3D Gaussian Solar Radiative Transfer Comparison

Monte Carlo: Nphot=1E6 Nscat=25 SHDOM: no cell splitting, solacc=1E-5 Mean: Flux up: 0.0350 Flux down: 0.9616 Rad1: 0.0064 RMS Abs Diff/Mean RMS frac diff Nx Ny Nz Nmu Nphi Fup Fdn Rad1 Rad1 10 10 6 8 16 .0313 .0071 .0212 .0247 20 20 11 8 16 .0265 .0042 .0217 .0083 40 40 21 8 16 .0249 .0032 .0254 .0100 10 10 6 16 32 .0326 .0068 .0042 .0249 20 20 11 16 32 .0115 .0038 .0065 .0027 40 40 21 16 32 .0068 .0026 .0099 .0032 20 20 11 2 4 .4220 .0544 .0991 .0380 20 20 11 4 8 .1398 .0210 .0132 .0094 20 20 11 6 12 .0765 .0088 .0252 .0112 20 20 11 8 16 .0265 .0042 .0217 .0083 20 20 11 12 24 .0109 .0038 .0121 .0041 20 20 11 16 32 .0115 .0038 .0065 .0027 Comparisons are made for 60 locations (half plane) on the top or bottom boundaries. The primary measure of accuracy is the rms difference between Monte Carlo and SHDOM results divided by the mean over the 60 locations. Fup is upwelling flux, Fdn is downwelling flux, and Rad1 is nadir upwelling radiance. There is a general trend toward increasing accuracy as the angular and spatial resolution of SHDOM is increased. The Monte Carlo results appear to be accurate only to 0.5% to 1.0% for the solar case (based on the results for 3E5 photons). The Nx=20 Nmu=8 case took 7 iterations and under 1 minute CPU time (on an HP 715/75).

3D Gaussian Thermal Radiative Transfer Comparison

Monte Carlo: Nphot=1E5 Nscat=10 SHDOM: no cell splitting, solacc=1E-5 Mean: Flux up: 24.94 Flux down: 4.24 Rad1: 8.40 RMS Abs Diff/Mean Nx Ny Nz Nmu Nphi Fup Fdn Rad1 10 10 6 8 16 .0085 .0357 .0021 20 20 11 8 16 .0087 .0351 .0027 40 40 21 8 16 .0081 .0395 .0029 10 10 6 16 32 .0021 .0122 .0007 20 20 11 16 32 .0024 .0075 .0002 40 40 21 16 32 .0020 .0103 .0001 20 20 11 2 4 .1691 .2553 .0387 20 20 11 4 8 .0408 .0982 .0173 20 20 11 6 12 .0168 .0653 .0077 20 20 11 8 16 .0087 .0351 .0027 20 20 11 12 24 .0038 .0160 .0003 20 20 11 16 32 .0024 .0075 .0002