Summary

The atmospheric neutrino fluxes are calculated in the 3D-scheme with a huge statistics.
For the neutrino oscillation analysis, thr `3D feature' is not important. When they are averaged over azimuth directions, 3D calculation agree with 1D results well above a few GeV. 
(Rather absolute normalization at ~100 GeV is crucial.)

From a theoretical interest, the statistics is poor yet for the horizontal directions. We will simulate 100,000,000,000 cosmic ray, or equivalently 0.1 micro-second, for a better  statistics.