Date of Award
2016
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Earth System Science
Committee Chair
Robert Griffin
Committee Member
Andrew C. Molthan
Committee Member
Udaysankar Nair
Subject(s)
Electric power failures, Visible infrared imaging radiometer suite (VIIRS), Remote sensing
Abstract
Severe meteorological events such as thunderstorms, tropical cyclones, and winter ice storms often produce prolonged, widespread power outages. Actions taken by disaster response agencies like the Federal Emergency Management Agency (FEMA), the American Red Cross, and the U.S. Department of Defense to provide support to communities during the recovery process need accurate and timely information on the extent and location(s) of power disruption, which is not often available. Observations made by the Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB), which provide daily, nighttime measurements of light sources, however, can be used to detect and monitor power outages caused by these meteorological events. Building on inherent nighttime capabilities, this study benefited from a new DNB product which incorporates additional VIIRS bands (as QA flags) and a lunar BRDF correction algorithm during post-processing. This information allows for the use of nighttime observations during all moon phases and the ability to identify image features which may impact analysis, such as clouds and snow. The frequency of clouds and snow over the continental U.S. (CONUS) was first assessed to identify locations with the highest likelihood of high quality observations throughout the year. Next, mean nighttime light variability was analyzed at varying spatial scales (e.g. individual pixels and entire city extents) over nearly four years of DNB observations. Results from these efforts were then applied to the Hurricane Sandy severe weather event of October 2012, which produced widespread and prolonged power outages. Replicating and expanding upon methods used by previous studies, qualitative, false color RGB composites as well as quantitative, percent of normal and percentile products were produced to estimate the degree of light reduction caused by this event. Furthermore, an image texture analysis extended previous efforts by introducing an alternative approach to power outage estimation. Finally, both quantitative outage products (percent of normal and percentile) were validated against state-based, U.S. Department of Energy (DoE) power outage reports to assess the accuracy of DNB-based outage estimations. Results showed general agreement to outage and recovery trends reported by the DoE, for state-wide, cloud-free conditions. Future work will refine methodologies and thresholds used in power outage detection. Furthermore, a pixel-level outage dataset will be sought to improve DNB-based outage estimates.
Recommended Citation
Cole, Tony, "Characterization of nighttime light variability for power outage detection" (2016). Theses. 175.
https://louis.uah.edu/uah-theses/175