SIParCS 2016 - Xiaoying Pu

Xiaoying Pu, Bucknell University

Visualizing Intermodel Comparison of Climate Simulations

(Slides)  (Recorded Talk)

We explored using multidimensional scaling (MDS) to visualize intermodel comparisons for climate simulations. Due to a large number of models in CMIP5 and their visual complexity, we wished to present the similarity of the models more intuitively than the status quo. With simplicity and effectiveness in mind, we chose structural similarity (SSIM) as the distance metric, treating model grid cells as image pixels. The distance matrix was computed from pairwise SSIM, and the metric multidimensional scaling produced a 2D scatter plot, whereby a point represented an ensemble member. We analyzed CMIP5 members (~200) alongside with observational datasets from CVDP outputs. Clustering on the scatter plot revealed development lineages in the underlying climate models. In addition to the intermodel comparison, we applied a novel visualization method, time curve, to characterize temporal similarity in the sea ice (SIC) data.

Mentors: Rick Brownrigg and Bill Ladwig, CISL