Exploring VIIRS Night Light Long-Term Time Series with CNN/SI for Urban Change Detection and Aerosol Monitoring

Abstract

There is a great need to study the decadal long-term time series of urban night-light changes since the launch of Suomi NPP, NOAA-20, to future JPSS-2, 3, and 4 in the next decades. The recently recalibrated and reprocessed Suomi NPP VIIRS/DNB dataset overcomes a number of limitations in the operational data stream for time series studies. However, new methodologies are desirable to explore the large volume of historical data to reveal long-term socio-economic and environmental changes. In this study, we introduce a novel algorithm using convolutional neural network similarity index (CNN/SI) to rapidly and automatically identify cloud-free observations for selected cities. The derived decadal clear sky mean radiance time series allows us to study the urban night light changes over a long period of time. Our results show that the radiometric changes for some metropolitan areas changed on the order of 29% in the past decade, while others had no appreciable change. The strong seasonal variation in the mean radiance appears to be highly correlated with seasonal aerosol optical thickness. This study will facilitate the use of recalibrated/reprocessed data, and improve our understanding of urban night light changes due to geophysical, climatological, and socio-economic factors.