Azimuthal visualization of Artificial Light at Night (ALAN) based on 3D Gaussian Splatting

Abstract

Modeling nocturnal light environments is challenging due to the heterogeneous and strongly directional nature of artificial light emission. In this presentation, we propose a method for describing nocturnal lighting conditions based on 3D Gaussian Splatting, which enables individual light sources to be represented as continuous Gaussian primitives with parameters describing intensity, RGB color, spatial extent, and emission anisotropy. The resulting set of Gaussian splats provides a smooth and scalable representation of spatial light distribution without requiring full geometric reconstruction of the scene. In the second stage, the 3D model is projected into an azimuthal brightness profile, visualized as a circular representation corresponding to the observer’s horizon. Each viewing angle reflects the cumulative perceived light from a given direction, incorporating both luminance and RGB color distribution, thereby forming a compact and intuitive depiction of the nocturnal light signature. This approach enables identification of dominant emission directions, comparison of different lighting configurations, and delineation of critical azimuthal sectors characterized by elevated light exposure. The present study was tested in a rural area, as this setting captures a wide range of artificial light emission intensities, and the results demonstrated suitability for application across diverse contexts. By using this method, it is possible to delineate areas of influence of light pollution to support decision making from a socio environmental perspective, enabling illumination only where it is necessary and with minimal associated impact. By balancing lighting requirements with the maintenance of natural environmental conditions, the model provides critical information to avoid or mitigate documented ecological impacts, including altered animal behavior and movement, interference with reproduction and foraging, increased predation risk, as well as shifts in species interactions and community structure. The proposed framework bridges modern scene-representation techniques with environmental analysis, offering a practical and transparent tool for assessing nocturnal light environments.