Large-scale analysis of cirrus clouds from AVHRR data: Assessment of both a microphysical index and the cloud-top temperature

An algorithm that allows an automatic analysis of cirrus properties from Advanced Very High Resolution Radiometer (AVHRR) observations is presented. Further investigations of the information content and physical meaning of the brightness temperature differences (BTD) between channels 4 (11 mu m) and 5 (12 mu m) of the radiometer have led to the development of an automatic procedure to provide global estimates both of the cirrus cloud temperature and of the ratio of the equivalent absorption coefficients in the two channels, accounting for scattering effects. The ratio is useful since its variations are related to differences in microphysical properties. Assuming that cirrus clouds are composed of ice spheres, the effective diameter of the particle size distribution can be deduced from this microphysical index. The automatic procedure includes first, a cloud classification and a selection of the pixels corresponding to the envelope of the BTD diagram observed at a scale of typically 100 x 100 pixels. The classification, which uses dynamic cluster analysis, takes into account spectral and spatial properties of the AVHRR pixels. The selection is made through a series of tests, which also guarantees that the BTD diagram contains the necessary information, such as the presence of both cirrus-free pixels and pixels totally covered by opaque cirrus in the same area. Finally, the cloud temperature and the equivalent absorption coefficient ratio are found by fitting the envelope of the BTD diagram with a theoretical curve. Note that the method lends to the retrieval of the maximum value of the equivalent absorption coefficient ratio in the scene under consideration. This, in turn, corresponds to the minimum value of the effective diameter of the size distribution of equivalent Mie particles. The automatic analysis has been applied to a series of 21 AVHRR images acquired during the International Cirrus Experiment (ICE'89). Although the dataset is obviously much too limited to draw any conclusion at the global scale, it is large enough to permit derivation of cirrus properties that are statistically representative of the cirrus systems contained therein. The authors found that on average, the maximum equivalent absorption coefficient ratio increases with the cloud-top temperature with a jump between 235 and 240 K. More precisely, for cloud temperatures warmer than 235 K, the retrieved equivalent absorption coefficient ratio sometimes corresponds to very small equivalent spheres (diameter smaller than 20 mu m). This is never observed for lower cloud temperatures. This change in cirrus microphysical properties points out that ice crystal habits may vary from one temperature regime to another It may be attributed to a modification of the size and/or shape of the particles.