Ionospheric scintillations, in particular at equatorial and polar latitudes, are responsible of GNSS receiver loss of lock and accuracy decreasing. To improve the future GNSS systems, an understanding of the effect of the ionospheric irregularities is necessary, completed by an accurate propagation modeling across this layer. The inhomogeneous ionospheric layer responsible for the scintillation effects is classically represented by its spectral density function (or spectrum) in propagation modeling. The inhomogeneity spectrum is anisotropic due to Earth magnetic field influence and to induced ionospheric currents [6]. Propagation across this inhomogeneous layer can be modeled by asymptotic methods based on Rytov theory when the ionospheric turbulence is weak [8][9], and by numerical approach as Parabolic Wave Equation (PWE) resolution associated with Multiple Phase Screen (MPS) [5][3]. The PWE-MPS technique is valid in strong scattering regime and can consider a variability of ionospheric turbulence characteristics along the path. As PWE-MPS technique in 3D can be time and memory space consuming, some authors assume a dimensional reduction of the problem from 3D to 2D [1]. The latter is assumed to be valid in equatorial region, where the irregularities are highly elongated along the earth magnetic field, mainly perpendicularly to the Line Of Sight direction for earth satellite links [2][7]. Nevertheless, the validity of this approximation for other configurations, as for instance in polar region where the morphology of irregularities is different, is still open. This paper proposes to quantify the consequences of the dimensional reduction on the prediction of log-amplitude and phase variances from 2D numerical schemes.