A method for determining reservoir architecture using modeling of a non-equilibrium distribution of at least one analyte in reservoir fluids. The analyte(s) of the analysis preferably has (have) significant compositional variation in the reservoir. For example, the analyte can be a later charging single gas component (such as methane, carbon dioxide, or hydrogen sulfide) in a multi-component fluid system. In this case, the model can assume that the components of the early charge are in a stationary state or in equilibrium, whereas the later charge is in a state of non-equilibrium. The non-equilibrium distribution of the analyte(s) derived from the model is compared to the distribution of the analyte(s) derived from downhole or laboratory fluid analysis of reservoir fluid, and the architecture of the reservoir is determined based upon such comparison.