Conditional Simulation for Downscaling Large Spatial Data

• Activity Number: 687
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Pulong Ma* and Emily L. Kang and Amy Braverman and Hai Nguyen and Noel Cressie
• Companies: University of Cincinnati and University of Cincinnati and Jet Propulsion Laboratory and Jet Propulsion Laboratory and University of Wollongong
• Keywords: conditional simulation ; downscaling ; spatial mixed effects model

Abstract

A conditional simulation model is built to generate high spatial resolution data using large spatial outputs from a computer model. This proposed method is based on a spatial mixed effects (SME) model with parameters calibrated to coarse-scale computer model output. In particular, the spatial dependence present in the coarse-scale output is inherited in the SME model and thus in simulation. Moreover, the simulated values at high spatial resolution are generated through conditional simulation so that when aggregated they are consistent with the coarse-scale model output. In addition, the dimension is reduced in the SME model so that large spatial output can be dealt with, while the SME model is also able to incorporate nonstationary spatial dependence. We demonstrate our approach by producing downscaled high-resolution fields from output of atmospheric carbon dioxide (CO2) from the PCTM/GEOS-4 atmospheric model.