The importance of the flow unit approach to reservoir description has been recognised recently, but its application to predict porosity, permeability and water saturation from well logs has not been attempted in previous studies. This Paper describes a genetic approach to reservoir description, which combines lithofacies analysis with discriminant analysis and probability field simulation for the identification and characterisation of flow units on the basis of core and log data. Lithofacies with distinct depositional, diagenetic and petrophysical characteristics, which essentially act as lithohydraulic flow units, have been identified from cores. A set of discriminant functions is then computed using log data from cored wells to identify lithofacies from wireline logs in uncored wells. Each lithofacies has been found by regression analysis to possess a distinct porosity and permeability relationship. The lithofacies-specific relationships between sonic travel time and core porosity is also established by regression analysis. Porosity and permeability values predicted from regression analysis lack variability when compared to actual core data. Hence, probability field simulation is applied to add fine-scale variation to the values predicted from regression analysis. The techniques described here can be applied to any type of reservoir. The application of these techniques has resulted in an improved prediction of porosity, permeability and water saturation for a shaly, glauconitic reservoir in the North West Shelf area of Australia, where traditional log analysis has been proved to be difficult to apply.