The Scale Enhanced Wild Bootstrap Method for Evaluating Climate Models Using Wavelets

• Activity Number: 280 - Climate Statistics: Studies on the Physics and Impacts of Climate Change Using Data Science
• Type: Topic Contributed
• Date/Time: Tuesday, August 1, 2017 : 8:30 AM to 10:20 AM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Ansu Chatterjee* and Amy Braverman and Megan Heyman and Noel Cressie
• Companies: University of Minnesota and Jet Propulsion Laboratory and Rose-Hulman Institute of Technology and University of Wollongong
• Keywords: Wavelet ; resampling ; Indian monsoon

Abstract

There is ambiguity about how to transform weather data to climate signals, which in turn creates ambiguity about how to assess the performance of Physics-based climate models and reanalysis objects in modeling observed climate phenomena. We propose a definition of climate signal based on the coarse levels from a wavelet decomposition of weather data. Using that definition, we propose a new resampling-based methodology for testing if a Physics-model or reanalysis object captures the observed climate signal. We establish consistency of the resampling procedure for the hypothesis testing problem. Simulation results, and a real data application on precipitation in the Indian subcontinent, are provided to illustrate the performance and efficacy of our proposal.