High-Performance Time-Series Quantitative Retrieval From Satellite Images on a GPU Cluster

The quality and accuracy of remote sensing instruments continue to increase, allowing geoscientists to perform various quantitative retrieval applications to observe the geophysical variables of land, atmosphere, ocean, etc. The explosive growth of time-series remote sensing (RS) data over large-scales poses great challenges on managing, processing, and interpreting RS ``Big Data.'' To explore these time-series RS data efficiently, in this paper, we design and implement a high-performance framework to address the time-consuming time-series quantitative retrieval issue on a graphics processing unit cluster, taking the aerosol optical depth (AOD) retrieval from satellite images as a study case. The presented framework exploits the multilevel parallelism for time-series quantitative RS retrieval to promote efficiency. At the coarse-grained level of parallelism, the AOD time-series retrieval is represented as multidirected acyclic graph workflows and scheduled based on a list-based heuristic algorithm, heterogeneous earliest finish time, taking the idle slot and priorities of retrieval jobs into account. At the fine-grained level, the parallel strategies for the major remote sensing image processing algorithms divided into three categories, i.e., the point or pixel-based operations, the local operations, and the global or irregular operations have been summarized. The parallel framework was implemented with message passing interface and compute unified device architecture, and experimental results with the AOD retrieval case verify the effectiveness of the presented framework.

I. Introduction

In recent decades, the volume of Earth observation data has increased rapidly. The datacenters of the China National Satellite Meteorological Center (NSMC) have archived 4.126 PBs of data, and the China center for Resources Satellite Data and Application (CCRSDA) achieved more than 16 million scenes of remote sensing (RS) images up to August 2017 as reported [1]. As a typical example, data from the moderate resolution imaging spectroradiometer (MODIS) instruments onboard the satellite TERRA and AQUA have been used to study the properties of land, atmosphere, and ocean widely since been launched in December 1999 and May 2002, respectively. Each MODIS sensor produces 70 GB of raw data per day [2], from which large amount of higher level products have been generated by national aeronautics and space administration's MODIS adaptive processing system (NASA's MODAPS). Great efforts have been made to develop quantitative remote sensing models to estimate various geophysical parameters such as normalized difference vegetation index (NDVI) and aerosol optical depth (AOD), and link these massive data to different aspects of dynamic earth [3], [4]. Time-series RS data covering large areas have been extensively used to monitor the temporal and spatial changes and patterns of Earth [5], [6]. For instance, MODIS AOD dataset are used to portray the global, regional, and seasonal distribution of the AOD [7], [8]. However, the large amount of archived RS data and the relatively poor performance of compute-intensive algorithms can reduce the data utilization and delay the response [9], [10], which raise the need for effective processing methods of RS data [11].