We have carried out a detailed analysis of radiative transfer and observational errors for the high-resolution infrared radiation sounder 2 (HIRS2) channel 9 centered in the 9.6 μm ozone absorption band. Several previous studies have shown significant differences between the total column ozone derived from HIRS2 and the ultraviolet (UV) radiometers. Here we use collocated ozone profiles derived from the solar backscatter ultraviolet (SBUV) spectrometer to isolate errors in HIRS2 channel 9. Radiative transfer in the 9.6 μm band is complicated because it is affected by atmospheric ozone, temperature, humidity, and the surface skin temperature and emissivity. We examine the accuracy of three fast radiative transfer algorithms. We validate current models for weak water vapor continuum absorption at 9.6 μm. In addition, we develop a method to correct for errors in older continuum models, in ozone transmittances, and in the channel 9 spectral response function. We also identify errors resulting from the spectral dependence of the surface emissivity and propose a correction method. Using UV-derived reflectivity, we have detected instances where IR cloud-detection and cloud-clearing algorithms have apparently failed. Our results show that if appropriate corrections and quality control are applied, it is possible to compute HIRS2 channel 9 brightness temperatures with an root-mean-square (RMS) accuracy of better than ~0.6 K and a bias less than 0.1 K in clear skies and RMS of better than ~0.9 K and a bias less than 0.1 K in partially cloudy skies. We plan to use the methods described here to improve IR ozone retrieval algorithms. ¿ 1998 American Geophysical Union