High-Resolution Nonstationary Weather Simulation

• Activity Number: 14
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): William Kleiber*+
• Companies: University of Colorado
• Keywords: Nonstationary ; Simulation ; Spatial process ; Weather generator

Abstract

Stochastic weather generators (SWGs) are used in many scientific studies, including model downscaling, climate impact assessments and seasonal resource planning. The fundamental requirement of a stochastic weather generator is the ability to generate simulated realizations of plausible weather patterns. In recent years, focus has shifted to developing spatially-consistent SWGs. Unless the region of interest is relatively small, or has homogeneous topography, the SWG will necessarily require simulation of nonstationary spatial fields. However, high resolution simulation of a nonstationary process is difficult, typically requiring a Cholesky decomposition of a matrix whose dimension equals that of the desired simulation resolution. We introduce an approach to high resolution nonstationary process simulation by exploiting ideas very similar to Sampson and Guttorp (1992), relying on spatially deforming geographical space to achieve approximate stationarity, then using fast stationary simulation algorithms, followed by an inverse transformation back to the nonstationary plane. We illustrate the algorithm on simulated and real datasets.