Sliced-Elastic Distances for Climate Model Validation
Conference
64th ISI World Statistics Congress - Ottawa, Canada
Format: IPS Abstract
Wednesday 19 July 2 p.m. - 3:40 p.m. (Canada/Eastern)
Abstract
Global climate model validation is integral for ensuring climate models produce realistic climatologies. However, many post-hoc statistical evaluation methods rely on simplifying models that discard information and fail to distinguish between different sources of variability. Here, we introduce a functional data analysis approach for computing Sliced-Elastic distances between spatiotemporal processes. By incorporating spatial information using convolutions of radial projections and accounting for temporal variation using time warping, our method precisely and rigorously quantifies shape differences between climate model outputs. This allows us to compare the underlying dynamics of each model in terms of phase (timing variability) and amplitude (magnitude after accounting for phase). Using the Sliced-Elastic distances, we rank outputs from CMIP6 historical climate models by similarity to historical surface temperature and precipitation from the ERA5 reanalysis. We repeat the ranking process for CMIP5 in order to assess the progression of climate model performance.