Cloud amounts derived from Landsat digital data over 22 regions with various cloud types are compared to cloudiness derived for collocated, nearly simultaneous 4¿8 km GOES visible and infrared data using a hybrid bispectras threshold method (HBTM). The mean and rms differences between the Landsat and GOES cloud amounts are -0.008 and 0.075, respectively. A sensitivity analysis showed that underestimation of approximately 0.01 in clear-sky reflectance by the HBTM increased the mean GOES cloud amount by 0.06, more than twice the decrease in cloud amount obtained by an equivalent increase in clear-sky reflectance. Accuracy limits were estimated for the Landsat analyses by considering the fraction of cloud-edge pixels in a scene. For scenes having relatively homogeneous backgrounds, the Landsat cloudiness has an estimated accuracy of -0.02 to +0.04, while cloud amount in scenes with variable background reflectance is more susceptible to underestimation, having accuracy limits of -0.02 to +0.12. The lower bound for the Landsat cloud cover is reliable when considered in terms of most working definitions of cloud cover.

The upper bound will remain uncertain until an absolute definition of cloud cover is established which sets the bounds on the optical properties which dictate when a cloud does or does not exist. Many of the scenes contained a substantial proportion of partially cloud-filled GOES pixels. Landsat imagery and cloud properties derived from the Landsat data are used to explain how the partially cloud-filled GOES pixels were treated by the HBTM. Most of the rms difference occured over two regions: one where many of the GOES pixels were partially cloud-filled with very bright cloud cells and one where the GOES pixels were mostly overcast with both bright (thick) and very dim (thin) clouds. It is concluded that the HBTM accounts for the effects of partially cloud-filled fields of view in most of the cases in this study. The reflectant limits used to constrain the HBTM appear to be reasonable but cannot account for the effects of anomalously bright or dark cloud fields. Techniques are discussed for improving the HBTM's retrieval of cloud fraction in the more difficult cases. This study, though only the beginning of a validation of the HBTM, demonstrates the potential for using Landsat data to improve routine cloud parameter retrievals from the low-resolution sensors on meteorological spacecraft. Since the scenes used in this study represent only a limited number of cloudiness conditions and viewing and illumination situations, there remains much more work ahead in order to completely validate the HBTM or any other satellite cloud retrieval algorithm. Future cloud studies using Landsat should include both visible and infrared channels. ¿ American Geophysical Union 1988