Online Program Home
My Program

Abstract Details

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
Abstract #306987
Title: Implementing Spatial Statistical Methods for Massive Data
Author(s): Dorit Hammerling* and Huang Huang and Lewis Blake
Companies: National Center for Atmospheric Research and National Center for Atmospheric Research and Colorado School of Mines
Keywords: Spatial Statistics; HPC; Massive data

With increasing amounts of data being produced (e.g., by remote sensing instruments and numerical models), the statistical and computational techniques to handle these data sizes of millions of observations have historically lagged behind. While a variety of statistical methods have been developed theoretically to tackle this problem, readily available computational implementations that work with irregularly-spaced observations are still rare. We introduce a set of computational implementations for the Multi-resolution Approximation (MRA), a recently developed spatial statistical method that lends itself particularly well to massive parallelization. The implementations range from having fairly simple parallelization strategies, targeting small computing units such as laptops, to sophisticated implementations in C++ with OpenMP and MPI leveraging high performance computing infrastructure. We show and compare results for millions of observations and discuss practical challenges that arise with such massive data sets.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program