This study was undertaken to develop a method to reliably predict the performance of high relief carbonate reservoirs in the Rainbow Field (Alberta) -when produced by gravity controlled internal and external gas drives. These processes were invest-igated in the laboratory and a practical way of hand-ling the results for field predictions has been established for the external gas drive. For a medium GOR oil (800 scf/STB), primary depletion resulted in a complex diphasic flow with thermodynamical exchanges, which immiscible num-erical simulations were unable to match. INTRODUCTION Recent progress in laboratory technology has resulted in development of laboratory experiments which tend to reproduce the actual field conditions: (a) Actual core samples assembled in such a fashion as to make up a physical model representative of the reservoir under-study; (b) Reservoir fluids at reservoir pressure and temperature; (c) Fluid velocity in model similar to velocity in reservoir. Due to the low velocities and the complexity of the technology, the duration of such exper-iments ranges from several weeks up to about one year. The data f rom each experiment are analyzed with a numerical model in the same manner as for an actual field history. Since the porous media and fluid. properties and boundary conditions are well known, the adjusting parameters concern only the physics of the diphasic flow. They are the basic parameters which also control the productic,n mechanisms in the actual reservoir.Two experiments were conducted to simulate the gas pressure maintenance process and one the primary depletion process. Since gravity seg-regation controls the production in the high relief Rainbow reservoirs, all the experiments were perf'onned vertically with the ldroduction fluid withdrawr., from the bottom of the model. PHYSICAL MODELLING OF THE GAS PRESSURE MAINTENANCE (G.P.M.) PROCESS (a) EQUIPMENT AND PROCEDURE Figure 1 presents a diagram of the apparatus used in this study. Two physical models were used: a short model