Matern-Based Nonstationary Cross-Covariance Models for Global Processes

• Activity Number: 445
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics and the Environment
• Author(s): Mikyoung Jun*+
• Companies: Texas A&M
• Keywords: Climate model output ; Cross-covariance model ; Global processes ; Matern covariance function ; Nonstationary process

Abstract

Many spatial processes in environmental applications, such as climate variables and climate model errors on a global scale, exhibit complex nonstationary dependence structure, in not only their marginal covariance but also their cross-covariance. Flexible cross-covariance models for processes on a global scale are critical for an accurate description of each spatial process as well as the cross-dependences between them and also for improved predictions. We propose various ways to produce cross-covariance models, based on the Mat´ern covariance model class, that are suitable for describing prominent nonstationary characteristics of the global processes. In particular, we seek nonstationary versions of Mat´ern covariance models whose smoothness parameters vary over space, coupled with a differential operators approach for modeling large-scale nonstationarity. We compare their performance to the performance of some existing models in terms of the aic and spatial predictions in two applications problems: joint modeling of surface temperature and precipitation, and joint modeling of errors in climate model ensembles.