A decision support system for scientists by processing large-scale satellite images on a distributed computing environment

Data-driven research strategy has been widely accepted in many scientific areas. In the strategy, which consists of several processing steps including data collection, processing, analysis and discovery, data processing step is so time-consuming that it delays the overall time of scientific discovery. This also happens in the field of satellite remote sensing. As SeaDAS, a popular software package used for processing satellite remote sensing images, has been initially devised to work on a single site; it is inevitable to take unendurable time for processing massive remote sensing images. To rectify this issue, this article introduces a data-driven analysis system that rapidly processes a massive volume satellite remote sensing images. Unlike the conventional program for remote sensing images, our system is developed by virtue of both a distributed array-based DBMS and a high-throughput computing facility in order to process massive remote sensing images in a parallel and distributed manner. Consequently, it allows scientists to perform their analysis workloads more quickly and efficiently. Through extensive experiments performed with real-world dataset on 10 computing nodes, we show that our system is up to 27.5 times faster than the conventional SeaDAS package.