Resumen: In this Master Thesis, we propose a physically-based method to render the atmosphere of Earth in real time from any arbitrary viewpoint. Atmosphere rendering is a difficult problem to solve due to its inherent complexity. The physics equations that describe the scattering of light feature multiple nested integrals that must be computed numerically or approximated. Moreover, the atmospheric medium is strongly dependent on the wavelength of incoming light, which makes spectral rendering a requirement to faithfully represent the colors of sky. Our main contribution relies on an approximation to spectral rendering, where instead of computing dozens of spectral samples, only four carefully chosen samples are used to evaluate the final color. This makes our method exceptionally cheap and accurate, as it is able to produce images that are perceptually indistinguishable from those generated by a fully-fledged spectral renderer, at a fraction of the computational cost. We also propose an approximation to multiple light scattering that significantly reduces the computational cost of evaluating the light transport equations. We tackle the issue of developing an atmospheric model that is highly configurable and able to represent a wide variety of atmospheric conditions, while still being physically-based and suitable for real-time applications. Finally, we implement our method on Shadertoy to prove its high portability and ease of implementation, and we demonstrate how the results compare to a ground truth path tracing simulation.