Spectroscopic study of the light-polluted night sky in Hong Kong

TL;DR

This study is the first comprehensive urban sky spectroscopy of light pollution, using low-cost spectrometers to collect over 12,000 zenith spectra in Hong Kong from 2021–2023 across urban and suburban sites. By defining spectral bands aligned with ALAN SPDs and applying continuum removal with trapezoidal band integration, the authors identify major contributors to skyglow, notably CFL and HPS, with expanding presence of LED and MH in urban centers, and a detectable Ne component from a neon-sign billboard. A decade-long comparison reveals a transition toward solid-state lighting and a decline in MH, reflecting evolving urban lighting practices and policy impacts. The work demonstrates the viability of accessible spectroscopy for monitoring light pollution and informs mitigation strategies such as shielding and earlier switching-off of nonessential external lighting.

Abstract

Spectroscopic study of the night sky has been a common way to assess the impacts of artificial light at night at remote astronomical observatories. However, the spectroscopic properties of the urban night sky remain poorly documented. We addressed this gap by collecting more than 12,000 zenith sky spectra with compact spectrometers at urban and suburban sites from 2021 to 2023. Here, by examining the intensity variations of the spectral features that represent characteristic emissions from common artificial light sources, we show that the skyglow is predominantly shaped by artificial emissions, including compact fluorescent lamps and high-pressure sodium lamps. Contributions from commercially controlled lighting, including those for floodlighting and advertising adopting light-emitting diode and metal halide technologies, were more pronounced in urban areas during the hours leading up to midnight. We also documented direct evidence of the impact of a neon sign located on top of a commercial tower, illustrating how a single light source can significantly influence the surrounding environment. Compared with observations made a decade ago at the same location, our findings indicate a growing popularity of light-emitting diode lighting for external use, consistent with the existing literature. This first comprehensive spectroscopic investigation of light pollution in an urban environment emphasizes the evolving patterns of outdoor lighting and highlights the critical and unique role of spectroscopic measurements. The results provide essential information for the development of effective strategies and policies to mitigate light pollution in urban areas and at sites of astronomical importance.

Figures (9)

• Figure 1: Selected SPDs and spectral bands for analysis. The dashed lines indicating the ranges of the chosen bands. The banding is based on the SPDs of common artificial light sources derived from field measurements and online databases. For SPDs displaying multiple representative emissions, such as HPS, MH, and CFL, the central wavelengths are labelled for clarity.
• Figure 2: The top three panels display averaged and normalized sunlight-free zenith sky spectra observed at various locations, illustrating the relative contributions of different light usages in comparison to HPS lighting. Major peaks are labelled to align with those identified in Fig. \ref{['fig:spectral_band']}. The sky spectra are binned to the nearest integer wavelength. S/N, as defined in stoehr:2008, are 11.4, 16.9 and 15.8 for Urban 1, Urban 2 and Suburban spectra respectively, for our bands between 440 and 645 nm. The bottom five panels, which replicate Fig. \ref{['fig:spectral_band']}, present the normalized SPD of common artificial light sources alongside the selected bands. This allows for a comparison between the spectral signatures of the night sky and those emitted by various artificial lighting sources.
• Figure 3: Average Urban 1's sky spectra for observations conducted before 23:00 (upper panel, S/N = 11.4) and after 00:00 (middle panel, S/N = 11.5). The bottom panel presents the residual (early-late) in which major peaks are labeled to align with those identified in Fig. \ref{['fig:spectral_band']}, signifying changes in the sky's spectral composition after midnight. Spectra are binned to the nearest integer wavelength.
• Figure 4: Histograms of $R$ in different spectral bands as observed at Urban 1 during moonlight-free period, quantifying nightly variations of ALAN spectral features. The vertical dashed lines represent $R=0$, i.e., same early evening and late night intensities.
• Figure 5: Urban 1's NSB light curves in the upper panel illustrate the sharp transitions in cloud coverage at midnight (indicated by the vertical dashed line) for the Clear→ Overcast and Overcast→ Cloudy Nights. These transitions highlight the significant changes in sky conditions and their subsequent effects on the darkening ratios $R$ (see Section \ref{['sec:result_exceptions']}). In contrast, the light curves for two Overcast Nights in the bottom panel show relatively little variation in cloud amounts after 23:00 (indicated by another vertical dashed line), enabling us to isolate the impact of artificial lighting from fluctuations in cloud conditions (see Section \ref{['sec:result_casestudies']}). The NSB measurements were conducted using a SQM-LE at HKU (Urban 1), as part of the Globe at Night - Sky Brightness Monitoring Network. These measurements are plotted on the same scale, allowing for easy comparison of relative changes in sky brightness across the different nights.