For studies of Antarctic climate change, the Advanced Very High Resolution Radiometer (AVHRR) offers a time series spanning more than two decades, with numerous overpasses per day from converging polar orbits, and with radiometrically calibrated thermal infrared channels. However, over
the Antarctic Plateau, standard multispectral application of AVHRR data for cloud optical property retrieval with individual pixels is problematic due to poor scene contrasts and measurement uncertainties. We present a method that takes advantage of rapid changes in radiances at well-defined
cloud boundaries. We examine a transect of AVHRR-measured radiances in the three thermal infrared channels across a boundary between cloudy and cloud-free parts of the image. Using scatter diagrams, made from the data along this transect, of the brightness temperature differences between channels
3 and 4, and channels 4 and 5, it is possible to fit families of radiative transfer solutions to the data to estimate cloud effective temperature, thermodynamic phase, and effective particle radius. The major approximation with this method is that along such a transect, cloud water path has
considerable spatial variability, while effective radius, phase, and cloud temperature have much less variability. To illustrate this method, two AVHRR images centred about the South Pole are analysed. The two images are chosen based on their differing contrasts in brightness temperature between
clear and cloud-filled pixels, to demonstrate that our method can work with varying cloud top heights. In one image the data are consistent with radiative transfer simulations using ice cloud. In the other, the data are inconsistent with ice cloud and are well simulated with supercooled liquid
water cloud at 241.5 K. This method therefore has potential for climatological investigation of the radiatively important phase transition in the extremely cold and pristine Antarctic environment.