Abstract Underground combustion operations were initiated in a 60-acre Bartlesville sand "shoe-string" reservoir in Allen County, Kans., in 1956. Tests in separate patterns were conducted using various combinations of air and recycle gas to propagate combustion fronts from the injection toward the producing wells. These patterns were made up of 6 injection and 20 producing wells. Gas and liquid production from each pattern was measured on an individual well basis, and comparisons were made between the three patterns to ascertain the relative elects of injected gas composition on production behavior. Breakthrough of the combustion front at a well was characterized by an increase in water production from the well followed by an increase in bottom-hole temperature to approximately 250F. After burning fronts had broken through at five producing wells, operations were terminated in 1960. From the total project approximately 79,000 bbl of oil were produced during thermal operations at a cumulative produced GOR of 23 Mcf/bbl. No appreciable change in the character of the produced crude was observed. Combustion in the reservoir was maintained with injected gas compositions ranging from 6 per cent oxygen in recycle gas to 100 per cent air. Injection of large quantities of recycle gas resulted in higher producing GOR's from offset wells than were measured from a pattern into which straight air was injected. The air required to move the combustion front through 1 acre-ft of reservoir was computed to be 20 MMscf. This value was found to be relatively independent of the quantities of recycle gas injected. The recovery efficiency from the swept area was estimated to be about 59 per cent. Areas swept were similar in shape to those obtained with a laboratory potentiometric model. Samples of sand taken from behind the burning front by coring indicated almost total oil removal from the sand. Petrographic analysis of the core samples indicated that the sand had been heated to a peak temperature of about 1,200F. No significant difference in peak temperature was found in two areas where compositions of injected gas were quite different. Compression costs for thermal recovery were estimated to be $1.20/bbl of produced oil. Introduction The use of the "forward combustion" process as an oil recovery method has received a great deal of attention. This method involves ignition of the formation in an injection well, followed by propagation of a combustion front through the reservoir. Combustion is maintained by the injection of an oxygen-containing gas to react with reservoir hydrocarbons. As the flame front progresses through the reservoir, oil and formation water are vaporized, driven forward in the gaseous phase and recondensed in the cooler part of the formation. In turn, the condensed fluids push oil into the producing wellbores. Completed field tests of the process were first reported by Kuhn and Koch, and by Grant and Szasz. Results from other tests have since been reported by Walter, by Moss, white and McNeil, and by Gates and Ramey. Each of these tests essentially utilized a single injection well surrounded by four or more producing wells. Sinclair Research, Inc., elected to do field experimental work using a number of test patterns in a single field in order that comparisons between various operating schemes could be made. The site selected and purchased in 1955 for this experimental work was a 60-acre Bartlesville sand reservoir located in Allen County, Kans. Combustion operations were initiated in mid–1956. Between that time and termination of the project in mid–1960, combustion fronts were propagated from injection wells to producers in three separate well patterns, using different mixtures of air and recycle gas. The test was terminated before sweep of the three patterns was complete so that information about the effect of combustion on the swept areas could be obtained by coring. Results from the test in the form of injection and producing well performance have been carefully recorded, and these form the general basis for this paper. DESCRIPTION OF RESERVOIR The reservoir in which the combustion tests were conducted is a Bartlesville sand "shoe-string", typical of a number of small reservoirs in Southeastern Kansas. Average reservoir characteristics are shown in Table 1. Fig. 1 is an isopachous map of the producing sand showing the reservoir to be approximately 400-ft wide and 2,500-ft long. Maximum net productive sand thickness is 21 ft. Fig. 2 shows a typical core analysis obtained by coring with water-base mud. The reservoir has no appreciable dip and is closed on the sides by degradation of sand into shale. The main body of sand is heavily laminated with shale stringers, which are not continuous between wells. The main reservoir is overlain by 30 to 40 ft of laminated low-permeability sand and shale streaks. JPT P. 671^