We report ground-based microwave measurements of chlorine monoxide (ClO) mixing ratio profiles above Scott Base, Antarctica (latitude=77.85 ¿S), during the late winter and early spring of 1996, 1997, and 1998 and analyze in detail the 1996 data. The instrument is permanently installed at the site, permitting continuous measurements throughout the ozone hole period, starting before the end of the polar night. The development of the lower stratospheric ClO layer is strikingly similar in all three years, with the peak (day-night) ClO mixing ratio at 22 km rising from 0.1 to 1.9--2.0 ppbv in the 21 days from August 13 to September 4. The altitude of the peak drifts downward from 22 to 17--18 km by the end of September. In 1997 and 1998 the chlorine remained in active species until about September 29, while in 1996 there was a rapid and sharp decline in ClO beginning on September 12. We compare the daily average ClO profiles for 1996 with the output of a three-dimensional off-line chemical transport model. The model output is sampled to match the data and is used as an a priori profile for retrieval of the measured profiles from the data, allowing a rigorous comparison of the consistency between the model and measurements without bias from an a priori. The model captures all of the major features of the measurements, including the rapid rise in August, the sharp drop on September 13 due to the arrival of warmer, ClOx-poor air over Scott Base, and the deactivation of ClOx into the reservoir species, HCl and ClONO2, during the last half of September. However, the model using recommended rate coefficients consistently underestimates the measured ClO, suggesting that the ratio of the ClO dimer photolysis rate to the dimer formation rate, j/kf, needs to be increased by a factor of 1.8¿0.3. This has important implications for model calculations of ozone loss. ¿ 2000 American Geophysical Union