Rainfall fields exhibit extreme variability over wide range of space-time scales which make them complex to characterize, model and even measure. Another basic feature, which is also observed for most geophysical fields, is a strong anisotropy. Fortunately, scaling anisotropy has been developed for a few decades to generalise scaling in an anisotropic framework, e.g., in the simplest case iso-surfaces become self-affines ellipsoids instead of self-similar spheres. This is particularly straightforward for continuous in scale cascades (Schertzer and Lovejoy, 1987). For them, as well as for discrete in scale cascades, Universal Multifractals (UM) have been widely used to analyse and simulate such geophysical fields with the help of a very limited number of physically meaningful parameters. In order to remain in the simple framework of discrete cascades while partly overcoming their well know and often neglected no translation invariance issue, blunt cascade were introduced (Gires et al. 2020, 2023). It basically consists in geometrically interpolating over moving windows the multiplicative increments at each cascade steps. The size of the moving window in tailored according to the cascade step to remain in a scale invariant framework. Here we suggest to incorporate observed anisotropic features in 2D and 3D (space and time) blunt discrete cascade simulations. The first step consists in characterizing anisotropy in UM framework. This is achieved by performing a 1D analysis along various directions, considering each column as a different “sample” of the process. Such methodology is implemented on high resolution space-time rainfall data collected with help of a dual polarisation X-band radar operated by HM&Co-ENPC. Changes in UM parameters with the angle of the chosen direction are observed. These variations are more or less pronounced depending on the intrinsic anisotropy of the studied time steps. In a second step, anisotropy features are incorporated into blunt extension of discrete UM cascades simulations. This is tentatively done by using moving window shaped as ellipses instead of squares. It was found that tuning the eccentricity and orientation of the ellipses enables to introduce various levels of anisotropy within the stochastically simulated fields and to retrieve with a good level of approximation the multifractal behaviour previously observed on actual rainfall data. Applications to downscaling of rainfall field will finally be discussed. Authors acknowledge the RW-Turb project (supported by the French National Research Agency - ANR-19-CE05-0022), for partial financial support.