Project Team: Global Disasters Team
Team Location: Jet Propulsion Laboratory, Pasadena, California

Radar-derived surface change pixels were overlain on building footprints, road network, and a regional base layer of Tacloban City, Philippines, to determine damaged structures after Typhoon Haiyan made landfall in November 2013.

Authors:
Judy Cheng (University of California, Los Angeles)
Sara Lafia (California State Polytechnic University, Pomona)
Rodell Allan Zorilla (University of California, Los Angeles)

Mentors/Advisors:
Benjamin Holt (DEVELOP Mentor, Jet Propulsion Laboratory)
Dr. Sang-Ho Yun (Processing System Lead of ARIA, Jet Propulsion Laboratory/California Institute of Technology)

Abstract:

Year after year, natural disasters such as earthquakes, typhoons, and floods claim the lives of many people and cause economic losses that can total in the billions of dollars. In the aftermath of such events, an accurate and comprehensive assessment of damage is needed for rapid rescue response to minimize loss of life and begin the recovery process.

NASA’s Jet Propulsion Laboratory (JPL) and the California Institute of Technology (Caltech) developed a prototype damage-detection algorithm that uses Interferometric Synthetic Aperture Radar (InSAR) coherence change to produce damage proxy maps (DPM) of disaster-affected regions. These DPMs indicate areas that have undergone changes in surface conditions. Once the initial DPM is produced, it is handed off to responding agencies. However, false positives such as vegetation or anthropogenic changes are detected and included in the initial DPM; therefore, it is necessary to validate true positives, which show structural damages.

Currently, the validation of DPMs is a time-consuming process that can take days to complete. This project developed tools and a Web application to expedite the validation process, reducing the latency of product delivery to hours. A masking tool, automated using ArcGIS models and Python scripts, removed false-positive pixels by extracting DPM polygons that intersect building footprints. If building footprints were not available, land classifications derived from a supervised classification were applied instead to extract DPM polygons that intersected areas with buildings. The masked DPM was then uploaded to the online Web map-validation application where it was layered with pre- and post-event optical imagery, allowing volunteers to quickly validate a DPM with a click of a button.

Back to VPS page.