We examine the uncertainty of the calculation of the atmospheric neutrino flux and present a way to reduce it using accurately measured atmospheric muon flux. Considering the difference of the hadronic interaction model and the real one as a variation of hadronic interaction, we find a quantitative estimation method for the error of the atmospheric neutrino flux calculation from the residual of the reconstruction of the atmospheric muon flux observed in a precision experiment, by the study of atmospheric neutrino and muon fluxes response to the variation of hadronic interaction. However, the efficiencty of this method is largely dependent on the observation site of the atmospheric muon flux, as the relation of the error of the atmospheric neutrino flux calculation and the residual of the reconstruction of the atmospheric muon flux is also largely dependent on the muon observation site, especially for the low energy neutrinos. We calculate several observation sites, near Kamioka at sea level, same but 2770m a.s.l.., Hanle India (4500m a.s.l.), and at Balloon altitude ($\sim$ 32km). Then we estimate how stringently can the atmospheric muon reduce the error in the calculation of the atmospheric neutrino flux. We also discuss on the source of error which is difficult to reduce by only the observation of atmospheric muon.

In this paper, we extend the calculation of the atmospheric neutrino flux~\cite{hkkm2004,hkkms2006,hkkm2011} to the sites in polar and tropical regions. In our earliest full 3D-calculation~\cite{hkkm2004}, we used DPMJET-III~\cite{dpm} for the hadronic interaction model above 5~GeV, and NUCRIN~\cite{nucrin} below 5~GeV. We modified DPMJET-III as in Ref.~\cite{hkkms2006} to reproduce the experimental muon spectra better, mainly using the data observed by BESS group~\cite{BESSTeVpHemu}. In a recent work~\cite{hkkm2011}, we introduced JAM interaction model for the low energy hadronic interactions. JAM is a nuclear interaction model developed with PHITS (Particle and Heavy-Ion Transport code System)~\cite{phits}. In Ref.~\cite{hkkm2011}, we could reproduce the observed muon flux at the low energies at balloon altitude with DPMJET-III above 32 GeV and JAM below that better than the combination of DPMJET-III above 5~GeV and NUCRIN below that. Besides the interaction model, we have also improved the calculation scheme according to the increase of available computational power, such as the "virtual detector correction" introduced in Ref.~\cite{hkkms2006} and the optimization of it in Ref.~\cite{hkkm2011}. The statistics of the Monte Carlo simulation is also improved at every step of the work.

We present the calculation of the atmospheric neutrino fluxes for the neutrino experiments proposed at INO, South Pole and Pyh\"asalmi. Neutrino fluxes have been obtained using ATMNC, a simulation code for cosmic ray in the atmosphere. Even using the same primary flux model and the interaction model, the calculated atmospheric neutrino fluxes are different for the different sites due to the geomagnetic field. The prediction of these fluxes in the present paper would be quite useful in the experimental analysis.