Nonstationary Spatial Data: Think Globally Act Locally

• Activity Number: 331 - Advances in the Analysis of Massive Space-Time Data Sets Using High Performance Computing
• Type: Topic Contributed
• Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Douglas William Nychka*
• Companies: NCAR
• Keywords: Spatial fields

Abstract

Large spatial data sets are now ubiquitous in environmental science. Fine spatial sampling or many observations across large domains provides a wealth of information and can often address new scientific questions. However, the richness and scale of large datasets often reveal heterogeneity in spatial processes that add more complexity to a statistical analysis. A strategy for handling larger problems is to rely on separate local analyses of the data but with a view to combine the results into a seamless global model.In this talk two examples are presented for handling the simulation and uncertainty quantification of non-stationary Gaussian processes. The global model in this case is a process convolution of a white noise field where the convolution function varies across space. Such a model is difficult to implement explicitly for large spatial fields.