2018pcpi banner

4: Assessing model performance in the polar regions

Co-leads: Hugues Goosse (Université catholique de Louvain, Belgium) and Jennifer Kay (National Center for Atmospheric Research, USA)

The goal of theme 4 is to perform comprehensive evaluation coupled models in the polar regions. This assessment of model performance must go beyond identifying model biases against trusted observational benchmarks and quantifying inter-model spread. The activities within this theme must ask "Why" questions to assess model performance.  Why do models behave in a different ways? Why do they or don't they match observations?  Answering "why" questions requires understanding important climate forcings and feedbacks and their process representation in models.  This theme seeks to mitigate problems that arise from the frequent disconnect between model development and model evaluation timescales and barriers from limited personnel.  As a result, a process-level link between model evaluation and improvement is a backbone of this initiative.  Model evaluation and improvement should address the mean climate state but also changes in response to internal variability and external forcing.

Our first step will be to engage a small group of people to identify "state-of-the-art" evaluation and improvement of models in polar regions.  We will rely on the latest assessment performed by IPCC as well as peer-reviewed and grey literature, but also discuss gaps as they exist now.

This group of people will then be invited to a workshop in which a synthesis will be performed, leading to a review paper if we consider that enough information has been gathered. We expect this workshop of theme 4 to be held in early 2014, maybe in conjunction with the February meeting of the Polar Climate Working Group in Boulder, USA.  The goal of this workshop will also be to identify best practices and key diagnostics that should be performed in order to have a useful evaluation of the models.  By "useful", we mean evaluation that has the potential to lead towards model improvements and process-level understanding.