DCFP’s Terry O’Kane, Courtney Quinn and Dylan Harries present at ANZIAM

February 14th, 2020

DCFP’s Terry O’Kane, Courtney Quinn and Dylan Harries gave talks at the Australia and New Zealand Industrial and Applied Mathematics (ANZIAM) conference in early February in the Hunter Valley.

Courtney Quinn presented a talk on the recently published paper “Application of local attractor dimension to reduced space strongly coupled data assimilation for chaotic multiscale systems”, authored by herself and CSIRO colleagues Terry O’Kane and Vassili Kitsios.  Her presentation, although focused on a conceptual model of atmosphere-ocean interaction, highlighted the mathematical challenges of implementing strongly coupled data assimilation in high dimensional multiscale operational climate forecasting systems such as CAFE.

Terry O’Kane gave a presentation on “Coupled data assimilation and ensemble initialization with application to multiyear climate prediction”, which introduced the set-up and output of our CAFE60 reanalysis product, showing some of the atmospheric and oceanic diagnostics.

The final talk of the session was given by Dylan and titled “Identification and evaluation of causal networks in CMIP5 models”.  Dylan presented work being done with Terry O’Kane in collaboration with USI Lugano’s Illia Horenko from, on inferring causal relationships between climate teleconnections. This project aims to identify potential causal interactions between some of the main teleconnection modes by learning dynamic Bayesian networks that describe the relationships between modes. By performing this analysis both on reanalysis data and CMIP5 historical runs, they are able to compare how well models capture the observed relationships between teleconnections, and to assess model biases and weaknesses. The method is built on a non-stationary clustering framework, which is necessary to allow secular trends and regime transitions, as might arise under anthropogenic forcing for example, to be modelled.