ABSTRACT

The Black Marble Product Suite, derived from the Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB) onboard Suomi-NPP and NOAA-20, capture nocturnal lights over visible and infrared bands, enabling detection of a wide variety of natural and human-driven phenomena at night. However, extracting these events from the decade-long record can be challenging due large spatial and temporal variations in the signals. This talk will highlight current efforts in tailoring machine learning approaches to leverage large volumes of data for extracting these nighttime signals of interest. These include multispectral anomaly detectors to extract nighttime thermal anomalies and nighttime light time-series analysis to monitor changes in urban areas. For both applications, we train our models from the naturally occurring variations in the data and demonstrate the use of machine learning-based data labeling, that minimizes the annotation effort from domain experts. We will discuss future directions to study a wider class of nighttime events using Black Marble and on the generalizability of the approaches to other remote sensing datasets.

About the Speaker:

Srija Chakraborty is an Associate Scientist at the Earth from Space Institute (EfSI), Universities Space Research Association and focuses on tailoring machine learning methods for NASA's Black Marble Product Suite for studying the Earth at night. She volunteers on the NASA Science Mission Directorate Artificial Intelligence Machine Learning working group for Earth Science and is a technical co-lead in IEEE Geoscience and Remote Sensing Society's Image Analysis and Data Fusion working group. Prior to joining EfSI, USRA, she was a NASA Postdoctoral Program Fellow at Goddard Space Flight Center with the Black Marble Science Team studying machine learning approaches for nighttime remote sensing. She received a Ph.D. in Computer Engineering from Arizona State University in 2019, studying machine learning and statistical signal processing algorithms for analyzing remotely sensed Earth and Planetary observations. Her research interests lie at the intersection of machine learning and remote sensing, with an emphasis on unsupervised learning, time-series analysis, anomaly detection, explainable machine learning, and onboard data analysis methodologies.