Scaling constituent algorithms of a trend and change detection polyalgorithm

Layne T. Watson

Abstract

Earth observation satellites (EOS) such as Landsat provide image datasets that can be immensely useful in numerous application domains, by extracting information via time series analysis. While the literature is replete with algorithms, the size of the datasets itself is prohibitive, currently of the order of petabytes and growing, which makes them computationally unwieldy --- both in storage and processing. An EOS image stack typically consists of multiple images of a fixed area on the Earth's surface (same latitudes and longitudes) taken at different time points. Meaningful time series analysis on one such interannual, multitemporal stack with existing state of the art codes can take several days on multicore servers. This work lays the foundation for a polyalgorithm based on two change detection algorithms, EWMACD and BFAST, for time series analysis of satellite image stacks, and presents speedup results for those two algorithms.

Rishu Saxena, Layne T. Watson, Valerie A. Thomas, Randolph H. Wynne:
Scaling constituent algorithms of a trend and change detection polyalgorithm. SpringSim (HPC) 2017: 6:1-6:12