Validating Climate Models with Spherical Convolutional Wasserstein Distance

Authors: Robert Garrett, Trevor Harris, Zhuo Wang, Bo Li

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We apply this method to evaluate the historical model outputs of the Coupled Model Intercomparison Project (CMIP) members by comparing them to observational and reanalysis data products. Additionally, we investigate the progression from CMIP phase 5 to phase 6 and find modest improvements in the phase 6 models regarding their ability to produce realistic climatologies.
Researcher Affiliation Academia 1University of Illinois Urbana-Champaign 2Texas A&M University
Pseudocode Yes Algorithm 1 Spherical Convolutional Wasserstein Distance Approximation
Open Source Code Yes the code to reproduce the SCWD calculations for CMIP5 and CMIP6 to the reference datasets is provided in the supplemental material.
Open Datasets Yes To serve as references for climate model evaluation, we collect the European Centre for Medium Range Weather Forecasts (ECMWF) Reanalysis 5th Generation (ERA5) [Hersbach et al., 2020] as well as the Reanalysis-2 data from the National Centers for Environmental Protection (NCEP) [Kanamitsu et al., 2002] reanalysis datasets.
Dataset Splits No The paper uses observational and reanalysis datasets as "references for climate model evaluation" but does not specify training, validation, or test dataset splits in a traditional machine learning context.
Hardware Specification Yes All computations were performed using R version 4.3.2 on an Ubuntu operating system with an Intel i5-9600k processor (6 cores), 32GB RAM, and 3TB hard disk space.
Software Dependencies Yes All computations were performed using R version 4.3.2
Experiment Setup Yes All SCWD calculations in this section use the Wendland kernel with range parameter of 1,000km for slicing.