Flow unit classifications can be used in reservoir characterization and modelling of heterogeneous carbonate reservoirs where there is uncertainty and variability in the distribution of porosity and permeability. A flow unit classification requires the integration of geological and petrophysical data, together with reservoir engineering and production data. In this study, cores and thin sections from the upper part of the Cretaceous Sarvak Formation at the Dehluran field, SW Iran, were studied to identify flow units which were then used in reservoir modelling. Eight flow units were defined based on a classification of depositional environments and diagenetic processes and an evaluation of porosity and permeability. In lagoonal deposits, two flow units were distinguished in terms of dissolution effects (i.e. low or high values of vuggy porosity). In shoal/reef deposits, three flow units were distinguished in terms of cementation volumes and grain frequency. In open-marine deposits, two flow units were identified with different degrees of dissolution; while intrashelf basinal deposits were characterized by a single flow unit with no observable reservoir potential. Each flow unit was characterized by unique values of porosity, permeability, water saturation and pore throat distribution. Grain-supported deposits from high energy depositional environments (shoals) had the highest porosities and permeabilities. However, these rocks were frequently cemented with a consequent reduction in porosity and permeability. By contrast, low permeability mud-supported deposits had undergone dissolution, forming highly permeable flow units. Capillary pressure curves from mercury injection were used to determine the distribution of pore throat sizes and the pore characteristics of the flow units, and were used to give an indication of the productivity of each flow unit. Flow units were modelled using a pixel-based modelling tool. Modelled reservoir characteristics were mainly controlled by facies changes in the vertical direction, and by diagenetic variations in the horizontal direction. Input values for the geometry of the flow units were based on information from geological and diagenetic models of the reservoir, and from thickness maps of the flow units derived from well data.