Stochastic models of petroleum reservoir geological attributes are used in reservoir studies to: (i) generate effective reservoir properties at the reservoir gridblock scale; and (ii) assess uncertainty in reservoir performance forecasting. The present paper formalizes the methodology in terms of transfer functions and introduces an alternative implementation of the sequential indicator simulation algorithm based on relative indicator variables. In addition, the determination of effective block permeabilities from stochastic images of point support-scale permeability fields is presented in the context of generalized power averages. Applications of the above are demonstrated in the simulation of reservoir lithofacies and gridblock permeabilities. The effects of stochastic imaging and reservoir characterization in assessing reservoir forecasting are illustrated.