Physically-based simulation of twilight phenomena

Jörg Haber, Marcus A. Magnor, Hans-Peter SeidelACM Trans. Graphics 2005 · DOI

TL;DR

This paper presents a physically-based rendering approach to simulate the complex colors of the sky during twilight. By incorporating detailed atmospheric models and radiative transfer simulations, the authors achieve photo-realistic twilight sky visualizations that are computationally efficient enough for practical use.

What Problem Does This Paper Address?

Rendering realistic outdoor scenes, especially during the dynamic and visually rich twilight period, has been a long-standing challenge. While daytime skies can be approximated with parametric models or photographs, twilight presents unique difficulties due to its rapidly changing colors, wide dynamic range, and the subtle interplay of light scattering and atmospheric composition. Previous approaches often relied on simplified physical models or empirical approximations, failing to capture the full spectrum of twilight phenomena accurately. This paper aims to bridge this gap by providing a physically grounded method that synthesizes twilight skies based on actual atmospheric conditions.

Methods

The core of the paper is a physically-based simulation that models light scattering and transfer within a detailed atmospheric model. Key components include:

  • Atmosphere Model: A layered atmosphere up to 35 km, with each layer characterized by optical properties (aerosol scattering/extinction, air scattering, refractive index) that vary with altitude and climate. These properties are derived using the OPAC (Optical Properties of Aerosols and Clouds) software.
  • Radiative Transfer Simulation: The simulation solves the 3D rendering equation by considering both direct sunlight and indirect, multiply scattered light.
  • Direct Illumination: Accounts for solar irradiance, ozone absorption, and atmospheric refraction.
  • Multiple Scattering: This is the most computationally intensive part. The authors employ optimizations to reduce computation time, including exploiting symmetries and subsampling shells. They differentiate between the first scattering event (anisotropic for aerosols) and subsequent scattering events (approximated as isotropic).
  • Sky Texture Generation: The computed spectral radiance is converted into an RGB "sky texture" or environment map, suitable for rendering in 3D scenes.
  • Optimization Techniques: To achieve interactive frame rates (under two hours on a conventional PC), the authors introduce several algorithmic optimizations, such as exploiting symmetries in the scattering process and subsampling atmospheric shells during multiple scattering passes.

Datasets

The paper utilizes the publicly available OPAC (Optical Properties of Aerosols and Clouds) software package to obtain atmospheric optical properties for various climate conditions, aerosol compositions, and humidity levels. They also reference spectral solar irradiance data from Kurucz et al. [1984]. No new datasets are introduced.

Key Insights and Ideas

  • Twilight's Complexity: The paper highlights that twilight colors are not just a simple gradient but arise from a complex interplay of Rayleigh scattering (air molecules), Mie scattering (aerosols), multiple scattering events, and atmospheric refraction.
  • Importance of Multiple Scattering: Figure 7 visually demonstrates that single scattering alone is insufficient to capture the richness of twilight; multiple scattering is crucial for the characteristic halo and color gradients.
  • Humidity's Impact: Figure 8 shows that humidity significantly influences twilight colors, particularly by affecting the size and hygroscopic nature of aerosol particles, leading to more pronounced orange-red hues.
  • Climate Variation: Figure 9 illustrates how different climate conditions (continental, urban, maritime) produce distinct twilight appearances due to variations in aerosol composition and height distribution.
  • Computational Efficiency: The authors successfully balance physical accuracy with computational feasibility through clever algorithmic optimizations, making their approach practical.

Key Results

  • The system can generate photo-realistic hemispherical twilight sky textures in less than two hours on a conventional PC (3GHz Intel Pentium4).
  • The rendered sky textures can be used for real-time rendering once generated.
  • The simulation accurately replicates the influence of sun position, climate conditions, and humidity on twilight colors.
  • Validation against actual photographs (Figure 12) shows a strong qualitative agreement, particularly in the characteristic orange-red band along the horizon.

Limitations and Future Work

  • Terrain and Clouds: The current system does not account for terrain or clouds, which can significantly affect the visual appearance of a sunset. The authors suggest that shadowing effects from terrain and clouds could be incorporated.
  • Mesopic Vision: While the system generates realistic sky radiance, the authors note that the psycho-physics of mesopic vision (the intermediate range between photopic and scotopic vision) could be further investigated using their simulations, potentially leading to more perceptually accurate tone mapping operators.
  • Automatic Parameter Determination: The paper suggests that future work could focus on developing methods to automatically determine atmospheric model parameters by comparing rendered images to actual twilight recordings, which could also serve as a technique for atmospheric pollution probing.
  • Cloud Appearance: The authors propose that cloud appearance could be realistically modeled by illuminating them in 3D using the system's atmosphere cells' radiant power distribution.

Summary updated

BibTeX
@article{haber2005physically,
  author = {Jörg Haber and Marcus A. Magnor and Hans-Peter Seidel},
  title = {{Physically-based simulation of twilight phenomena}},
  journal = {ACM Trans. Graphics},
  year = {2005},
  doi = {10.1145/1095878.1095884},
}
Figure