In this paper we study the mean and standard deviation of concentrations using random walk models. Two-particle models that take into account the space correlation of the turbulence are introduced and some properties of the distribution of the particle concentration are studied. In order to reduce the CPU time of the calculation a new estimator based on reverse time diffusion is applied. This estimator has been recently introduced by Milstein et al. (Bernoulli 10(2):281-312, 2004). Some numerical aspects of the implementation are discussed for relatively simple test problems. Finally, a realistic application to predict the spreading of the pollutant in the Dutch coastal zone is described.