Henning Rust Seminar - "Drift in Decadal Predictions -- Ideas for Parametric Post-processing"

Event type: 
Seminar
Date: 
27 February 2017
Time: 
2.00pm - 3.00pm
Location: 

Climate Change Research Centre seminar room, level 4 Mathews Building, Kensington Campus, Sydney

Presenter: 
Henning Rust
Freie Universität Berlin
Host: 
Climate Change Research Centre, UNSW Australia

A climate projection is a climate model simulation with prescribed boundary conditions, such as a pathway of radiative
forcing, and starting from an arbitrary climate state on the model's attractor. The idea of decadal climate predictions is instead to use
the same boundary conditions but initialise the climate model with a climate state close to the observed state. As this is not necessarily
a state on the climate model's attractor, simulations show an artificial behaviour thought to be the model's attempt to get back to
it's preferred states. This is artificial in the sense that it is neither a reaction to changing boundary conditions nor part of
natural climate variability.

From a data analysis perspective, this behaviour is conceived as a model bias changing with forecast lead time. The term frequently used
for this behaviour is "drift". This talk is about a concept to quantify drift due to initialisation and parametric strategies to reduce it's
influence on the prediction by post-processing either the ensemble mean ("drift correction") or the full probabilistic prediction
("re-calibration").

 

Brief Biography: Henning Rust is a Professor in Statistical Meteorology at the Freie Universitat Berlin, Institut für Meteorologie, Berlin, Germany. He leads the working group Climate diagnostics and extreme meteorological events (CliDia). Henning main research interests are stochastic precipitation models and parameterisations, extreme convective events, decadal prediction, forecast verification and statistical modelling.