A Distributed Implementation of the Multi-resolution Approximation for Very Large Spatial Data
With data of rapidly increasing sizes in the environmental and geosciences such as satellite observations and high-resolution climate model runs, the spatial statistics community has recently focused on methods that are applicable to very large data. One such state-of-the-art method is the multi-resolution approximation (MRA), which was specifically developed with high performance computer architecture in mind.
Overview
Abstract
With data of rapidly increasing sizes in the environmental and geosciences such as satellite observations and high-resolution climate model runs, the spatial statistics community has recently focused on methods that are applicable to very large data. One such state-of-the-art method is the multi-resolution approximation (MRA), which was specifically developed with high performance computer architecture in mind. Attractive features of the MRA are the ability to capture variability from very large to very small scales and to approximate any covariance function. We will present an implementation of the MRA method on the National Center for Atmospheric Research’s super computer and demonstrate how this method can be applied to geophysical data sets of tens of millions of observations. This is joint work with Matthias Katzfuss.