||The emission spectrum of a light-pollution source is a determining factor for modelling artificial light at night. The spectral composition of skyglow is normally derived from the initial spectra of all artificial light sources contributing to the diffuse illumination of an observation point. However, light scattering in the ambient atmosphere imposes a wavelength-specific distortion on the optical signals captured by the measuring device. The nature of the emission, the spectra and the light-scattering phenomena not only control the spectral properties of the ground-reaching radiation, but also provide a unique tool for remote diagnosis and even identification of the emission spectra of the light-polluting sources. This is because the information contained in the night-sky brightness is preferably measured in directions towards a glowing dome of light over the artificial source of light. We have developed a new method for obtaining the emission spectra using remote terrestrial sensing of the bright patches of sky associated with a source. Field experiments conducted in Vienna and Bratislava have been used to validate the theoretical model and the retrieval method. These experiments demonstrate that the numerical inversion is successful even if the signal-to-noise ratio is small. The method for decoding the emission spectra by the light-scattering spectrometry of a night sky is a unique approach that enables for (i) a systematic characterization of the light-pollution sources over a specific territory, and (ii) a significant improvement in the numerical prediction of skyglow changes that we can expect at observatories.