SMART-G, a Monte Carlo GPU Radiative Transfer code

SMART-G (Speed-Up Monte Carlo Advanced Radiative Transfer code with GPU) is a radiative transfer solver. It was initially co-developed by HYGEOS for CNES to simulate the radiation received by a satellite sensor after the exact and precise propagation of sun light in the coupled atmosphere-ocean system. SMART-G1 is now continuously upgraded by HYGEOS.

Now considered like a reference, SMART-G is used for feasibility studies of future spatial missions, to verify common assumptions in operational processing chains and, above all, it is the heart of the SolaRes® service, providing near real time and historical archives of solar resource estimate.

Contact at HYGEOS: Didier Ramon

(1) Ramon, D., F. Steinmetz, D. Jolivet, M. Compiègne and R. Frouin, “Modeling polarized radiative transfer in the ocean-atmosphere system with the GPU-accelerated SMART-G Monte Carlo code”, Journal of Quantitative Spectroscopy and Radiative Transfer, 222–223, p89-107 (2019) https://doi.org/10.1016/j.jqsrt.2018.10.017

SMART-G features

  • Coupled Ocean/Atmosphere Monte Carlo Radiative Transfer solver
    • GPU accelerated using the CUDA framework (massively parallel processing)
    • Coded in Python and PyCUDA
  • Physical processes
    • scattering (molecules, water, aerosols, clouds)
    • absorption (molecules, water, aerosols) and with extended band models (K distribution, …)
    • reflection/refraction (Fresnel, atmospheric refraction)
    • polarization
  • Plane-parallel or spherical-shell atmosphere
  • Surface
    • wind roughned sea surface
    • bidirectional reflectance distribution function
    • horizontal variation of surface albedo
  • Fast spectral computation
  • 3-Dimensional characteristics
    • 3D optical properties of atmosphere
    • 3D objects

Applications

Simulation of radiance in limb geometry with atmospheric refraction

Vertical profiles of limb diffuse radiance as a function of the tangent height for different Relative Azimuth to the Sun, wavelengths and Solar Zenith Angles.
RGB composite of the limb radiance for a low Sun in the forward directions

Limb diffuse radiance for an observer at 120km altitude as a function of the tangent height for different relative azimuths to the Sun (ΔΦ = 0, 30, 60, 90°), 3 wavelengths (430, 660 and 840 nm) and 4 solar zenith angles (SZA = 60, 80, 91 and 94°).

The computation time with 105 photons/height/azimuth is about 1s with a GeForce RTX3090 graphics card.

 

RGB composite (430nm, 660nm, 840nm) of the limb radiance for a low Sun (SZA=94°) in the forward directions up to ΔΦ = 20°.

Simulation of topographic and 3D objects influence on radiance

Radiometric calibration of in situ spectro-radiometer at the top of a mountain. The radiometer is viewing a horizontal plate of known bidirectional reflectance with the Sun disk unobscured and masked, allowing a determination of the direct component of downwelling planar irradiance.

The SMART-G simulation is done with a proper k-distribution at the same spectral resolution as the instrument.

Radiance by a radiometer and simulated by SMART-G

Simulation of 2D adjacency effects

Scheme of the adjacency effects
Simulation of TOA spectra of S3/OLCI sensor for a black ocean pixel as a function of the distance to the coastline

Simulation of TOA spectra of the Sentinel-3/OLCI sensor for a black ocean pixel as a function of the distance d to the linear coastline separating ocean and land covered by snow and ice. The principal plane is parallel to the coastline.

Simulation of 3D clouds

Surface reflectances simulated by SMART-G at 710 nm
TOA reflectances simulated by SMART-G at 710 nm for a cloudy scene.

Maps of surface (left) and TOA (right) reflectances at 710 nm for a cloudy scene, simulated using Large Eddy Simulation microphysical outputs and for a continental aerosol polluted atmosphere. The horizontal resolution is 66 m and there are 53 vertical layers for a total of atmospheric cell of 100x100x53. The surface is a lambertian reflector whose spectral albedo is obtained from Sentinel-2 Land Cover map.