Fast Fourier-transform calculation of artificial night sky brightness maps

TL;DR

This paper introduces a fast Fourier-transform based method to efficiently compute detailed global night sky brightness maps from satellite data, significantly reducing computation times for light pollution monitoring.

Contribution

It demonstrates that light pollution map calculations can be reformulated as 2D convolutions evaluated with FFT, enabling rapid processing of large-scale satellite data.

Findings

Computation times per pixel are reduced to less than 10^-6 seconds.

The method applies to various light pollution functions, including zenithal brightness.

Efficient global night sky brightness mapping becomes feasible with this approach.

Abstract

Light pollution poses a growing threat to optical astronomy, in addition to its detrimental impacts on the natural environment, the intangible heritage of humankind related to the contemplation of the starry sky and, potentially, on human health. The computation of maps showing the spatial distribution of several light pollution related functions (e.g. the anthropogenic zenithal night sky brightness, or the average brightness of the celestial hemisphere) is a key tool for light pollution monitoring and control, providing the scientific rationale for the adoption of informed decisions on public lighting and astronomical site preservation. The calculation of such maps from satellite radiance data for wide regions of the planet with sub-kilometric spatial resolution often implies a huge amount of basic pixel operations, requiring in many cases extremely large computation times. In this paper we show that, using adequate geographical projections, a wide set of light pollution map calculations can be reframed in terms of two-dimensional convolutions that can be easily evaluated using conventional fast Fourier-transform (FFT) algorithms, with typical computation times smaller than 10^-6 s per output pixel.