Evaluating Climate Model Cloud Representation through a Machine Learning Framework
Alex Schuddebooom, PhD Candidate
School of Physical and Chemical Sciences
Time & Place
Tue, 18 Jun 2019 10:00:00 NZST in Room 701, Level 7, WEST
All are welcome
Climate models are an essential tool for understanding future climate and preparing for the challenges of climate change. As such, they are constantly being evaluated and improved to give us the best possible projections. There are several long standing errors that exist within these models with one of the largest being an excess of solar radiation localized over the Southern Ocean. My thesis is focused on examining the relationship between Southern Ocean clouds and this error.
I use a machine learning technique known as a self organizing map to classify clouds into different types. This enables us to directly link the representation of different types of clouds in the model to radiation errors. This classification is extensively explored throughout my work and used to identify more nuanced differences between the model and satellite data. The classification scheme also enables an investigation of the role that compensating errors play in model evaluation.