Abstract To interpret and optimize field performance of in-situ combustion, dominant chemical and physical mechanisms comprising the displacement process must be understood, and the influence of reservoir characteristics and operating procedures in determining swept areas of the reservoir must be established. This paper addresses important aspects of these factors via a numerical simulation study. Vaporization of the oil, changes in oil volatility due to coking, and changes in oil properties due to low temperature oxidation are shown to have considerable impact and need to be adequately modelled. The extent of fingering of the burned region due to high permeability layers and the dependence of fingering on vertical permeability are investigated. Various operating procedures that are evaluated include: reinjection of exhaust gases procedures that are evaluated include: reinjection of exhaust gases rich in carbon dioxide, simultaneous water-oxygen injection to control override, post-burn waterflooding, and selection of well spacing. Introduction Prediction of in-situ combustion field performance requires: Prediction of in-situ combustion field performance requires:understanding the relative importance of the constitutive chemical and physical processes that determine oil displacement in those areas of the reservoir that are swept, andmarket opportunities for the application of energy technologies and systems. understanding how fluid flow and reservoir characteristics determine swept volume. Although researchers have derived much information from one-dimensional combustion tube studies, results cannot be scaled up to predict field performance. Due to differences in length and time scales and difficulty in reproducing reservoir conditions, the relative importance of various mechanisms in a combustion tube is not the same as in a reservoir. Although bench-scale experiments have been used to investigate individual processes known to occur, the extent to which each process can affect field results, such as oil produced or oxygen required, remains to be determined. Although numerous field studies have been performed, results are far from comprehensive, reservoir peculiarities cloud results, and extrapolation of results to other fields is difficult. Numerical simulation programs have been developed to help understand and predict in situ combustion performance. The cost of running a predict in situ combustion performance. The cost of running a simulator that contains all applicable physics, however, is prohibitive, and much required experimental input is unavailable. prohibitive, and much required experimental input is unavailable. The goal of the present investigation is to determine which physical, chemical, and fluid flow processes can significantly physical, chemical, and fluid flow processes can significantly affect oil production, oxygen consumption, and time required for in-situ combustion. Such information will facilitate execution and interpretation of field projects and help determine which phenomena should be included in in-situ combustion reservoir simulators and/or deserve additional investigation. To do this, oil production via in-situ combustion was phenomenologically examined to determine which processes are occurring. To determine the relative importance of these processes, simulations have been performed, using ARCO Oil and Gas Company's in-situ combustion reservoir simulator, in which comparisons have been made between predicted field performance with and without the effects of each process. Processes addressed include vaporization of the oil, coking, low Processes addressed include vaporization of the oil, coking, low temperature oxidation, exhaust gas re-injection, gas override and possible control via water injection, and instabilities due to possible control via water injection, and instabilities due to vertical variations in horizontal permeability. Using this approach, an improved understanding of in-situ combustion has emerged. The current findings are for a light oil subjected to high oxygen concentration injection gases; generalization of the results, however, is expected to be straightforward.