Spatiotemporal Characterization of VIIRS Night Light

The Visible Infrared Imaging Radiometer Suite (VIIRS) Day Night Band (DNB) on board the Suomi NPP satellite now provides almost a decade of daily observations night light. temporal frequency sampling, without degree averaging annual composites, makes it necessary to consider distinction between apparent changes light related imaging process and actual in underlying sources being imaged. most common approach change detection involves direct attribution observed phenomenon interest. Implicit this is assumption that other forms source are negligible or non-existent. An alternative characterize spatiotemporal variability prior deductive causation so can be made context full range spatial variation. primary objective study over scales provide for interpretation both subannual interannual time scales. This analysis based combination moments, correlation Empirical Orthogonal Function (EOF) analysis. A key result pervasive heteroskedasticity VIIRS monthly mean Specifically, monotonic decrease with increasing brightness. Anthropogenic remarkably stable while background luminance varies considerably. variance partition from eigenvalues covariance matrix 88, 2 2% spatial, seasonal (respectively) diverse region Earth (Eurasia). Heteroskedasticity composites; present all areas months year, suggesting much, if not most, month-to-month may otherwise subjected multiple aspects varying time. Given skewed distribution arising radial peripheral dimming bright subject atmospheric scattering, even aggregate metrics using thresholds must interpreted fact much larger numbers more variable low pixels statistically overwhelm smaller higher cause as sources.