A technique is presented for deriving the column abundance of O relative to N2 (designated as O/N2 and referenced to an N2 depth of 1017 cm-2) from DE 1 imager dayglow data. For the band-pass filter of interest, the data are dominated by O I 130.4 nm with minor contributions from O I 135.6 nm and N2 Lyman-Birge-Hopfield. The technique relies on a first-principles calculation of the intensity I1P for the given band pass. Independent variables are solar zenith angle S, look angle D, and O/N2. There is a degree of nonuniqueness between I1P and O/N2(~10%), which is accounted for in our error analysis. A bias exists between I1P and a given data set owing to errors in cross sections, the assumed solar EUV flux, and calibration. The bias is removed by scaling I1P to agree with quiet time data as close in time as possible with the data set of interest. A comparison of the behavior of I1P versus S and D is made with the empirical model of Craven and colleagues [Craven et al., 1994; Nicholas et al., 1997; Immel et al., 1997>, which shows satisfactory agreement for S>40¿ but some significant departures at smaller angles. Sequences of O/N2 images over 2-day periods are presented for geomagnetic storms commencing on days 267 and 295, 1981. Both storms produced major O depleted patches at midlatitudes to high latitudes, with extreme values of O/N2 more than a factor of 2 smaller than undisturbed values. The results also show longitudinal motions of the patches over the reporting periods. ¿ 1999 American Geophysical Union