Wildfire propagation is a non-linear and multiscale system in which multiple physical and chemical processes are involved. One critical mechanism in the spread of wildfires is so-called fire-spotting: a random phenomenon that occurs when embers are transported over large distances by the wind, causing the start of new spotting ignitions that jeopardize firefighting actions. Due to its nature, fire-spotting is usually modelled as a probabilistic process. In this work, the physical parametrization of fire-spotting RandomFront has been implemented into the operational wildfire spread simulator PROPAGATOR, which is based on a cellular automata approach. In the RandomFront parametrization, the downwind landing distribution of firebrands is modelled by the means of a lognormal distribution, which is parameterized taking into account the physics involved in the phenomenon. The considered physical parameters are wind field, fire-line intensity, fuel density, firebrand radius, maximum loftable height, as well as factors related to atmospheric stability and flame geometry. The results are compared against an already established fire-spotting empirical submodel for cellular automata-based wildfire models. Preliminary results show that the RandomFront parametrization on the one hand reproduces the main spotting effects given by the available literature model, while on the other hand, it generates a variety of fire-spotting situations as well as long range fluctuations of the burning probability. The physical parametrization allows for complex patterns of fire spreading in this operational simulator context.