A Conditional Simulation Approach to Future Precipitation Scenario Generation

• Activity Number: 89
• Type: Invited
• Date/Time: Sunday, August 9, 2015 : 8:30 PM to 9:15 PM
• Sponsor: Korean International Statistical Society
• Author(s): Won Chang* and Michael L. Stein and Elisabeth Moyer and Jiali Wang and Rao Kotamarthi
• Companies: The University of Chicago and The University of Chicago and The University of Chicago and Argonne National Laboratory and Argonne National Laboratory
• Keywords: climate change ; climate simulation ; precipitation ; non-Gaussian spatial data ; high-dimensional spatial data

Abstract

Future precipitation scenarios are important input for the development of adaptation strategy for climate change. Climate models allow physically consistent simulations of future precipitation changes, but their application for future scenario generation is complicated due to their inability to reproduce the current precipitation patterns. Our goal is to propose a conditional simulation approach, which modifies the observed precipitation based on the changes quantified by statistical analysis of climate model output. In order to capture the highly-localized behavior of precipitation, we use high-resolution ensemble from the WRF regional climate model to derive the changes in precipitation patterns and radar-based observations from NCEP Stage IV data to have highly accurate baseline precipitation observations. This problem poses nontrivial computational and inferential challenges due to the non-Gaussian and high-dimensional nature of the data. We apply various image processing techniques and random objects analysis to mitigate these challenges.