Mapping Urban Nighttime Lights using EnMAP: Las Vegas and Riyadh

Kurzfassung

When observing artificial lights at night, these not only differ by intensity, but also exhibit characteristic spectral features in the spectral range visible to the human eye (VIS). There are additional narrow-band emissions from some sources of artificial light in the near infrared (NIR) and short-wave infrared (SWIR), as well as broad-band emissions from high-thermal radiations. These spectral features allow to identify lighting types and are usable, e.g., as indicators of urban growth and technological progress.

These signatures have been analyzed in the past using field and laboratory spectroscopy, whereas from space mapping lighting types is more challenging, because of the relatively coarse spatial resolution and low Signal-to-Noise-Ratios at low signals of imaging spectrometers. Hence, many established spectral identification and classification methods provide unsatisfactory results. Nevertheless, this contribution illustrates, based on EnMAP products, how they are usable to identify and map artificial nighttime lights, using Las Vegas and Riyadh as test sites. Well-established spectral unmixing algorithms are applied, and allow recognizing different light sources at sub-pixel level, although neither the EnMAP sensor nor processors are designed for such low signals.

First, the spectrally extreme, and thus potentially pure spectral signatures, are detected using Sequential Maximum Angle Convex Cone and Spatial-Spectral Endmember Extraction. Second, to identify these endmembers (EM), a spectral signature mapping is applied using the known emission lines based on laboratory measurements.

Based on the EMs, a large variety of light sources are detectable, with spectrally pure EM signatures related to various broad- and narrow-band emissions from LED, High-Pressure-Sodium, and Xenon lamps, but also high-temperature events like fires. Using spectral unmixing algorithms, mapping of the lighting types was possible, although the algorithms are not applied to estimate the fractional surface occupied by pure materials. Thus, the methods do not allow imposing certain constraints (e.g., sum-to-one), here, and other constraints are not considered (e.g., atmosphere, reflections). Consequently, abundances are to be considered relative and not absolute. As the spectral signatures of the different lights are diagnostic, the transferability of the approach to other images with high illumination levels are achievable, as illustrated by the results from Las Vegas and Riyadh.