SIParCS 2016 - Delilah Feng

Delilah Feng, University of Wyoming

Evaluating Similarity Across Different Climate Model Simulations with Structural Similarity Index

(Slides)  (Recorded Talk)

Working with the Data Visualization & Analysis Tools (DVAT) division, we present an approach using image processing techniques to evaluate the similarity of a large number of climate model simulations. Our target data set is the hundreds of output files and images of dozens of key variability metrics generated by the Climate Variability Diagnostics Package (CVDP). CVDP was developed at NCAR by the Climate Analysis Section. It is an analysis tool that documents the major modes of climate variability in models and observations. However, while model performance metrics are calculated, there currently is no attempt to rank the models. The structural similarity (SSIM) index is widely used in the evaluation of the perceived quality in digital images and videos. It consists of the weighted comparisons of intensity, contrast, and structure of image pairs. We treated the climate model grids as images, developed a modified SSIM algorithm in NCAR Command Language (NCL), and tested it on Coupled Model Intercomparison Project Phase 5 (CMIP5) data. As a result, we generated structural similarity maps and difference maps comparing the models to the observational data for the key variability metrics. We also applied the concept of parallel coordinates to visualize the ranking of the model simulation results with respect to the calculated structural similarity indices.

Mentors: Rick Brownrigg and Bill Ladwig, CISL