An overview of mesoscopic models suitable for studying fluid-surface interactions is given. These models may be continuum or discrete and stochastic or deterministic. A spatially adaptive coarse-grained Monte Carlo method is introduced as a discrete, stochastic simulation tool that can reach large length and time scales with high accuracy. This method is applied to diffusion-reaction in a microporous membrane. It is shown that aside from concentration profiles and rates, coarse-graining gives the correct thermal fluctuations, an essential aspect for noise-controlled phenomena such as nucleation and nonlinear dynamics. An illustrative example of nanopatterning in reacting systems is provided to highlight the effect of noise at the mesoscale. Finally, it is shown that mesoscopic models are accurate even for nonlinear chemistry, which induces very short-range correlations, when diffusion is sufficiently fast. (C) 2004 Elsevier Ltd. All rights reserved.