Principal Investigator (PI): David Bekaert, NASA’s Jet Propulsion Laboratory

Interferometric Synthetic Aperture Radar (InSAR) is a precise geodetic technique that has revolutionized how we study our dynamic planet by providing deformation observations critical for improving our understanding of the solid Earth (tectonic and volcanic) and of hydrological, cryospheric, and vertical land motion processes. Although InSAR is identified in the 2017 Decadal Survey as the presumed measurement technique for the surface deformation and change targeted observable, it remains a greatly underutilized capability, as high-level data production remains a specialized task both in processing expertise and the necessary computational infrastructure.

Objectives and Implementation

The SAR community is not yet accustomed to the idea of processing data within the cloud. The existing processing algorithms are not well suited to data access handling and are designed to run on local computers, requiring the physical download of the SAR data. The volume of data moved outside NASA's Distributed Active Archive Centers’ (DAACs) AWS S3 storage — egress volume — is typically around 500–700 TB/month with December 2019 reaching a peak of 1.3 PB. Unless processing is moved alongside the data, processing capacity, download bottlenecks, and egress costs will only become greater upon the launch of the NASA-Indian Space Research Organization SAR (NISAR) mission.

This project addresses the primary challenges of performing SAR analysis in the cloud by

  • Improving DAAC utilities for data discovery, in-place access to discovered data (without requiring data movement), and pre-processing of identified data in the cloud. These utilities include a Python API, tools for the identification and extraction of individual tiles from Sentinel-1 SAR data, mosaicking functionality for large-scale processing efforts, and updated tools for in-place data access.
  • Updating the data access and handling in open-source SAR software tools (ISCE, ARIA-tools, MintPy) to utilize virtual and direct NumPy access removing the need for data movement when processing in AWS.
  • Preparing the community for processing in the cloud through interactive Jupyter notebook tutorials and by providing training workshops to ensure dissemination of our cloud-enabled processing.

In addition, the project will demonstrate the scientific value of our developed technology along with AWS scalability by processing in the cloud higher-level InSAR products over extensive areas along the U.S. west coast to support science use cases for the tectonic, volcanology, and hydrological communities.

Significance and Data Growth Outlook

Over the last years the volume of SAR has grown to 9 PB, predominantly from the recent Sentinel-1 satellite with 7 PB of data. This volume is expected to grow at an unprecedented rate of approximately 86 TB per day with the launch of NISAR in May 2022. Currently, only a few attempts have been made in the SAR community to move in this direction.

Last Updated
Sep 24, 2020