Perspectives on Climate Evaluation

BO LI – WASHINGTON UNIVERSITY, ST. LOUIS 

ABSTRACT

Evaluating climate models by comparing their historical simulations to observed data is crucial for identifying systematic errors and guiding further model development. Climate models can be evaluated from various perspectives, each requiring specific methodologies tailored to the particular objectives. In this work, we present two examples of climate model evaluation. The first example assesses the marginal extreme behavior of a regional climate model against reanalysis data. We developed a multiple testing procedure based on the characteristics of spatial extremes to identify areas where the marginal extreme value distribution and return levels diverge. The second example focuses on ranking general circulation models of the Coupled Model Intercomparison Project (CMIP) by comparing the distribution of their daily temperature and precipitation data to observations. To achieve this, we introduced the Spherical Convolutional Wasserstein Distance (SCWD), which accounts for spatial variability through convolutional projections and quantifies local differences in the distribution of climate variables. We additionally applied SCWD to evaluate the progression from CMIP Phase 5 to Phase 6 in terms of their ability to produce realistic climatologies.

Related Papers:

• https://arxiv.org/abs/2401.14657v1
• https://www.tandfonline.com/doi/full/10.1080/01621459.2020.1775613
• https://www.tandfonline.com/doi/abs/10.1080/09332480.2023.2179273