In an ensemble of climate predictions each member then produces different stochastic realizations. This provides an estimate of the possible pathways the real climate system might take, quantifying the likelihood of future climate events.

In this talk I will introduce and discuss stochastic parametrizations in global climate models, including the atmosphere, the oceans, and sea ice. I will show how stochastic parametrizations can provide model uncertainty estimates that improve the reliability of weather and seasonal predictions. Furthermore, the probabilistic approach to parametrization development offers an opportunity to provide models with more accurate sub-grid scale process variability, leading to reduced climatological biases and forecast errors, and an improved representation of climate variability. I will conclude by discussing current plans for probabilistic climate model development.