This paper aims to propose a new method to generate turbulent fluctuations of wind velocity and scalar such as temperature and pollutant based on the Cholesky decomposition of time-averaged turbulent fluxes tensor of momentum and scalar. The artificially generated turbulent fluctuations satisfy not only the prescribed profiles of turbulent fluxes of wind and scalar but also the prescribed spatial and time correlations. Following the method proposed by Xie and Castro (2008), two-dimensional random data are filtered to generate a set of two-dimensional data with the prescribed spatial correlation. Then, these data are combined with those from previous time step by using two weighting factors based on an exponential function. The method was validated by applying it to a LES computation of contaminant dispersion in a half channel flow.