Detecting Exceedance Regions for Spatial and Spatio-Temporal Data

• Activity Number: 477
• Type: Invited
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Joshua French*+ and Stephan R. Sain
• Companies: University of Colorado Denver and NCAR
• Keywords: spatial data ; spatio-temporal data ; hotspots ; exceedance regions ; excursion sets ; large data sets

Abstract

The exceedance region of a spatial or spatio-temporal process is the set of locations where the response exceeds some threshold of interest. Identification of these regions is particularly important in environmental and health applications, as these regions are frequently the ones most impacted by an environmental event or where there is greatest concern for public health. A new method for confidently identifying exceedance regions with known confidence will be presented. The method utilizes conventional geostatistical tools but combines them in a novel way to obtain the desired confidence regions. Both theoretical and practical aspects of this method will be discussed. Extra care must be taken when applying this method to very large spatial and spatio-temporal data sets. Additional discussion will focus on how this methodology can be extended with flexible spatial models suitable for inference using very large spatial and spatio-temporal data sets. These methods will be applied in assessing the scope of impact for recent unusual climate events in the United States.