We present a simple stochastic-dynamical approach to estimate the turbulent vertical diffusivity (D) in the lower stratosphere from routinely observed ozone profiles. First, an observed ozone profile is reconstructed using three-dimensional back trajectories obtained from analyzed winds and initializing the trajectories with ozone values from the output of a chemistry-transport model. Assuming that diffusion in the vertical follows a simple random walk leading to a Gaussian probability distribution for the particle displacements, we perform Monte Carlo simulations with ensembles of particles originating along each point in the vertical profile. By choosing different values of D as input in the calculations, we generate different profiles that are smoothed through diffusion. Comparing with the observed profile, we can identify that value of D which is in best agreement at an intermediate range of vertical length scales as an upper bound of the actual D. For northern midlatitude lower stratospheric conditions during winter over a period of 12 days, the best estimate is D ≈ 0.1 m2 s-1 or slightly larger. The present results are discussed in the context of comparable estimations of vertical diffusivity in the literature.